Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement IX Dr.-Ing. Carsten Hoff Dipl.-Ing. (Univ.) Ottmar Sirch (Hrsg.) und 71 Mitautoren Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement IX Haus der Technik Fachbuch Band 148 Herausgeber: Prof. Dr. Werner Klaffke · Essen Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http: / / dnb.dnb.de abrufbar. © 2019 · expert verlag GmbH Dischingerweg 5 · D-72070 Tübingen Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Alle Informationen in diesem Buch wurden mit großer Sorgfalt erstellt. Fehler können dennoch nicht völlig ausgeschlossen werden. Weder Verlag noch Autoren oder Herausgeber übernehmen deshalb eine Gewährleistung für die Korrektheit des Inhaltes und haften nicht für fehlerhafte Angaben und deren Folgen. Internet: www.expertverlag.de eMail: info@verlag.expert Printed in Germany ISBN 978-3-8169-3464-6 (Print) ISBN 978-3-8169-8464-1 (ePDF) Haus der Technik Fachbuch Herausgeber der Reihe Prof. Dr. Werner Klaffke Geschäftsführendes Vorstandsmitglied des Hauses der Technik e.V. Die Konkurrenzfähigkeit einer rohstoffarmen Volkswirtschaft hängt ganz wesentlich vom Faktor „Wissen“ ab. Verbunden mit kreativem Gestaltungswillen wird aus Wissen Kompetenz. Kompetenzvermittlung ist der zentrale Aspekt des Hauses der Technik, die weit über 80 Jahre schon praxisorientiert und disziplinenüberschreitend durch Tagungen, Symposien, Seminare und Workshops qualitativ hochstehend dargestellt wird. Damit arbeiten wir an den Grundlagen für neue Produkte und Dienstleistungen, deren Vermarktung zu Innovationen und damit zu Wertschöpfung führen. Mehr als 70% der erfolgreichen Innovationen, ob inkrementell oder radikal, entstehen aus der Verknüpfung häufig bereits bekannter Elemente, weshalb es geradezu essentiell ist, akademische Schubladen zu verlassen und die Elemente der Kompetenzen intelligent und bedarfsorientiert zu kombinieren. Das geschieht in branchenübergreifenden Innovationsnetzwerken und Technologieclustern, die sich in neuen Wertschöpfungsketten zusammenfinden. Neue Elemente der Netzwerkbildung belebt durch die zunehmende Digitalisierung der Arbeitswelt gesellen sich zu den traditionellen Informationsquellen, zu denen auch die vorliegende Publikation gehört. Die bewährten Haus der Technik Fachbücher befassen sich mit den wichtigen Themen der Technik, der Wirtschaft und angrenzender Gebiete, wie Medizintechnik, Biotechnik und neue Medien. Das Beste, das oft mühsam und mit viel Aufwand von den Veranstaltungsreferenten zusammengetragen wurde, wird damit einem größeren Fachpublikum zugänglich gemacht. Die Haus der Technik Fachbücher dienen den Teilnehmern als nützliches Nachschlagewerk und anderen Interessenten beim Selbststudium zu beruflichem Nutzen und Erfolg. Contents Preface.....................................................................................................1 Carsten Hoff, Ottmar Sirch Overview................................................................................................37 Magnet-Free Electric Machines & Drives for Electric Vehicles .............. 37 John Grabowski, Andrea Colognese A study of semiconductor and sensors for vehicle electrification ........ 47 Dexin Chen, Richard Dixon Electric Charging - Components, Systems and Infrastructure........54 High Power Charging - Consideration of the cost parameters of fast charging from vehicle to charging infrastructure ........................ 54 Sebastian Rickert Funded Project for smart charging infrastructure for EVs up to 22kW .................................................................................................. 62 Detlev Endner, Michael Kahlstatt, Roland Matthé Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen ...................................................................................... 70 Uwe Hauck, Michael Leidner, Michael Ludwig, Helge Schmidt, Marco Wolf 48V Power Supply Architectures and Components ..........................83 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs................................................................................... 83 Julian Taube, Laurenz Tippe, Joachim Fröschl, Hans-Georg Herzog Enabling Technologies.........................................................................95 Enabling Technologies - die attach and substrate technologies for power electronics ....................................................................................... 95 Louis Costa Future Packaging Technologies in Power Electronic Modules ............ 107 Christoph Friedrich Bayer, Zechun Yu, Hoang Linh Bach, Jonas Müller, Andreas Schletz Contents Li-Ion Batteries in 48V Vehicles.........................................................127 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle.... 127 Guoqiang Ao, Lilly Han Power Electronics System Engineering ...........................................140 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles ............................................................ 140 Ayman Ayad, Martin Brüll, Andreas Greif, Sebastian Rogge, Matthias Töns Voltage Ripple on Electric Vehicle System Test benches .................... 154 Jozsef Gabor Pazmany, Samuel Siegel, Klaus Rechberger, Bernard Bäker Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles................................... 168 Andreas Bendicks, Tobias Dörlemann, Stephan Frei, Norbert Hees, Marc Wiegand Safety of Vehicle Power Supply ........................................................175 Ring Structures in Automotive Power Nets: Idea and Implementation.......................................................................... 175 Laurenz Tippe, Julian Taube, Joachim Fröschl, Hans-Georg Herzog Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen................................. 185 Christian Sültrop, Thomas Lang, Franz Rohlfs, Jan Helfrich, Pedro Bossay de Almeida Nogueira, Helga Weber, Uwe Prüfer, Dominik Bergmann, Jürgen Gebert Towards reliable power supply for highly automated driving .............. 205 Kay Klobedanz, Stefan Grösbrink, Sebastian Kahnt Power Electronics Applications ........................................................220 High Power Density Modular Six Phase Drive Inverter ......................... 220 Niklas Langmaack, Günter Tareilus, Markus Henke Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation................................................................................................... 234 Teresa Bertelshofer, Andreas März, Mark-M. Bakran Contents Design of Reliable Power Net Systems ............................................249 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge....... 249 Janis Lehmann, Benjamin Löwer, Mohamed Ayeb, Ludwig Brabetz Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen ..................................................................... 264 Stefan Schwimmbeck, Quirin Buchner, Hans-Georg Herzog Model-Based Analysis of Transient Processes in Highly Available Automotive Energy Systems ................................................................... 284 Martin Baumann, Christoph Weissinger, Hans-Georg Herzog Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications .............................................................................................. 302 Fabian Schipperges, Stefan Schumi, Stefan Hörtling, Bernard Bäker Poster...................................................................................................316 Anwendungsmöglichkeiten von all-solid-state Zellen in Niedervolt- Autobatterien ............................................................................................ 316 Ines Miller, Varvara Sharova, Robert Stanek The Authors.........................................................................................320 Preface Last year, for the first time more than 1 million electric cars were sold in the largest automotive markets, according to the analysis of PwC. In China, USA and Europe (Germany, Great Britain, France, Spain and Italy) about 1.1 million electric vehicles were registered. This is an increase of 70%. Including Hybrid and Plug-In Hybrid vehicles, in 2018 2.6 million electrified vehicles were registered. The market share reached 2.8% at a level of 96 million vehicles world-wide. 20% growth rate of Mild and Full Hybrid vehicles was significant lower than the growth rate of Plug-In Hybrid vehicles (64%) and BEVs. (Source: PwC - Press Release 02/ 07/ 2019) Automotive Engineering of the next decade will be affected by e-Mobility, Digitalization and Autonomous Driving. Some Megatrends will directly or indirectly influence the automotive world in general and the design of future E/ E architectures:  Digitalization, Cloud, Smart Vehicle, Connected Vehicle  Mobility Services, Car2Go  Driver Assistance, Autonomous Driving, Telematics  Zero Emission, Sustainability, Electrification Drivetrains with combustion engines - gasoline and Diesel - will lose their dominant market position facing the requirements for reduction of emissions and the upcoming restrictions of access to cities or their centers. The introduction of high power recuperation based on 48 Volt mild Hybrid Systems and the electrification of ancillaries will exceed the lifetime of combustion engine vehicles, but will not assure their long-term existence. Electric drives are far superior to the combustion engine drives in terms of efficiency, comfort and dynamics. In the near future they will be the preferred drive technology for locally emission-free and exciting mobility, as soon as the capacity of high voltage batteries will allow a sufficiently high range and a fast and comfortable charging function. Hybrid drives, especially plug-in hybrids, offer a compromise in terms of range and CO 2 reduction and enable temporary emission-free driving within a restricted area. Charging systems which charge the batteries in any location and with the required energy in a short time and with high performance are one of the central factors for the success of e-Mobility. Other alternative drive concepts such as e.g. Fuel cell systems can be launched in the long term. 1 Regardless of the kind and performance of the electrification, the number of additional components which have to be integrated into the vehicle must decrease. An ongoing integration of dedicated components is inevitably necessary and unavoidable to fulfill the continuously rising requirements and to meet the future cost targets. Networking and connectivity provide an almost infinite space for new services and solutions in the vehicle and its environment and open up new business areas. 1 Today Automation is one of the most important drivers of change in the mobility industry. It has the potential to increase the safety of mobility for all road users and to permanently improve the flow of traffic. Step by step the way will be paved for automated 1 Preface up to autonomous driving - starting with driver assistance features for more safety and comfort, via partially or highly automated driving which will relieve the driver significantly, to fully automated driving and parking in a city or driving on a highway, and at least autonomous vehicles which will cruise around like robots. 2 In the next few years, automated driving will become reality. Assistance systems already help the driver to reach his destination safely and comfortably. Soon highly and fully automated vehicles will be cruising on country roads, highways and even in the more complex city traffic. 2 At least, an automated vehicle must perform even better than a human driver. First, it has to be able to identify and to understand its environment. For this purpose it uses environment sensors, just as humans use their senses. Secondly, it must process all the information gained and plan its driving strategy. This task is done by the vehicle computer using software and intelligent algorithms. And thirdly, it has to use its propulsion, steering and braking powers to move its wheels in order to turn the planned driving strategy into action. 2 (Source : 1 Continental Corporation Website, 2 Robert Bosch GmbH Website) The focus items of engineering will be charging with all its aspects for higher power and its link to a smart grid as well as the use of higher voltages like 800 Volt and the requirements of legislation. This has a high impact on Battery Management Systems with respect to State-of-Health, safety, quality, reliability and IoT (Internet-of-Things). The development of 48 Volt is still ongoing. After the recent launches of the first vehicles starting in 2017 there is enough room for further steps. Beginning with the migration of high power loads like pumps, heaters and fans from 12 Volt to 48 Volt or changing from mechanical to electrical systems, more loads may follow in the next generations of E/ E architectures, which may finally end up with a Single-Voltage Power Supply with 48 Volt in the future. And Light Electric Vehicles with low weight and low propulsion power of 20 to 30 kW may use 48 Volt instead of high voltages, too. Enabling technologies like Wide-Band Gap-Semiconductors, Packaging and Modules, Thermal Management and innovative Passive Components will step into the focus of the engineers´ interests and will be a major part of the discussion in the E/ E community. Power Electronics technologies and its systems engineering as a key issue was, is and will be in the center of all discussions. As a new focus item the E/ E architectures for reliable power nets and their design methods have entered the discussion during the last few years, too. These subjects gained more and more interest in the community. The book in hand was printed on the occasion of the 8 th conference “Electric & Electronic Systems in Hybrid and Electric Vehicles and Electrical Energy Management” organized by Haus der Technik e.V. Essen May 22 nd and 23 rd , 2019, in Bad Nauheim. Subjects and articles about the overall system of electrics/ electronics in Hybrid and Electric Vehicles and the concepts of electrical energy management will be pointed out and discussed in detail. Approaches for Hybrid, Plug-In Hybrid and Electric Vehicles and E/ E architectures, charging systems, power electronics, and low voltage accumulators will be covered. 2 Preface We want to thank all the authors and speakers who supported with their interesting technical contributions to the edition of this book. Furthermore, we thank Mr. Bernd Hömberg and his staff from Haus der Technik e.V. in Essen for planning and organizing the event, Mrs. Anita Koranyi and Mr. Robert Narr from expert verlag GmbH in Tübingen for publishing the book and Mrs. Dr. Vera Lauer, Mr. Dirk Balzer, Mr. Prof. Dr. Ludwig Brabetz, Mr. Prof. Dr. Stephan Frei, Mr. Friedrich Graf, Mr. Prof. Dr. Hans- Georg Herzog, Mr. Dr. Jan Lichtermann, Mr. Dr. Marc Nalbach, Mr. Dr. Dieter Polenov, Mr. Dr. Hartmut Pröbstle, Mr. Dr. Tomas Reiter, Mr. Prof. Dr. Dirk-Uwe Sauer, Mr. Peter Schmitz, Mr. Dr. Marc Thele, Mr. Andreas Hinderlich and Mr. Dr. Jochen Langheim for their support and engagement in the scientific advisory board and program committee. Bad Nauheim, May 2019 Ottmar Sirch Dr. Carsten Hoff 3 Preface The new Voltage Level 48 Volt for Vehicle Power Supply T. Dörsam, Daimler AG, Böblingen, S. Kehl, Porsche AG, Weissach, A. Klinkig, Volkswagen AG, Wolfsburg, München, O. Sirch, BMW Group, München, A. Radon, Audi AG, Ingolstadt ATZelektronik, January 2012. The increasing requirements to the vehicle power supply in conventional cars have taken the 12 Volt power supply to its limits. The car manufacturers Audi, BMW, Daimler, Porsche and Volkswagen were aware of this subject very early and worked out a first specification for a second voltage level at 48 Volt. The OEMs explain the conditions and backgrounds for deriving the voltage range and the concepts for the tests and test conditions for any components which will be applied in this voltage level. Additionally, the car manufacturers describe the probably next milestones for standardization and the standard of the voltage level as well as some potential E/ E architecture concepts. History The advantages of a second voltage level in vehicles have been already discussed once during the initiative for 42 Volt. At that time the initiative focused on increasing the power and enabling new functions. Aspects of CO 2 emissions played a minor role, and the existing battery technologies limited the progress. In order to achieve the maximum CO 2 reduction it is reasonable to define a system with the highest nominal voltage below the limit for the shock protection at 60 Volt. Today existing cell-based accumulators are scalable and fulfill the requirements [6]. Existing and new high power loads can be applied to this voltage level. With respect to the requirements for CO 2 reduction and supply of high power loads the 48 Volt power supply competes directly with the high voltage (HV) system (voltage class B). These HV systems are dedicated for electrical power of 15 kW upwards, definitely upon the power limits of 48 Volt systems. An evaluation of 12 Volt, 48 volt and HV systems in terms of power, CO 2 reduction and costs shows that a 48 Volt system is an interesting alternative #1. #2 shows the relation between current and electrical power at different voltages with the simplifying assumption of a constant operating voltage. Stationary currents of 200 to 300 A can only be realized with highly technical efforts due to the experience in practice. Therefore values of more than 200 A are considered as borderline and more than 300 A as very critical. Thus the maximum stationary electrical power at 12 Volt is about 3 kW, at 24 Volt about 6 kW and at 48 Volt about 12 kW. If the limit of 12 kW will not be exceeded dramatically, a voltage range below 60 Volt is reasonable [1, 2, 3]. In this case the voltage level at 48 Volt fits to the requirements according to the specification which was commonly defined by Audi, BMW, Daimler, Porsche und Volkswagen. This common specification as a delivery specification is the base for the OEM specific standards for 48 Volt voltage range [4, 5] and defines the electrical requirements for nay components which will be integrated at this voltage. A dual voltage power supply system based on the existing 12 Volt voltage range and the new 48 Volt voltage range will be developed in parallel to the recently established Hybrid or Electric Vehicle power supply systems. A minimum system consisting of an electric motor, an accumulator and coupling device as well as a more complex system with the components mentioned before and further sources and loads are conceivable. Conditions for the definition of the voltage range Some presumptions had to be made to define the voltage range and were constituted in the following assumptions, #3: 4 Preface  For DC voltages below 60 Volt a shock protection is not required. This is valid up to ripples of maximum 10% rms.  For AC voltages the requirements of ISO 6469-3 have to be considered. Up to AC voltages U eff ≤ 30 V a shock protection is not required.  Generating or regenerating components must not cause a transition into the overvoltage range.  During regular operation of the 48 Volt power supply any voltages above the shock protection limit must not occur.  There is a common ground for 12 Volt and 48 Volt. The contacts have to be locally separated.  Ground for 12 Volt and 48 Volt control units have to implemented separately.  Loss of ground of a 48 Volt component must not cause a disturbance or a destruction of the communication networks and the electrical networks  All data for voltages and currents are related to the I/ Os of the component and do not include any voltage drops caused by the wiring harnesses in the vehicle.  The base is existing Li-Ion accumulator technology (like LiFePO 4 , LiNiMnCo und Li-Titanium). The voltage range shall be valid for further and new technology developments.  There will not be a reverse polarity protection at 48 Volt. Adequate measures have to be applied in order to exclude any reverse polarity. A jump start at 48 Volt is not scheduled.  A single failure must not cause a short circuit between 48 Volt and 12 Volt (or 24 Volt) systems. Voltage Range The voltage range which is specified in the LV148 is as hereafter shown: Fig. 1: Definition of the voltage range [graph replaced] 5 Preface  Shock Protection Range: according to ECE-R 100 a shock protection for DC voltages is required.  Overvoltage Range: The overvoltage range between U 48max,high,limited and U 48r includes all tolerances. The overvoltage protection shall be active in this range and any voltages above U 48max,high,limited have to be stored in the failure memory. The range between U 48r und U 48shprotect includes a contingency reserve.  Upper Operating Range with functional Degradation: The range between U 48max,unlimited and U 48max,high,limited is dedicated for the calibration of the accumulator and the absorption od recuperation energy.  Operating Range without Degradation: The range between U 48min,unlimited and U 48max,unlimited is dedicated to operate the components without any functional degradation.  Lower Operating Range with functional Degradation: operating in the range between U 48min,low,limited and U 48min,unlimited is only temporarily allowed. Countermeasures have to be taken to return into the operating range without functional degradation.  Undervoltage Range: Voltages below U 48min,low,limited are defined as undervoltages and have to be stored in the failure memory. The accumulator protection voltage is defined at U 48stoprotect . Tests and test conditions The new specification LV148, which was defined by the car manufacturers Audi, BMW, Daimler, Porsche und Volkswagen as a common base of the OEM specific standards for components at the 48 Volt voltage range, is aligned to the existing LV124 for 12 Volt and is revised and completed in terms of presumptions und assumptions. The objective of LV148 is a standard specification of the voltage range and the required tests to achieve a maximum of industrial building blocks for components and parts [6]. The requirements, tests and test conditions for electric, electronic and mechatronic components and systems are defined for the application in vehicles at 48 Volt, #4. Approach The specification of the 48 Volt voltage range and the tests for the components which will be applied there have been developed by the car manufacturers as listed before commonly. The revision of the DIN 72552 was requested by the mentioned OEMs via VDA. For the new 48 Volt voltage level the following terminals are proposed:  “Klemme 40“ is the positive wire of 48 Volt power supply  “Klemme 41“ is ground wire of 48 Volt power supply Further steps for standardization or international standards are not explicitly planned. Outlook In order to meet the constantly increasing requirements to reduce CO 2 emission and fuel consumption in conventional vehicles with combustion engines a significant contribution can be made to by increasing the recuperation power to a range of 5 to 12 kW. The supply of the vehicle with electrical energy has to be realized without using any fuel as possible. During deceleration the combustion engines shall be switched off. The Charging and discharging periods due to the velocity profiles and the state of 6 Preface the accumulator are shown in #5. Furthermore, a sailing mode can be applied while the combustion engine may be switched off at constant speed. The electrical energy which is required for a driving cycle can only be generated by recuperation, if the electrical power is increased, respectively, because the time slot for generating is reasonably lower [6]. In addition a 48 Volt electrical motor offers the possibility to shift the operating point of the combustion engine due to its generating and driving modes. An electric slow motion may be possible depending on the design. The new requirements for higher electrical power will affect and change future power supply architectures in terms of topology, overall current consumption and functional characteristics. The independently working power supplies can pretend the interaction between the high power loads. Furthermore, the displacement of high power loads from 12 Volt to 48 Volt can be discussed. The reduction of cross-sections in the wiring harnesses offer potentials of weight. #6 shows as an example an overview of electrical systems with respect to power supply. #7 shows a possible migration of the configuration in #6. The following high power loads of the existing 12 Volt system are dedicated for a transition: The electric cooling fan with up to stationary electrical power consumption of 1 kW is the maximum load. Electric heating systems like PTC are in the same range. Additionally, the proposed voltage level enables the electrification of functions, which can be transferred from mechanical to electrical drive, like an electric climate compressor for driving operation and stand-still. The new voltage level 48 Volt offers a wide range of applications from boost functions and CO2 reduction in small vehicles up to the supply of large scale of functions in fully equipped luxury cars, so that a broad adoption can be expected. Literature: [1] O. Sirch, J. Fröschl, Dr. H. Pröbstle, Energiebordnetz und elektrisches Energiemanagement der Zukunft. 1. Braunschweiger Symposium zu Elektrische Leistungsbordnetze und Komponenten von Straßenfahrzeugen, Oktober 2008 [2] S. Wolff, O. Sirch, M. Schmid, G. Immel, H. Pröbstle, R. Neudecker, J. Fröschl: Anforderungen aus Sicht der Energiesysteme der Zukunft an das Fahrzeugbordnetz. 29. Tagung Elektronik im Kraftfahrzeug, Juni 2009 in Dresden. [3] O. Sirch, G. Immel, H. Pröbstle, R. Neudecker, J. Fröschl, Future Electrical Powernet - From Optimization Concerning Energy to a new Concept, 14. Internationaler VDI- Kongress Elektronik im Kraftfahrzeug, Oktober 2009. [4] LV148, AUDI AG, BMW AG, Daimler AG, Porsche AG und Volkswagen AG, Juli 2011. [5] T. Dörsam, D. Grohmann, S. Kehl, A. Klinkig, A. Mai, O. Sirch, A. Radon, J. Winkler, Die neue Spannungsebene 48 Volt, 15. Internationaler VDI-Kongress Elektronik im Kraftfahrzeug, Oktober 2011. [6] R. Falsett. T. Dörsam, Elektrische Energiespeicher im 48V Bordnetz, 15. Internationaler VDI-Kongress Elektronik im Kraftfahrzeug, Oktober 2011. 7 Preface Dual Voltage Power Supply System with 48 Volt J. Fröschl, H. Pröbstle, O. Sirch, BMW Group EEVC, Dresden, June 2012 and Societé des Ingénieurs d´Automobiles Magazine, March 2013 Automotive electrics/ electronics have just reached a period of tremendous change. High voltage systems for Hybrid, Plug-In Hybrid or Battery Electric Vehicles with high power electric motors, high energy accumulators and electric climate compressors will be introduced in order to achieve the challenging targets for CO 2 emissions and energy efficiency and to anticipate the mobility of the future. Additionally, innovations and the continuous increase of functionality for comfort, safety, driver assistance and infotainment systems require more and more electrical power of the vehicle power supply at all. On the one hand side electrified vehicles will certainly achieve a significant market share, on the other hand side they will increase the pressure to conventional vehicles with combustion engines for fuel consumption and CO 2 emissions. These vehicles will be enabled to keep their competiveness by new functions and the optimization of their electric systems. A dual voltage power supply with 48 Volt and 12 Volt will be one of the key technologies to realize these requirements. The power capability of the existing 12 Volt power supply has reached its limits. Further potentials can only be admitted by the introduction of 48 Volt. For this reason the car manufacturers Audi, BMW, Daimler, Porsche and Volkswagen started very early on this item and developed a common specification of the new voltage range. Now, it is necessary to indentify the probable systems at this voltage range and to start the developments. 1 Introduction 150 years ago as the automobile was invented there was a strong competition between combustion engine, electric motor and steam engine. Independently, electrification was there right from the beginning. 1769 the French Military Engineer Nicholas Cugnot invented a steam engine driven car which was planned to pull heavy ordnance. This was demonstrably the invention of an automobile, without any electrification yet. More than 100 years later the first electric vehicles appeared on the streets. Gustave Trouvé (1881) developed his electric vehicle and drove through the streets of Paris (see fig. 1). At nearly the same time, in 1882 Ayrton and Perry in England presented an electric vehicle with electric light. 8 Preface Fig. 1: Gustave Trouvé, Paris 1881 The first vehicle with a combustion engine (see fig. 2) invented by Carl Benz in 1886 had an electric ignition due to the problems which occurred at windy conditions as long as he tried to use a flame for this function. Fig. 2: Carl Benz Vehicle, 1886 The next generation of electric vehicles came up. Andreas Flocken, Coburg in Germany, built up a four wheel vehicle with an electric motor (see fig. 3) in 1888. 9 Preface Fig. 3: Andreas Flocken, Coburg 1888 In 1903 Clyde J. Coleman, USA, invented an electric cranking system as a next big step of electrification to get rid of the cranking by muscles at all weather conditions. Beside these very early electrification efforts a competition between different drive train technologies took place. At the end of the19 th century steam cars, electric vehicles and vehicles with combustion engines have been seen on the streets. Oberbaurat a. D. Klose, President of the Middle-European Association of Motor Vehicles (Mitteleuropäischer Motorwagen-Verein) stated on September 30 th , 1897: “There are 3 kinds of vehicles which carry their energy for locomotion onboard of particular importance: vehicles moved by steam, by oil engines and by electricity. The first kind will be used in the future mainly for rail-based coaches and heavy vehicles on the road, while the great range of the wide country will be passed by oil engine driven vehicles and the smooth asphalt surface of the big cities will be populated by charged-electricity driven vehicles.” („Als Motorfahrzeuge, welche ihre Energie zur Fortbewegung mit sich führen, machen sich zur Zeit drei Gattungen bemerkenswert, nämlich: durch Dampf bewegte Fahrzeuge, durch Oelmotoren bewegte Fahrzeuge und durch Elektrizität bewegte Fahrzeuge. Die erste Gattung dürfte voraussichtlich in Zukunft hauptsächlich für Wagen auf Schienen und schwere Straßen-Fahrzeuge in Betracht kommen, während das große Gebiet des weiten Landes von Oelmotorfahrzeugen durcheilt werden und die glatte Asphaltfläche der großen Städte wie auch die Straßenschiene von mit Sammlerelektrizität getriebenen Wagen belebt sein wird.“) 2 New Requirements The requirements concerning the electrical vehicle power supply in conventional cars which are mainly driven by the advancing electrification and reduction of CO 2 and fuel consumption are constantly increasing. This leads today´s established 12 Volt power supply to its limits. Additionally, measures have been implemented, to reduce the current consumption of each function, system and actuator. Thus the step to a more complex power supply 10 Preface system can be postponed as long as possible. For this step to a new architecture will affect the costs seriously and has to be balanced by significantly functional increases. Fig. 4: step to a new architecture by increasing power demands The head space, which is created by the step-by-step implementation of measures to reduce the current consumption of existing functions [9], can be used to install new functions, as shown in fig. 4. The higher the reduction is at a certain point, the higher is the potential for the installation of new functions. But if the power demand of a new function exceeds the threshold significantly, the step into a more powerful architecture cannot be avoided anymore. Current assumptions expect this step to be performed during this decade. The requirements which are affected by new functionalities like recuperation, expansion of engine start-stop, electric assist or boost as well as electrification of ancillary units. Recuperation is directly linked to the demand of electric energy which has to be cumulated during a deceleration phase of the vehicle to supply the vehicle electric system during the next phase of acceleration and constant speed. This operating strategy enables to reduce the amount of fuel to be employed to generate the electric energy. The size of the new generator is proportional to the electric power required to generate the electric energy and is limited by the maximum torque which can be provided as a function of the mechanical design. Engine start-stop functionality will be extended in the future. The engine will be switched off at higher speeds during deceleration and partially during constant speed. Subsequently the engine has to be cranked very quickly. Therefore the generator has to perform as an electric motor, too. 11 Preface The motor capability of the generator can be used in some cases to assist or boost the power train electrically in order to optimize the operating point of the combustion engine. The electrification of ancillary units like climate compressor, water pump or turbo charger is a new opportunity to get rid of fixed mechanical link between the engine and the ancillary unit. This transfer allows a tailoring on demand of the power supply and to improve the efficiency. 3 The new Voltage Range The car manufacturers Audi, BMW, Daimler, Porsche and Volkswagen were aware of this subject very early and worked out a first specification for a second voltage level at 48 Volt [5]. The OEMs explained the conditions and backgrounds for deriving the voltage range and the concepts the tests and test conditions for any components which will be applied in this voltage level. Additionally, the car manufacturers describe the probably next milestones for standardization and the standard of the voltage level as well as some potential E/ E architecture concepts. Some presumptions had to be made to define the voltage range and were constituted in the following assumptions: • For DC voltages below 60 Volt a shock protection is not required. This is valid up to ripples of maximum 10% rms. • For AC voltages the requirements of ISO 6469-3 have to be considered. Up to AC voltages Ueff ≤ 30 V a shock protection is not required. • Generating or regenerating components must not cause a transition into the overvoltage range. • During regular operation of the 48 Volt power supply any voltages above the shock protection limit must not occur. • There is a common ground for 12 Volt and 48 Volt. The contacts have to be locally separated. • Ground for 12 Volt and 48 Volt control units have to implemented separately. • Loss of ground of a 48 Volt component must not cause a disturbance or a destruction of the communication networks and the electrical networks • All data for voltages and currents are related to the I/ Os of the component and do not include any voltage drops caused by the wiring harnesses in the vehicle. • The base is existing Li-Ion accumulator technology (like LiFePO4, LiNiMnCo und Li- Titanium). The voltage range shall be valid for further and new technology developments. • There will not be a reverse polarity protection at 48 Volt. Adequate measures have to be applied in order to exclude any reverse polarity. A jump start at 48 Volt is not scheduled. • A single failure must not cause a short circuit between 48 Volt and 12 Volt (or 24 Volt) systems. The voltage range which is specified in the LV148 is shown in fig. 5: 12 Preface Fig. 5: Definition of the voltage range 48 Volt  Shock Protection Range: according to ECE-R 100 a shock protection for DC voltages is required.  Overvoltage Range: The overvoltage range between U 48max,high,limited and U 48r includes all tolerances. The overvoltage protection shall be active in this range and any voltages above U 48max,high,limited have to be stored in the failure memory. The range between U 48r und U 48shprotect includes a contingency reserve.  Upper Operating Range with functional Degradation: The range between U 48max,unlimited and U 48max,high,limited is dedicated for the calibration of the accumulator and the absorption od recuperation energy.  Operating Range without Degradation: The range between U 48min,unlimited and U 48max,unlimited is dedicated to operate the components without any functional degradation.  Lower Operating Range with functional Degradation: operating in the range between U 48min,low,limited and U 48min,unlimited is only temporarily allowed. Countermeasures have to be taken to return into the operating range without functional degradation.  Undervoltage Range: Voltages below U 48min,low,limited are defined as undervoltages and have to be stored in the failure memory. The accumulator protection voltage is defined at U 48stoprotect . The new specification LV148, which was defined by the car manufacturers Audi, BMW, Daimler, Porsche und Volkswagen as a common base of the OEM specific standards for components at the 48 Volt voltage range, is aligned to the existing LV124 for 12 Volt and is revised and completed in terms of presumptions und assumptions. The objective of LV148 is a standard specification of the voltage range and the required tests to achieve a maximum of industrial building blocks for components and parts [6]. 13 Preface The requirements, tests and test conditions for electric, electronic and mechatronic components and systems are defined for the application in vehicles at 48 Volt. 4 System Design Basically, coming from the today´s existing system design with 12 Volt (Fig. 6 A) there are several opportunities to build up a dual voltage vehicle power supply. Most common are either load sided (Fig. 6 B) or generator sided (Fig. 6 C) extensions of the established 12 Volt architectures. Depending on load characteristics and efficiency requirements some basic design rules can be made out for the decision of a specific architecture: 1. For singular loads with low average power consumption, e.g. peak power loads like electric power steering or chassis control functions, the load sided 48 Volt architecture (B) may be the best choice. A small DC/ DC converter (< 1 kW) and energy store technologies like double layer capacitors are nearly sufficient to meet the requirements for electric energy and power supply. 2. Higher energy demands either by a large number of 48 Volt loads or generators for high power recuperation require a generator sided extension (C). The typical power output of the 48/ 12 Volt DC/ DC converter is up to 3 kW; appropriate energy store technologies are large scaled double layer capacitors as well as lithium ion accumulators with a high charge acceptance over a wide temperature range like LTO (lithium titanate) based systems. In this architecture a major focus on high efficiencies in generating, storing and delivering of recuperated energy is required in order to achieve the highest CO 2 -benefit. Fig. 6: Typical vehicle power supply architectures 14 Preface 5 Components The generator, the accumulator and the DC/ DC converter are the key components for the generator sided extension. Additionally, existing stationary as well as transient high power loads will be transferred to the more powerful 48 Volt power supply and new functions which could not be realized with 12 Volt may be applied. A first possible set of these 48 Volt loads may consist of some stationary high power loads like the cooling or electric heating and transient high power loads like electric power steering or roll stabilization. All these additional components will not be discussed in detail in this paper. A quite complete set of these components and the benefits and consequences will be sketched in chapter 7. 5.1 Generator and Starter The generator consists of an electric machine and a DC/ AC-inverter. A mechatronic integration of both components seems to be the best solutions. As a multi-phase electric machine combined with a power inverter the generator allows to be used like motor, e.g. starting the combustion engine (cranking), assisting or boosting. The resulting requirements for this complex component are mainly defined by the maximum torque characteristics, maximum and continuous generating power, mechanical parameters, ambient temperature range, minimum and maximum operating voltage and the maximum inverter currents. Different types of electric machines have to be discussed with respect to the system requirements like in [11]. Fig. 7: Power limitations of a synchronous machine externally excited [11] However the essential question is the system integration of the electric machine into the drive train. In a first approach a replacement of the existing 12 Volt generator driven in the belt of the combustion engine is reasonable. But the limitations by torque and volume do not allow exceeding the range of more than 4 kW at least for a longer period. The maximum 15 Preface torque is given by the mechanical design of the belt and the sensitivity of torque capability of the crank shaft and the drag torque of the combustion engine itself which reduced the recuperation during deceleration. In order of overcome all these limitations and to achieve to requirements of a high recuperation and the application of high power loads the generator has to be integrated in the ideal position in the power train and with some flexibility in terms of speed compared to engine and drive train speed. The transmission is one of the promising positions which may fulfill these requirements. 5.2 Accumulator The requirements and the present status of 48 Volt accumulator is discussed in [12] in detail. An additional 48 Volt power supply can cope with power peaks and control higher electric power. This leads to new requirements of the energy storage system lifetime for reliability and cycling resistance. Depending on the requirements and the architecture the design criteria of the accumulator are different with respect to the overall power capability of the overall system. The typical usable energy of the accumulator depending on the targeted functions is in the range of 70 Wh to 1000 Wh. Referred to the voltage range as described in fig. 5, an accumulator based on LFP technology will be built with 14 or 15 cells in series, see table 1. If LTO technology is applied, 19 or 20 cells are used with respect to the lower cell voltage. Table 1: Possible configurations of 48 Volt accumulators with different technologies [12] LTO LFP Number of cells V_min (20% SOC) V_nom (80% SOC) V_end-ofcharge V_min (20% SOC) V_nom (80% SOC) V_end-ofcharge 1 2,1 V 2,4 V 2,7 V 3,15 V 3,3 V 3,6 V 14 44,1 V 46,2 V 50,4 V 15 47,25 V 49,5 V 54,0 V 19 39,9 V 45,6 V 51,3 V 20 42,0 V 48,0 V 54,0 V LTO cells have a significantly lower internal resistance, approximately half the value of LFP. But in terms of voltage stability LFP shows a better performance. A final validation in terms of voltage stability cannot be performed yet, because 48 Volt accumulators have to be built up and have to be analyzed at high currents. These measurements have to cover the whole operating range and all the environmental conditions. The present status allows to postulate two major requirements: The discharge capability of LFP have to be improved significantly and the prices of LTO accumulators have to be decreased dramatically. 5.3 DC/ DC Converter The third key component is the DC/ DC converter which is the electrical link between 48 Volt and 12 Volt power supply. 16 Preface The variety of requirements in different applications lead directly to a platform of DC/ DC converters at voltage levels up to 60 Volt, enabling an efficient implementation of power converters with different power requirements even with high product diversity. Particular attention is paid to a DC/ DC converter for future dual voltage power systems using 48 Volt and 12 Volt, as well as additional implementations of the converter platform. There is a major focus on feasibility of scalable power classes as well as on efficiency, thermal management and industrialisation potential of the concept. The major requirements which are discussed in more detail in [12] are  Scalability: Due to the spectrum of applications the power transmission has to be scalable. This can be realized by a design of multi phase converters with qualified devices, circuits and software blocks.  Industrialization: Established production processes and approved technologies for assembly und interconnect shall be the base for DC/ DC converters in order to meet the requirements for reliability and quality.  Availability: Single failures may not cause a total shutdown of the DC/ DC converter which a central part of the vehicle power supply system. The design of the converter has to be able to handle a defect in one phase and to perform a limited operation.  Protection: The converter has to be protected against low and high voltages, reverse polarity (12 Volt side) and high temperatures.  Thermal Management: Target is air cooling. Optional liquid cooling.  Operation modes: Primary operation is buck, secondary operation is boost. The operating voltage ranges are defined by LV124 for 12 Volt side and LV148 for 48 Volt side. The current on the 12 Volt side is in the range of 45 A to 220A, peak currents up to 270 A. The efficiency over a wide operating range has to be more than 95%. Fig. 8 Architecture of the 48 Volt / 12 Volt DC/ DC converter [10] control logic BN2 BN1 current- monitor- I 1 ist GND GND FET driver EMC Filter EMC Filter & Protection PWM 48V 12V 17 Preface In order to meet the requirements in terms of efficiency in the wide operating range a switch-off capability of every single phase is a promising solution. Fig. 9: Optimization of efficiency vs. output current [10] 5.4 Possible 48 Volt Loads The electric cooling fan with a stationary electrical power consumption of about 1 kW is one of the maximum loads. The higher thermal isolation of the engine compartment as well as new functions coming up like cooling of the exhaust manifold will require higher cooling capability in the future. An electric heating system like a PTC (Positive Thermal Coefficient) is in the range of more than 1 kW stationary electric power. High performance electric power steering demands power peaks up to 2 kW. The resulting current peaks affect too high voltage drops on a 12 Volt system. Intermediate solutions with a 24 Volt power supply came up in the past. Thus this application fits perfectly to a generator sided extension with 48 Volt. Electric roll stabilization exceeds the power consumption of 2 kW and has a high regeneration in terms of bad roads. Therefore a capsulation is mandatory, if integration into a 12 Volt vehicle is aspired. An additional electric power supply has to be implemented. In the case of a dual vehicle power supply with 48 Volt the electric roll stabilization can easily be applied to the 48 Volt system. 80% 82% 84% 86% 88% 90% 92% 94% 96% 98% 100% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Efficiency Output current related to maximum output current Efficiency optimization at reduced loads Phasen-geschaltet Alle-Phasen 18 Preface 6 A flexible Energy and Power Management The physical side of a power supply network including components like generator, accumulator, DC/ DC convertors or consumption units as discussed before is complemented by a management side. This shows figure 9 and is discussed in [14]. Fig. 10: Architecture of a dual voltage power network including a management level A flexible energy and power management as shown in [6] has to coordinate the 12 Volt power supply network, the 48 Volt power supply network and at least the interoperability of both. The coordination means that the energy flow, the power distribution and the operational states of the components should be hold within tolerable limits and that the power affected by the loads is not able to destabilize the single network. Therefore a power distribution management with a primary purpose of voltage stabilization is discussed in [13]. To guarantee an optimized use of the components discussed in chapter 5 a different strategy is implemented in each power supply network. This is necessary because of the different electric loads in the two power supply networks. If in future a load is displaced from one side to the other side of the combined power supply network it may be necessary to change the strategy of operational behavior to enable this migration in a flexible way of development. The coupling of the two power supply networks is discussed in [14]. The feature of coupling different power supply networks is a chance but includes the risk of destabilization of both networks caused by the interoperability with a DC/ DC converter. This converter can work at different operational strategies like current regulation mode, voltage regulation mode or power regulation mode. It can also be used to realize a transfer of predefined value of energy from one side to the other side of the power supply network. The strategy of coupling has to be defined by the management system based on the monitoring of both sides of the power supply network. The targets of the coupling strategy is the energy and power flow between the different power supply networks, the guarantee of stability on both sides and at least a efficient way of interoperation between the coupled power supply networks. 19 Preface 7 Summary and Outlook In order to meet the constantly increasing requirements in the future to reduce CO 2 emission and fuel consumption in conventional vehicles with combustion engines a significant contribution can be made by increasing the recuperation power to a range of 5 to 12 kW. The idea is to supply of the vehicle with electrical energy with less fuel as possible. During deceleration the combustion engine shall be switched off. The charging and discharging periods due to the velocity profiles and the state of the accumulator are shown in fig. 11. Furthermore, a sailing mode can be applied while the combustion engine may be switched off at constant speed. Fig. 11: Approach for a high recuperation system The electrical energy which is required for a driving cycle can only be generated by recuperation, if the electrical power is increased, respectively, because the time slot for generating is reasonably lower [3]. In addition a 48 Volt electrical motor offers the possibility to shift the operating point of the combustion engine due to its generating and driving modes. An electric slow motion may be possible depending on the mechanical design of the drive train. The new requirements for higher electrical power will affect and change future power supply architectures in terms of topology, overall current consumption and functional characteristics. The independently working power supplies can reduce the interaction between the high power loads. Furthermore, the displacement of high power loads from 12 Volt to 48 Volt can be discussed. The reduction of cross-sections in the wiring harnesses offer potentials of weight. Fig. 12 shows a possible migration of the configuration. 20 Preface Fig. 12: Approach for a future dual voltage power supply system The new voltage level 48 Volt offers a wide range of applications from boost functions and CO2 reduction in small vehicles up to the supply of large scale of functions in fully equipped luxury cars, so that a broad adoption can be expected. Additionally, the proposed voltage level enables the electrification of functions, which can be transferred from mechanical to electrical drive, like an electric climate compressor for driving operation and stand-still. Literature [1] O. Sirch, J. Fröschl, Dr. H. Pröbstle, Energiebordnetz und elektrisches Energiemanagement der Zukunft. 1. Braunschweiger Symposium zu Elektrische Leistungsbordnetze und Komponenten von Straßenfahrzeugen, Oktober 2008. [2] H. Pröbstle, R. Neudecker, O. Sirch, G. Immel: Topologien zukünftiger Energiebordnetze, Energiemanagement & Bordnetze III, November 2009 in Würzburg. [3] S. Wolff, O. Sirch, M. Schmid, G. Immel, H. Pröbstle, R. Neudecker, J. Fröschl: Demands by Tomorrows Energy Systems on the Vehicle Electric System. 7. Symposium Steuerungssysteme für den Antriebsstrang von Kraftfahrzeugen, Juni 2009 in Berlin. [4] O. Sirch, G. Immel, H. Pröbstle, R. Neudecker, J. Fröschl, Future Electrical Powernet - From Optimization Concerning Energy to a new Concept, 14. Internationaler VDI- Kongress Elektronik im Kraftfahrzeug, Oktober 2009. [5] H. Pröbstle, R. Neudecker, O. Sirch, Power Supply in Future Start-Stop-Systems. 4 th International Conference Energy Management Wire Harness Systems, March 2011. 21 Preface [6] T. Kohler, J. Fröschl, H.-G. Herzog, Systemansatz für ein hierarchisches, umweltgekoppeltes Powermanagement. 2. Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen, März 2010. [7] J. Fröschl, H.-G. Herzog, R. Neudecker, H. Pröbstle, O. Sirch, Considerations and approaches for a Dual Voltage Power Supply System with 48 Volt, EEVC Juni 2011. [8] T. Dörsam, D. Grohmann, S. Kehl, A. Klinkig, A. Mai, O. Sirch, A. Radon, J. Winkler, Die neue Spannungsebene 48 Volt, 15. Internationaler VDI-Kongress Elektronik im Kraftfahrzeug, Oktober 2011. [9] R. Falsett. T. Dörsam, Elektrische Energiespeicher im 48V Bordnetz, 15. Internationaler VDI-Kongress Elektronik im Kraftfahrzeug, Oktober 2011. [10] A. Körner, M. Nalbach, C. Hoff, O. Sirch, Leistungselektronik für die stabile Energieversorgung für Micro-Hybrid-Fahrzeuge der nächsten Generation, Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement, Haus der Technik, April 2012. [11] W. Hackmann, A. Rudorff, Y. Günsayan, Design of a 48V-Belt driven Starter Generator-System drawing special system requirements into account, Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement, Haus der Technik, April 2012. [12] R. Falsett, T. Dörsam, Elektrische Energiespeicher im 48V Bordnetz, Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement, Haus der Technik, April 2012. [13] T. Kohler, G. Kiener, J. Fröschl, A. Thanheiser, D. Bücherl, H.-G. Herzog, Voltage Stabilization in Vehicle Power Nets by Power Distribution Management, EVS26, Los Angeles, California, May 2012. [14] J. Fröschl, T. Kohler, A. Thanheiser, H.-G. Herzog, Intelligente Bordnetzkopplung am Beispiel Zweispannungsbordnetze mit 12 V und 48 V, Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und Elektrisches Energiemanagement, Miesbach, April 2012. 22 Preface Die Herausforderungen für die Elektrik/ Elektronik in der Automobilindustrie durch die Einführung einer 48 Volt Versorgungsspannung The challenges for electrics/ electronics in the automotive industry introducing a 48 Volt power supply Dipl.-Ing. Ottmar Sirch, Dr.-Ing. Georg Immel, Dipl.-Ing. Joachim Fröschl, Dipl.-Ing. Alexander Mai, Dr. rer. nat. Hartmut Pröbstle, BMW Group, München VDI Kongress Elektronik im Fahrzeug, Baden-Baden, Oktober 2013 Kurzfassung Die ständig steigenden Anforderungen an die elektrische Energieversorgung in konventionellen Kraftfahrzeugen, getrieben durch die fortschreitende Elektrifizierung sowie CO 2 - und verbrauchsreduzierende Maßnahmen, haben die Einführung von 48 Volt als zusätzliche Versorgungsspannung im Kraftfahrzeug in absehbare zeitliche Nähe rücken lassen. Bei Audi, BMW, Daimler, Porsche und Volkswagen wurde die Thematik frühzeitig aufgegriffen und eine erste Spezifikation für eine zweite Spannungsebene mit 48 Volt in Form der Liefervorschrift LV148 erarbeitet. Auch andere Automobilhersteller beschäftigen sich derzeit intensiv mit neuen Systemkonzepten auf Basis von Zwei-Spannungs-Bordnetzen mit 12 Volt und 48 Volt. Die Gestaltung und Ausprägung solcher Systeme und deren Architekturen hängen in erster Linie mit den CO 2 -Zielsetzungen zusammen und erfordern von den Lösungskonzepten für das Zwei-Spannungs-Bordnetz ein ausgewogenes Verhältnis zwischen Effizienz, Gewicht, Bauraum und Kosten. Zusätzlich soll Freiraum für funktionale Mehrungen und Innovationen geschaffen werden, die den Kundennutzen des Systems und damit des Fahrzeugs deutlich steigern. Neben neuen funktionalen Anforderungen und den daraus resultierenden Architektur- und Dimensionierungsaufgaben ergibt sich eine Vielzahl an technischen Fragestellungen in der Elektrik/ Elektronik, die frühzeitig im Rahmen der Vorentwicklungsphase zu beantworten sind: Elektromagnetische Verträglichkeit, Lichtbögen, Kurzschlüsse zwischen den Spannungsebenen und die Darstellung der Vernetzung in gemischt versorgten Systemen seien an dieser Stelle als Beispiele genannt. Die Herausforderung an das Management von elektrischer Energie und Leistung besteht einerseits im Management der Teilbordnetze mit 12 Volt und 48 Volt und andererseits in der Koordination der Teilbordnetze unter prädiktiven Aspekten mittels einer Umweltkopplung. In diesem Beitrag werden die Anforderungen an das 2-Spannungs-Bordnetz, die neuen Herausforderungen und der korrespondierende Lösungsraum vorgestellt. Anhand von Fallbeispielen werden die technischen Details erläutert. Abstract The continuously increasing requirements for vehicle power supplies in conventional automobiles pushed by the progressing electrification as well as the reduction of CO 2 emissions and fuel consumption are bringing the launch of 48 Volt as an additional power supply voltage into the near future. Audi, BMW, Daimler, Porsche and Volkswagen had been aware of this new item and started a specification of the second voltage level early with LV148. Also other automotive manufacturers are intensively working on new concepts based on dual voltage power supplies with 12 V and 48 V. The design and characteristics of these systems and their architectures primarily depend on the future CO 2 targets and require balanced solutions in terms of efficiency, weight, volume and costs. Additionally, there has to be enough expandability for novel 23 Preface innovations which increases significantly the customer´s benefits of the system and finally of the vehicle. Beside the new functional requirements and the challenges for architecture and dimensions there are a lot of open questions regarding Electrics/ Electronics which has to be answered early during the predevelopment phase: Electromagnetic interference, arcing, short circuits between the voltages levels and communication concept of a mixed power supplied system are some examples of items which have to be discussed. A further very important challenge is the management of electrical energy and power in the branches with 12 Volt and 48 Volt and the coordination of the branches with respect to prediction via an environmental interface. The requirements for a Dual Voltage Power Supply, the challenges and the solution area will be discussed. Technical details will be explained by examples. 1. Einleitung Seit Mitte der 90er Jahre hat sich der Energie- und Leistungsbedarf im Bordnetz von konventionellen Kraftfahrzeugen mit Verbrennungsmotor überproportional erhöht. Neue E/ E-Kundenfunktionalitäten führen zusammen mit fortschreitender Elektrifizierung von Antriebsfunktionen und Nebenaggregaten zur Erreichung der angespannten CO 2 -Ziele das elektrische Energiebordnetz im Kraftfahrzeug an die Grenzen seiner Belastbarkeit. Besonders hervorzuheben sind an dieser Stelle die Auto-Start- Stopp-Funktion, die Bremsenergierückgewinnung und die Elektrifizierung mechanisch betriebener Systeme, wie z.B. die elektrische Lenkkraftunterstützung. Neben dem zunehmenden Energiebedarf einzelner Funktionen stellt vor allem auch die Überlagerung mehrerer Hochstromverbraucher eine besondere Herausforderung an die Leistungsfähigkeit zukünftiger Bordnetzarchitekturen dar. Auch wenn absehbar ist, dass in Zukunft Mobilität über elektrifizierte Antriebsstränge definiert wird [5], bleiben effiziente verbrennungsmotorische Antriebe weiterhin eine unverzichtbare Option. Die Effizienz von Verbrennungsmotoren lässt sich u.a. durch Reduktion der Verlustleistung seiner Aggregate, wie z.B. der Generatorverluste massiv steigern. Ermöglicht wird dies z.B. durch „starke“ Bordnetze, die einen generatorischen Betrieb vorwiegend in den Brems- oder Verzögerungsphasen des Fahrzeugs benötigen. Dies bedingt wiederum zyklenfeste Speicher und leistungsfähige Generatoren, die das vorhandene Rekuperationspotenzial effizient ausnutzen. Es ist hier wichtig zu erwähnen, dass sich der aktuell oftmals propagierte „Segelbetrieb bei abgestelltem Motor“ unter der Randbedingung konventioneller motorgebundener Generatoren nachteilig auf das Rekuperationspotenzial auswirkt. Ein abgekoppelter Generator kann in einem solchen Fall die Bremsenergie nicht in wertvolle Rekuperationsenergie umwandeln. „Segeln“ bei abgeschalteten Motor macht daher aus Energiebordnetzsicht nur dann Sinn, wenn gleichzeitig der „generatorische Druck“ in den Verzögerungsphasen aufrechterhalten bleibt, was neue Antriebsarchitekturen mit z.B. getriebegebundenen Generatoren erfordert. Die Energiemenge, die bei einem Bremsvorgang rekuperiert werden kann, hängt bei vergleichbarer Antriebsarchitektur vom Fahrzeuggewicht und der Fahrzeuggeschwindigkeit ab. Bei großen Fahrzeugen ist die Verzögerungsleistung und damit das theoretische Rekuperationspotenzial höher. In der Realität wird dieses Potenzial durch die Leistungsfähigkeit der verbauten Systemkompenten wie Generator und Energiespeicher begrenzt. Hier bietet z.B. die Anhebung der Systemspannung von 12 V auf 48 V Potenziale. Erste Untersuchungen haben gezeigt, dass bei kleinen Fahrzeugen mit „starken“ 12-Volt-Bordnetzen bereits eine Kompensation des Bordnetzeigenbedarfs über Rekuperation möglich ist. Bei großen Fahrzeugen mit hoher Ausstattung 24 Preface und typischerweise hohem Bordnetzbedarf bieten 48-V-Rekuperationsarchitekturen die Möglichkeit einer 100%-igen Eigenbedarfsdeckung. Mit einer 48-Volt-Elektromaschine eröffnen sich zudem weitere Möglichkeiten zur CO 2 - Einsparung, da sie sowohl im generatorischen als auch im motorischen Betrieb den Lastpunkt des Verbrennungsmotors in einen günstigeren Betriebspunkt verschieben kann. Die neuen Anforderungen nach höherer elektrischer Leistung werden zukünftige Energiebordnetz-Architekturen hinsichtlich Topologie, Gesamtstromverbrauch und funktionaler Ausprägung maßgeblich beeinflussen und verändern. Als Beispiel hierfür zeigt Bild 1 ein Zwei-Spannungs-Bordnetz klassischer Ausprägung. Durch die Entkoppelung der Bordnetzzweige können die von Hochleistungsverbrauchern verursachten Wechselwirkungen zwischen den verschiedenen Bordnetzzweigen verringert oder sogar verhindert werden. Bild 1: Schematische Darstellung eines Zwei-Spannungs-Bordnetzes Weitere Ausprägungen des Zwei-Spannungs-Bordnetzes sind denkbar und werden in Kapitel 3 behandelt. 2. Änderung der Anforderungen Bis 2020 ist eine signifikante Veränderung der Rahmenbedingungen für die Bewertung des CO 2 -Ausstoßes von Kraftfahrzeugen in Europa zu erwarten. Der Normverbrauchszyklus soll vom Neuen Europäischen Fahrzyklus NEDC auf den Worldwide Harmonized Light Vehicle Test Procedure WLTP umgestellt werden. Die unterschiedlichen Geschwindigkeitsprofile von NEDC und WLTP sind in Bild 2 dargestellt. Bild 2: Geschwindigkeitsprofile NEDC (oben) im Vergleich zu WLTP (unten) 25 Preface Während der NEDC über einen hohen Anteil an Stopp-Phasen verfügt, ist der zukünftig von der Europäischen Union präferierte WLTP durch erheblich weniger Stopp-Phasen gekennzeichnet. Beim WLTP werden im Gegensatz zum NEDC Sonderausstattungen, allerdings keine Einschaltprofile von Komfortfunktionen berücksichtigt. Eine deutliche Erhöhung der Verzögerungsanteile von 31% auf 43% begünstigt die Gewinnung von elektrischer Energie durch Rekuperation. Tabelle 1: Vergleich zwischen NEDC und WLTP NEDC WLTP Wegstrecke [km] 11,013 23,141 Dauer [s] 1180 1800 Mittl. Geschwindigkeit [km/ h] 33,6 46,3 Max. Geschwindigkeit [km/ h] 120 131 Max. Beschleunigung ]m/ s2] 1,04 1,88 Max. Verzögerung [m/ s2] -1,39 -1,52 Standzeiten [s] 280 227 Anteil Standzeiten [%] 23,7 12,6 Anteil Verzögerung [%] 31,0 43,1 3. Architekturen für Zwei-Spannungs-Bordnetze mit 12 Volt und 48 Volt Das elektrische Energiebordnetz im Fahrzeug hat in den letzten 10 Jahren einen erheblichen Anstieg der Anforderungen, jedoch keine grundlegende Veränderung in Architektur und Topologie erfahren. Die Integration neuer Funktionen wurde dabei stets über die Dimensionierungen von Klauenpol-Generator, Bleibatterie, Kabelbaum und Stromverteiler im „standardisierten“ 12-Volt-Bordnetz (Bild 3) umgesetzt, was auf einen Mehrbedarf an  elektrischer Energie durch stationäre Stromverbraucher (z.B. Heizsysteme, Motorlüfter),  elektrischer Leistung durch transiente Stromverbraucher (z.B. Lenkkraft-unterstützung, Fahrwerksregelsysteme, Warmstart) und  Speicherzyklisierung durch Energiedurchsatz (z.B. Bremsenergierückgewinnung, Vor-/ Nachlauf, Versorgung in der Motorstopphase) zurückzuführen ist. Bild 3: Schematische Darstellung der heutigen Energiebordnetzarchitektur 26 Preface Durch ständige Optimierungen und Verbesserungen sowohl bei der Generierung als auch bei der Speicherung von elektrischer Energie kann die Leistungsfähigkeit dieser Bordnetzarchitektur weiter gesteigert werden. Dies geschieht auch durch die Einführung von Effizienzmaßnahmen, die den Stromverbrauch von Funktionen, Systemen und Aggregaten reduzieren, und somit indirekt die Energie- und Leistungsverfügbarkeit erhöhen. Ein anstehender Architektursprung, wie z.B. die Einführung eines Zwei- Spannungs-Bordnetzes, kann unter Umständen dadurch hinausgezögert werden. Bild 4: Architektursprung durch zunehmende Leistungsanforderungen Durch die sukzessive Einführung von Maßnahmen zur Reduzierung des elektrischen Verbrauchs bestehender Funktionen [1], wie in Bild 4 dargestellt, kann Freiraum zur Einführung neuer Funktionen geschaffen werden. Je höher die Reduzierung zu einem bestimmten Zeitpunkt ist, desto höher ist das Potenzial für die Einführung neuer Funktionen. Wird allerdings durch den hohen Leistungsbedarf einer neuen Funktion die Schwelle zu einer leistungsfähigeren Architektur überschritten, so kann der Architektursprung nicht länger vermieden werden. Da Architektursprünge mit erheblichen Kosten verbunden sind, sollten sie über kundenwerte Funktionsmehrungen legitimiert werden. Aufbauend auf die einleitend diskutierte und in Bild 3 dargestellte klassische Basisarchitektur bieten sich verschiedene Architekturvarianten in Abhängigkeit von den erweiterten Anforderungen an. Sämtliche Varianten an elektrischen Energieversorgungsarchitekturen lassen sich auf  verbraucherseitige,  generatorseitige und  speicherseitige 27 Preface Erweiterungen der in Bild 3 gezeigten Basisarchitektur zurückführen. Eine ausführliche Beschreibung hierzu findet sich in [2]. Bei Überschreitung der stationären und dynamischen Leistungsgrenzen des 12-Volt- Bordnetzes bietet sich die Integration einer oder mehrerer Funktionen in Form einer verbraucherseitigen Erweiterung auf 48 V an. Bild 5 zeigt das Prinzipschaltbild einer verbraucherseitigen 48-V-Erweiterung, die vor allem zur Stützung von Hochleistungsverbrauchern auf ein entkoppeltes Teilbordnetz zurückgreift. Die verbraucherseitige Erweiterung ist jeweils mit oder ohne Energiespeicher ausführbar, je nach Energie- oder Leistungsbedarf der versorgten Funktionen. Verbraucherseitige Erweiterungen ohne 48-Volt-Speicher sind für stationäre 48-Volt- Verbraucher, die eine höhere Betriebsspannung benötigen (z.B. auf Beschichtung basierende Frontscheibenheizung) geeignet, sofern die Leistungsabnahme die Belastung der 12-V-Seite nicht überschreitet. Für dynamische Kurzzeitverbraucher wie z.B. Lenk- oder Wankstabilisierungssysteme werden vorzugsweise Doppelschichtkondensatoren mit hoher Leistungsfähigkeit und Zyklenfestigkeit eingesetzt. Bild 5: Architekturprinzip der verbraucherseitigen Erweiterung VEW Signifikante energetische Mehrbedarfe sollten vorzugsweise über Speicherarten wie z.B. Lithium-Ionen-Batterien versorgt werden, um eine bedarfsorientierte Versorgung mit minimaler Belastung des Basisbordnetzes zu gewährleisten (z.B. im Start-Stopp- Betrieb). Stationär erhöhte Energiebedarfe (z.B. Klimafunktionen) oder Rekuperationsanforderungen mit Spitzenleistungen über 5 kW sollten vorzugsweise über die in Bild 6 gezeigte generatorseitige Erweiterungen bedient werden. In dieser Leistungsklasse bietet sich zudem ein elektromotorischer Betrieb an, der über die Boost-Funktion CO 2 günstige Lastpunktverschiebungen des Verbrennungsmotors ermöglicht. 28 Preface Bild 6: Architekturprinzip der generatorseitigen Erweiterung GEW 4. Spezifikation des Spannungsbereichs Der Spannungsbereich für 48 Volt ist in der LV148 spezifiziert. Bild 7 zeigt die detaillierte Aufteilung des statischen Spannungsbereichs, der für jede 48-Volt-Komponente gilt. In der LV148 sind außerdem die Komponententests detailliert beschrieben. Bild 7: Definitionen der statischen Spannungsbereiche Berührungsschutzbereich In diesem Bereich ist für Gleichspannungen ein Berührungsschutz erforderlich (siehe ECE-R 100). Überspannungsbereich Zwischen U 48max,high,limited und U 48r liegt der Überspannungsbereich inkl. aller Toleranzen. In diesem Bereich soll der Überspannungsschutz aktiv sein und Spannungen größer U 48max,high,limited müssen durch einen Fehlerspeichereintrag protokolliert werden. Der Bereich zwischen U 48r und U 48shprotect beinhaltet die Sicherheitsreserve. Oberer Betriebsspannungsbereich mit Funktionseinschränkung Der Bereich zwischen U 48max,unlimited und U 48max,high,limited ist für die Kalibrierung des Speichers und die Aufnahme von Rückspeiseenergie vorgesehen. Limited Operation Unlimited Operation Undervoltage Protection Against Electrical Shock Overvoltage Limited Operation 24V - U 48min,low,limited 36V - U 48min,unlimited 52V - U 48max,unlimited 48V - U 48n 54V - U 48max,high,limited 58V - U 48r 60V - U 48shprotect 20V - U 48stoprotect Storage Protection 29 Preface Betriebsspannungsbereich ohne Funktionseinschränkung Der Bereich zwischen U 48min,unlimited und U 48max,unlimited lässt den Betrieb der Komponenten ohne Funktionseinschränkung zu. Unterer Betriebsspannungsbereich mit Funktionseinschränkung Der Betrieb im Bereich von U 48min,low,limited bis U 48min,unlimited ist nur temporär zulässig. Gegenmaßnahmen sind zu ergreifen, um in den Betriebsspannungsbereich ohne Funktionseinschränkung zurückzukehren. Unterspannungsbereich Alle Spannungen unter U 48min,low,limited sind als Unterspannung definiert und müssen als Fehlerspeichereintrag protokolliert werden. Bei U 48stoprotect befindet sich die Speicherschutzspannung. Speicherschutzspannungsbereich Alle Spannungen unter U 48stoprotect. Tabelle 2: Abkürzungen zu Spannungen und Strömen Abkürzung Bezeichnung Wert U 48shprotect Berührungsschutzspannung. Abgeleitet aus der Forderung der Einhaltung des Grenzwertes für den Berührungsschutz von Gleichspannungen (siehe ECE- R 100). 60 V U 48r 2 V Sicherheitsreserve bis zur Berührungsschutzspannung. 58 V U 48max,high,limited Maximale Spannung des oberen Betriebsbereichs mit Funktionseinschränkung. 54 V U 48max,unlimited Maximale Spannung des Betriebsbereichs ohne Funktionseinschränkung 52 V U 48n BN48-Nennspannung 48 V U 48min,unlimited Minimale Spannung des Betriebsbereichs ohne Funktionseinschränkung 36 V U 48min,low,limited Minimale Spannung des unteren Betriebsbereichs mit Funktionseinschränkung 24 V U 48stoprotect Speicherschutzspannung 20 V U 48pp Spitze-Spitze-Spannung U 48rms Effektivwert einer Spannung U 48max Maximalspannung, die während einer Prüfung auftreten kann U 48min Minimalspannung, die während einer Prüfung auftreten kann U 48test BN48-Prüfspannung U 12test BN12-Prüfspannung GND48 Geräte-Masse (Kl. 41) 5. Herausforderungen Die Einführung einer zweiten Spannungsebene deutlich größer als 12 Volt bringt eine Vielzahl an Herausforderungen an die elektrischen Komponenten und deren Integration mit sich. Einige dieser Anforderungen, die sich als besonders wichtig und systemrelevant herauskristallisiert haben, sollen im Folgenden näher erläutert werden. 30 Preface 5.1 Elektrisches Energie- und Leistungsmanagement Die Herausforderung an das flexible Management von elektrischer Energie und Leistung besteht einerseits im Management der Teilbordnetze mit 12 Volt und 48 Volt und andererseits in der Koordination der Teilbordnetze unter prädiktiven Aspekten mittels einer Umweltkopplung, wie schematisch in Bild 8 dargestellt. Bild 8: Architektur eines Zwei-Spannungs-Bordnetzes und der zugehörigen Managementstruktur Bei einem Ein-Spannungs-Bordnetz besteht die Aufgabe des flexiblen Energie- und Leistungsmanagements darin, sowohl die Zustände von Energiehaushalt, -reserve, und -tendenz als auch die Zustände von Leistungshaushalt, -reserve, und -tendenz in einem stabilen Arbeitsbereich zu halten [11]. Werden zwei Ein-Spannungs-Bordnetze gekoppelt, so gelten diese Aufgaben gleichermaßen für beide Teilbordnetze. Eine ergänzende Aufgabe ergibt sich jedoch aus der Koordination der Teilbordnetze [4]. Diese besteht darin, die beiden Bordnetze in einem ausgeglichenen Zustand zueinander zu halten. Der Ausgleich muss gleichermaßen für Energie und Leistung unter Berücksichtigung von Stabilität und Effizienz erfolgen. Unter Berücksichtigung von Informationen aus der Systemumwelt, wie beispielsweise einer Navigationskopplung, kann dieses Gleichgewicht situationsgerecht modifiziert werden [12]. Es muss aus Sicht der koordinierenden Betriebsstrategie entschieden werden, ob beispielsweise für eine Rekuperationsphase zunächst prädiktiv eine Konditionierung, d.h. Absenkung des Speicherladezustands , für ein oder beide Teilbordnetze erfolgen sollte, um anschließend den einen oder beide elektrischen Speicher rekuperativ zu befüllen. Dabei muss für beide Teilbordnetze die Stabilität der Betriebsspannung zu jedem Zeitpunkt gewährleistet sein. Das Managementsystem muss weiterhin die oben gezeigten Architekturvarianten optimal bedienen können. Dabei ist eine standardisierte Anbindung der Komponenten [12] einerseits von wirtschaftlicher Bedeutung und andererseits von technischem Vorteil hinsichtlich der Austauschbarkeit und Migrationsfähigkeit innerhalb des architektonischen Lösungsraumes. Die Betriebsstrategie muss dabei den hinsichtlich Effizienz optimalen Einsatz der verwendeten elektrischen Komponenten als Gesamtsystem mit dem Ziel der bestmöglichen Stabilität der einzelnen Systemspannungen sicherstellen. Ebenso ist der Einsatz des DC/ DC-Wandlers innerhalb seiner Leistungsgrenzen aber 31 Preface auch möglichst lange im Bereich seiner maximalen Effizienz sicherzustellen. Hierzu bieten sich die Möglichkeiten der Erhöhung des Angebots und der Degradation der Verbraucher in beiden Teilnetzen an. Eine Prädiktion der Erzeugungs- und Verbrauchssituation mittels Umweltkopplung unterstützt dabei die Entscheidungen der Betriebsstrategie hinsichtlich der Steuerung des Energie- und Leistungstransfers des Wandlers. Ein weiterer Aspekt ist die Behandlung von Fehlern. 5.2 Lichtbögen Bereits bei Bordnetzspannungen über 12 Volt können stabile Störlichtbögen entstehen. Daher muss dieses Phänomen bei allen Spannungsklassen unter 60 V genau analysiert werden. Sowohl in Personenkraftwagen mit 12 Volt als auch in Nutzkraftfahrzeugen mit 24 Volt tritt das Phänomen heute sporadisch auf und wird akzeptiert. Mit der Einführung von 48 Volt kommt die bereits bei 42 Volt geführte Diskussion erneut auf. Die physikalischen Zusammenhänge in Abhängigkeit von verwendeten Leitungsmaterialien und anliegenden Spannungen sowie weiteren Parametern sind in [6] ausführlich erläutert. Für die weiteren Betrachtungen wird zwischen parallelen und seriellen Lichtbögen unterschieden. 5.2.1 Parallele Lichtbögen Parallele Lichtbögen können durch Berührung zwischen einer spannungsführenden 48-Volt-Leitung und Masse hervorgerufen werden. Fehler dieser Art sind konstruktiv, z.B. durch Leitungsverlegung und Auswahl geeigneter Isolation, zu vermeiden. 5.2.2. Serielle Lichtbögen Das Trennen unter Last, hervorgerufen durch Abziehen oder Aufstecken des Steckers oder durch Leitungsabriss, kann zur Bildung eines Lichtbogens führen. Einflussfaktoren sind hier Spannung, Stromstärke und Trenngeschwindigkeit. Da der Lichtbogen- Mindeststrom sehr klein ist, kann sich auch bei Spannungsquellen geringer Leistung bei Überschreiten der Mindestspannung (je nach Leitermaterial, z.B. Kupfer 13 V) ein stabiler Lichtbogen entwickeln. Der minimal notwendige Haltestrom für Lichtbögen liegt bei 0,4 A [6]. Durch den Einsatz eines Kondensators am 48-Volt-Eingang eines Verbrauchers kann die Bildung von seriellen Lichtbögen unter normalen Bedingungen verhindert werden. Dazu muss der Kondensator so dimensioniert werden, dass bei einer Trennung von der elektrischen Versorgung die Spannungsdifferenz an der Trennstelle 12 Volt nicht überschreitet. Dies gilt innerhalb der Zeitdauer, die eine typische Trennung benötigt, um einen ausreichenden Abstand von 0,02 mm zu erreichen. Die Dimensionierung des Kondensators hängt von der Größe des Laststroms ab und ist über folgenden Zusammenhang zu ermitteln: 𝜏 𝐶 𝑅 𝑙𝑛 𝑈0 𝑈1 𝜏: 𝑍𝑒𝑖𝑡𝑘𝑜𝑛𝑠𝑡𝑎𝑛𝑡𝑒 𝐶: 𝐾𝑎𝑝𝑎𝑧𝑖𝑡ä𝑡 𝑑𝑒𝑠 𝑍𝐾 𝐾𝑜𝑛𝑑𝑒𝑛𝑠𝑎𝑡𝑜𝑟𝑠 𝑅: 𝑟𝑒𝑠𝑢𝑙𝑡𝑖𝑒𝑟𝑒𝑛𝑑𝑒𝑟 𝑊𝑖𝑑𝑒𝑟𝑠𝑡𝑎𝑛𝑑 𝑑𝑒𝑟 𝐿𝑎𝑠𝑡 𝑈0: Startwert Eingangsspannung 𝑈1: Endwert Eingangsspannung 32 Preface Ist die sich einstellende Distanz in der Trennstelle innerhalb der Zeitdauer größer oder gleich 0,02 mm, so tritt anschließend bei Zunahme der Spannungsdifferenz über der Trennstelle kein Lichtbogen mehr auf, da die kritische Feldstärke für die Ionisation von Luft (30 kV/ cm) nicht mehr erreicht werden kann. 5.3 Kurzschlüsse zwischen den Spannungsebenen In einem Zwei-Spannungs-Bordnetz gelten die gleichen Anforderungen hinsichtlich Kurzschlüssen wie im 12-Volt-Bordnetz. Kurzschlüsse nach Masse von 48 Volt nach Masse müssen genauso sicher erkannt und abgeschaltet werden. Der Kurzschluss von 48 Volt nach 12 Volt ist ein neuer Fehlerfall, der zu einer Zerstörung von 12-Volt-Komponenten führen kann. Er ist durch geeignete konstruktive Maßnahmen, z.B. Leitungsverlegung und Isolation, auszuschließen. 5.4 Kommunikation in gemischten Netzwerken Die Herausforderungen für Kommunikationssysteme in gemischten Netzwerken liegen vor allem in Signalintegrität und Robustheit [9]. Auch wenn in diesen Bordnetzen die elektrischen Massen beider Spannungsebenen elektrisch miteinander verbunden sind, stellt sich die Frage, welche Auswirkungen dies auf die Fahrzeugkommunikationsnetzwerke hat, die diese beiden Spannungswelten logisch verbinden. Im Beitrag [9] werden im Detail die zu erwartenden Systemkonstellationen bei der Verwendung verschiedener Bussysteme erörtert. Gezeigt werden dort denkbare Systemzustände und deren Auswirkungen. In Kommunikationssystemen, basierend auf CAN oder FlexRay, die eine differentielle Übertragung nutzen, haben die Bus-Transceiver trotzdem einen gemeinsamen Massebezug. Im Falle eines Masseverlusts eines 48-Volt-Busteilnehmers treten bei nur 2 Busteilnehmern, die identische Bustransceiver einsetzen, folgende Spannungsdifferenzen an den Transceiver-Eingängen auf: Am mit Masse verbundenen Transceiver fallen U Bat48 / 2 ab, und am nicht mit Masse verbundenen Transceiver liegt eine Spannungsdifferenz von -U Bat48 / 2 an. Je mehr Transceiver im gleichen Netzwerk noch eine Masse haben, desto größer wird die Spannungsdifferenz zum Transceiver ohne Masse bis nahe an -U Bat48 . Wenn der Masseabriss während der aktiven Kommunikation stattfindet, kann auch der masselose Transceiver noch eine gewisse Zeit weiter senden und Pegel bis nahe an U Bat48 auf den Bus einprägen. Daraus leiten sich folgende Anforderungen an die verwendeten Transceiver ab [9]:  In 48-Volt-Komponenten eingesetzte Transceiver müssen robust für negative Spannungen bis +/ -58 V sein.  In 12-Volt-Komponenten eingesetzte Transceiver müssen eine Spannungsfestigkeit bis 58 V aufweisen.  Harte Kurzschlüsse zwischen 48 Volt und einer der Busleitungen müssen durch konstruktive Maßnahmen verhindert werden.  In Komponenten mit beiden Versorgungsspannungen muss die Anforderung „Querstrom kleiner 1 µA“ (gem. LV 148 Test E48-20) erfüllt werden. 5.5 Power-Up und Power-Down Für die Inbetriebnahme und die Beendigung des Betriebs von 48-Volt-Systemen sind folgende Randbedingungen zu beachten. 33 Preface  Inbetriebnahme: Der 48-Volt-Speicher ist im abgestellten Zustand einpolig abgeschaltet. Vor dem Schließen des Schalters müssen die Zwischenkreise der 48-Volt-Komponenten vorgeladen werden, damit wegen der entladenen Kapazitäten keine hohen Ladeströme fließen. Die Inbetriebnahme muss innerhalb der Aufstartsequenz des Gesamtfahrzeugs so frühzeitig erfolgen, dass es zu keinem Fehlverhalten in den Kommunikationsabläufen der Netzwerke kommt.  Beendigung: Nach einer in Abhängigkeit von Nachlauffunktionen zu definierenden Zeit ist der Schalter des 48-Volt-Speichers zu öffnen und die Zwischenkreise sind auf einen definierten Spannungswert zu entladen. Vorher müssen die relevanten Kommunikationsabläufe definiert beendet werden. 5.6 Elektromagnetische Verträglichkeit Die elektromagnetische Verträglichkeit (EMV) spielt bei der Zulassung neuer Fahrzeuge eine wichtige Rolle. Elektromagnetische Störungen, die durch das 48VoltSystem erzeugt werden, müssen den gleichen Richtlinien und Grenzwerten genügen, die für herkömmliche Fahrzeugkonzepte mit 12 Volt gelten [8]. Diese Anforderungen sind bereits beim Entwurf der elektrischen Architektur des Fahrzeugs und der Komponenten zu berücksichtigen, um das System EMV-technisch ordnungsgemäß zu gestalten. Die hohen Spannungs- und Stromgradienten du/ dt und di/ dt erfordern eine entsprechende Berücksichtigung bei der Auswahl der Topologie und geeignete Filtermaßnahmen in den Komponenten. Hier bedarf es bereits in frühester Phase einer intensiven Zusammenarbeit zwischen Fahrzeughersteller und Komponentenlieferanten, um ein Optimum zwischen Filtermaßnahmen und Bordnetzgestaltung zu finden. 5.7 Umgang mit Spannungen < 60 Volt Der Spannungsbereich wurde gezielt so spezifiziert, dass ein Überschreiten der Berührungsschutzgrenze von 60 Volt_DC ausgeschlossen ist. In der LV148 ist folgende Prämisse postuliert: Es ist kein Berührungsschutz für Gleichspannungen ≤ 60 V erforderlich. Dies gilt für Gleich-spannungen bis zu einer Bordnetzwelligkeit von maximal 10% RMS. Bei Wechselspannung bis U eff ≤ 30 V ist kein Berührungsschutz notwendig (ISO 6469- 3). Ferner gilt die ECE-R100. Ein entsprechender Änderungsantrag zur ECE-R100 wurde in die entsprechenden Gremien eingesteuert und wird bis Ende 2013 dort bearbeitet. 6. Chancen durch die Einführung der neuen Spannungsebene Neben der höheren Leistungsfähigkeit und den damit verbundenen funktionalen Erweiterungsmöglichkeiten bietet das Zwei-Spannungs-Bordnetz mit 48 Volt weitere Potentiale hinsichtlich Gewicht des Kabelbaums und der Effizienz der elektrischen Subsysteme. 6.1 Gewichtsreduzierung des Kabelbaums Bei der Auslegung physischer 12-Volt-Bordnetze geht aus Gewichtsgründen der Trend aktuell von Kupfer zu Aluminium. Nachteilig wirken sich hier in Einzelfällen der höhere Leitungsquerschnitt bei gleichem Strom und die schwerer beherrschbare Verbindungstechnik zu kupferbasierten Teilkomponenten aus. 34 Preface Das 48-Volt-Leistungsbordnetz kann mit Kupfer oder Aluminium als Leitungsmaterial ausgelegt werden. Das Leitungsgewicht wird dort unabhängig vom Leitungsmaterial durch den geringeren Leiterquerschnitt, verursacht durch den geringeren Strom, deutlich reduziert, und bei nicht vollständiger Ausschöpfung der Querschnittsreduzierung können außerdem die Leitungsverluste verringert werden. Das Optimum aus Gewicht und Verlusten ist im Einzelfall zu ermitteln und in Abhängigkeit von Bauraumvorgaben zu betrachten. Tabelle 3: Vergleich am Beispiel eines Verbrauchers mit 600 W 12 V 48 V Laststrom 50 A 50 A 12,5 A 12,5 A Leitungsmaterial Kupfer Aluminium Kupfer Aluminium Leitungsquerschnitt 10 mm 2 17 mm 2 1,5 mm 2 2,5 mm 2 Gewicht/ Länge 108 g/ m 74 g/ m 17 g/ m 11 g/ m Verlustleistung/ Länge 4,5 W/ m 3,8 W/ m 1,8 W/ m 1,6 W/ m 6.2 Effizienzsteigerung durch Verlagerung von Verbrauchern Bei generatorseitigen Erweiterungen ist aus Effizienz- und Dimensionierungsgründen eine Verlagerung von leistungsstarken Verbrauchern auf die 48-Volt-Seite sinnvoll. Damit werden die Wandlungsverluste großer Leistungsflüsse von 48 Volt nach 12 Volt über den DC/ DC-Wandler in Höhe von 3% bis 5% vermieden. Ist die Verlagerung von 12-Volt-Verbrauchern auf die 48-Volt-Seite geeignet gewählt, kann die DC/ DC- Wandler-Größe bei ausgewogener Konfiguration der Leistungsverbraucher in den verschiedenen Betriebssituationen (z.B. Sommervs. Winterbetrieb), deutlich reduziert werden. Für die Verlagerung von existierenden 12-Volt-Hochleistungsverbrauchern eignen sich der elektrische Motorlüfter, dessen stationäre Leistungsaufnahme bis zu 1 kW erreichen kann, elektrische Heizsysteme (z.B. PTC), deren stationäre Leistungsaufnahme heute bereits über 1 kW liegt, und verschiedene elektrische Pumpen. Aus dem Bereich Fahrdynamiksysteme sind z.B. die elektrische Lenkkraftunterstützung und die elektrische Wankstabilisierung zu nennen, die durch ihre hohen transienten Leistungsaufnahmen das heutige 12-Volt-Bordnetz destabilisieren. 7. Ausblick Mit den genannten Herausforderungen sind wichtige Themen für die Elektrik/ Elektronik von Zwei-Spannungs-Bordnetzen mit 48 Volt andiskutiert und erste Lösungsansätze beschrieben. Weitere Themen, die diskutiert werden müssen, sind z.B. Crash- Abschaltung, Zuverlässigkeit von elektrischen Kontakten und Bordnetzstabilität im Falle transienter Hochleistungsverbraucher sowie die Relevanz hinsichtlich funktionaler Sicherheit und der Umgang in Produktion, Montage, Transport und Service. Langfristig ist zu erörtern, welche Verbraucher letztendlich auf 48 Volt verlagert werden können und welche neuen Funktionen, die heute mit 12 Volt nicht elektrifiziert werden können, dort dargestellt werden. 35 Preface Literatur [1] Ottmar Sirch, Georg Immel, Hartmut Pröbstle, Rupert Neudecker, Joachim Fröschl, Zukunft Energiebordnetz - Von der energetischen Optimierung zum neuen Gesamtkonzept, VDI 14. Internationaler Kongress Elektronik im Kraftfahrzeug, Baden-Baden, Oktober 2009. [2] Hartmut Pröbstle, Rupert Neudecker, Ottmar Sirch, Power Supply in Future Start-Stop- Systems, Energiemanagement und Bordnetze IV, München, März 2011. [3] T. Dörsam, Daimler AG, Böblingen, D. Grohmann, Daimler AG, Sindelfingen, S. Kehl, Porsche AG, Weissach, A. Klinkig, Volkswagen AG, Wolfsburg, A. Mai, BMW Group, München, O. Sirch, BMW Group, München, A. Radon, Audi AG, Ingolstadt, J. Winkler, Audi AG, Ingolstadt, Die neue Spannungsebene 48 Volt, VDI 15. Internationaler Kongress Elektronik im Kraftfahrzeug, Baden-Baden, Oktober 2011. [4] Fröschl, T. Kohler, A. Thanheiser, H.-G. Herzog, Intelligente Bordnetzkopplung am Beispiel Zweispannungsbordnetze mit 12 V und 48 V, Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und Elektrisches Energiemanagement, Miesbach, April 2012. [5] Herausforderungen und Potenziale von 48-V-Startsystemen, Michael Timmann, Martin Renz, Oliver Vollrath, ATZ, März 2013 [6] Peter Meckler, Störlichtbögen in Automotive-Bordnetzen, Fahrzeugelektronik 6/ 2012 [7] J. Bast, M. Kilger, W. Galli, A. Eiser, H.-W. Vaßen, I. Kutschera, Die Chancen des Antriebsstrangs durch das 48V-Bordnetz, 34. Internationales Wiener Motorensymposium 2013 [8] Franck Briault, Salah Benhassine, Marco Klingler, Michel Maher, Naguib Rezkalla, EMC Simulation Approach for 48V Systems Integration, eehe 2013 [9] Matthias Muth, Bus Systems within 48V Vehicle Networks - New Challenges for the Physical Layer, eehe 2013 [10] Paul Boucharel, Alexander Nuss, Thomas Knorr, Tobias Galli, Reduction of CO2 in a Low-Voltage Mild Hybrid Vehicle - Conditions, Challenges, and Realization, Electric & Electronic Systems in Hybrid and Electric Vehicles and Electrical Energy Management, Bamberg, April 2013 [11] Joachim Fröschl, Dominik Ilsanker, Hans-Georg Herzog, Investigation of a State Triggered Energy Management System, Electric & Electronic Systems in Hybrid and Electric Vehicles and Electrical Energy Management, Bamberg, April 2013 [12] Tom P. Kohler, Joachim Fröschl, Hans-Georg Herzog, Systemansatz für ein hierarchisches, umweltgekoppeltes Powermanagement, Elektrik/ Elektronik in Hybrid und Elektrofahrzeugen, München, März 2010 36 Magnet-Free Electric Machines & Drives for Electric Vehicles John Grabowski, Andrea Colognese Abstract The use of rare earth magnets in electric vehicle motor designs is becoming less desirable. Electric vehicle motor manufacturers are being forced to consider alternative designs without the use of rare earth materials that will impact the cost and complexity of future designs. Three electric machine alternatives and corresponding drive inverter topologies required to support these motor drive systems are briefly described in this publication. In addition, this publication highlights the operation of alternative circuit topologies in the context of their respective circuit diagrams. A novel field coil winding transformer system is also examined. 1 Freedom from Magnets for Electric Vehicles High power density electric machines for electric and hybrid electric vehicles (xEV) have traditionally been constructed using rare earth material based permanent magnet motors. The increased scarcity of rare earth materials such as Neodymium, Samarium and Dysprosium are causing the cost of manufacturing electric machines to increase at an alarming rate. As a result, many manufacturers are considering alternate designs that do not contain the expensive materials. Currently, these magnets are constructed using the rare earth elements neodymium, samarium, and dysprosium predominantly mined in China (>95%). Exports are being restricted as a result of an expanding domestic market and a policy of relocating magnet manufacturing to China, thereby multiplying the costs of raw materials for magnet manufacturers in Europe. In accordance with EU objectives to remove, or greatly reduce, the need for heavy rare earths in permanent magnets, the ROMEO initiative has been initiated to research and develop several novel microstructural-engineering strategies that will dramatically improve the properties of magnets based purely on light rare earth elements. In addition, it is commissioned to develop a totally rare-earth-free magnet; aiming to drastically reduce Europe’s dependence on Chinese imports while shifting emphasis in magnet manufacturing from a raw-materials-dependent business to one that is essentially knowledge-based, and flourishing in Europe [1]. Due to the increasing complexity in obtaining rare earth magnetic materials, the industry is considering the use of alternative electric machine types. The induction machine, the separately wound rotor machine, and the synchronous reluctance machine are becoming attractive alternatives to motors built with permanent magnetics. Since automobile manufacturers are now actively researching the use of these alternative electric 37 Magnet-Free Electric Machines & Drives for Electric Vehicles machine designs, many new system level design considerations will need to be addressed. The alternative machine types require different configurations of sensing, control and electrical excitation. These system level design considerations will be addressed along with the alternate configurations of the power inverters used to drive these machines. 1.1 Magnetic Material Availability Electric Vehicles (EV’s) are now being manufactured in record numbers. Previously, advanced battery system materials have been the major topic with respect to the availability of specialized materials. At the present moment, these specialized materials have been added to a list of materials with questionable availability for the manufacture of EV’s. The rare-earth crisis is particularly critical for heavy rare earths such as dysprosium that are currently required to assure the high temperature performance of the magnets. China has invested heavily in the rare earth metals process but its crackdown on mining, smelting and other polluting industries is forecast to slow supply. It already helped the neodymium price to a two-year high. “Rare earth production is as bad as you can get in terms of environmental damage,” the trader said. Global demand of 31,700 tons for neodymium last year already outstripped supply by 3,300 tons, he said. Demand was expected to climb to 34,200 tons this year and 38,800 tons in 2019, leaving larger deficits [2]. It has often been said - "Rare Earths are the 'vitamins" required for the shift from a carbon based economy to the new 21st century electron economy." This is why China has focused on them in the past, restricted their export, and continue to dominate rare earths production. China also dominates rare earths resources and use, as they have successfully moved up the end use supply chain [8]. There are 17 rare earth elements either heavy rare earths or light rare earths depending on their atomic weight. China dominates the rare earth industry with 105,000 MT of production in 2016 (NB: Australia is second with a mere 14,000 MT). China has recently cracked down on illegal rare earth mining and is expected to put an annual limit on its rare earths production beginning in 2020. Both of these events help rare earth pricing, and should help support non-China rare earth companies [7]. 1.1.1 Not so rare Rare earth elements are a group of metallic materials that are grouped together on the periodic table. They are not actually rare, but are often found together in the same minerals. Despite their name, rare earths are found in many places around the world, but the process of extraction is difficult and expensive since it requires separating multiple different metals from a single deposit. This is unlike the much simpler process, for example, of recovering copper from ore. They don’t collect into large deposits, like iron or copper, they are relatively difficult to extract from ore and are thus expensive. They are used in small amounts in everything from specialty glass for medical lasers (gadolinium, yttrium) to super magnets (neodymium, dysprosium, praseodymium, and samarium). Although rare earth elements can be found in many places around the world, 38 Magnet-Free Electric Machines & Drives for Electric Vehicles parts of China - including Inner Mongolia - are the primary source for the rare earth materials used today [2]. The supply of rare earth elements wasn’t always like this. In the 1960s through the 1980s, the Mountain Pass Rare Earth Mine on the California and Nevada border was the world’s leading producer. By the 1990s, however, Chinese companies began buying up production, eventually controlling the world’s production of the useful minerals. Meanwhile, the company that owned the Mountain Pass Mine in California went bankrupt and the mine was purchased by a Chinese company in 2017 [3]. 1.1.2 Security concerns Having a monopoly on rare earth elements has allowed the Chinese government to control the market. Pricing on rare earth elements spiked when China imposed a 40% cut in its production quota in July 2010. In September 2010, China instituted a temporary ban on shipments of rare earth elements to Japan in retaliation for the Japanese Coast Guard detaining a Chinese fishing boat captain [3]. Because rare earth elements are in use in so many high technology products, control of their production by a single government raises definite national security concerns. Aside from the magnets in electric motors, the US military’s need for rare earth elements includes use in communications, devices, weapons systems, guidance, targeting, and control systems - just about everything having to do with modern warfare [3]. The auto industry and the emerging electric vehicle market are also particularly sensitive to the price and availability of rare earth elements. A typical EV has around 2 kilograms of rare earth elements in the permanent magnets of its drive motor(s). In addition to use in traction motors, other smaller electric motors - like those powering the air conditioning, power steering, and even the electric windows - all potentially use magnets made with rare earth elements. China’s export production quotas are in place for raw materials, but not for finished products. This means that Chinese manufacturers have been able to increase their capacity, building and selling powerful magnets and electronic devices on the world market [3]. 1.2 Alternative Electric Machine Types The Interior Permanent Magnet (IPM) electric machine is currently the most popular choice for the majority of modern EV’s. In the search for applicable alternatives there are three main types identified. The AC Induction Machine (IM), the Separately Excited Synchronous Machine (SESM), and the Switched Reluctance Machine (SRM). These three alternatives provide magnet-free design choices as an alternative to IPM type electric machines for EV’s. Each of these choice have their pros and cons for use in these applications. 1.2.1 Induction Machines for EV The IM is a popular magnet-free machine currently used by many EV manufacturers. It is primarily chosen for its low cost, excellent reliability, and favorable fault modes. 39 Magnet-Free Electric Machines & Drives for Electric Vehicles The IM is simple and inexpensive due to the absence of slip rings, brushes, and commutators. It has a wide range of acceptable efficiency over the majority of the operational speed range. Unfortunately, it has very low power factor, especially at low speeds, which require additional control effort to improve. This results in poor low speed efficiency, higher starting current, and low starting torque. Its configuration is quite simple consisting of a wound three-phase stator and a squirrel cage rotor as shown in Figure 1. Figure 1. The Induction Machine One fundamental difference between Permanent Magnet (PM) motors and the IM is that the IM must use electric power to produce the rotor magnetisim. When the stator winding is energized with a three-phase sinusoidal voltage, it produces a current which builds up a magnetic field in the stator. The stator field induces a field into the rotor, which in turn generates currents in the shorted rotor windings, resulting in the magnetic rotor flux. The stator and rotor magnetic fields repel to create the force that produces torque and causes the machine to rotate. As the EV market has matured, the IM performance has been continuously improved. Most recently, Tesla has achieved significant improvements in their IM using a proprietary rotor design fabricated with copper. These enhancements have made IM a popular choice in the construction of traction drive machines for electric and hybrid electric vehicles. 1.2.2 Separately Excited Synchronous Machines for EV The Separately Excited Synchronous Machine (SESM) is another candidate considered as a substitute for electric vehicle PM machines. This machine type is new to the EV market, but has been common in industrial applications for many years. It is a synchronously operated motor that consists of a wound three-phase stator and a wound single-phase rotor. The rotor is energized by the use of a pair of slip rings. It is known for being very durable, having high efficiency, producing a constant power curve over a very wide speed range, and having very large peak torque. Another major advantage is the ability to control the power factor over a wide speed range. This is due to the 40 Magnet-Free Electric Machines & Drives for Electric Vehicles direct control the rotor flux controlled by the rotor field current. This facilitates the machines ability to run in a weakened flux mode necessary when operating the machine at high speeds. The configuration of a typical SESM is shown in Figure 2. In typical PM machines a large magnetic flux is required for high torque operation, but it hampers the machine’s ability to operate at high speeds due to the large back-emf. In PM machines, these large flux levels must be reduced through the use of field weakening for high speed operation. Field weakening is a process where the stator creates a counter field to reduce the field created by the rotor magnets. This causes a decrease in PM machine efficiency, in the high speed range, where the vehicle spends a great deal of time. Since the SESM has direct control of the rotor current, and thus rotor flux, it can easily reduce the flux levels as required for high speed operation providing an efficiency advantage over the PM machine in an important speed range. In the event of certain system failures, it is desirable to de-energize the machine as quickly as possible. With its direct control of rotor flux, the SESM can quickly remove the rotor excitation and quickly reduce the machines output torque. An undesirable aspect of the rotor windings is their requirement for slip rings, required to conduct the rotor current across the rotating boundary. Unfortunately the use of slip rings significantly reduce the machine’s reliability. “Contactless rotary transformer” designs have been proposed to improve the reliability. These transformers have been implemented using axial airgap type transformers. When these transformers are used they also require additional support circuitry to allow them to function. The addition of a primary-side converter is needed to couple the current across the air gap boundary. In addition, since a transformer coupled solution can only deliver an AC voltage to the secondary, a secondary-side rectification circuit is also required. Figure 2. The Separately Excited Synchronous Machine 41 Magnet-Free Electric Machines & Drives for Electric Vehicles 1.2.3 Switched Reluctance Machines for EV With the definite advantages of low material cost, robust structure, high efficiency, and satisfactory power density the SRM motor drives have been considered as a promising candidate for EV applications. Without any installation of PM material or windings on its rotor, the SR motor drives enjoy the higher cost-effectiveness and wide-speed operation range, as compared with its counterparts . To be specific, unlike the induction and PM motor drives, SRM motor drives can relieve the mechanical problems caused by the centripetal forces at high-speed operation [4]. The drawing of a SRM is shown in Figure 3. The SRM has a solid salient-pole rotor and a stator constructed of silicon steel laminations with wound stator coils. Its simple operation is based on reluctance torque through the creation of flux paths from stator to rotor and switching the applied stator current to guide the rotor in a rotational direction. The constantly changing flux path causes a large nonlinear effect in the machines inductance which can make the analysis and control of the SRM difficult. Another advantage of the SRM is an improved fault operation. If one of the phase coils or coil drive circuit fails the machine can still be operated in a limited operational mode using the remaining coils. The SRM has some disadvantages in vibration, acoustic noise, and torque ripple. The majority of the electromagnetic forces that are generated within a conventional SRM do not contribute to the motion. In fact a significant part of these forces will initiate undesirable vibrations that have been identified as a major drawback for SRM drives [5]. Attempts have been made to alleviate the vibrations and torque ripple by implementing a double-stator designed SRM. The rotor position must be known (or predicted) with a high degree of accuracy, and the current control loop for each winding must be very fast, much faster than in an AC induction motor inverter, to get the best performance from the SRM. Also, the effects of the change in the resistance of the windings with temperature must be accommodated. All of these demands add up to a very computationally-intensive control strategy for the SRM, which more or less explains why they have been sitting on the dusty shelves since the 1840s. There simply wasn’t enough computing power in programmable logic or microcontroller ICs to operate them until very recently [6]. Figure 3. The Switched Reluctance Machine 42 Magnet-Free Electric Machines & Drives for Electric Vehicles 1.3 Alternative Inverter Drive Topologies One factor common to all of the alternate magnet-free machine designs is the requirement for specific electronic drives needed to power them. Each machine type has specific stator and rotor windings that need to be excited with the correct voltage, current, and frequency. These excitations are formed by using the proper circuit topologies specific for the particular machine type. 1.3.1 Induction Machine Inverter Drive The IM inverter drive consists of a three-phase power bridge usually populated with Insulated Gate Bipolar Transistor (IGBT) devices. Its schematic is shown in Figure 4, using paralleled combinations of the ON Semiconductor FGY160T65S_F085 devices for the power bridge. In general, the bridge devices are typically rated at 600V to 1200V depending on the vehicles battery voltage. The purpose of this three-phase power bridge is to control the conduction of the currents in each of the three motor phases. These currents are monitored by Hall-effect sensors which condition and route the current signals to an inverter motor controller. Also, the position of the motor rotor is measured by a high resolution absolute position sensor and sent to the inverter motor controller. The inverter motor controller has the job of calculating the IGBT drive signals based on the phase currents, rotor position, motor temperature, and driver acceleration demands. The controller must also monitor the system inputs to determining the existence of any faults and take the appropriate actions. Figure 4. The Induction Machine - Inverter Topology 1.3.2 Separately Excited Synchronous Machine Inverter Drive The SESM inverter drive has a considerably different configuration from that of the IM inverter drive. On the stator side, there is a three-phase power bridge, shown schematically in Figure 5, which is very similar to that of the IM drive. However there is additional circuitry required to provide the current needed to create the rotor field flux. The additional circuitry is needed to properly energize the rotor field coil. Since the machine output torque is proportional to the rotor flux level and rotor field current, it 43 Magnet-Free Electric Machines & Drives for Electric Vehicles must be precisely controlled. The SESM inverter rotor field drive circuit, shown in Figure 5, shows a half-bridge converter circuit. The half-bridge circuit is used to provide positive current to the field coil when energizing and to rapidly remove stored energy when switched off. The circuit as shown uses the ON Semiconductor AFGHL40T65SPD IGBT’s and FFH50US60S-F085 diodes. Also these devices could be included in an “Exciter” module which would perform the entire field excitation function in a single package. The half-bridge circuit is controlled by the inverter motor controller which has the job of creating a Pulse Width Modulated (PWM) signal used to regulate the current level in the rotor field coil. The PWM signal causes the transistors to turn-on and turn-off at regular intervals, with varying duty-cycles. This action drives current into the rotor coil which is measured and used to correct the duty-cycle command, in a closed loop fashion. The resulting average rotor coil current is thus maintained at the desired operating level. When the transistors switch to the off-state the energy that is contained in the rotor coil need a path to dissipate, the diodes perform this function. If a motor drive system failure occurs that requires a rapid de-energization of the motor, this circuit will quickly remove any stored energy from the rotor coil. Figure 5. The Separately Excited Synchronous Machine - Inverter Topology Some SESM designs endeavor to eliminate the use of slip rings. Instead of having a physical connection to the rotor field coil, a transfromer coupled method is employed. In this method two additional circuits are needed to properly convert the signal. To excite the priimary winding an AC voltage is required. This can be achieved by the use of a full bridge converter circuit, as shown in Figure 6. Its function is to convert the DC bus voltage into a form suitable for driving the transformer primary coil. Once the AC voltage has reached the secondary coil of the transformer, it must be converted back 44 Magnet-Free Electric Machines & Drives for Electric Vehicles into a DC voltage suitable for exciting the rotor field coil. One method of accomplishing this is by using a bridge rectifier circuit, as seen in Figure 6. Figure 6. SESM Rotor Field Converter Topology 1.3.3 Switched Reluctance Machine Inverter Drive Of all of the magnet-free motor drives, the SRM has the most unique inverter drive configuration. The example shown in Figure 7, is the schematic of a three-phase SRM motor drive. The three stator coil pairs are given as L1, L2 and L3. Each of these pairs consist of two single coils that are wound around opposing legs of the machine stator core that are electrically connected in series. When current is conducted into each of the coil pairs, each pair generates flux that combines to create each pole of the stator field. Each of the other two coil pairs do a similar job in creating the stator fields for the other stator poles. Each phase of the drive circuit is constructed with a pair of IGBT’s and pair of diodes. The IGBT’s are used to conduct positive phase current into each coil pair while the diodes are used to carry the recirculating current. The drive is operated in a current control mode using PWM signals to drive each of the IGBT devices. The phase current is measured and is modulated by switching the devices on and off to regulate its average value. The drive signals are generated by a controller that uses the rotor position and other system inputs to compute the instantaneous switching requirements. Figure 7. The Switched Reluctance Machine - Inverter Topology 45 Magnet-Free Electric Machines & Drives for Electric Vehicles 1.4 Conclusions In conclusion it is apparent that there are sufficient numbers of electrical machine types which could be considered as replacements for PM based machines for EV applications. The main issue becomes just how much additional system complexity and cost would be incurred to realize these alternate solutions. Many of the alternative designs have attractive attributes which might make them an improved choice over IPM type machines. The proof of this will only become apparent after the necessary research and development has been done. One thing for sure is that the field of EV and HEV vehicle engineering is still evolving. References [1] European Commission Project - ROMEO, Grant ID 309729. https: / / cordis.europa.eu/ project/ rcn/ 105901/ factsheet/ en [2] Reuters - Environment Online, (Mar 12, 2018). Tesla’s electric motor shift to spur demand for rare earth neodymium. www.reuters.com/ article/ us-metals-autos-neodymium-analysis/ teslas-electric-motor-shift-to-spur-demand-for-rare-earth-neodymium-idUSKCN1GO28I [3] Clemens, Kevin. (August 16, 2018). Analysis: Trouble may be brewing as China’s monopoly of the critical materials needed for electric vehicles and national security may be impossible to break. https: / / www.designnews.com/ electronicstest/ analysis-growing-impact-rare-earth-elements/ 175978578959272 [4] Lee, Christopher H.T., Kirtley, J. L. Jr., Angle, M. (June 21, 2017). Switched Reluctance Motor Drives for Hybrid Vehicles. / / www.intechopen.com/ books/ switched-reluctance-motor-drives-for-hybrid-electric-vehicles [5] Bouiabady, M. M., Aliabad, A.D., Amiri, E. (June 21, 2017). Switched Reluctance Motor Topologies: A Comprehensive Review. / / www.intechopen.com/ books/ switched-reluctance-motor-concept-control-and-applications/ switched-reluctance-motor-topologies-a-comprehensive-review [6] Jenkins, Jeffery. (January 25, 2013). A Closer Look at Switched Reluctance Motors. / / chargedevs.com/ features/ a-closer-look-at-switched-reluctance-motors/ [7] Roos, Dave. (Sept. 6, 2017). China’s Monopoly on Rare Earth Elements Tightens with Purchase of US Mine. www.seeker.com/ tech/ materials/ chinas-monopolyon-rare-earth-elements-tightens-with-purchase-of-us-mine. [8] Bohlsen, Matt. (Nov. 14, 2017). A Look at the Rare Earth EV Magnet Metals and Their Miners. / / seekingalpha.com/ article/ 4124575-look-rare-earths-ev-magnetmetals-miners 46 A study of semiconductor and sensors for vehicle electrification Dexin Chen, Richard Dixon Abstract This paper discusses the major impact that electrification will have on underlying components for EVs and HEVs. IHS Markit analysis finds that 5-6 new powertrain modules result from even moderate forms of electrification. The implication of these modules - motor generator inverter, battery management system, DC-DC converter, on-board charger, and sound generator - on the market uptake of current, voltage, and temperature sensors, and including power semiconductors, is very significant. Design challenges for semiconductor and sensors are discussed and new technologies to address them are also presented. 1 Automotive market 2018 Worldwide vehicle production (defined as passenger cars and light commercial vehicles with gross vehicle weight less than six tons) amounted to 95.2 million in 2017 and has declined almost 1% year-on-year in 2018 to 94.2 million [1]. The main reason is a combination of macro economy impacts including slowing down of the Chinese car market, US-Sino trade tension and regulations like the new vehicle emissions testing procedure or Worldwide Harmonized Light Vehicle Test Procedure (WLTP) in Europe. In contrast, the total market for automotive semiconductors and sensors has grown in sales by 9.6% in 2018 despite the aforementioned vehicle shipments slowdown [1]. Clearly accelerated vehicle electrification is one of the factors leading to a much higher electronic content per car and drives the automotive semiconductor market forward. IHS Markit believes that this trend will continue for the coming years, helping to propel the total automotive semiconductor and sensor market to $68 billion by 2024, up from 43 billion of US dollar in 2018 - a Compound Annual Growth Rate (CAGR) of 8.3% over this period [3]. In particular, as new vehicle architectures are adopted, new design challenges and requirements will emerge that will open doors for new technologies, and certain components will see higher growth rate than others. Among these are power semiconductors especially power modules, temperature, current and position sensors, in addition to galvanic isolation ICs 47 A study of semiconductor and sensors for vehicle electrification Figure 1: Global Automotive Semiconductor Revenue increased 9.6% in 2018 The new requirements will also bring chances for new technologies such as devices based on xMR sensors - Anisotropic Magnetoresistance (AMR), Giant magnetoresistance (GMR) and Tunnel Magnetoresistance (TMR), not to forget power discrete or power modules based on wide band gap materials like Silicon Carbide (SiC) or Gallium Nitride (GaN). The latter will enjoy higher adoption rates despite the fact their higher expense. 2 Power semiconductors As the power source for a vehicle is shifting gradually from fossil fuel to electricity, semiconductors that control power flow will ship at higher growth rates in the coming years. This is not only because of the ramping up of hybrid electric vehicle (HEV) and electric vehicle (EV) production, but also because of the increased number of subsystems per vehicle. For example, there could be more than one e-motors per vehicle. Indeed, the power semiconductor revenue in 2018 was $9.1 billion in 2018 and is forecasted to reach 13.7 billion in 2023, a CAGR of 8% [4][2] as Figure 2 shows. From technology standpoint, as the power rating of e-motors increases, SiC or GaN devices start to be adopted by the car industry. These devices, when compared to their silicon counterparts, have higher switching frequency and thus enable a smaller system design with lighter magnetic components and system mass, which is very important for the mileage of electric vehicles and the overall efficiency of HEVs. 48 A study of semiconductor and sensors for vehicle electrification Figure 2: Automotive Power Semiconductor Revenue (2018-2023) The potential applications for SiC or GaN devices are the main powertrain inverter, on-board charger and DC-DC converter. DC-DC converters and on-board chargers using SiC Schottky barrier diodes began in mass production in 2014 and sales from a variety of suppliers are growing rapidly, although the extensive use of SiC power modules in powertrain inverters will not happen that soon due to its relatively higher price compared to Silicon based IGBT [5][2]. However, at the current pricing of SiC devices (e.g. hundreds of dollars for a full SiC module), it is argued that while the SiC device itself is still expensive, the overall system cost will be lower. 3 Sensors An electrified vehicle, compared to cars powered by internal combustion engines, require different physical signal measurements and therefore emphasize different sensing technologies. The biggest sensor category for ICE is the pressure sensor, but in a battery based EV powertrain it will be current sensor [6][5]. The HEV, which is deemed as a transition between the ICE and EV, will keep the sensor contents from both powertrain architectures and sensor content will be additive. 3.1 Current sensing Current sensors are a very fast-growing beneficiary of vehicle electrification, with as many as 15 - 30 such devices being deployed depending on level of electrification (mild, full, plug-in, etc.) and the number of motors in the architecture [5]. This does not include the EV charging infrastructure at street level. 49 A study of semiconductor and sensors for vehicle electrification From a technology standpoint, there are different solutions for current sensing, including shunt resistors, Hall ICs, xMR ICs and fluxgate sensors. The application of different current sensing solutions depends on the current measurement range and other requirements such as size, accuracy and frequency of measurement. A breakdown of the technology adoption is shown in the figure below. Figure 3: Current sensing technology used at different current ratings (IHS Markit) In the low current range up to 10 A such as module-level battery monitoring, shunt solutions will prevail because of their low cost and precision. Suppliers include Isabellenhütte and Vishay. A simple Hall IC also starts to be adopted when the current increases above 10 A. Allegro Microsystems, Melexis, and TDK-Micronas are examples of suppliers of Hall-based current ICs. xMR sensors also show potential here due to a combination of high bandwidth and accuracy. Typical suppliers of xMR sensors are Sensitec and in future AKM and Aceinna (previously MEMSIC). When the current range exceeds 100A, Hall based sensors are mostly preferred because at such current levels, shunt solutions will have power dissipation issue and the galvanic isolation is also difficult. Applications with current range 100-400A include DC-DC converters, auxiliary inverters (compressor for HVAC, etc.), electric power steering, active suspension and Integrated Starter Generator (ISG). The main inverter requires the highest current measurement range - peak ratings of up to 4000 A - and an increased switching frequency especially with the adoption of SiC or GaN power devices. Fluxgate technology has a higher precision and lower temperature drift compared to Hall and xMR sensors, but its bandwidth is much lower. Normally, this kind of device is used for battery pack busbars or on-board chargers for leak detection. Suppliers include Texas Instruments and TE Connectivity. 3.2 Position sensing The position encoding slot for the main traction is an interesting technology battleground for position sensing. Today, this is a market for resolvers from companies like Tamagawa and Minebea. Resolvers measure rotational angles as two-phase AC voltages with analog signals. However, resolvers are non-electronic and not subject to ISO26262. This is where other solutions like magnetic technology or inductive technology come to play. While xMR sensors from the likes of Infineon or TDK could 50 A study of semiconductor and sensors for vehicle electrification be used, it seems that the inherent advantages of inductive sensors regarding stray fields means this is a strong contender for this application in new platforms at European OEMs. Compared to resolvers, inductive and XMR devices are smaller and offer advantages in terms of cost and flexibility regarding functional safety. Potentially, resolvers, inductive sensors and magnetic devices could co-exist for functional safety reasons. Another consideration for choice of technology is the magnetic stray-field noise (electromagnetic interference or EMI) incurred by large electric motors and the proximity of large current carrying cables to magnetic sensors. Today, a typical solution is to use shielding or differential sensing to eliminate unwanted noise. Inductive technology has some promise as no magnets are involved. Many companies are considering this as one of the potential routes for EMI-free sensors electrified vehicles. 3.3 Temperature sensing Temperature sensors are a key protection device needed in the modules deployed in xEVs. Several types of technology exist depending on the requirements on accuracy and temperature of operation. Typically, the applications needed for xEVs are up to a maximum of 150°C and are therefore well served by negative temperature coefficient (NTC) thermistors - a set of devices that change electrical resistance with temperature based on a ceramic or polymer material and with an accuracy of ±2-3°C across the operating range. Typical of the applications in question are DC-DC converters, on-board chargers, motor generator inverters and associated traction motors, inverters and especially battery monitoring systems (see Figure 4). The features of an NTC include small size and low cost but suffer from non-linearities and require an ADC for digital output. Typical suppliers are TDK-EPCOS, Amphenol, and Littelfuse, and prices range from $0.20 to over $1 depending on accuracy required by the end use. NTCs make up most of the market volume for electrification applications today. An external ADC is needed to turn the signal into a digital output. Silicon-based ICs on the other hand cover the same range of temperature of 90°- 130°C and have higher accuracy, but this comes at a price. These devices are used in automotive applications such as the external air temperature probe place in the front bumper, or as interface ICs for other temperature sensors, e.g. thermocouples used on the exhaust. Suppliers of this kind of device include Texas Instruments, Microchip, and Analog devices, etc. Specifically, 1-2 NTC devices are used depending on the number of electric motors in the propulsion system and are used to protect the motor from overheating. Prices tend towards the higher end of the spectrum because of the robust packaging requirement (i.e. a ceramic chip inside either glass or epoxy package), and this precludes the use of simple NTC elements. Inverters need 1-2 sensors because power electronics tend to get hot. 51 A study of semiconductor and sensors for vehicle electrification However, the xEV battery monitoring systems needs anywhere from 2-10 temperature sensors as there is a need to measure as many points as feasible, although this is a limit of the battery design. Accuracy is important here. The on-board charger also features 1-2 temperature sensors, while the motor generator inverter uses an NTC inserted into the cooling fin. Figure 4: a plug-in hybrid vehicle can feature as many as 20 temperature sensors depending on the configuration (IHS Markit) 4 Galvanic isolation As new powertrain architectures introduce multiple power rails with voltage higher than traditional 12-volt board net, it is very important at the component level to have a proper galvanic isolation solution for data transmission and power supply between the high voltage and low voltage domains. For data transmission, the solution should be able to provide sufficient data rates, high channel counts, a good timing specification, low power consumption, high temperature operation and low degradation over time. Two kinds of solution are available: an optocoupler and a non-optic solution based on capacitive isolation or magnetic isolation. The non-optic solution is relatively new and is taking bigger share in the higher performance market, which was previously dominated by the optocoupler [7]. Capacitive isolation is provided by companies like Texas Instruments and is identified as the only one with large volumes of sales. Other companies like Infineon, Analog Devices and STMicroelectronics are using magnetic isolation, which is a kind of chipscale transformer or two coils close to each other. Capacitive isolation has the advantage of higher frequency operation, but compared to magnetic isolation, at a relatively lower maximum voltage. 52 A study of semiconductor and sensors for vehicle electrification Magnetic isolation can withstand higher voltages but has a problem of EMI susceptibility at higher frequency. Both solutions are replacing the optocoupler in applications like main inverters where temperature is high and aging effects should be considered. 4 Conclusion The market for both semiconductor and sensor components looks very compelling as a result of vehicle electrification. Not only the increasing penetration of HEVs and EVs, but also the large number of associated modules per vehicle requiring protection and measurement, is assuring this market. Particularly shipments of power modules, but also temperature and current sensors, will thus accelerate very strongly in the coming years, proving excellent chances for component makers focused on this space. Literature [1] IHS Autoinsight, Light Vehicle Production Forecast Januray 2019, https: / / autoinsight.ihs.com/ [2] IHS Technology, Semiconductor Competitive Landscape CLT Intelligence Service, Q4 2018, https: / / technology.ihs.com/ Services/ 548450/ semiconductorcompetitive-landscape-clt-intelligence-service [3] IHS Technology, Automotive Semiconductor Market Tracker Q4 2018, https: / / technology.ihs.com/ 529008/ automotive-semiconductor-market-tracker [4] Kevin Anderson, Power Semiconductors in Automotive Report 2018, https: / / technology.ihs.com/ 602939/ power-semiconductors-in-automotive-report- 2018 [5] Richard Eden, SiC & GaN Power Semiconductors Report 2018, https: / / technology.ihs.com/ 601312/ sic-gan-power-semiconductors-report-2018 [6] Richard Dixon, Magnetic Sensors for Automotive Report 2018, https: / / technology.ihs.com/ 608224/ magnetic-sensors-for-automotive-report- 2018 [7] Jamie Fox, Optoelectronic components report 2015, https: / / technology.ihs.com/ 518989/ optoelectronic-components-2015 53 High Power Charging - Consideration of the cost parameters of fast charging from vehicle to charging infrastructure Sebastian Rickert 1 Market development, drivers and business models of HPC infrastructure eMobility is a clear trend in the automotive industry and will soon dominate new vehicle registrations. The providence of a sufficient electric range is one of the major challenges on the way there. An enabler that is associated with this issue is battery technology. With higher energy densities, achieved through advancing developments in battery technologies, costs are shrinking and electric vehicles can be equipped with more battery capacity. In 2030, about 34% of new vehicle registrations in the EU will have an electric powertrain 1 . A second, no less important enabler is the availability and performance of charging infrastructure. With higher charging power and a dense charging network, electric range can be artificially extended. Therefore, in the following, the scope of DC High Power Charging 2 (HPC) with focus on total costs will be presented. Due to the prevailing 400V system topology of battery electric vehicles (BEVs), the standard fast charging power used to be 50kW. Nowadays, leading BEV vehicle manufacturers (e.g. Tesla, Audi) achieve power levels of up to 150kW at this system voltage level. From here on, technological obstacles have an influence on the implementation of such systems. An increased charging power at a constant voltage level inevitably generates a higher current according to the power principle. The resulting heat generation represents a challenge for passively cooled charging systems. As a solution, an actively cooled cable and plug is implemented to control the thermal situation and improve the mechanical characteristics (thin and light cables). An additional increase of the system voltage to 800V leads to a further leap of charging performance to up to 350kW. In this context, the design of cell chemistry and safety measures, such as specific battery cooling concepts, are key for a successfully operating High Power Charging process. By using innovative battery and charging technologies as well as implementing >800V charging voltage, recharging stops can shrink to well under 20 minutes for 400km. 1 Calculation based on EU regulations for Original Equipment Manufacturers (OEMs) regarding CO 2 emissions (P3 CO 2 -Tool) 2 A charging power of 150kW or higher is defined here as “High Power Charging” (HPC) 54 High Power Charging Consideration of the cost parameters of fast charging from vehicle to charging infrastructure *Assumptions: 23kWh/ 100km highway consumption, charging without degradation Figure 1: Charging times of different power levels (Source: P3) In order to give EV drivers a central access to charging stations of different providers, so-called eRoaming platforms are implemented. The aim is to reduce transaction costs and the effort for the customer by replacing contracts between EV drivers and a large number of stakeholders of charging infrastructure, such as CPOs, MSPs or energy suppliers. Instead of many individual contracts, a user now only has one single contract with a single provider, which considerably increases customer satisfaction. As a link between Mobility Service Providers (MSPs) and Charge Point Operators (CPOs), the platforms thus reduce the technical effort for all parties, since an MSP only needs one contract with such a platform. The platforms provide a link between the MSP's customer data and the CPOs' charging station data. However, due to the competitive situation between the platforms, the current cost situation is highly intransparent for the customer. For maximum coverage, each charging station operator or MSP must conclude a contract with both networks, which in turn results in double costs. For this reason, market players must carefully consider which platform is the most beneficial for their own customers. Figure 2: Schematic cash-flow landscape in Germany (Source: P3) 55 High Power Charging Consideration of the cost parameters of fast charging from vehicle to charging infrastructure 2 State-of-the-art technology and future technology costs The DC fast power charging process is enabled through several on-board components in the vehicle as well as external charging equipment. With the new HPC requirements, both areas face challenges that will result in development costs and higher expenses. For this reason, a rough cost estimation will be given (cost indication). On the vehicle side, the battery cell level, the pack level and the system level of the battery as well as the HV booster must be considered. High C-rates are causing thermal stress for the battery cells, which increases the risk of aging, degrading or even damaging the battery. Therefore, modified anode and cathode materials are required. Moreover, battery cells need to be integrated into intelligent battery systems. Cost indication: high As mentioned before, HPC rates require very effective cooling of the battery system. There is no optimal solution yet for automotive HV battery applications, which means that each cell format leads to an individual design for thermal battery management. Cost indication: high Safety components like switches, contactors and fuses as well as electric components like bus bars and insulations need to be upgraded for higher current and voltage rates. An upgrade for these components is manageable. Cost indication: low Charging an EV with an 800V-system topology at a 400V-charging station, can not be done without technical adjustments. One possible way is the use of an onboard HV-booster, which transforms the incoming voltage of 400V from the charging station to a higher level to feed into the battery. Whereas the technology is already realizable, the use of the HV booster depends on the vehicle architecture and the opportunity costs regarding alternative solutions. Cost indication: medium Figure 3: Exploded view of a battery cell (Source: CCI) Figure 4: Battery module (Source: VW) Figure 6: Basic topology of an HV-Booster (Source: Porsche) Figure 5: Setup of an HV-fuse (Source: Autoliv) 56 High Power Charging Consideration of the cost parameters of fast charging from vehicle to charging infrastructure At the off-board side, the charging cable and plug, the AC/ DC converter, the charging station peripherals and the grid connection are relevant hardware components. As mentioned above, the charging cable and plug must be active cooled to counteract occurring thermal stress during charging processes. While suppliers already have developed products with an integrated liquid-cooling cycle, long-term and real-life validation are still ongoing. Cost indication: medium Technologies for AC/ DC converting are already existing and well known in the field of power electronics. A feasible challenge here is to introduce standardized power levels to the market aiming to cover all power level with a platform setup to achieve economies of scale effects. Cost indication: low Besides the converter, charging cable and plug, safety components of the charging station must withstand higher current and voltage ratings. Furthermore, additional components for DC metering and thermal sensors are required. However, since off-board components do not have strict packaging and weight restrictions, these are manageable requirements. Cost indication: low A far bigger challenge is the grid connection for HPC charging parks. The extreme power demand, which can occur at times of high utilization, must be guaranteed by a sufficiently high installed capacity of the grid connection. Although this can be ensured by a connection to the medium-voltage grid, it represents a major cost driver due to the expensive grid connection fees. An indication of the costs incurred will be shown in the next section. Cost indication: high Figure 7: Exploded view of a charging plug (Source: Phoenix Contact) Figure 8: Block diagram of a DC charging system configuration by use of converter modules (Source: Benning) Figure 9: Example of HPC-stations (Source: ABB) Figure 10: Example of an HPC-park (Source: P3) 57 High Power Charging Consideration of the cost parameters of fast charging from vehicle to charging infrastructure 3 Total cost structure and possible cost levers Offering charging infrastructure is a complex business with several issues from grid connection to efficient operating. This section is intended to give an approximate indication of average costs for an 150kW-HPC charging station with two charging points installed. It is important to emphasize that the values shown below have been significantly simplified and are only giving an average cost indication. Factors like the location have a serious impact on the profitability/ business case. For ease of reference, a distinction is made between one-time investment costs (CAPEX) and annual operating costs (OPEX). A cost breakdown of the CAPEX can be divided into five items:  Charging Hardware (Charging Station), which consists e.g. of power electronics, charging cable, plug and the user interface.  One or more transformers to reduce the grid voltage from medium to an appropriate level.  Mounting and commission costs to set up the charging station and get it ready for operation.  The above-mentioned grid connection is the second biggest cost item summing up to about 40,000 Euro in average.  Planning and Permission, for implementing the charging station depending on its location. Representing the biggest cost item, the charging hardware also feature the biggest potential for cost savings. This relates to the future technology leaps in the area of power electronics. P3 estimates cost saving potential of 20 to 30% for the next generation of power electronics (in 5-7 years). Since the implementation of more than one charging station at a single location leads to economies of scale effects, another cost potential can be achieved in the area of planning and permission as well as mounting. In conclusion, the average CAPEX for an HPC-station sums up to about 160k€, with a cost saving potential of 15k€ within the next years. Figure 11: Cost breakdown of average CAPEX (Source: P3) 58 High Power Charging Consideration of the cost parameters of fast charging from vehicle to charging infrastructure On the other side, OPEX are divided in the following items:  Costs for maintenance and repair as biggest operating cost driver owing for example to the regular exchange of air filters.  Grid usage fee to the grid operator for the providence of energy at all times.  Site rental for the ground on which the charging station was built on.  Service and IT backend costs for the execution of the charging process, invoicing, etc. When not considering depreciation and energy costs, the average annual operating costs amount to about 13,000€. This corresponds to a comparably high share of about 10% in relation to the CAPEX. A cost potential for service and IT backend can be expected with more providers joining the market. In addition, an ideal charging park consists of further elements, such as stationary battery storage systems, photovoltaics systems and intelligent charge control, which, although they represent a further investment, can reduce operating costs. PV systems and stationary storage systems reduce electricity procurement costs, especially in peak load situations. The use of such elements depends on the business case, in which the location of the charging station plays an essential part. At the end of this paper a simplified cost calculation for a single charging process at an HPC charging station with 150kW shall be carried out. The first step is to establish a suitable cost structure. Besides the already known average CAPEX and OPEX, costs for electricity, billing, taxes and overheads have to be considered. In addition, various static assumptions are made, which imply a certain degree of blurriness, but are required for the price calculation at different points in time. Figure 12: Cost breakdown of average OPEX (Source: P3) 59 High Power Charging Consideration of the cost parameters of fast charging from vehicle to charging infrastructure  Considered points in time: snapshot in 2018 and 2025  Period of depreciation: 10 years  Billing: no fee to an aggregator  Average charging power per event: 35kW (80kW in 2025)  Average consumption per event: 40kWh (50kWh in 2025)  Charging station utilization: 5% (10% in 2025)  Electricity price: Industrial tariff - - - With the given assumptions, the costs of a kWh per charging event in 2018 will be around 1.85€. In the year 2025, they could be reduced by 64% through decreasing shares of CAPEX and OPEX. The reasons for this can be summarized in three effects: 1. Technology improvements - Technological developments and economy of scale effects in power electronics the investment costs (CAPEX). 2. Charging Capabilities - EVs in 2025 having bigger battery capacity and are capable of charging with a higher power. These developments will lead to the distribution of CAPEX and OPEX to a higher power demand. 3. BEV Sales - In 2025 the share of new BEV registrations per year is about 34% eventually leading to a higher charging station utilization, further decreasing the CAPEX and OPEX. 4 Conclusion This paper highlighted the area of High Power Charging with focus on total costs. While the market conditions were described initially, technological challenges of different HPC-component were described afterwards. Subsequently, cost structures regarding one-time investments, operational costs and a cost breakdown of a charging event were presented. In summary, the following statements were made: 1. The EV market is growing massively within the next decade. In 2030 about 34% of European new vehicle registrations will have an electric powertrain. Figure 13: Cost structure per charging event & kWh (Source: P3) 60 High Power Charging Consideration of the cost parameters of fast charging from vehicle to charging infrastructure 2. To enable HPC on-board components (especially the battery cell and module), as well as external charging equipment must meet challenging requirements. 3. The construction of HPC infrastructure (≥150kW) is costly and must therefore be carried out as required along the vehicle availability with increasing penetration rates and grid coverage. 4. The EV market ramp-up, developments in charging technology and the implementation of technical solutions can optimize the cost structure and result in high cost savings. 61 Funded Project for smart charging infrastructure for EVs up to 22kW Detlev Endner, Michael Kahlstatt, Roland Matthé Abstract The inclining electro mobility is accompanied by an increasing demand for charging infrastructure. This rising demand results into a challenge for the local energy supply companies and their distribution networks. In order to better understand the needs of grid expansion for electro mobility and to avoid supply network expansion, Opel Automobile GmbH has launched several research projects on intelligent charging infrastructure with partners from industry and the public sector. The article shows examples of intelligent load management and the development of an energy storage container as a supply network buffer. The presentation will show the status of the recently launched projects. Kurzfassung Mit der Zunahme der Elektromobilität geht ein gesteigerter Bedarf an Ladeinfrastruktur einher. Dieser gestiegene Bedarf stellt eine Herausforderung an die lokalen Energie- Versorgungsunternehmen und deren Verteilnetze dar. Um den Bedarf des Netzausbaus für die Elektromobilität besser verstehen bzw. Netzerweiterungen vermeiden zu können, hat die Opel Automobile GmbH mehrere Forschungsprojekte zum Thema intelligente Lade Infrastruktur mit Partnern aus Industrie und öffentlicher Hand gestartet. Im Beitrag werden Beispiele zum intelligenten Lastmanagement gezeigt, sowie die Entwicklung eines Batterie Containers als Energiepuffer. Im Vortrag wird der Stand der soeben angelaufenen Förderprojekte dargestellt. 1 OPEL wird elektrisch Opel wird bis 2020 vier Modelle in elektrischen Varianten auf den Markt bringen. Die nächste Generation des Corsa wird es also auch als eine rein batterieelektrische Version geben, den Nachfolger des Transporters Vivaro und das Nachfolgemodell des Kompakt-SUVs Mokka X, sowie den Grandland X als Plug-In-Hybrid. 2024 werden alle unsere Modelle in einer elektrifizierten Variante verfügbar sein. Opel wird dann eine elektrische Pkw-Marke sein! 62 Funded Project for smart charging infrastructure for EVs up to 22kW 2 EV Charging Projects Die besten E-Autos sind kaum etwas wert ohne eine funktionierende Infrastruktur. Der Kunde möchte sein Auto fast jeden Tag aufladen können. Auch wenn es nicht die Kernaufgabe eines Automobilherstellers ist für Ladeinfrastruktur zu sorgen, sind wir uns in dieser immer noch frühen Phase der Technologie unserer Verantwortung bewusst. Auch wir haben unseren Beitrag zu leisten, um die E-Mobilität voranzubringen. So benötigen wir auch im Entwicklungszentrum in Rüsselsheim und im Testzentrum Dudenhofen eine entsprechende Ladeinfrastruktur, damit wir ab 2020 ein zeitgemäßes Umfeld für unsere Fahrzeugentwicklung abbilden können. Wenn die Entwicklung einen gewissen Reifegrad erreicht hat, kommen der Qualitätsbereich, die interne Testflotte und die produktionsbegleitenden Bereiche ins Spiel, die zusätzliche Lademöglichkeiten benötigen. Nicht zu vergessen sind dann natürlich die Nutzer von Firmenwagen auf dem Opel Gelände und die Mitarbeiter, die eine Elektrofahrzeug kaufen oder leasen werden und auf dem Mitarbeiterparkplatz während der Arbeitszeit laden möchten. Summa Sumarum rechnen wir mit ca. 1.100 Ladepunkten, die benötigt werden von der Entwicklung bis hin zu den Mitarbeiter-Parkplätzen. Opel ist daher in verschiedenen spannenden, innovativen Projekten engagiert, vor allem an unserem Hauptsitz in Rüsselsheim. 3 REALLABOR E-MOBILITÄT Bild 1: Reallabor E- Mobilität Aktuell ist Opel in zwei Förderprojekten aktiv, die unser Vorhaben unterstützen: Das E-Mobility Lab Hessen und Electric City Rüsselsheim (CLEVER). 63 Funded Project for smart charging infrastructure for EVs up to 22kW 3.1 E-Mobility Lab Hessen Das sogenannte E-Mobility-Lab Hessen‘ Projekt wird vom Hessischen Ministerium für Wirtschaft, Energie, Verkehr und Landesentwicklung mit Mitteln des Europäischen Fonds für regionale Entwicklung (EFRE) gefördert. Im Entwicklungszentrum Rüsselsheim und im Testzentrum Dudenhofen forschen wir gemeinsam mit der Universität Kassel und der Flavia IT am optimalen Aufbau des Stromnetzes der Zukunft und dem intelligenten Laden an verschiedenen, miteinander vernetzten Ladepunkten. Die Installation der ersten Ladesäulen auf dem Opel-Gelände in Rüsselsheim sowie in unserem Testzentrum in Rodgau-Dudenhofen ist schon gestartet. Insgesamt werden mehr als 160 Ladepunkte entstehen, mit denen die E-Automobilflotte unseres Entwicklungszentrums künftig geladen wird. Der Clou: Die Dichte von Elektrofahrzeugen in unserem Engineering-Zentrum wird eine Mobilitätssituation abbilden, wie sie im Jahr 2035 erwartet wird. So werden wir gemeinsam mit unseren Projektpartnern wertvolle Erkenntnisse über das Ladeverhalten und die Anforderungen an das Stromnetz der Zukunft erhalten. Mit Hilfe des intelligenten Steuersystems ‚Gridware‘ werden Ladestrom und -Zeitpunkt an die jeweilige Nutzung der Fahrzeuge angepasst. Mit dieser smarten Infrastruktur wird realisiert, dass die gesamte Flotte, trotz des hohen Energiebedarfs der Elektrofahrzeuge, jederzeit bedarfsgerecht geladen ist. Und dies bei minimalem Ausbau des bestehenden Stromnetzes. Die Ziele dieses spannenden Projekts: Opel wird in einem sogenannten Real-Labor erforschen, wie man ganze Ladefarmen und einzelne Ladepunkte intelligent steuern kann, um den Ausbau der Infrastruktur auf das Nötigste zu reduzieren. Diese Erkenntnisse sind enorm wichtig für einen ökonomischen Ansatz zum Ausbau der Infrastruktur im öffentlichen Raum. Das wiederum ist ein entscheidender Erfolgsfaktor für die Elektromobilität. Projektpartner: Universität Kassel, Hessischen Ministerium für Wirtschaft, Energie, Verkehr und Landesentwicklung, Flavia IT, House of Energy, Plug’n Charge Dauer: 3 Jahre 64 Funded Project for smart charging infrastructure for EVs up to 22kW 3.2 Elektrik City Rüsselsheim Bild 2: Rüsselsheim Electric City Dieses ehrgeizige Projekt ist eine Partnerschaft mit der Stadt Rüsselsheim und der Hochschule RheinMain, die das Projekt wissenschaftlich begleitet, und wird vom Bundeswirtschaftsministerium gefördert. Gemeinsam gestalten wir den Aufbruch in das Zeitalter der Elektromobilität und treiben massiv den Ausbau der Ladeinfrastruktur voran. Bis zum Jahr 2020 werden am Opel-Stammsitz 1.300 Ladepunkte installiert im gesamten Stadtgebiet und auf dem Opel-Gelände. Davon befinden sich ca. 600 auf den Mitarbeiterparkplätzen von Opel und knapp 350 auf dem Opelgelände für Firmenfahrzeuge. Damit hat die Stadt künftig mit die höchste Dichte an Ladestationen in der gesamten EU und wird so zur „Electric City“ Rüsselsheim. Fakten und Projektpartner: Projektname: CLEVER-(Charging Low Emission Vehicles in Rüsselsheim) im Rahmen des Sofortprogramm saubere Luft 2017-2020 Partner: Stadt Rüsselsheim, Entega Energie GmbH, Bürgerenergie Untermain eG, Hochschule Rhein Main, Gewobau Rüsselsheim, B2M Software GmbH Dauer: 2018-2020 Volumen: 1,9 Mio.€ 65 Funded Project for smart charging infrastructure for EVs up to 22kW 4 LASTMANAGEMENT Die Ladesäulen, verschiedener Hersteller, werden mit unterschiedlichen Lastmanagement Strategien betrieben. Man muss nicht zwingend kostspieligen Netzausbau betreiben, um den massiven Anstieg der Zahl an Elektroautos und deren Lade-Anforderungen Herr zu werden. Durch intelligentes Lade-Management und die Kappung der Lastspitzen, unterstützt durch dezentrale Energiespeicher, entstehen zwei Ebenen, die netzdienlich wirken. Dadurch kann sehr viel Geld durch die Vermeidung von Netzausbau gespart werden. Das intelligente Lastmanagement System muss in der Lage sein, verschiedene Verbraucher-Topologien zu bedienen und zu steuern. Die gesamte Kette vom Transformator bis zum einzelnen Ladepunkt muss betrachtet werden, inklusive aller anderen angeschlossenen Verbraucher, die einen Einfluss auf die verfügbare Leistung haben. Ebenso die leistungsbeschränkenden Kabelquerschnitte und Absicherungen in den Haupt- und Unterverteilungen, sowie der Ladefarmen und Ladepunkte. Die Infrastrukturauslegung basiert auf dem Prinzip, dass nicht alle Ladepunkte gleichzeitig belegt sind. Im Falle einer Überbelegung regelt das Lastmanagementsystem den Energiefluss der Ladepunkte so, das keine Überlastung im Versorgungsnetz auftritt. Dies kann durch zeitliche Verschiebung der Ladevorgänge oder durch Reduzierung der absoluten Ladeleistung erfolgen. Bild 3: Lastmanagement Normalfall Im Schaubild 3 ist eine Lastmanagement Simulation einer Lade-Farm mit 6 Ladepunkten (CPx - Charge Point) gezeigt. Das angewendete Prinzip in dieser Simulation ist eine Reduzierung des Ladestroms in Abhängigkeit der Ladepriorität für ein Fahrzeug. Eine Leistungsregulierung erfolgt in diesem Beispiel am Ladepunkt 2, nachdem ein zusätzliches Fahrzeug am Ladepunkt 5 angeschlossen wird und der Maximal Strom von 150A dieser Lade Farm erreicht ist. Weitere Leistungsregulierungen erfolgen nach Anschluss eines weiteren Fahrzeugs am Ladepunkt 6. Lastmanagementsysteme im Überblick Version B, Normalfall Time-[min] Priority 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 #-Chargepoint(CP) CP-1 1 32 32 32 32 32 32 32 32 32 CP2 2 32 32 32 32 22 32 32 32 0 32 32 32 CP3 2 32 32 32 22 22 32 32 32 22 32 32 32 32 32 CP4 2 32 32 32 0 22 32 32 32 22 32 32 32 32 32 32 CP5 2 32 32 32 0 22 32 32 32 32 32 32 32 32 32 32 32 CP6 3 32 32 32 22 32 32 32 32 32 32 32 32 32 32 32 32 Current,-Total 64 64 96 128 150 150 150 150 150 150 150 150 128 128 128 128 96 96 64 64 32 32 0 0 Current,-Maximum 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 CP-Start CP1/ 2-Start CP3-Start CP4-Start CP5-Start CP6-Start CP-End CP1-End CP2-End CP3-End CP4-End CP5-End CP6-End Loadmanagement- active/ inactive LM‐Start LM‐End ZURÜCK ZEITKONSTANTE: -30min Chargecurrent [A] 32 32 32 32 32 32 32 32 32 32 32 32 32 22 32 32 32 0 32 32 32 32 32 32 22 22 32 32 32 22 32 32 32 32 32 32 32 32 0 22 32 32 32 22 32 32 32 32 32 32 32 32 32 0 22 32 32 32 32 32 32 32 32 32 32 32 32 32 32 22 32 32 32 32 32 32 32 32 32 32 32 32 0 20 40 60 80 100 120 140 160 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 Achsentitel Achs entitel xxx CP -1 CP2 CP 3 CP4 CP5 CP6 Current,-Total 66 Funded Project for smart charging infrastructure for EVs up to 22kW Es ist ebenso zu erkennen, dass die Leistungsregulierung zurück genommen wird, nachdem das Fahrzeug am Ladepunkt 1 und 2 den Ladevorgang beendet haben. Bild 4: Lastmanagement Sonderfall Das Schaubild 4 zeigt die Simulation eines dynamischen Lastmanagement Systems. In diesem Fall, steht der installierten Ladefarm ‚temporär‘ weniger Energie zur Verfügung, da andere Verbraucher, die am gleichen Transformator/ Hauptverteilung angeschlossen sind, ihren Leistungsbedarf voll ausschöpfen. Die Leistungsregulierung des Lastmanagements wird in diesem Fall dynamisch auf einen neuen Referenzwert gesetzt. Die dynamische Regelung wird vielfältige Anwendungsfälle haben, egal ob große oder kleine Ladefarmen. Immer dann, wenn es weitere große Verbraucher im System gibt, die zeitlich begrenzt das Versorgungsnetz belasten. Dies kann in der Industrie, aber auch im privaten Nutzungsbereich der Fall sein. 5 BATTERIE SPEICHER Ein wichtiger Teil des E-Mobility-Labs wird im Opel Testzentrum in Dudenhofen entstehen: Ein modularer Batteriespeicher, in dem gebrauchte Fahrzeugbatterien aus dem Opel Ampera der ersten Generation wieder verwendet werden, um elektrische Energie vorübergehend zu speichern. Der stationäre Batterieeinsatz im sogenannten „Second Life“ dient dazu, Verbrauchsspitzen auszubalancieren und das Stromnetz zu puffern. Im Falle des Aufstellungsortes im Testzentrum handelt es sich um ein Gebiet mit limitierter Netzversorgungsleistung und einer Leitungslänge von ca.1,5 km bis zur Einspeisung aus der lokalen Stromverteilung. Um an diesem abgelegenen Ort ausreichende Ladeleistung für 8 Ladepunkte à 22 kW zur Verfügung stellen zu können, würden sehr hohe Kosten für den Netzausbau entstehen. Wenn man eine grobe Beispielrechnung machen würde, und ca. 500€/ m für die Leitungsverlegung berechnet, dann summieren sich die Kosten für den Netzausbau auf Time-[min] Priority 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 #-Chargepoint(CP) CP-1 1 32 32 32 32 32 32 32 32 32 CP2 2 32 32 32 32 30 30 24 18 18 24 24 24 CP3 2 32 32 30 30 24 18 18 24 24 24 30 30 30 30 CP4 2 32 29 29 24 18 18 24 24 24 30 30 30 30 32 32 CP5 2 29 29 23 17 17 24 24 24 30 30 30 30 32 32 32 32 CP6 3 23 17 17 24 24 24 30 30 30 30 32 32 32 32 32 32 Current,-Total 64 64 96 128 150 150 150 120 120 120 120 120 120 120 120 120 96 96 64 64 32 32 0 0 Current,-Maximum 150 150 150 150 150 150 150 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 CP-Start CP1/ 2-Start CP3-Start CP4-Start CP5-Start CP6-Start CP-End CP1-End CP2-End CP4-End CP4-End CP5-End CP-6-End Loadmanagement- active/ inactive LM‐Start LM‐End ZURÜCK Lastmanagementsysteme im Überblick Version A, Sonde Chargecurrent [A] 32 32 32 32 32 32 32 32 32 32 32 32 32 30 30 24 18 18 24 24 24 32 32 30 30 24 18 18 24 24 24 30 30 30 30 32 29 29 24 18 18 24 24 24 30 30 30 30 32 32 29 29 23 17 17 24 24 24 30 30 30 30 32 32 32 32 23 17 17 24 24 24 30 30 30 30 32 32 32 32 32 32 0 20 40 60 80 100 120 140 160 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 Achs entitel Achs entitel xxx CP -1 CP2 CP3 CP4 CP5 CP6 Current,-Total 67 Funded Project for smart charging infrastructure for EVs up to 22kW ca. 750.000€. Die Gesamtkosten eines Batterie Containers liegen in der Berechnung mehr als 50% darunter. Durch die Bestückung mit wiederverwendeten Traktionsbatterien entstehen nur geringe Batteriekosten. Damit ist die Entscheidung eindeutig für den Batteriecontainer gefallen. Dieser wird von Opel im Entwicklungszentrum konstruiert und aufgebaut. 5.1 Container Konzept Für den Aufbau des modularen Batteriespeichers wurde ein 40 ft Container ausgewählt. Am Aufstellungsort entstehen nur geringe Kosten für die Installation vor Ort und die Netzanbindung. Die geplanten 8 Ladepunkte mit max. 22kW je Ladepunkt, werden außen am Batterie Container installiert und kommunizieren mit unserem Backend, so wie alle anderen Ladepunkte am Standort. Der Batteriespeicher kommuniziert ebenfalls mit dem Backend, um eine Lastmanagement Steuerung der Ladepunkte in Abhängigkeit von der verfügbaren Energie im Speicher zu ermöglichen. Eine APP für den Betreiber ist geplant, um Information über den Betriebszustand des Batterie Containers zu übermitteln. Ergänzt wird diese durch ein ‚Early Warning System‘, das frühzeitig Diagnose-Informationen einzelner Batterien bis hin zum Gesamtsystem, in einzelnen Warnstufen übermittelt. Damit werden Service-Aktivitäten eingeleitet und so kritische Betriebszustände vermieden. Zusätzlich ist die Ergänzung mit einer Photovoltaik-Anlage angedacht, um regenerative Energien zu nutzen. 5.2 Leistungsdaten Durch die Fahrzeug-Rücknahme-Verpflichtung der Automobilhersteller werden zukünftig wertvolle Hoch-Volt (HV) Batterien verfügbar sein, die einer Second-Life Nutzung zugeführt werden können. Insgesamt können bis zu 18 Fahrzeugbatterien in dem Batteriespeicher wiederverwendet werden, mit deren Speicherkapazität ein Vier-Personen-Haushalt einen Monat lang mit Strom versorgt werden könnte. Die Netz-Umrichter Leistung im Container beträgt bis zu 200kW. Es kann je nach Alterungszustand der Batterien bis zu 300kWh an Energie gespeichert werden. Jede Batterie hat ihren eigenen bidirektionalen DC/ DC Wandler der die Batteriespannung von 300 V bis 400 V in einen DC Zwischenkreis mit 750 V Spannung wandelt. Vom DC Zwischenkreis mit 750 V Gleichspannung wird mit einem bidirektionalen DC/ AC 3-Phasen-Umrichter in 400V Drehstrom gewandelt. Damit wird entweder der Speicher mit Drehstrom geladen oder Dreh - oder Wechselstrom für bis zu 8 Ladestationen an der Container Außenwand zur Verfügung gestellt. Alle Batterien werden temperiert durch ein zentrales Thermalsystem. 68 Funded Project for smart charging infrastructure for EVs up to 22kW Die Steuerung des Batteriecontainers wird von Opel selbst entwickelt und aufgebaut. Diese beinhaltet eine Gesamtsystemsteuerung, den Batterie Clustermanager der die Ladezustände ermittelt, sowie eine Steuerung des Thermalsystems. 5.3 Die verwendeten Batterien Im Container werden Li-Ion Batterie-Systeme der ersten Generation des Opel Ampera verbaut. Dieses Fahrzeug, ein Elektrofahrzeug mit „Range -Extender“, war damals seiner Zeit voraus und wurde ausgezeichnet mit „Car of the Year 2012“ Award. Die Batteriespannung liegt im Betrieb zwischen 300V und 400V. Ein Batteriesystem im Neuzustand hat eine maximale Speicherenergie von 16 kWh. Im Container werden gebrauchte Batteriesysteme verwendet, diese werden mit einem maximalen Energieinhalt von 9-16 kWh entsprechend ihres Alterungszustandes verwendet. Im Fahrzeug müssen Batteriesysteme im Betriebsbereich bis 110 kW Spitzenleistung für Beschleunigungsvorgänge liefern. In der stationären Anwendung als Speicher, wird die Entlade- und auch Ladeleistung jedoch auf 10 kW begrenzt. 5.4 Betrieb des Batterie Containers Am Aufstellort kann der Container netzseitig mit bis zu 22kW leitungsbegrenzt aus dem Stromnetz geladen werden und kann bis zu 8 mal 22 kW an die Ladepunkte liefern. Zusätzlich besteht die Möglichkeit einer Sonderstromversorgung, mittels CEE Steckdose, für den Betrieb von externen, mobilen Verbrauchern. Die AC Verschaltung erlaubt es, optional, eine Photovoltaik Anlage sowie einen 150kW Schnellader an den Container zuzuschalten. Aufgrund der örtlichen Gegebenheiten kommuniziert die Steuerung über Mobilfunk mit dem Backend zur Leistungsregelung und den Leistungsmessgeräten in der Zuleitung, um eine Leitungsüberlastung der Zuleitung zu vermeiden. Auch die Kommunikation zur lokalen Brandmeldezentrale sowie die Fernüberwachung der Steuerung erfolgen via Mobilfunknetz Ferner musste dafür gesorgt werden, dass weder die Leistungselektronik noch die Batterien thermisch überlastet werden. Dazu werden zwei Klimaanlagen installiert. Die Batterien werden in stabilen Gestellen getrennt und sicher gelagert. Durch große Tore sind die Batteriefächer von außen gut zugänglich und die Batteriesysteme können einfach und sicher mit Hilfe eines Gabelstaplers getauscht werden. 69 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Uwe Hauck, Michael Leidner, Michael Ludwig, Helge Schmidt, Marco Wolf Abstract Faster vehicle charging will be a key enabler of greater consumer acceptance of battery electric vehicles (BEV) or electric vehicles (EV) especially in megacities where few drivers have their own private parking spaces and ability to charge overnight. The aim for high power charging (HPC) is to charge a 300 km range in less than 10 minutes, but the related thermal loads would be far higher than found in any normal electric vehicle operation. Traditionally the power ratings of terminal and connector designs were derived from derating modelling, measuring current loads over time, originally designed to test the limitations of relay and fuse technology. Ostensibly these models attempted to simulate current load peaks and their time duration, however they were based on discrete rooted mean square (RMS) profiles. This paper is presenting a new approach, creating a link between thermal and electrical models and analyzing the relationship of the temperature profile to the current profile in each component. TE Connectivity (TE) collaborates with the ZVEI, one of the most important German electrical and electronic manufacturers’ association, to develop harmonized thermal models used with electrical simulation tools to align with the future needs of high-power connectivity. Kurzfassung Um Elektromobilität der batterieelektrischen Fahrzeuge (BEV) oder engl. Electric Vehicles (EV) erfolgreich zu machen, müssen Kaufhemmnisse wie der hohe Preis und die begrenzte Reichweite beseitigt werden. Die Grundlage für lange Fahrten, sowie die notwendige schnelle Ladung von Fahrzeugen in Ballungszentren, die keinen eigenen Zugang zum Netz haben, schafft das Superschnellladen mit Gleichstrom. Bei dem Laden mit Gleichstrom und hoher Ladeleistung von künftig beispielsweise 350 kW beim High Power Charging (HPC) werden in 10 Minuten bis zu 300 km Reichweite „nachgetankt“. Dies stellt jedoch eine Spitzenbelastung für das elektrische System eines Fahrzeugs dar, wie sie die in keiner anderen Betriebssituation auftritt. Die bisherige Vorgehensweise zur Auslegung der elektrischen Komponenten entlang eines Hochvolt-Strompfades (HV-Pfad) basiert auf statischen Lastpunkten (ursprünglich aus Anforderungen zur Bemessung von Relais und Sicherungen), die mit statistischen Verfahren gemittelt und in ihrer Häufigkeit und Bedeutung gewichtet werden. Daraus resultiert ein sogenannter Effektivwert (Rooted Mean Square, RMS). In diesem Artikel wird eine alternative Auslegungsmethode für die systemische Wärmesimulation von HV-Komponenten auf der Grundlage von Ersatzschaltbildern vorgestellt. TE Connectivity (TE) arbeitet mit dem ZVEI, dem Zentralverband Elektrotechnik- und 70 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Elektronikindustrie zusammen, um eine einheitliche Vorgehensweise zur systemischen Modellierung und thermischen Simulation zu schaffen, die die gesamte Wertschöpfungskette zwischen Tier 2, Tier 1 und OEM abdecken kann. 1 Vorbemerkung Die individuelle Mobilität wird derzeit von drei großen weltweiten Trends geprägt: Vernetzung, Autonomisierung, Elektrifizierung. Alle drei wirken sich stark auf das elektrische Bordnetz und die elektrische und elektronische Architektur (E/ E Architektur) künftiger Fahrzeuge aus. Vernetzte, automatisierte und elektrifizierte Fahrzeuge werden viel mehr Daten produzieren, verarbeiten und kommunizieren als heutige Fahrzeuge. Drahtlose Vernetzung über die Luftschnittstelle (z.B. 5G, V2X) schafft zudem die Voraussetzungen für die Kommunikation mit anderen Fahrzeugen, mit der Infrastruktur und auch für Software Updates Over-the-Air (OTA). Gleichzeitig werden in elektrifizierten Fahrzeugen hohe Ströme und große Leistungen fließen. Schon heute verfügen Elektroautos über 120 kW und mehr an Antriebsleistung. Die hohen Ströme für solche Antriebe erzeugen starke elektromagnetische Felder, so dass benachbarte Signalleitungen und Elektroniken vor Einstrahlungen und Störungen geschützt werden müssen (hohe Datenraten bis zu 20 Gbit vs. hohe Leistung). Somit bewirken alle drei weltweiten automobilen Megatrends höhere Anforderungen an die physikalische Schicht als Grundlage für die kommende Funktionalität sowie deren Zuverlässigkeit und Verfügbarkeit. TE Connectivity als Spezialist für Verbindungstechnik, Schaltungstechnik und Sensorik liefert technische Entwicklungen und Innovationen für Signal und Leistung in künftigen Fahrzeugen. 2 Rahmenbedingungen für die Elektromobilität Die Nutzung von Strom als Antriebsenergie im Fahrzeug dient dazu, den Einsatz fossiler Brennstoffe für das Fahren trotz global wachsendem Mobilitätsbedarf zu reduzieren. Gleichzeitig lassen sich die immer strengeren Grenzwerte für den Ausstoß des Klimagases CO 2 pro Kilometer Fahrstrecke mittel- und langfristig nur mit einem elektrifizierten Antrieb erreichen, also mit Hybriden und reinen Elektrofahrzeugen. Noch ist die Marktdurchdringung mit Elektrofahrzeugen sehr gering, allerdings mehren sich die Anzeichen, dass dieser bisher langsame Trend inzwischen Fahrt aufnimmt: Fast die Hälfte aller derzeit in Deutschland zugelassen Elektrofahrzeuge (53.861 am 1.1.2018) wurden 2017 gekauft [1]. Um Elektromobilität erfolgreich zu machen, gilt es auch weiterhin einige Hürden zu überwinden. Kaufhemmnisse wie der hohe Preis und die begrenzte Reichweite verlieren nur langsam an Bedeutung. Positiv dazu trägt bei, dass die Batteriekosten pro kWh weiter sinken, während die Kapazität der Batterie und damit die Reichweite des Fahrzeugs steigen. Gleichzeitig wird inzwischen merklich in den Ausbau der Ladeinfrastruktur investiert, so dass auch lange Fahrten mit einem EV möglich werden. Die Grundlage für lange Fahrten schafft das Superschnellladen mit Gleichstrom (DC- Laden) und einer hohen Ladeleistung von künftig beispielsweise 350 kW beim High- Power-Gleichstromladen (High-Power-Charging, HPC DC). Zum Vergleich: Die meisten EV sind heute für das Wechselstromladen (AC-Laden) mit 2,3 kW (1-phasiger Haushaltsstrom) und bis zu 22 kW (3-phasiger Wechselstrom) ausgelegt. Vor allem Premiummodelle verfügen aktuell auch über die Möglichkeit zum DC-Laden bis zu 150 71 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen kW mit der Alternative, (langsames) AC-Laden ebenfalls zu nutzen, wenn keine DC- Ladesäule erreichbar ist. Generell ist die Sorge, mit einem EV womöglich liegen zu bleiben, unverändert ein ernst zu nehmender psychologischer Faktor. 3 Bedeutung des HPC Bisher wurde bei der Elektromobilität in der Regel mehr über das Fahren nachgedacht als über das Laden. Dahinter stecken erst allmählich reifende Geschäftsmodelle zweier verschiedener Industriezweige: Fahrzeughersteller (OEM) und Energiewirtschaft. Aktuell gilt: - Global gesehen unterscheiden sich die Nutzerprofile für EV. Während europäische EV-Fahrer grundsätzlich auch die Möglichkeit haben wollen, lange Strecken mit ihrem Fahrzeug zurückzulegen, nutzen EV-Fahrer in Asien (auch bedingt durch einen besser entwickelten intermodalen Verkehrsverbund) ihre Fahrzeuge eher für kürzere Strecken in Mega Cities. Mit der Möglichkeit zum HPC DC eignen sich EV optimal für alle denkbaren Nutzerprofile. - Der Ausbau der innerstädtischen AC-Ladeinfrastruktur ist allein nur bedingt zielführend, weil die geringe Ladeleistung eine lange Standzeit bedeutet, so dass Ladestationen lange belegt sind. Bei vielen EV würde das den Bedarf an Ladepunkten nach oben treiben. Dies gilt insbesondere für „heimatlose“ Fahrzeuge („Laternenparker“), für die der Halter keinen eigenen Stellplatz und damit keine eigene Lademöglichkeit hat. - Für AC-Ladestationen spricht, dass sie eine bidirektionale Nutzung der angeschlossenen Fahrzeuge erlauben. Während DC-Ladestationen reine Energiequellen für das EV sind, können Fahrzeuge, die länger als die eigentliche Ladezeit (zu Hause oder am Arbeitsplatz) an einer AC-Ladestation hängen in Spitzenlastzeiten auch als dezentrale Energiequelle für das Stromnetz dienen und damit einen ökonomischen Nutzen generieren, der sich monetär positiv für den Halter auswirken kann. Damit sind beide Ladetechniken sinnvoll. - Steigende Batteriekapazitäten (und damit größere Reichweiten) lassen sich nur dann sinnvoll nutzen, wenn größere Batterien nicht zu noch längeren Ladezeiten führen. - Neue Nutzungsmodelle für EV, wie etwa autonom fahrende Robotaxis basieren in ihrer Wirtschaftlichkeit darauf, dass sie möglichst 24/ 7 fahren, denn nur im Betrieb amortisieren sie sich. Auch in diesem Fall ist ein langsames AC-Laden nicht gut mit dem Einsatzzweck vereinbar. Mit einer Ladeleistung von 350 kW ließen sich in maximal 10 Minuten bis zu 300 km Reichweite „nachtanken“. Damit werden „Tankstopps“ beim EV zu vertretbar kurzen Pausen (ganz ähnlich wie beim Fahrzeug mit Verbrennungsmotor), und DC- Ladepunkte werden sehr schnell wieder frei für das nächste Fahrzeug. Allerdings bedeuten 350 kW Ladeleistung mit Stromstärken bis zu 500 A eine Spitzenbelastung für den gesamten Strompfad vom Ladepunkt bis zur Fahrzeugbatterie. Der hohe Stromfluss entlang dieser Strecke bewirkt durch die physikalisch unvermeidlichen elektrischen Widerstände aller Komponenten (Steckverbinder, Kabel) eine hohe Verlustwärme. Diese Wärme muss bei der Auslegung aller elektrisch leitenden Komponenten berücksichtigt werden, um entweder Überlastungen / Überhitzungen zu vermeiden, oder (wie heute teilweise üblich) eine kontrollierte Reduktion des Ladestroms (Derating) 72 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen planen zu können, wenn beispielsweise die Batterie beim Ladevorgang zu warm wird. Ein solches Derating schützt zwar die Batterie, aber es verlängert auch die Ladezeit. Insofern gilt es, diesen Zielkonflikt optimal zu lösen, indem das Wärmemanagement den genauen Zustand aller Komponenten in deren Struktur zu jeder Zeit kennt oder vorhersagen kann. 4 Herausforderung des High Power Charging Das HPC DC stellt eine Spitzenbelastung für das elektrische System eines EV dar. In keiner anderen Betriebssituation fließt für einen längeren Zeitraum derart viel elektrische Energie vom Ladepunkt zum Fahrzeug und im Fahrzeug. Selbst bei sportlichem Fahren mit hohen Lastanforderungen durch den Fahrer werden dauerhaft keine vergleichbaren Ströme erreicht. Mit dem hohen Ladestrom geht auch eine starke Erwärmung der stromführenden Komponenten einher, was im Stillstand kritischer ist als beim Fahren, weil im Stand keine Konvektion für die Kühlung verfügbar ist. Will man ein DC-Schnellladen also aus den eingangs genannten Gründen ermöglichen, so muss das elektrische System vom Ladepunkt bis zur Fahrzeugbatterie dafür elektrisch und thermisch ausgelegt sein. Vor allem eines erweist sich dabei als Herausforderung: je höher der Strom, desto größer muss der Kabelquerschnitt bei gleichbleibender Spannung sein, um diese Leistung zu transportieren, ohne dabei zu überhitzen. Im Fahrzeug ist das in erster Linie eine Frage des Gewichts und des verfügbaren Bauraums. Es macht beispielsweise einen deutlichen Unterschied, ob man mit 50 mm 2 Leiterquerschnitt vom fahrzeugseitigen Anschluss für den Stecker des Ladekabels (Inlet) zur Batterie auskommt, oder doch auf bis zu 95 mm 2 verstärken muss. Deshalb bietet es sich auch an, die Spannung zu erhöhen, da sich bei höherer Spannung dieselbe Leistung bei reduziertem Strom transportieren lässt (Leistung ist das Produkt aus Strom und Spannung). So erklärt sich der Übergang einzelner OEM von 400 V Systemspannung auf 800 V, denn damit lassen sich größere Kabelquerschnitte vermeiden. Was im Fahrzeug unerwünschte zusätzliche Masse ist, stößt auch bei fest an die Ladeinfrastruktur angeschlossenen Ladekabeln (Mode 4 Anschlüsse) an Gewichtsgrenzen: Will man ein HPC DC realisieren, muss man jede Überdimensionierung des Kabels und aller beteiligten elektrischen Kontaktstellen vermeiden. Aus Gewichts- und Bauraumgründen gilt das auch für das EV, denn zusätzliche Masse bedeutet einen höheren Energiebedarf und damit geringere Reichweite. Allerdings gehen systemische Optimierungen oft mit der Erhöhung von Komplexität einher und müssen daher stets sorgfältig abgewogen werden. 5 Auslegung elektrischer Komponenten heute Die bisherige Vorgehensweise zur Auslegung der elektrischen Komponenten entlang eines Hochvolt-Strompfades (HV-Pfad) basiert ursprünglich auf Annahmen, die für hoch dynamische Fahrprofile und das Anforderungsprofil des HPC DC im Fahrzeug weniger geeignet sind. Geltende Normen und Standards basieren auf statischen Lastpunkten (und ursprünglich auch auf elektrischen Anwendungen zur Bemessung von Relais und Sicherungen), die mit statistischen Verfahren gemittelt und in ihrer Häufigkeit und Bedeutung gewichtet werden. Daraus resultiert ein sogenannter Effektivwert (Rooted Mean Square, RMS), der die statischen Bedingungen abbildet (Bild 1). 73 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Bild 1: Quantifizierungslogik eines herkömmlichen Stromprofils Für dieses in der Realität nichtzutreffende - Lastprofil werden mit einem Sicherheitszuschlag von beispielsweise +20 % elektrische Anschlusskomponenten ausgelegt. Im EV unterscheidet sich das tatsächliche Lastprofil jedoch dramatisch von bisherigen Fahrzeuganwendungen und von den Effektivwerten (Bild 2). Bild 2: Fahrprofil mit Lastprofil beim HPC zum Vergleich 0 20 40 60 80 100 0 100 200 300 400 500 Ladezustand-[%] Strom-[A]-/ -Spannung-[V] Zeit-[s] Typisches-Batterie‐ Ladeprofil- Spannung Strom Ladezustand 74 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Bild 2 zeigt sehr deutlich, warum die thermische Auslegung im Hinblick auf das Laden so wichtig ist. Während sich beim Fahrbetrieb eine sehr dynamische Stromkurve ergibt, die von einem Lastwechsel zwischen hoher und niedriger Last charakterisiert ist, wird die hohe Dauerlast beim HPC DC von der aus dem Fahrprofil abgeleiteten Kurve überhaupt nicht abgedeckt. Will man eine Spitzenbelastung mit 350 kW Ladeleistung ermöglichen, so benötigt man eine andere Herangehensweise an die Auslegung der elektrischen Komponenten. Während die in der Batterie eines EV gespeicherte Leistung beim Fahren typischerweise im Verlaufe mehrerer Stunden abgerufen wird, fließt während des HPC DC die 3-4-fache Leistung in wenigen Minuten in die Batterie. Deshalb muss man den gesamten HV-Pfad systemisch in seiner Verhaltensweise während des Ladens betrachten (Bild 3). Gemittelte Werte sind aus den genannten Gründen dafür nicht sinnvoll. Bild 3: Strompfad von der HPC Ladesäule bis zu Batterie inklusive möglicher Kühlstrategien im Thermomanagement Entscheidend ist zu wissen, wo bei Dauerlast Übertemperaturen entstehen, die zu einem kritischen Zustand führen können. Dieser thermische Aspekt bedarf genauerer Betrachtung. Das ist mit den heute verwendeten Verfahren nicht möglich. Mit dem Ergebnis, dass aktuelle Systeme aus Sicherheitserwägungen um einen Faktor X statisch überdimensioniert werden. Bei 350 kW Ladeleistung kann man sich das aus Gründen des Gewichts, des Bauraums und der Handhabbarkeit prinzipiell nicht mehr leisten. TE Connectivity ist durch aktive Mitarbeit im ZVEI engagiert in der Entwicklung eines neuen Ansatzes zur Bewertung. Dabei geht es um eine Methodik, die es dynamisch ermöglicht, den Temperatureintrag durch Komponenten und die Entwärmung im System mit etablierten Simulationsmethoden (bekannt aus elektrischen Systemen) zu jedem Zeitpunkt zu ermitteln und damit die Auslegung der verwendeten Komponenten frühzeitig untersuchen und das Verhalten dieser im Betrieb vorhersagen zu können. Zu beachten ist: es geht nicht darum, den Sicherheitsfaktor zu verringern. Im Gegenteil. Mit einer neuen Auslegungsmethodik auf der Basis einer realitätsnahen systemischen Wärmesimulation entlang des gesamten HV-Pfades soll gerade erst ein sicherer Langfristbetrieb bei gleichzeitig gut handhabbarer Dimensionierung ermöglicht werden. Die systemische Wärmesimulation (Model-based Thermal Management) schafft 75 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen eine überprüfbare Grundlage für künftige Lastprofile und damit den Nachweis der Sicherheit, Zuverlässigkeit und Verfügbarkeit aller Verbindungskomponenten entlang des HV-Pfades. 6 Begründung der Wärmesimulation Bedingt durch die Physik der Stromübertragung, entstehen entlang eines kabelgebundenen Stromflusses sogenannte Verlustleistungen in Form von Wärme. Ursache dafür ist der elektrische Widerstand (gemessen in Ohm/ Ω) aller metallischen Leiterelemente. Dieser Widerstand ist für den Ausgangszustand jedes Element des HV-Pfades bekannt, er ändert sich jedoch mit der Erwärmung im Betrieb. Es ist berechenbar, welche Verlustleistung bei einem bestimmten Strom, einer bestimmten Spannung und einer bestimmten Temperatur an einer Komponente auftritt. Damit ist auch die entstehende Wärme berechenbar - bisher allerdings nur im stationären Zustand, wenn alle Wärmetransporte im ausgeglichenen Zustand sind. Es gab bisher wenig praktikable Verfahren, mit denen sich das Verhalten des gesamten Systems „HV-Pfad“ dynamisch vorausberechnen ließ. Wollte man dazu die bekannten Methoden - Methode der Finiten Elemente (FEM) oder Flow - anwenden, so müsste man in schneller Folge für jeden Betriebspunkt separat rechnen. Eine betriebsbegleitende thermische Berechnung in Echtzeit (im Fahrzeug) erfordert ein anderes Verfahren, das mit viel weniger Rechenkapazität auskommt. Ein Teil der Herausforderung liegt darin, dass die Wärmeflüsse in einem HV-Pfad ein verhältnismäßig träges System bilden. Je nachdem, welche Masse eine Komponente hat und worin die nächsten erreichbaren Wärmesenken bestehen, reagieren einzelne Komponenten unterschiedlich auf schwankende Lastprofile. Komponenten mit geringer Masse und nur begrenzter Möglichkeit zur Wärmeabfuhr können daher zu einem Engpass für das Wärmemanagement werden. Wenn für die erzeugte Wärme keine genügende Ableitung mehr besteht, so verhält sich die Komponente vorübergehend als adiabatisches Element (= ein Zustand ohne Wärmeaustausch mit der Umgebung), dessen Erwärmung von außen nicht mehr beeinflussbar ist. Solche Wärmeengstellen müssen in ihrem Verhalten genau verstanden werden, um ein System nicht unnötigerweise zu begrenzen oder zu stressen. Hinzu kommt, dass die Wärmeableitung auf unterschiedlichen Wegen erfolgt: Neben dem Wärmetransport im Material (Wärmeleitung) gibt es noch die Anteile für Wärmeabstrahlung und die Wärmeabfuhr durch kühlende Luftbzw. Kühlmittelströme (Konvektion). Für jede Komponente entlang des HV-Pfades setzen sich diese drei Faktoren unterschiedlich zusammen. So sind die Bedingungen für die Wärmeabfuhr am Inlet vergleichsweise günstig, weil die Verlustwärme hier beim HPC DC über die aktive Kühlung des CCS-Steckers mit abgeführt werden kann - der Batteriestecker dagegen hat diese aktive Wärmesenke in vielen Fällen nicht. Damit gilt für die Leitung zwischen Inlet und Batterie: Die Bedingungen für die Wärmeabfuhr sind an einem Ende anders als am anderen Ende. Mit der Erhitzung elektrischer Komponenten geht ein Alterungsprozess einher, der die elektrischen (und/ oder mechanischen) Eigenschaften der Komponente über die Zeit verändert. Je stärker die Erwärmung, desto schneller der Alterungsprozess und desto geringer die Restleistungsfähigkeit der Komponente. Auf die typische Lebensdauerannahme eines Fahrzeugs (300.000 km / 15 Jahre / 8000 Betriebsstunden) gesehen, wird die Alterung jeder Komponente also von den tatsächlichen Lastprofilen beein- 76 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen flusst. Addiert man 30.000 h Ladezeit (kombiniert AC und DC laden) über die Fahrzeuglebensdauer hinzu, so bietet die systemische Simulation eine Lösung für die umfangreichen Testprofile. Erkannte Aufgabenstellung: Hier ist also dringend eine andere Vorgehensweise gefordert, um frühzeitig zu einer sicheren, wirtschaftlich darstellbaren Konstruktion des Strompfades für das HPC DC zu kommen und dafür auch den Sicherheitsnachweis antreten zu können. Mit dem Mittel einer bewährten systemischen Wärmesimulation ist es kein Problem, automatisiert eine nahezu beliebige Zahl an denkbaren Lastprofilen vorab abzuprüfen. Mögliche Engpässe im thermischen System werden dabei sichtbar und können konstruktiv behoben werden. So entfällt der Aufwand für die Fehlersuche ex-post. Dieser investigative Aufwand ist aktuell beträchtlich, gerade weil das thermische System so komplex ist und die eigentliche Fehlerursache möglicherweise gar nicht direkt in der diagnostizierten Komponente liegt, sondern im Verhalten einer benachbarten Komponente im Wärmepfad. 7 Systemisches Simulationsverfahren Das hier vorgestellte systemische Simulationsverfahren für die Verlustleistungen entlang des HV-Pfades unter dynamisch wechselnden Lastbedingungen basiert im Kern auf den Regeln von Kirchhoff. Seine aus der Elektrotechnik bekannte Maschenregel und Knotenregel beispielsweise besagt, dass die Summe aller Ströme in einem Knoten und die Summe aller Spannung entlang einer Masche gleich NULL sein muss. Dabei gilt auch, dass die Energie stets erhalten bleibt. Der Anteil, der durch den elektrischen Widerstand in Wärme umgesetzt wird (= die Verlustwärme) geht also nicht verloren, sondern dieser Wärmeanteil entspricht exakt der Differenz zwischen eingeleiteter elektrischer Energie und am Zielsystem verfügbarer elektrischer Energie. Die Ersatzschaltbilder nutzen den direkten und linearen Zusammenhang zwischen elektrischem und thermischem Verhalten (Bild 4). Elektrisch Thermisch Strom I P Wärmestrom Spannung U T Temperatur Widerstand R R th thermischer Widerstand Kapazität C C th Wärmekapazität Bild 4: Entsprechung zwischen elektrischen und thermischen Größen als Grundlage für Ersatzschaltbilder 77 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Bild 5: Ersatzschaltbild für die thermische Simulation: Widerstände repräsentieren die drei Wärmetransportformen So dienen elektrische Ersatzschaltbilder (Bild 5) dazu, um das gekoppelte elektrische und thermische Verhalten zu simulieren. So wie eine Spannung einen Strom durch einen Widerstand treibt, erzeugt eine Temperaturdifferenz einen Wärmetransport. Die physikalisch unterschiedlichen Transportformen (Wärmeleitung, Konvektion, Strahlung) werden jeweils als Widerstand abgebildet. Hinterlegte mathematische Formeln im Komponentenmodell berechnen laufend die entstehende Wärme je nach anliegendem Strom und anliegender Spannung sowie der Umgebungstemperatur. Auf dieser Basis (= Wärmeentstehung) werden die verschiedenen Möglichkeiten der Wärmeableitung im Ersatzschaltbild durch Widerstände (= thermische Barrieren) und thermische Massen/ Kapazitäten dargestellt, welche die zeitlich aufgelöste Wärmeleitung im Leiterwerkstoff, durch Abstrahlung und Konvektion repräsentieren. Mit diesem vergleichsweise einfachen Verfahren lassen sich sowohl einzelne Kontakte (etwa eine Kontaktfeder), ganze Komponenten (etwa ein Steckverbinder, wie in Bild 6) und auch ein HV-Pfad simulieren, weil Wärmeentstehung und Wärmeableitung durch die Maschenbildung berechenbar sind. 78 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Bild 6: elektrisches / thermisches Ersatzschaltbild eines HV Steckverbinders Wo es von Kabelherstellern inzwischen entsprechende Modelle gibt, lassen sich auch die Zwischenstrecken berechnen. Die Einbindung von Komponenten unterschiedlicher Hersteller (im Bordnetz der Regelfall) ist kein Problem, weil nur die jeweils herstellerspezifischen elektrischen Parameter bereitgestellt und eingegeben werden müssen. Anschließend erfolgt im Modell die Übergabe an die Mathematik, die für die Berechnung nach den Kirchhoffschen Regeln sorgt. Das Modell beschreibt also die Wärmeentstehung und den Wärmeaustausch mit der Umgebung. Die Modellierung beantwortet Fragen wie: Wo sind Wärmequellen und -senken? Ab wann werden die Temperaturen für die Komponente gefährlich, bzw. verkürzen deren Lebensdauer? Wie sieht das Ganze im Cluster aus? Wo entstehen adiabatische Zustände und wie wirken sie sich aus? Im Zuge der Modellentwicklung wurde in Iterationszyklen zwischen Simulation und Test (Rohdaten aus dem Laborversuch) der mathematische Anteil des Modells optimiert, bis die Vorhersagegenauigkeit der Simulation den Testergebnissen entsprach. Mit diesem Verfahren lassen sich bei minimaler Rechenleistung dynamische Lastprofile für jede einzelne Komponente und für den HV-Pfad abprüfen. 8 Sicherheitszuwachs Die erforderliche Rechenleistung für die Thermosimulation auf Basis von elektrischen Ersatzschaltbildern ist gering, so dass es ohne weiteres möglich ist, diese Berechnung als Routine in einem fahrzeugtypischen elektronischen Steuergerät (Electronic Control Unit, ECU) mitlaufen zu lassen. Somit lassen sich die real herrschenden Lastprofile in Echtzeit berechnen und diese Simulation verbessert die Datenbasis für die Funktionssicherheit durch eventuell fehlerbehafteten Sensordaten. Die Systemfunktion der ECU oder die übergeordnete Diagnosefunktion können überlastete Komponenten erkennen und eine Reduzierung (engl. Downrating) der Strombelastung anordnen um kritische 79 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Zustände zu vermeiden. Eine dynamische Eingriffsgrenze ist schon deswegen sinnvoll, weil Grenzsituationen bei extremen Umgebungstemperaturen schneller erreicht werden als bei der durchschnittlichen Nutzungsbedingung. Eine weitere Möglichkeit des Eingriffs liegt in der aktiven Kühlung kritischer Komponenten. Hier übernehmen Thermomodelle nicht nur die Überwachung, sondern auch die Stellgröße einer Temperaturregelung. Simulation und Sensordaten können sich hier gegenseitig als inhomogene Diagnostikroutinen ergänzen. Nicht nur in automatisierten Fahrzeugen, bei denen aus Sicherheitsgründen eine mehrfache Redundanz gefordert ist, kann das ein Beitrag zum Sicherheitskonzept sein. 9 Auslegung von HV-Komponenten Mit der klassischen Methode der Finiten Elemente lassen sich geometrische und designrelevante Größen wie thermische Widerstände oder die Verteilung der Stromdichte beschreiben und simulieren. Sie bilden die wichtigste Informationsquelle für die Modellbeschreibung im Ersatzschaltbild, insbesondere wenn es noch keine experimentell ermittelten Stromerwärmungskurven gibt. Bild 7 zeigt die Temperaturverteilung im Inlet und die Stromdichten innerhalb der HV Kontakt Komponenten. Bild 7: Simulation der Temperaturverteilung und der Stromdichteverteilung mittels FEM Mit der systemischen Wärmesimulation wird die belastungsgerechte Auslegung von HV-Komponenten für das Fahrzeug wesentlich realitätsnäher. Die gekoppelte Simulation von elektrischem und thermischem Modell ermöglicht, Fahrdaten aus Versuchs-, Prüfstand oder anderen Simulationen direkt auf die Modelle anzuwenden. Bild 8 zeigt exemplarisch das Stromprofil einer Motoranwendung, bei der Fahrsituationen auf dem Prüfstand nachgebildet werden. Temperatur Informationen von der Motoranschlussleitung können im Modell unabhängig simuliert werden, während die Ergebnisse des 80 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen Prüfstandes durch die Abwärme anderer Komponenten verfälscht sein können. Dies ermöglicht die Optimierung der HV Komponenten nahe an der technischen Belastungsgrenze mit geringstem Bauraum und Gewicht. Gleichzeitig ist der Kostendruck in der Automobilanwendung enorm. So müssen Komponenten mit kleiner Masse (z.B. die Kontaktfedern) und Oberflächen mit minimalem Materialeinsatz hohe Ströme tragen. Bild 8: Stromprofil einer Motoranwendung und die simulierte Kontakterwärmung mittels thermischem Ersatzschaltbild Unter diesen Randbedingungen ist es von hohem Wert, wenn man das Verhalten des Produktes schon in der Entwicklung vorhersagen kann, weil die systemische und dynamische Wärmesimulation die Auswirkungen von realem Verschleiß im späteren Betrieb aufzeigen kann. Dies gilt insbesondre für die Kontaktelemente, die durch die geringe Masse sehr dynamisch reagieren und eine thermische Überbeanspruchung nicht nur die Kontaktnormalkraft, sondern auch die Langzeitstabilität der Kontaktoberfläche reduziert. Da beide Effekte den Übergangswiderstand im Kontakt erhöhen, kann die zusätzliche Verlustwärme bei Überlast zu einem thermisch selbstbeschleunigten „Durchbrennen“ führen. So lässt sich ein komplexes System wie der HV-Pfad simulieren und sein Verhalten vorhersagen. Die Simulation kann dabei eine Breite abdecken, die im Test kaum realisierbar wäre. 81 Thermosimulation für das High Power Charging (HPC) von Elektrofahrzeugen 10 Fazit Ein HPC DC stellt im EV ein ganz außerordentliches Nutzungsprofil dar, das es sonst in keinem anderen Betriebszustand des Fahrzeugs gibt. In dem komplexen und trägen Wärmesystem des HV-Pfades können beim Laden ganz unterschiedliche Temperaturprofile an einzelnen Komponenten auftreten. Um 350 kW Ladeleistung sicher nutzen zu können, muss man den gesamten HV-Pfad mit dynamischen Lastprofilen systemisch simulieren, um mögliche Wärmeengstellen unter realen Betriebsbedingungen erkennen und in ihrer Wirkung beurteilen zu können. Eine Simulation muss außerdem übergreifend funktionieren und die gesamte Wertschöpfungskette zwischen Tier 2 - Tier 1 - OEM abdecken. Die systemische Wärmesimulation von HV-Komponenten auf der Grundlage von Ersatzschaltbildern liefert Daten für ein optimiertes Design, bei dem mittels realer dynamischer Simulation sichtbar wird, wie oft Komponenten an der Temperaturgrenze betrieben werden dürfen, um die geforderte Langzeitstabilität und Zuverlässigkeit des Gesamtsystems zu erreichen. Diese Datengrundlage erlaubt bei optimierter Auslegung des HV-Pfades einen größeren Sicherheitszuwachs, weil die simulierte thermische Belastung die Realität und mögliche Alterungsprozesse berücksichtigen kann. Es wird das thermische Verhalten der Komponenten im Bordnetz simuliert, wobei akkumulierte Testprofile auf Basis dynamischer Lasten als Basis für den Komponententest dienen. Ziel ist es, die Komponenten des HV-Pfades so auszulegen, dass sie die kurzzeitige dynamische Last beim HPC DC (10 Minuten) mit hoher Leistung über den gesamten Lebenszyklus hinweg sicher tragen, eine statische Überdimensionierung jedoch vermieden wird. Durch die Simulation werden Hot Spots (vor allem passive Komponenten mit geringer thermischer Masse) sichtbar und können frühzeitig durch konstruktive Maßnahmen optimiert werden. Damit leistet die systemische Thermosimulation von HV-Komponenten in der Entwicklung auch einen großen Beitrag zur Verbesserung der Validierungsgenauigkeit. Literatur [1] Lebedew, A.: Eine Million E-Autos - aber erst 2020. Stuttgarter Zeitung, Stuttgart, 20.9.2018, Seite 14 82 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs Julian Taube, Laurenz Tippe, Joachim Fröschl, Hans-Georg Herzog Abstract Due to a steadily increasing power demand in the power net of modern vehicles and increasing demand for CO2 reduction, the power net gets a growing role in the development of vehicles. The introduction of a 48 V Voltage level is imminent to accommodate these challenges [1]. The traditional 12 V voltage will remain for low power components to reduce the development costs and complexity. The future power net will therefore consist of two parallel voltage levels. The common approach is to double up the wiring harness, one for 48 V and one for 12 V, which leads to significant wiring effort in the harness and system complexity. A central DC/ DC converter converts the power between the two voltage levels. For fail operability in the context of autonomous driving, this converter needs to be highly reliable. The proposed topology is instead based on a central power rail. Apart from simply distributing the electric power in the vehicle, the proposed supply rail gives the possibility to integrate ECUs, Power converter and other components directly into the supply rail, so an additional wire harness to the components is not needed. Thereby a reduction of the total wire length and weight of the power and communication harness is expected. Kurzfassung Wegen der stetig wachsenden Leistungsanforderungen von modernen Fahrzeugen und den steigenden Anforderungen für eine CO2 Reduktion erhält das Energiebordnetz, bei der Entwicklung von Fahrzeugen eine zunehmende Bedeutung. Die Einführung eines 48 V Bordnetzes steht bei vielen OEMs kurz bevor, bzw. ist schon geschehen, um diese Herausforderungen zu meistern [1]. Das traditionelle 12 V Bordnetz wird für Kleinverbraucher weiterhin bestehen bleiben, um die Kosten und Komplexität einer Neuentwicklung der Komponenten zu reduzieren. Das zukünftige Bordnetz wird daher aus zwei oder mehr parallelen Spannungsebenen bestehen. Der meist verbreitete Ansatz ist es, das bestehende Bordnetz mit der zweiten Spannungsebene aufzudoppeln, was zu einer hohen Komplexität und Gewicht führt. Ein zentraler DC/ DC Wandler verbindet beide Spannungsebenen. Für fehlertolerantes Verhalten im Zuge des autonomen Fahrens, muss dieser Wandler hochverfügbar sein. Die hier vorgestellte Topologie basiert im Gegensatz dazu auf einer zentralen Versorgungsschiene. Neben der Verteilung der Energie durch das Fahrzeug, sollen Steuergeräte, Wandler und andere Komponenten direkt in die Schiene eingesteckt werden können, sodass kein konventioneller Kabelbaum für diese Komponenten notwendig ist. Dadurch wird eine Reduzierung der Gesamtlänge und damit des Gewichts des Kabelbaums erwartet. 83 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs 1 Introduction The power rail consists of two power busses with 48 V and 12 V. Additional rails with a ground potential can be included to reduce ground shifting and help with EMI considerations. Multiple communication busses are integrated in the center of the power rail to accommodate for the communication between the components. Furthermore, the rail makes use of power line communication to achieve a redundant communication between the components. To achieve a high updateability and modularity design, the power rail is divided into slots of different width. Each slot consists of connectors for the power rails and a connector for the communication busses that are implemented as a plug-in system. This accomplishes a plug-and-play system, where new components can be simply plugged into the rail without any wiring afford. The power conversion between the two voltages is achieved with a distributed DC/ DC converter, which consists of multiple DC/ DC converter modules in parallel. The power capacity of each converter may be smaller than the power consumption of the entire system. By running them in parallel, the total power demand can be covered. If one of the converter fails, the rest of the converter can compensate for this failure, making the converter system highly reliable. Using cybernetic control strategies, these converter can achieve high efficiency and voltage stabilization over a wide range of operation [5]. The modularity of the converter enables a big economy of scale, since different numbers of the same converter can be integrated for different derivatives or depend on the equipment variance. The need for a liquid cooling of the converter can be eliminated using highly efficient wide band gap semiconductor based DC/ DC converter integrated in the power rail. 2 Objectives This Article describes a concept for a power supply rail with an integrated communication rail for automotive applications. First, the concept for the rail itself is given and possible applications are shown. The different modules, which can be integrated into the rail are explained in more detail. Next, an electrical and thermal simulation model of the proposed architecture will be explained. To validate the model, a test bench is necessary, which will be shown in chapter 6. Finally, a conclusion and outlook is given. 3 Hardware Concept The main goal of the supply rail is to be able to integrate ECUs, so called inlay-ECUs and other components directly into it, without the need for further wiring, thereby reducing wiring effort and increasing modularity and flexibility. This is achieved, by combining power supply busses with a communication rail, which includes the necessary communication busses, as seen in figure 1. The power rail can have a single voltage level e.g. 12 V or two voltage levels e.g. 48 V and 12 V or two 12 V busses for redundancy. This paper focusses on a supply rail with two voltage levels namely 48 V and 12 V and a communication rail with various busses. To ensure modularity, a slot system with three different slot sizes as depicted in figure 2 is proposed. The different sizes of the slot system incorporates the different sizes of ECUs in a vehicle. Components can be plugged into each of these slots and are di- 84 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs rectly connected to power and data lines, resulting in a plug-and-play system. Components, which are plugged into the rail can be supplied by only one voltage level or two levels simultaneously, to achieve a redundant power supply. Furthermore, the power rails serve as heat sinks for the components. With additional airflow through the rail housing an active air cooling for the components can be achieved. Possible modules for the supply rail are discussed in more detail in chapter 4. The width of the supply rail is proposed as 10 cm, which corresponds to the standard Eurocard size for PCBs. 3.1 Power Bus Bars The power bus bars of the supply rail are made from aluminum 1050A which offers high electric and thermal conductivity with lower weight and cost than traditional copper conductors. Compared to copper, aluminum has 35% lower electric conductivity, so the diameter of the power bars has to be larger compared to copper. The density of aluminum on the other hand is 70% lower than copper, so even with an increased diameter the weight loss is approx. 50%. To achieve an EMI considerable design, the bus bars are layered with a thin electric isolator between them, as seen in figure 1. By layering the bus bars, they form a capacitor, which acts as an EMC filter [2]. The isolation layer between the bus bars should be thermally conducting, to achieve an even temperature distribution in both power bars and have a high permittivity to increase the capacitive effect. One option is to Fig. 1: Sectional View of the Supply Rail Fig. 2: Slot System of the Supply Rail 85 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs anodize the aluminum power bars, to achieve an aluminum oxide (Al 2 O 3 ) layer, which is isolating and has a high permittivity. To achieve an even higher degree of EMI shielding, the pack of layered power bars can additionally be wrapped with aluminum foil. The most challenging aspect of the power bars is to design the contact system, so ECUs can be easily plugged in. One possible solution for this challenge is to punch a flap into each of the bars and bend it to the top (Fig. 3). Thereby forming a traditional flat plug. Instead of a traditional flat plug, the flaps could also be bend 180°, so ECUs can contact to the bus bar, by pressing a contact area against the flaps to form an abrasive contact, as seen in figure 3. Since the current density around the punched out area is higher, a thermal hotspot is formed. Due to the high thermal conductivity of aluminum though, the temperature difference of this hot spot compared to other areas of the bus bar is quite small. Other options for contacting the ECUs with the power rail include flaps at the edges of the bus bars, as seen in [7]. The most challenging part of the contact system of the aluminum bus bars is the oxidization of aluminum in air. Therefore, within a few minutes, an isolating oxide layer is formed. To circumvent this challenge, the contact area can be electroplated with a thin tin or silver layer to protect the contact area from oxidization, maintaining a low contact resistance. 3.2 Communication Rail ECUs are plugged into the supply rail and are directly connected to power and communication busses. Since the supply rail itself provides structural stability, the communication busses can be made out of very thin wiring. One option is to use twisted ribbon cables with IDC connector at the points, where the ECUs are plugged in (Fig. 4a). In this fashion, approx. eight wires / cm can be integrated into the rail. Another option is to use Flat Flex cables. They offer a high wire density of approx. 20 wires / cm and a significantly lower weight as ribbon cable. The main disadvantage of these wires is, that a twisted pair configuration is harder to achieve. It is possible to fashion a quasi-twisted pair configuration in a flexible printed circuit (FPC) (Fig. 4b), but the drawback of this solution is signal reflections in the wire, which makes it unusable for high frequency signal transmission. A compromise solution is to use two layered flat flex cables (FFC), where the differential signals lay on top of each other (Fig. 4c), thereby reducing the influence of EMI in the signal with similar efficiency as with Fig. 3: Flat Plug (Left) and Planar (Right) Connection of Power between ECU and Power Busses (Length in mm). 86 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs twisted pair wiring. In this way up to 20 pair signals per centimeter width of the communication rail are possible. To further improve robustness against external EMI, the communication rail can be wrapped with a shielding foil, such as aluminum. At each slot in the supply rail is a small PCB, which routes the signals to a PCB connector. Additionally this small PCB can also contain active routing components, such as gateways between busses or switches for Ethernet connectivity. Alternatively, spring contacts can be used to contact directly on pads on the PCB, FCC or FPC. 3.3 Varieties The modularity and scalability of the described approach offers a flexible design and integration into the vehicle. This achieves a high degree of freedom and ensures the migration to this new topology is handable. One scenario could be, that only a short segment of supply rail is integrated in an area with high ECU density, such as behind the dashboard (Fig. 5a). Another possibility is to put supply rails in each of the vehicles corners, where most of the ECUs are placed, and connect these segments with e.g. a traditional wiring harness or conventional energy backbone (Fig. 5b). The architecture with the most potential is to integrate the supply rail in the central tunnel across the entire vehicle (Fig. 5c). One further option is to use the supply rail redundantly, by integrating it into an ISDD as described in [6] or intelligent current distributor, which is plugged into a larger supply rail (Fig. 5d). Fig. 4: Twisted Ribbon Cable (a),Twisted Pair Flexible Printed Circuit (b) Flat Flex Cable layered (c), 87 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs Fig. 5: Different possible Layouts of the Supply Rail inside a car. Short Rail e.g. Across Dashboard (a), Segmented Rails (b), Long Rail across entire vehicle (c) or integrated in an ISDD (d) 4 Components As described earlier, apart from ECUs, different other modules can be plugged into the supply rail, to achieve functionality. These will be explained in more detail in this chapter. 4.1 Converter The future automotive power net will contain multiple voltage levels. Due to the increased electric power demand, the traditional 12 V power net is getting to its limits. Therefore a second, higher voltage level standardization was introduced, namely the 48 V level [2]. The main power supply and battery in future vehicles, will be on the 48 V side. To accommodate migration periods and the supply of older and cheaper components, the traditional 12 V power net will coexist further. Therefore, DC/ DC converters are necessary to transfer energy from the higher voltage level to the lower. For automated driving operation, the converter has to be highly reliable, so a reliable power supply of the lower voltage level is permanent. This is increasingly important, if the energy storage on the low voltage side is reduced or eliminated. Increasingly strict emission regulations, result in an increased demand in overall system efficiency. This means, that the DC/ DC converter also have to be highly efficient. To keep the modularity and scalability of the proposed supply rail system, small parallel DC/ DC converters instead of one central converter are proposed, as described in [5] and depicted in figure 6. This results in a highly flexible choice of the number and thereby the total power rating of the overall system, without the need for redeveloping a new converter for each derivate. With parallel converters, a high level of redundancy can be achieved in case all converters are at least fail-safe. 88 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs 4.2 Segmentation Automated driving applications demand a highly available power supply even in case of a failure. Achieving this with the supply rail is possible by segmenting it into multiple segments, which in case of a failure can be separated. In this way a failure, e.g. a short circuit in the power bus can be isolated as depicted in figure 8. A possible solution for such a segmentation is shown in figure 7. It consists of power connector to contact the power busses with the segmentation device, semiconductor switches as switches and a controller, which supervises the current flow through the device and detects failures in the bus bars. The entire device has to have a very small conducting resistance, to reduce power losses. In addition, the switches have to be able to sustain high peak currents in case of a short circuit. This can be achieved by paralleling multiple high current switches. Since semiconductors have intrinsic body diodes, two switches have to be connected anti-serial to ensure disconnection in both directions. Another important feature of the segmentation is the compensation of the thermal deformation of the bus bars. With a temperature rise of 50 K, the power bus bars length increases approx. 2%. The longer each segment of the rail is the more deformation has to be compensated by the segmentation device. This is achieved by abrasive contacts. In case of a failure in one segment of one of the power busses, this segment is isolated. The power then has to be routed around the isolated segment. By using bidirectional DC/ DC converters, the power can be transferred over the not effected bus bar as shown in figure 8. Fig. 6: Paralleled DC/ DC converter for redundant conversion of power between 48 V and 12 V [5] 89 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs Fig. 7: Segmentation Module to segment parts of the Supply Rail in Failure Situations 4.3 Current Distributor The move to integrated ECUs inside the supply rail is a challenging and costly path, since all components have to be redeveloped. To ensure, that existing components can be connected to the supply rail, current and data distributors are necessary, which adapt the supply rail to a conventional wiring harness (figure 9a). This is also important, for components, which can not be integrated into the supply rail, because they are not location independent. This refers to almost all actuators in the system. The proposed current distributor offers the possibility to connect a traditional wiring harness with power and data lines to these actuators. With integrated electronic fuses (figure 9b) overcurrent protection as well as clamp control and further functions, such as autonomous degradation can be achieved. The electronic fuse can also be replaced with a power stage or DC/ DC converter (figure 9c), so that the output of the current distributor can be controlled. With that, decentralized functionality can be generated, by running the algorithms and functionality on a main ECU, which controls multiple such power stages. Thus increasing the number of equal parts, since the power stage can be used for a variety of functions. This ultimately results in a central controller architecture with distributed satellite actuators in which all functionality is computed on one or only a few ECUs. Fig. 8: Power flow in the Power Busses in normal and failure operation 90 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs 4.4 Energy Storage Voltage stability and quality is one important aspect of designing a power net. To achieve a stable and high quality voltage supply, filters and energy storages are needed, which buffer the transient current of various components such as switching power stages. Energy storages can also buffer the peak power demand from the consumer, so that generators and DC/ DC converters do not have to provide the transient peak power of the entire power net. To address these aspects, the supply rail offers the possibility to directly integrate energy storages. Depending on the topology and used components, different sizes and technologies of energy storages, such as capacitors or batteries can be plugged into the power busses. This potentially decreases the filtering effort, which all components have to implement to fulfill voltage regulations, such as the LV124 or LV148. 5 Simulation Understanding the working principals and the effect on the power net in an early development phase is crucial, for the development of any power net topology. Therefore, various simulation models were build, which model the electric and thermal behavior of the supply rail. 5.1 Dymola Simulation The first model was created using the popular multiphysical simulation environment Dymola. The model is based on the differential equations of different electrical and thermal components. The electric behavior of the power bars between two slots is assumed to be homogenous and is represented by one resistor and one inductor for each bus bar as depicted in figure 10. The thermal behavior in Dymola is modeled with a point mass model. It consists of a thermal conductivity and capacity between two slots, convection to the air and conductive heat transfer to the bottom side of the casing. Fig. 9: Current Distributor on 48 V Power Rail with a Traditional Fuse (a), Controlled Switch (b) or with a DC/ DC Converter (b) 91 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs Depending of the length of the supply rail and the number and size of the slots, multiple of these models are connected in series to form the model of the supply rail. The interconnection of two such section models is equivalent to the location of the power connector. 5.2 FEM Simulation In the Dymola model, the simplification has been made, that the current density and therefore heat generation across the power bus bars is homogenous. Since the connectors are proposed to be punched out as described in section 3.1, the effect of this assumption has to be analyzed. For this purpose an electro-thermal 3D model of the power rail is created and simulated with finite element methods (FEM). They show, that the current density around the punched out area is significantly higher than in the rest of the bus bar. An equivalent resistance for the section between slots is calculated from the FEM simulation and fed into the Dymola model, to incorporate this effect. The temperature difference on the other hand is quite small and therefore negligible. This is explained by the high thermal conductivity of the aluminum power bus bars. Additionally to the electro-thermal behavior of the rail, the EMC behavior of the power rail can be simulated using FEM. 6 Testbench Only verified models can be used for a successful development of a new system, therefore a test bench for the supply rail topology was created. In this test bench the electrical and thermal behavior of the system can be investigated and compared to the simulation models, to verify them. The test bench, which is schematically shown in figure 11, consists of multiple portable 19” racks, in which various actuators, such as electronic loads, or power supplies can be integrated. In this way the test bench is very flexible, so different configurations of the supply rail can be investigated. The electric consumers in the power net are emulated with switched resistors in case of simple consumers like heating elements or they are emulated with electronic loads. In this case, a physical model of the consumer is calculated in real time and the correspondent current is drawn by the load as described Fig. 10: Electrical and Thermal Modlica Model of a Segment of one Power Bus. 92 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs in [8]. With this approach, high dynamic consumers, such as the braking or steering systems can be emulated. The use of physical models, also results in the possibility to change a consumers behavior or size in software, so no hardware modification is needed. The battery and generator are emulated using bidirectional electronic loads or electronic power supplies with a high dynamic output. Alternatively, real batteries or a real 48 V starter generator can be coupled to the test bench. At each component, the voltage of the power busses as well as the current through the component is measured. The temperature is measured at 32 representative points in the supply rail and on components plugged into the rail. The measurement rate of this system test bench is 10 kHz. In total 80 Voltages and currents are measured, which can be analyzed in real time in a user interface, or after testing using an automated analysis. 7 Conclusion & Outlook This article proposes a modular approach for future power nets. The possibility to integrate ECUs directly into a supply rail as an inlay ECU, which is consequently supplied with power and data, can achieve a highly modular and scalable power net. Thereby reducing the complexity for future power net designs and increasing the economy of scale. With components such as current distributers, a migration path to the new topology is available, to reduce initial costs of development. By using paralleled DC/ DC converters a highly redundant and efficient converter system can be achieved, which is especially relevant for high level autonomous driving. With the help of simulation models, the new topology can be investigated and validated on the described test bench. This article described the overall concept of a supply rail system with integrated ECUs and power converters. The next step is to build a corresponding prototype to validate the physical models on the test bench. To effectively operate the system, a management system is necessary, which controls the different components e.g. the power converter. As the System is highly modular, scalable and complex, the management system must also be modular and scalable and be able to handle the complexity. Future work will therefore focus on a cybernetic approach for the management system. Fig. 11: Schematic overview of the Test bench with Constant loads (R), dynamic electronic loads (El. Load), DC/ DC converter (C) and Source / Sinks (Q/ S) 93 Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs References [1] DeMattia, N.: “BMW to debut 48-volt electrical systems by 2020”, 2018, https: / / www.bmwblog.com/ 2018/ 11/ 16/ bmw-to-debut-48-volt-electrical-systemsby-2020/ [Accessed: Feb. 15, 2019] [2] Bilo, J. et al.: „ZVEI: 48-Volt-Bordnetz - Schlüsseltechnologie auf dem Weg zur Elektromobilität“, 2015 [3] Tippe, L. et al.: “Introduction of Ring Structures in Future Car Generation’s Electrical Systems”, 2018, ESARS-ITEC 2018 [4] De Oliveira et al.: “Reduction of conducted EMC using busbar stray elements.” (2009) [5] Winter, W. et al.: "Using the Viable System Model to control a system of distributed DC/ DC converters" 2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Budapest, 2016, pp. 002768-002773. doi: 10.1109/ SMC.2016.7844658 [6] Tippe, L. et al.: “Ring Structures in Automotive Power Nets: Idea and Implementation”; Konferenzband der EEHE - Elektrik / Elektronik in Hybrid- und Elektrofahrzeugen 2019 [7] Dräxlmaier GmbH: “Lösungen zur dezentralen Bordnetzversorgung“; ATZ 4/ 2016, Page 48-53 [8] Winter, M. et. al.: "From Simulation to Testbench Using the FMI-Standard," 2015 IEEE Vehicle Power and Propulsion Conference (VPPC), Montreal, QC, 2015, pp. 1-5. 94 Enabling Technologies - die attach and substrate technologies for power electronics Louis Costa Abstract In this paper an overview is given on die attach and substrate technologies which are currently used in series mass production for power electronics in the field of emobility. Technologies can be classified based on application specific requirements for ambient temperature, voltage and power class. Based on real product examples the composition of material layers for functional power electronics is presented. Kurzfassung Ein Überblick über aktuell eingesetzte Aufbau- und Verbindungstechnologien für Leistungselektroniken im Bereich von E-Mobility Anwendungen wird vorgestellt. Die Technologien können über anwendungsspezifische Anforderungen für Einsatztemperaturen, Spannungs- und Leistungsklasse der Elektronik klassifiziert werden. Anhand von realen Produktbeispielen werden die angewendeten Materialschichtaufbauten von Leistungselektroniken aufgezeigt. 1 Power electronics for e-mobility E-mobility will change the way how people will be mobile in the future. The technology change is a global trend which is supposed to become reality between 2020 and 2030 for a very broad amount of consumers. Already for a few years now, hybridization of vehicles is progressing among the fleets of many car manufacturers and a continued broadening is expected. One of the keystones for this technology transition and most dynamical technology area is power electronics. The task to assemble and package power semiconductors in a highly efficient way both for lowest power dissipation and highest heat conduction offers the route for longer mobility ranges, faster electrical charging and many other applications. Today many automotive applications are on the market which can be classified by power density, see Fig. 1. Electrification of smaller auxiliary components in vehicles started already years ago. For example AB Mikroelektronik GmbH (AB) in Salzburg started already 2005 with development and production of electrical water pumps. Typically such auxiliary components can be designed for power densities up to several kilowatts and high voltage applications. Micro-hybrids are vehicles which have implemented a start-stop function which deactivates the combustion engine during idling. Such vehicles do not possess a battery or an electric motor. Associated power densities are smaller than 5 kilowatts and board net voltage is typically between 12 to 24 volts. Mild hybrids are vehicles with integrated batteries and which charge by recuperation. These batteries cannot be charged externally, but recuperation energy can be used for boost and acceleration. Full hybrid vehicles possess similar functions as mild hybrids but can be moved fully electrical. For plug-in hybrids an external 95 Enabling Technologies - die attach and substrate technologies for power electronics charging function is offered. Power densities increase typically to more than 60 kilowatts. A fully electric vehicle is moved solely by electrical energy and one of commercial selling points for such vehicles is the accessible range which can be achieved with one charging event. In order to achieve a safe operation many subsystems in an electrical vehicle need to be optimized individually but also as a system so that a maximum of efficiency can be acquired. Figure 1: Classification of electric applications for automotive vehicles and product examples for specific area In the following sections some real series product examples for power electronics are presented. Starting in section 2 with requirement guidelines for development of power electronics, section 3 covers a presentation of typical substrate technologies. AB follows the view to cover a broad range of power applications by applying solder or wire bonding technologies and adjusting substrate technology. Each substrate technology is limited to temperature or power requirements accordingly. This will be discussed in more detail within the next caption. Finally some new die attach technologies will be presented in section 4 which are currently introduced or under development. In section 5 finally a short summary follows. 2 Requirements for automotive power electronics For safe operation of a power electronic device several requirements need to be satisfied. So-called document AQG 324, a document published from an ECPE (European Center for Power Electronics e.V.) [1] working group, presents guidelines for qualification of automotive power electronics [2]. This guideline has been updated for a general automotive and European standard. In case for more specific information please refer to cited documents and norms [2-4]. A more general and international description of test standards is JEDEC, respectively test plans defined within AEC-Q [4]. In the following section a rough overview of qualification guidelines described within AQG 324 [2] is given. JEDEC, respectively DIN EN norms, are cited within the given citations for qualification guidelines. 96 Enabling Technologies - die attach and substrate technologies for power electronics 2.1 Module tests Module tests serve to assure electrical and mechanical functionality of modules. Typically these tests cover also specifications found in datasheets of module manufacturers. For example die attach technologies can be tested with ultrasonic microscopy, cross section polishing, scanning electron microscopy, X-Ray transmission and computer tomography to mention just a few of common methods. Void content of solder joints, delamination or crack generation can be determined. Basic electrical parameters are measured for semiconductor devices such as leakage currents, threshold voltages, nominal current and voltage values of a module. In order to switch semiconductor devices at high frequencies in the design and layout of current traces parasitic stray inductances need to be taken into account. A so-called double pulse experiment (IEC 60747-15: 2012) determines such stray inductances for high switching frequencies. For efficient cooling a low thermal resistance is necessary. Thermal resistance expresses the temperature rise of a module which is prone to power dissipation, and describes how fast thermal energy can be transferred from the heat dissipating device, i.e. power semiconductor, to the cooling medium. Thermal resistance is composed from contributions of each interlayer in the module assembly starting at the semiconductor, through die attach material, metallization, heat spreader, thermal interface materials, and connection to cooling medium. For measurement of low thermal resistance values the transient thermal impedance measurement method (Tr3ster) according to DIN EN 60747-15: 2012 can be applied. In Fig. 7 (lower right) a measurement of thermal resistance with two-contact method is shown. The crossing point of the curves shows the material interface which is changed within the twocontact method, and at which the thermal properties (resistance and capacitance) is changed. Short-circuit current capability is described by the junction temperature of the epilayer of the semiconductor, drain-source voltage, gate-source voltage (resp. collector-emitter and gate-emitter for IGBT based module) and short-circuit duration. This test is required to assure current switching capability of the module at any operation condition up to junction temperature. Typically power modules are isolated by thermal interface materials from heat sinks. An isolation strength test determines isolation voltages between heat sinks and the power electric circuit. Goal of isolation coordination DIN EN 60664 is to assure functional safety during operation of the power electronic modules. Guidelines for determination of relevant air and creepage distances in module design are given. 2.2 Environmental tests Environmental tests cover thermal shock, contactibility, vibration and mechanical shock tests. Robustness against passive temperature cycling during different storage conditions is tested. Depending on location of operation in the vehicle special attention needs to be addressed to contactibility and vibration tests. Appropriate die attach or contact technology needs to be defined in order to guarantee safe operation under different vibration profiles (DIN EN 60068-2-6). Vibration tests are typically carried out in a frequency range between 10 to 1000 Hz and an acceleration of 5 g. Mechanical shock tests are described within DIN EN 60068-2-27, and typical shock durations of 97 Enabling Technologies - die attach and substrate technologies for power electronics 6 ms and acceleration values of 30 m/ s 2 need to be achieved in each direction +/ -X, +/ -Y, +/ -Z. 2.3 Lifetime reliability tests Reliability tests are active tests of modules under current operation. These tests assure that all operational properties are conserved and held within a tolerance band over a specific time window. As power modules are installed for high electrical and thermal performance and need to satisfy functional safety requirements a reliable operation over lifetime need to be assured so that high costs or even human life risks can be omitted. Typical process or material uncertainties need to be taken into account during design stage so that maximum application requirements can still be achieved under most extreme production condition. E.g. during power cycling short (PC sec ) current pulses stress die attach or top contact technologies close to the active semiconductor chip or longer (PC min ) current pulses test contact technologies towards baseplate and heat sink. High-temperature reverse bias (HTRB), high-temperature gate bias (HTGB), high-humidity high temperature reverse bias (H 3 TRB) serve primarily to test short circuit capability, quality of the passivation layer and gate oxides. During testing leakage and short-circuit currents are monitored. High-temperature storage (HTS) and low temperature storage (LTS) assure quality of the housing of a power module under extremely high or low temperature storage. A FIT rate (failure in time) expresses the number of failures per operation hour. 3 Power electronics Depending on application specific requirements AB follows the strategy to adapt substrate technology rather than die attach technologies. Both die attach materials and substrate material layers need to be adjusted to one another for a perfect match. In order to complement, section 4 will give an outlook to current trends in die attach technologies. In this section real product examples will be given based on different substrate technology platforms. 3.1 FR4 technology platform FR4 materials are widespread and allow for very flexible designs. In combination with metal inlays FR4 substrate allows to integrate both logic/ driver and power stage in one assembly. For specific thermal management of high dissipative components laser drilled holes filled with copper allow a direct connection to embedded copper inlays (thermal vias). Numerous configurations and number of layers are possible. Up to 400 µm thick copper tracks can be realized with etching technologies. Thicker copper tracks need to be embedded as solid metal into the FR4 material. Despite all design flexibility FR4 solutions offer limitations with respect to current capability and thermal resistance performance, but continuous development efforts improve capabilities. Also due to lifetime reliability of breakdown voltage capability of resins certain design rules for isolation layers need to be satisfied. Standard resins possess a glass transition temperature of 135 °C, for high current capability there exist also materials 98 Enabling Technologies - die attach and substrate technologies for power electronics with a T G of up to 185 °C, or even higher. Typical resins offer thermal conductivities of 0.25 W/ mK, improved resins can provide up to 0.6 W/ mK. For an efficient design one needs to take into consideration these aspects. Within FR4 technology embedding technologies open up possibilities to design very short current/ commutation loops by minimizing parasitic inductances. Especially for high-frequency applications this can be exploited in combination with low-impedance FR4 substrates. Product example: Single board solution of a 3-phase inverter for 48 V Based on discrete power semiconductors a three-phase inverter is assembled with integrated microcontroller and gate drivers. This design is intended to present a benchmark for modules which apply FR4 technologies and gives a rough indication which typical power levels can be achieved, see Fig. 2. For thermal management 8 copper inlays with 105 µm thickness are applied. The module is actively cooled by two axial fans. Each semiconductor switch comprises four 100 V MOSFETs in parallel connection. On total 48 MOSFETs are integrated in the 3-phase inverter which are reflow soldered under vacuum condition. The DC link capacitor bank and current sensors are integrated within the heat sink. High current connections are made by pressed power contacts and screwed copper busbars. This module is tested under real conditions on a motor test bench and corresponding operation temperatures (see Fig. 2). Up to 255 A continuously and 510 A for 10 seconds are experimentally tested at an environment temperature of 75 °C and arising maximum temperatures at the chips are detected with thermography. For continuous operation with 255 A and active fan cooling a temperature rise of only 43 K within steady state can be observed. Total volume of the module is 2.11 liter which leads to a power density at steady state of 4.4 kW/ liter. Figure 2: Fully integrated 3-phase inverter with control and power part assembled on a copper multilayer FR4-PCB. Right side shows typical operation conditions on a lab motor test bench. 99 Enabling Technologies - die attach and substrate technologies for power electronics 3.2 Ceramic substrate technologies 3.2.1 Thick film technology Thick film technology describes methods to creating passive circuit elements and components such as conductor tracks, insulating layers and resistors. Usually the method employs screen printing of a viscous material on a rigid substrate, curing and subsequent firing at high temperature. Thick film inks for conductor tracks are made of silver, silver/ palladium/ platinum, gold or copper. Fine line printing and curing starts at 5 µm thickness and thick film printing of conductor tracks can be produced up to 500 µm thickness. Polymer thick film is mainly cured at lower temperatures. For insulating materials dielectric inks are applied. Resistor pastes consist of metals/ metal oxides in a glass matrix for different sheet resistivities. Rigid ceramic substrates are made of alumina (Al 2 O 3 ), aluminium nitride (AlN) or silicon nitride (Si 3 N 4 ). These materials distinguish primarily by its thermal conductivity values or by very low coefficients of thermal expansion (CTE) of e.g. 3 ppm/ K for Si 3 N 4 . Matching of CTE for different material layers in an assembly of power electronic devices determines reliability over lifetime prone to temperature or power cycling. Especially ceramic substrates offer a comparable matching of CTE to silicon (2.6 ppm/ K) or gallium nitride (3.2 ppm/ K). In addition ceramic substrates offer high isolation strengths, e.g. alumina 13.4 kV/ mm. A suspension of metal powder, an organic binder and a glass frit serves as thick film paste which is applied on a mesh and screen printed on a substrate. An inspection system controls quality of screen printing process. After a drying step, a firing profile is applied in an oven. Close to so-called eutectic point a phase condition is achieved where particles at the substrate/ paste interface are exchanged by diffusion processes. Organic binders are removed by thermal processes and the glass frit component sinks to the binding interface between paste and rigid substrate. After firing process the steps can be repeated as necessary in order to build thicker layers subsequently. Fig. 3 presents products realized with thick film screen printing technology at AB. Figure 3: Product examples of thick film screen printed ceramic substrates. Left side - High voltage cooling pump control unit up to 1.5 kW for FCEV. Right side - Air compressor control unit with 700 W. 100 Enabling Technologies - die attach and substrate technologies for power electronics 3.2.2 Direct bonded or brazed metal substrates Metal films of 300 µm thickness are directly bonded by a chemical oxidation step of the metal film (typically direct copper bonded, DCB) and subsequent sintering process on a ceramic substrate. Some manufacturers also provide aluminium bonded substrates with a nickel/ phosphor finish for soldering. Typically the solid metal layer of direct metal substrates is much denser than the metal layers which are created by screen printing and curing processes. AB observed that a sponge-like metal structure (e.g. as for copper ink sintered thick film hybrids, see Fig. 5) may provide a positive impact on CTE mismatch during extreme temperature cycling (0-350°C) since the metal structure is interrupted by pores. Product example: In-wheel electric drive power module As an example for a power module based on direct metal substrate an in-wheel electric drive is presented which integrates power electronics, control unit and brake in an 18 inch wheel for automotive vehicles, see Fig. 4. Maximum mechanical torque of 800 Nm and 510 Nm continuously is generated within a wheel of 31 kilogram weight. 85 % of kinetic energy can be recovered by recuperation. Power and control unit of module are separated between two different types of substrates. The control unit is realized by a standard FR4 PCB and the substrate for power electronics is chosen to be a high thermally conductive AlN thick film hybrid > 200 W/ mK. Bare die MOSFET chips are soldered on the drain side to the AlN direct metal substrate with a 400 µm thick aluminium metallization and Ni/ P finish. The substrate is directly soldered on an AlSiC baseplate with a fin structure. Within this assembly materials are chosen which offer very low CTE values. All thick material layers starting at the bare die silicon chips (2.6 ppm/ K) through an AlN ceramic substrate (~ 5 ppm/ K) and the baseplate AlSiC (~ 8 ppm/ K) offer similar CTE and delamination of the large solder areas during temperature cycle tests can be prevented by this choice. Top side of the chip is wire bonded with 300 µm thick wires at gate and source pads. Due to vibration and contactibility requirements bond connections are covered with a silicone potting. Figure 4: Intelligent electric drive power module fully-integrated in automotive wheel. 3D geometry of power module is adapted to specific application. 101 Enabling Technologies - die attach and substrate technologies for power electronics 3.3 Thick film technology with rigid metal substrates AB developed a new substrate technology based on screen printed copper conductor traces on rigid aluminium substrates. By this method aluminium is made solderable, and direct advantages from thermal and electric properties of aluminium can be exploited. Oven curing processes are carried out with temperatures just below melting point of aluminium substrate at about 650 °C where particles become mobile at the material interface through diffusion. Since at such elevated temperatures aluminium is changed to a softer state, special care needs to be addressed for substrate bending effects during burning. Multiple burn processes turn aluminium subsequently softer which offers some positive effects for additional processes for module assembly. On one hand robust connections of semiconductors to pure aluminium busbars can be achieved with both sided die attach technologies. On the other hand aluminium offers an elastic modulus which is roughly a factor of two lower than copper and allows for some bending of the module without stressing sensitive electronic components. In addition both sided cooled modules can be assembled with lowest thermal resistance values. Fig. 5 shows a focused ion beam cut through thick film layers deposited on rigid aluminium substrates. The illustration shows a scanning electron microscopy image of a cured thick film copper conductor trace. Porous structure on the conductor trace can be seen and the glass frit interface to aluminium can be detected. Figure 5: Left side - SEM image of a FIB cut through thick film copper metallization layers on rigid aluminium metal substrate. Right side - Thick film screen printed electrical layout with copper metallization and isolation layers. For high electrical conductivities between copper/ aluminium interfaces a special pretreatment of the rigid substrate is required. A standard SMD and vacuum reflow soldering process attaches passive components and power semiconductors from both sides to metallized aluminium substrates. Whole assembly is flat with typical height of less than five millimetres. The module is based on metal substrates with a high electrical conductivity. Therefore countermeasures need to be taken to isolate the module in a safe way. A current trend for power module packaging is to use low CTE epoxy materials for moulding. 102 Enabling Technologies - die attach and substrate technologies for power electronics These materials are typically processed with transfer moulding equipment. A high flow capability of compounds is achieved by typical temperatures of 180 °C and high transfer pressure of about 80 bar. This allows covering small arising gap sizes. Also silicone-free packaging materials are used and a bubble-free filling of the compound can be realized with vacuum suction system. In combination with a soft solder alloy (e.g. with silver or indium content) a high thermal cycling stability can be achieved of up to 2000 cycles from -40 to +125 °C with a change time of 15 seconds and a dwell time of 30 minutes. In addition Fig. 6 shows power cycling data of IGBT high voltage power switch modules with reliable stability. Rise of temperature difference between onand off-state and of on-voltage indicate aging effects of the power module. These effects are within tolerance window defined by guidelines LV 324. Vertical lines of the on-voltage measurements show a short time interval in which thermal resistance is measured by a Tr3ster thermal impedance measurement scheme. In addition to increasing on-voltage values, increasing thermal resistance values are also a signature for aging. Ultrasonic microscopy scans reveal material suppression of epoxy material close to semiconductor chip since at this location highest thermal expansion coefficient differences arise. Nonetheless only minor changes in the die attach material conditions of the module is found and it is full electrically functional after more than 2 ∙ 10 power cycle numbers for on-/ off temperature difference of 100 K. Figure 6: Left side - Power cycling test data of stacked high voltage power module. Right side - Scanning acoustic microscopy of power semiconductors before and after more than 2 ∙ 10 power cycles at T=100 K. Product example: Battery solid state protection switch 48 V In Fig. 7 a compact bidirectional battery solid state protection switch is presented. Curve tracer measurements reveal a low on-state resistance of the module of 0.75 m and a thermal resistance value of the module lower than 0.1 K/ W from thermal impedance measurements (single side cooling). Thermal energy coupling to the water occurs with a junction-to-water resistance of about 0.2 K/ W and a test current of 150 A. By combination of an ultralow dissipation factor and a high thermal conductivity of the device, a temperature rise of only 28 Kelvin for 300 A in continuous operation and maximum currents of 900 A for one second duration are possible with a liq- 103 Enabling Technologies - die attach and substrate technologies for power electronics uid cooling temperature of 45 °C. This device is transfer molded with an epoxy compound for isolation and mechanical stability reasons. Figure 7: Battery solid state protection switch 48 V. Low side - a combination of low on-state resistance and low thermal resistance allows driving high current profiles up to 900 A. Lower left side - Average of 20 curve tracer measurements revealing an on-state resistance of the bidirectional switch of 0.75 m. Lower right side - thermal impedance measurements involving different thermal interfaces of the module. Method reveals a thermal resistance value lower than 0.1 K/ W for module only. Coupling to cooling water is realized with 0.2 K/ W (single sided cooling). 4 Trends in die attach technology Beside already mentioned “traditional” die attach technologies with solder and wire bonding technologies there exist some newly arising methods for power electronic packaging. Especially wide bandgap semiconductors, mainly SiC and GaN, which allow for higher junction temperatures and switching frequencies require alternative 104 Enabling Technologies - die attach and substrate technologies for power electronics die attach methods. In this section different die attachment methods are presented. The choice of die attachment methods presented here does not claim to be complete. 4.1 Silver sintering Generally one discriminates between pressureless and pressure applied sintering. Attachment occurs by physical parameters time, pressure and temperature. By thermodynamic diffusion processes individual particles combine to partially solid phases. These sinter processes are dependent on metal powder properties like size, density and surface energy. Typically nanosilver particles are used, and temperatures higher than 200 °C, pressures higher than 5 MPa and a few minutes of processing time are applied for pressure-assisted sintering. Processing times for sintering without pressure are typically longer (3/ 4 to 3 hours depending on material/ process), but can be also achieved with slightly lower temperatures (e.g. 180 °C). Area size of sintered layers is typically limited, since thermodynamic process is volume dependent, but steady improvements have been achieved recently over the last years in which large area sintering has been shown. Die attachment to non-precious surface finishes like copper or tin have been realized or are under development. Directly comparing to solder technologies the sintered layer is high-temperature stable, basically limited to melting point of bulk silver at 967 °C, possesses a higher thermal conductivity (~ 100W/ mK for pressureless, 200 W/ mK pressure assisted versus ~ 60 W/ mK for solder). Dependent on the achieved density of the sintered layer the material can be quite stiff with a typical E-modulus of up to 50 GPa, but lies in a similar range as solder alloys. Especially important for high voltage applications die attachment by sintering is residual-free. 4.2 Transient liquid phase sintering This technology is also based on a low-temperature thermodynamic diffusion process. Corresponding pastes constitute of an organic matrix in which a binary metal system is embedded. Typical binary systems are silver-tin, copper-tin, gold-tin, nickel-tin, or indium-gold with different melting points. At a specific processing temperature, the metal with a lower melting point diffuses into the other metal, forming an alloy with a higher melting point. For example copper-tin system forms alloys (e.g. Cu Sn ) at temperatures similar for conventional reflow soldering (250 °C), and is stable up to 400 °C. Material is available as preforms, powder or paste, and can be processed under pressure or without pressure, similar to silver sintering. The main advantages are residual-free joining, lower processing temperatures than operation temperature, and a relatively homogeneous bond region structure. 4.3 Electrically conductive glue These materials constitute of a thermosetting resin system in which silver nanoparticles higher than 80 wt. % are embedded. Depending on the material system and weight percentage of silver content processing conditions like curing time, temperature profiles and environmental condition (e.g. nitrogen atmosphere) can be quite 105 Enabling Technologies - die attach and substrate technologies for power electronics different. Also arising properties of the cured die attachment layer such as electrical, thermal conductivity, mechanical properties can be adjusted by appropriate material preparation, and a wide range of settings is available. 5 Summary E-mobility currently drives innovation in different industry areas. Only if all areas starting from semiconductor, die attach, substrate materials match in their properties a successful transition to higher electrical, thermal efficiency, and last, but not least to a safe operation and cost effectiveness can be achieved. Primarily, material stress in each interlayer needs to be minimized. This can be achieved by high electrical conductivity (low thermal resistivity), so that thermal energy is coupled efficiently to a large thermal reservoir. Under typical operation conditions (time, current) the temperature range is kept small in this way. This is mainly achieved by double-side cooled power modules where die attachment area of the semiconductor is fully connected. Another direction is to find a material mixture which is perfectly matched within their mechanical properties (elastic modulus, CTE, etc.). In combination with small contact area of the interlayers of the module assembly, and even under large temperature change, the system can be stable and functional. Another trend is to design a module with very low parasitics so that very high switching frequencies of the electrical current can be achieved. This requires on one hand ultralow parasitic inductances (electrical current loops need to be minimized) and on the other hand very low parasitic volume resistivity of involved materials and corresponding interfaces. Last trend is to drive application to high voltage and therefore high electrical power, so that current flow and therefore heating by ohmic resistance can be minimized. This requires new materials for high voltage insulation for DC and AC mode. From material science and developer’s perspective design of power electronic devices will keep exciting. Literature: [1] https: / / www.ecpe.org/ [2] ECPE Guideline AQG 324, Qualification of Power Modules for Use in Power Electronics Converter Units (PCUs) in Motor Vehicles, V01.05 (2018). [3] http: / / www.aecouncil.com/ AECDocuments.html [4] Wintrich A. et al, Applikationshandbuch Leistungshalbleiter, Semikron International GmbH, ISLE Verlag (2015) 106 Future Packaging Technologies in Power Electronic Modules Christoph Friedrich Bayer, Zechun Yu, Hoang Linh Bach, Jonas Müller, Andreas Schletz Abstract This paper summarizes power electronics packaging trends and technologies that were identified at the Fraunhofer IISB as promising in future. Due to upcoming demands including further integration, minimization, modularization, high temperature and high voltage capability, listed packaging topics are recently addressed and researched. Besides a deeper insight into three main topics, high switching speed by RC-snubber devices and high current copper projection welding are presented. Main topics are ceramic embedding by subtractive manufacturing processes for harsh environments and best semiconductor electrical interconnect, selective sintering of power electronic devices on organic circuit carriers for higher process flexibility, higher powers and long lifetime, as well as direct bonding without sinter or solder material as a novel interconnection technology in power electronics. Kurzfassung In diesem Beitrag werden Trends und Technologien im Bereich der Leistungselektronik zusammengefasst, die am Fraunhofer IISB als zukunftsweisend identifiziert wurden. Aufgrund der anstehenden Anforderungen an die weitere Integration, Miniaturisierung, Modularisierung, Hochtemperatur- und Hochspannungsfähigkeit werden die aufgeführten Aufbau- und Verbindungstechnikthemen in jüngster Zeit behandelt und untersucht. Neben einem vertieften Einblick in drei Hauptthemen werden hohe Schaltgeschwindigkeiten durch RC-Snubber-Dämpfungsglieder und hochstromiges Kupfer-Buckelschweißen vorgestellt. Schwerpunkte sind die keramische Einbettung durch subtraktive Fertigungsprozesse für harsche Umweltbedingungen und beste Halbleiter-Verbindungstechnologie, das selektive Sintern von Leistungselektronikbauteilen auf organischen Schaltungsträgern für höhere Fertigungsflexibilität, höhere Leistungen und lange Lebensdauer sowie das direkte Bonden ohne Sinter- und Lotmaterial als neue Verbindungstechnologie in der Leistungselektronik. Introduction Packaging is one of the most important issues in power electronics. It has a significant influence on the functionality, lifetime and reliability of a power module. During operation, losses occur in the semiconductor chips, resulting in a large thermal load in overall material stack. In addition to module design, the selection of a suitable chip contacting technology is therefore decisive for optimum heat dissipation. This paper presents a summary of current interconnection technologies (Fig. 1) in power electronics and gives an overview of their use in innovative applications. The classical 107 Future Packaging Technologies in Power Electronic Modules soldering process, the promising silver sintering process and the novel direct bonding process are examined in more detail. Moreover, the publication concludes with three examples of different technologies that are ready to use in recent modules. Figure 1: Possible device interconnection technologies in power electronics packaging (red: research topics) Besides everlasting interest in further integration, minimization, modularization, high temperature and high voltage capability, new materials such as wide bandgap semiconducters come up and find their way into recent application. Thus it appears that SiC semiconductor devices are now used in power electronics mainly for switching high electrical currents (up to several 100 A) and high voltages (≥ 1 kV) in automotive drives, energy transmission and traction applications. Also due to its larger band gap (3.2 eV for SiC-4H), SiC is superior to Si in terms of high temperature capability (junction temperature of the power devices), high power density as well as high switching speed [1]. However, the approximately three times higher cost of the SiC wafer substrate, and thereby power devices, is a not negligible issue for the industry [2, 3]. In order to compensate the higher device fabrication costs, a cost optimization at the system level including all related relevant components is needed. Therefore, reducing the chip dimensions and number can be a cost effective solution. As a result, the power density and the power losses will be increased and can rise the operating temperature of the chip significantly. Basically, the properties of SiC semiconductor material itself allow them to withstand a higher temperature range up to 500 °C. However, most of the current packaging techniques are limited to operating temperatures of T max < 175 °C due to the insufficient thermal stability of the complete power module [1, 4]. Especially soft solder joints, aluminum bond wires, low cost organic potting compounds and housings are critical in terms of high thermal and chemical stability as well as thermomechanical cycle capability. Furthermore, next to temperature and performance capabilities, electrical properties such as fast switching, lifetime and reliability reasons, new power devices need to be packaged to their demands. Longer lifetime for smaller power devices, smaller com- Device- Interconnection- Technologies Soldering Soft-Soldering Brazing Diffusion-Bonding- (TLPB) Metal-Sintering Ag-Sintering Cu-Sintering Direct-Bonding Further-Possibilities- for-Lower-Powers Adhesive-Bonding Conductive-Adhesive- Bonding 108 Future Packaging Technologies in Power Electronic Modules mutation cells and thereby lower stray inductance for fast switching, hence lower voltage overshoot, as well as upcoming environmental resistance issues demand ongoing packaging research and development. 1 Interconnection Technologies 1.1 Soldering Soft soldering is one of the best-known chip contacting processes to date. It is at low cost, easy to process and comes with a broad variety of materials. Thanks to a long history, all kind of process equipment, analysis tools and methods, quality ensurance are available and accepted. Thanks to the melting and wetting of the solder the surface tension leads to a self alignment of the bond line and the joining partners. Two metallic components and the solder are usually brought into contact with each other. When the solder is heated above its melting point, an alloy of solder and the surface material of the joining partner (intermetallic phase) is formed. After the cooling phase, a material-locking connection of the components is generated. Important for achieving a high soldering quality is a good wetting and a acceptable voiding. On the market, solders are available in various forms, such as paste, preform, solder wire and solder ball. Depending on the shape, the solder can be applied by various methods, such as dosing, stencil printing or jet dispensing using solder pastes (Fig. 2). For the production of solders, at least one, often two or more metals are mainly mixed together (Tab. 1). The following components are most frequently found in solders: Sn (tin), Pb (lead), NI (nickle), Ag (silver), Au (gold), Cu (copper), Bi (bismuth), In (indium), Sb (antimony), Si (silicon), and Ge (germanium). Lead-containing solders are nowadays only used to a limited extent for industrial electronic applications due to the toxic properties of the heavy metal (see RoHS 2 or 2011/ 65/ EU). In power electronics, lead-free solders such as the SAC-305 (SnAgCu, melting point ~ 220 °C) are widely used. Gold-Germanium solders (e.g., Au88Ge12, melting point ~ 380 °C) with a top perfomance can be used for requirements at particularly high temperatures, but these have to high costs. The soldering process can be designed differently depending on the equipment and accessories. For the development of power modules, processes such as hot plate, vacuum, vapor phase, infrared, resistance and induction soldering are frequently used. The selection of a suitable soldering process depends on the material combination, thermal capacitance, transient soldification front and mechanical properties of the solder, module design and technical demands of the power module. Wide band gap semiconductors, especially SiC is not limited in the soldering processing temperature. Therefore it is possible to have (active) metal brazing as an alternative to soft solder. Brazing materials are high temperature capable, having excellent mechanical properties at the same time. In contrast to metal sintering, there is self-alignment due to melting and surface tension. 109 Future Packaging Technologies in Power Electronic Modules Figure 2: Overview of soldering technology for power modules Table 1: Selection of solder materials for power electronics applications [24] Solder composition T Liquidus in °C Electrical Conductivity in m/ Ωmm² Thermal Conductivity in W/ mK CTE in ppm/ K BiPb(32)Sn(15.5) 95 1.5 -- 16.6 BiSn(43) 139 -- 19 -- InPb(50) 215 3.3 22 24.4 SnPb(37) 183 7.4 70 24.3 SnAg(3.5)Cu(0.9) 217 -- 60 -- SnCu(1) 227 8.8 60 -- Sn(96.5)Ag(3.0)Cu(0.5) 217/ 221 -- -- -- Sn(96.5)Ag(3.5) 221 -- -- -- Sn(95)Ag(5) 221/ 245 -- -- -- Sn(99.3)Cu(0.7) 227 -- -- -- Sn(97)Cu(3) 227/ 300 -- -- -- Sn(100) 232 -- -- -- PbIn(19) 276 -- -- -- PbSn(5)Ag(2.5) 280 < 5 44 29 PbSn(5) 315 < 5 45 29 PbSn(2) 325 4.9 -- -- Pb(100) 327 -- 37 29.3 AuSn(20) 280 < 5 46 16 AuGe(12) 356 7 34 13.4 AuSi(3) 363 25 26 -- 1.2 Sintering The big advantages of metal sintering are its excellent mechanical properties which lead to excellent thermomechanical lifetime [35]. More than one order of magnitude lifetime improvement can be gained in comparison to SAC soft solder and aluminum wire bonds. On a system point of view, this can reduce the semiconductor cost by 25 %. Another benefit is the processing temperature and time which is nevertheless comparable to soft soldering. During sintering, the bond line is mechanical stable. There is only certain shrinkage in the thickness but no squeeze out like for molten Solder-Material Printing-Technology Solder-Technology Vapor-Phase Infrared Resistance-Brazing Induction-Brazing Preform Paste Stencil-Printing Screen-Printing Jet-Dispensing Balls/ Bumps 110 Future Packaging Technologies in Power Electronic Modules solder possible. Therefore sintering is a perfect solution for complex power module concepts like double sided cooled ones. On the other hand, there is only diffusion - no melting and only poor surface tension. Moreover, there is no self-alignment of the joining partners. The bond line is more or less dense, depending on the process parameters. Also not sintered material residues at the outer device edges can break away causing electric conductive particles. The biggest challenge is to connect the sinter bond line to the chip and substrate metallization. To distinguish between pass and fail is hard to detect and has to be ensured by process control. In the metal sintering process (usual sintering processes, Fig. 3), a compound layer is produced using micro or nano metal particles by diffusion. The particles are usually made of silver (Ag) and are mixed in a paste together with binders, solvents and other ingredients for preventing agglomeration [5] and making it processable. The risk of agglomeration exists due to the small particle size. This is used consciously, since thereby the surface energy is reduced, which is needed for the sintering process as driving force (minimization of the surface energy) [6]. The viscosity of the paste is controlled using organic solvents which is relevant for the processing of the paste. Stencil or screen printing and dispensing are commonly used for this purpose. However, the die transfer film (DTF) can be used alternatively. This technology will be discussed more detailed in section 2.2. The process flow is as follows. At first, the paste is applied and then dried in order to remove the organic material down to approximately 1 % remaining solvents due to manufacturer specifications (Heraeus). Subsequently, the device is manually placed or with an automatic placement machine. In addition, the applied temperature ( about 130 °C) ensures a sticky surface on which the die is attached. After that, the actual sintering process follows. A distinction is made here between pressure-assisted and pressureless sintering, which also differs in terms of the used sintering pastes. In pressure-assisted sintering, the substrate is bonded to the attached device with 10 MPa to 40 MPa attemperature from 200 °C to 300 °C. During pressureless sintering, on the other hand, the sintering phase takes place without external force in a controlled atmosphere (e.g. air or nitrogen). Furthermore, the paste is not dried after printing and the die is placed into the wet paste [7], however, sintering time is much longer than with preassure (about an hour compared to some minutes). The sintering process is similar for both pressureless and pressure-assisted sintering. It consists of the following: I) initiation, II) intermediate phase, and III) final phase. These process steps merge seamlessly. During I) the particles can first be reorganized into more stable positions. In addition, the formation of the sinter necks begins [8]. Further, II) is the main part of the sintering process. The sinter necks grow and a compaction takes place which reduces the pore sizes. Finally, the larger grains and the isolated pores shrink during III) until the theoretical maximum density is reached [8]. For dissolution after completed silver sintering, a temperature above the melting point of Ag, that is 961 °C, is required [9]. 111 Future Packaging Technologies in Power Electronic Modules Figure 3: Variety of metal sintering technology for die attach, e.g. mainly silver but also copper or copper-silver sinter paste 1.3 Direct Bond Talking about soldering and sintering, there is always the problem of how to conntect the tiny gate pad of SiC devices, if double sided interconnects without wires are desired. One alternative technology can be a direct bond technology without any additional process steps for the application of the bond line material. The right material mix comes with the metallization of the semiconductor. Without printing or placing a preform material, the tolerance problem gets simplified. The layers are thin, therefore no squeeze out occurs. Without sinter material, contamination by particles is banned. Direct bond technologies are easy to use, however, the drawback is that a very plane surface is needed. Next-generation power devices such as silicon carbide and gallium nitride devices have proven superior switching performance, while being capable of high-voltage and high-temperature operation [10]. Therefore, wafer-level and 3D integration concepts like chip-stacks, through silicon via (TSV) and silicon interposer design, which provide short interconnects with low parasitic effects [11], are becoming more and more interesting for power electronic applications. Interconnect and die-attach technologies in 3D power integration are currently limited by ultra-fine pitch devices [12]. Although many other bonding technologies, such as thermo-compression bonding [13], transient liquid phase bonding [14] and silver sintering [15] are extensively used to replace conventional soldering, relatively high bonding temperature and pressure, as well as complex manufacturing routine processes are required for a homogeneous bonding interface. Therefore, direct bonding (Fig. 4), which is also called stress migration bonding (SMB) [16, 17] has been investigated as an alternative die-attach technique for fine structures. This technology allows joining partners with a thick top side metallization together without an additional intermediate layer (Fig. 5). This process can be performed in air under different pressures (10 MPa - 35 MPa) and temperatures (240 °C - 300 °C). High temperature processing of the devices leaves the metallization with a large mechanical stress due to different coefficients of expansion (CTE) of the materials. Therefore, stress migration, which is driven by a stress gradient, often occurs in the metallization [18]. During this diffusion-controlled process, the so-called hillock formation and abnormal grain growth have been observed in the metal layer as a result of stress relaxation [19, 20]. The as-deposited film has equiaxed grains with sizes smaller than the layer thickness. The small, normal grains may continue to Sintering Particle-Size micro nano Application-of-the- Sinter-Material Stencil-Printing Dispensing Die-transfer-film- (DTF) Pressure‐assisted Pressureless-or Low-Pressure 112 Future Packaging Technologies in Power Electronic Modules grow but large abnormal grains grow at faster rates [21]. These abnormal grains grow generally by annihilation of surrounding normal grains until they impinge on each other [21]. By using numerous hillocks and abnormal grains on Ag films as bonding medium, high-strength bond lines can be achieved at low temperature and pressure. Figure 4: Direct bonding examples in power electronics - substrate materials rank as carriers of the needed metallization Figure 5: Schematics and cross-sectional SEM micrographs of the bonding interface; left: silver sintering, right: Ag stress migration bonding Direct-Bonding Film-Material- Combination-- Options Ag‐Ag Cu‐Cu Au‐Au Al‐Al Substrate- Materials Conductive Semiconductive Si GaN SiC Insulating Al 2 O 3 AlN Si 3 N 4 113 Future Packaging Technologies in Power Electronic Modules 2 Recent Research Examples Power modules that are recently produced cover traditionally build-up technologies like wire-bonding as well as devices that are mounted by soldering and sintering on a circuit carrier. Then the circuit carrier is attached to the heat sink, by soldering or a thermal interphase material. Beginning with these benchmark arrangements, new top side interconnection technologies came up such as leadframes instead of wirebonds [35, 36] as well as flexible printed circuit boards for the topside connection of the power devices [37, 38]. Furthermore, power modules with topside connection by other circuit carriers were developed for higher integration or better cooling options. The better cooling performance can go up to the maximum of the additional area that is mounted on the top side of the power devices [31, 39-42]. Besides such interesting and promising industrial approaches for power module concepts, the Fraunhofer IISB recently focused on the following exemplary introduced technologies in public founded projects. 2.1 Ceramic Embedding of Power Devices An innovative module design concept was presented in [22]. It is based on the embedding of power semiconductors in prefabricated ceramic circuit carriers (Fig. 6, ceramic embedding technology) such as DBC substrates (direct bonded copper). The packages are particularly suitable for high temperature applications above 300 °C due to the DBC ceramics such as Al 2 O 3 , AlN and Si 3 N 4 which have a very high temperature resistance and thermal conductivity compared to PCB (printed circuit board) and LTCC (low temperatur cofiered ceramic) technology. Furthermore, conventional bond wires and plastic housings are completely dispensed within this concept. Before the embedding process, the DBC substrate is structured by using laser technology to create cavities matching the geometry of the power semiconductor (Fig. 6). It is important to achieve an oxideand damage-free surface of the DBC. In the next step the chip contact is made by using a soldering process (SAC solder preform, 240 °C, nitrogen atmosphere). In the last step, the package is sealed with a copper lid. The finished package is shown as a demonstrator of four stacked SiC diode substrates in Figure 7. Further assemblies based on this concept are currently being developed, such as multichip packages (half-bridge and full-bridge circuits) with wide band gap power semiconductors. The possibility of stacking the embedded DBC packages brings considerable advantages in terms of high voltage capability. The hermetic sealing of the packages is one of the upcoming milestones, with the aim of increasing the reliability and lifetime of the modules. 114 Future Packaging Technologies in Power Electronic Modules Figure 6: Schematic of ceramic embedding technology [22] Figure 7: Sample produced by using ceramic embedding technology [22], four stacked embedded SiC diodes blocking over 12 kV without further insulation material 2.2 Selective Sintering on Organic Circuit Boards Silver sintering on ceramic circuit carriers is a mature process. However, this does not apply to sintering on organic PCBs. Due to the trend towards higher energy densities even for PCB designs (e.g., motor control, mixed signal and power electronics on one board), the applicability of the classic joining processes of soldering or conductive adhesives has reached its limits. Especially, this affects the components with significant power losses. Sintering all components onto the PCB is simply not possibly. The reason is that sintering needs special compontent’s metallizations. For some single devices like (shunt) resistors this might be possible. 115 Future Packaging Technologies in Power Electronic Modules But not for the big variety of a complex digital control with microprocessors, capacitors, inductors, resistors, analog to digital converters, power supplies and so on. The idea is to populate the PCBs in the traditional way by soldering or adhesives - except the power devices. These are placed afterwards. Since the SMDs (surface mounted device) and THDs (through hole devices) are already mounted at this point, no low cost stencial or screen printing can be applied anymore. This led to the goal of developing a selective Ag sintering process for bare dies [46]. Subsequently, a lifetime characterization using power cycling tests (PCT) was conducted showing excellent perfomance [45]. The development of the selective sintering process has already been completed and the process is described below (Fig. 8). The entire process takes place in die bonder. Before the start, the PCB is already populated with components (I). Then the prepared PCB is soldered or cured. In the next step (II), the device is picked up and pressed onto the DTF. The DTF is located on a special station and lies on a supporting silicone layer. The DTF is a sinter paste layer which has been pre-printed and dried on plastic film (sinter material: Alpha Argomax®). Thereby the printing and drying step is eliminated. Pressing the dies on the DTF cuts out the film. At the pickand place tool, temperature is applied. This supports the attachment of the DTF to the die. In the following (III), the die with the sinter film is attached with force onto the PCB. Temperature is applied to the tool (IV) and the die is sintered onto the PCB. Figure 8: Schematic of the selective sintering process, consisting of I) heating the PCB, II) cutting out and attaching the DTF, III) loading the dies, IV) sintering, and V) assembling with bond wires The assembled PCB (after step V, Fig. 9) with signal and power components, can now be processed to the final steup knwon from the state of the art. The development of the selective sintering process was followed by PCT and TCT and it is at least as good as recent joining technologies following Dresel et al.. The new technology of selective sintering on PCB enables the use of smaller chip sizes. This makes it possible to increase the maximum energy density on low-cost PCBs for applications such as motor controls. 116 Future Packaging Technologies in Power Electronic Modules Figure 9: Selectively sintered bare die on an organic circuit carrier 2.3 Direct Bonding of a Small Converter System GaN-based 48 V to PoL DC-DC converters are becoming more popular in numerous markets such as telecommunications, industrial and aerospace. A suitable packaging concept for this GaN-based power converter requires low inductance interconnects and improved thermal management, high reliability, and compactness [23]. Due to the requirements, a compact interconnection technology must be used to replace conventional wire bonding and soldering. Novel hybrid circuits can be enabled by heterogeneous integration of GaN power devices with an active silicon interposer. In order to achieve innovative stacking structures between two semiconductor devices, silver direct bonding is being developed. Figure 10: Planar packaging concept with vertical GaN power devices and lateral Si-Capacitors 117 Future Packaging Technologies in Power Electronic Modules Figure 11: SAM images (top) and shear strength (bottom) of Ag joints before and after TCT Based on a GaN H-bridge topology, a package concept with planar interconnections is shown in Figure 10. In a first step, two sets of decoupling capacitors are integrated directly in a silicon-based interposer. Then four vertical GaN MOSFETs are assembled face-down onto the silicon interposer with silver direct bonding without using any additional intermediate layers such as solder and sinter layer. This forms a GaN bridge switching cell that can be mounted onto a printed circuit board (PCB) to realize complex power and drive circuits. Temperature cycling test (TCT) was carried out on direct bonded samples in a twochamber thermal shock setup (+150 °C/ -55 °C; 15 min/ 15 min) to determine the resistance of the Ag joints to alternating temperature extremes. Three test groups, each with different metallization (Fig. 11) were cycled to compare the lifetime of the direct bonds. All samples were bonded using the same optimized parameters (300 °C, 35 MPa, 1 min). Scanning acoustic microscopy (SAM) was used to characterize the samples during TCT to identify voids and delamination in the silver joints. Moreover shear tests were carried out on the samples before and after TCT to measure the ageing of the bonding strength of Ag joints. Figure 11 shows the SAM images and shear strength of the initial and aged samples of the three test groups. The results show that neither delamination by SAM nor decrease in shear strength was detected for Cr/ Ni/ Ag metallized samples after 700 thermal cycles. 3 New Connection Technologies for Recent Modules In this section, three approved and useable technologies are summarized and referenced. These technologies are now ready to be integrated into recent production lines after a short adaption to the module process flow and were alredy published in former works in more detail. 3.1 Si Modules with Fraunhofer Add-on for SiC [25] “For extremely fast switching, it is necessary to reduce the effective switching cell inductance. This can be done by connecting a capacitance next to the SiC chip. But merely adding a capacitor will result in very large sizes acting as a second DC link capacitor, buffering energy over the whole switching period. Or it may be done on a1 a2 b1 b2 c1 c2 0 20 40 60 80 100 120 shear strength in MPa Cr/ Ni0.5/ Ag, initial Cr/ Ni0.5/ Ag, after 700 cycles Cr/ Ni1/ Ag,initial Cr/ Ni1/ Ag, after 700 cycles Ti/ Ag, initial Ti/ Ag, after 700 cycles 118 Future Packaging Technologies in Power Electronic Modules small size, buffering only during commutation, resulting in tremendous ringing through the terminal [27]. An overall very effective method is to use an impedance optimized highly damping RC-network acting as a decoupling filter. This network works very well, but results in significant thermal losses which cannot be handled by discrete parts inside the module. The novel silicon integrated high performance Si- RC network was recently introduced for reducing AC output ringing, lowering EMI issues on inverter modules while maintaining excellent reliability for a DBC mounted passive device [26-30].” [25] Figure 12: Equivalent circuit of conventional power module with integrated Si-RC chip [25] Figure 13: Voltage overshoot on SiC switch by (too) small R Gate compared to module including Si-RC, L DC-Link = 45 nH [25] This chip is also an excellent candidate to act as the introduced decoupling network, solely capable to handle the high ohmic damping of about 98 watt for a 45 kW module as shown in Figure 12. The silicon RC chip achieves an integration density of up to 0.3 nF/ mm 2 for 1200 V operating voltage. Heat removal of 200 W/ cm 2 is easily realized by a thermal chip to the DBC resistance similar to power semiconductor 119 Future Packaging Technologies in Power Electronic Modules devices. For the presented application, an RC chip of 12 nF/ 1 Ω with a chip area of 49 mm² is suitable. Additionally, these devices can be assembled using the same technology as the SiC power devices. The parasitic inductance of the Si-RC chip is dominated by the bond wire inductance of 3 nH to 4 nH while internal inductance is less than 100 pH. The impact of this Si-RC element on the switching waveform is shown in Figure 13. The thin curve is for a classic power module using SiC MOSFETs with R Gate = 2 Ω, resulting in a tremendous voltage overshoot. The dashed curve is state-of-the art technology, lowering switching speed by R Gate = 13 Ω, resulting in higher dynamic losses and thereby in larger SiC chip area. Additionally, the turn-on/ off is delayed. The bold curve is using a classic module with the described Si-RC decoupling chip, enabling the module to switch extremely fast (R Gate = 2 Ω) with limited voltage overshoot. This results in drastic SiC loss reduction and thereby saving a significant amount of SiC chip size and costs.” [25] 3.2 Double Sided Sintering (pressure-assisted and pressureless) [32] “Instead of soldering the chip onto DBC (Direct Bonded Copper) substrates pressure assisted silver sintering is already well known improving demonstrably the lifetime of the bottom bondline. Also the lifetime of top side chip contacts can be improved by pressure assisted silver sintering of silver leadframes instead of wire bonding [35]. Next to the stress relieving effects, the perforations (Fig. 14) also serve as diffusion channels, enabling the required oxidation of conventional pressureless silver sintering pastes. The concept shows the great benefit of this innovative pressureless sintering concept, which is not limited by the chip area using appropriate leadframe perforations. Comparable FEM simulations were performed to investigate the influence of different leadframe designs on the thermo-mechanical stress distribution. Active power cycling tests (PCT) were executed to compare the lifetime of pressure assisted with pressureless top side chip contacts (Fig. 14). Figure 14: Schematic cross section (top) and picture (bottom) of a top side pressureless sintered silver leadframe on IGBT and antiparallel diode [32] 120 Future Packaging Technologies in Power Electronic Modules The lifetime of this technology was compared with a pressure assisted sintered technology and with conventional wire bonded chip contacts within an active power cycling test. The test shows an improvement of the power cycling capacity of at least seven times in comparison to the wire bonded samples and an improvement of at least 2.5 compared to the pressure assisted sintered leadframes. The failure analysis of the failed samples indicates that the rather soft and elastic pressureless sintering layer works as a thermo-mechanical buffer between the silver leadframe and the chip. In contrast to this observation the brittle and inelastic pressure assisted sintered bondline transmits the thermo-mechanical stress directly into the top side of the chip, which provokes cracks propagating into the chip volume. But the failure analysis also showed the challenges of such a pressureless processing on great chip areas. Concerning manufacturing and cost issues the presented pressureless leadframe technology seems comparable to pressure assisted ones, as presented by Leicht et al. [43]. The time intensive processing of pressureless sintering pastes can be compensated by parallelized oven processes and the elimination of pre-drying process steps. Independent of the sintering technology, a change of the leadframe material for instance to Cu could reduce the material costs drastically and might make the leadframe technology even comprehensive to wire bonding.” [32] 3.3 Spot Welding for Terminals [44] “For the bonding of copper terminals onto DBCs a novel promising technology is resistance spot respectively resistance projection welding. In this work both processes were tested and projection welding was optimized for joining 0.5 mm thick copper terminals onto 0.63 mm Al 2 O 3 DBCs (0.3 mm Cu layers). Passive cycling tests were performed including welded and soldered AlN and Al 2 O 3 DBCs in order to compare the lifetime of both technologies. Single welding tests on 1 mm AlN DBCs with already soldered and wire-bonded semiconductors look promising. There was no influence on the functionality of the bare dies (Fig. 15). Though there are still some challenges in respect of the process control using projection welding. The structured top side of the DBC must be designed that way to position the blind electrode onto the same copper pad to ensure the welding current path. Considering process time optimization the blind electrode could also be used for welding a second terminal simultaneously. Also conceivable is it to use the terminal as blind electrode, in order to save the required contact area on the DBC. Moreover the projection pattern must consider that the current will flow directly to the blind electrode. Figure 15: Projection welded copper terminal on bare semiconductor module 121 Future Packaging Technologies in Power Electronic Modules Figure 16: Cross section of projection welded specimen [44] The investigations showed a good reproducibility without damaging 0.63 mm thick Al 2 O 3 ceramics of the DBCs. First attempts on AlN DBC substrates were also positive. Passive cycling tests showed no degradation after 300 cycles. This can be seen as an improvement compared to soldered terminals. Cross sections of the welded joints also showed no ageing effects. Further test and optimizations on AlN DBCs and on DBCs with thinner ceramics are ongoing. This bonding technology looks very promising.” [44] Conclusion and Outlook The focus of this publication is to give a short summary to soldering, sintering, and direct bonding technologies for power electronics. Furthermore, an insight in developing and researching topics was given such as ceramic embedding, selective sintering and die bonding of systems. However, there are already researched options to improve recent module fabrication by technologies such as a RC-snubber device on Si substrate in order to simplify the integration in any power module. The double sided sintering process with leadframes by a pressureless mounting step seems promising as well and a projection welding can easily be adapted to a production line process. Besides those recently investigated and already tested options of improving the power module performance and lifetime, there are still interesting outlooks for the future regarding research and application in packaging topics. One important topic, is a further development of ceramic embedding with a completely new, comparatively cheap and robust housing concept. Furthermore, FEM simulation becomes more and more integrated in the designing workflow of power electronics and power electronic components. Activities on topology optimization and automated computer-aided design steps will become more important. And to conclude the outlook on further interesting topics, environmental issues and lifetime question on corrosion and humidity in power components will be addressed more intensively. Literature [1] Amalu, E. H. & Ekere, N. N. & Bhatti, R. S., High Temperature Electronics: R&D Challenges and Trends in Materials, Packaging and Interconnection Technology. International Conference on Adaptive Science & Technology, 146-153, 2009 [2] Nielsen, R. Ø. & Török, L. & Munk-Nielsen, S. & Blaabjerg, F., Efficiency and cost comparison of Si IGBT and SiC JFET isolated DC/ DC. IECON, 695-699, 2013 122 Future Packaging Technologies in Power Electronic Modules [3] Su, M. & Chen, C. & Sharma, S. & Kikuchi, J. (2015). Performance and cost considerations for SiC-based HEV traction inverter systems. Workshop on Wide Bandgap Power Devices and Applications, 347-350, 2015 [4] Bayerer, R. (2018), Higher Junction Temperature in Power Modules - a demand from hybrid cars, a potential for the next step increase in power density for various Variable Speed Drives, PCIM, 2018 [5] K. S. Siow, “Are Sintered Silver Joints Ready for Use as Interconnect Material in Mircroelectronic Packaging? ”, Journal of Electronic Materials, Vol 43., No. 4, 2014 [6] M. W. Knörr, “Verbinden von Leistungshalbleiterbauelementen durch Sintern von nanoskaligen Silberpartikeln”, Doctoral Thesis, University of Erlangen, 2011 [7] N. Kim, R. Li, M. Machler, J. Burgers, S. Winkler, W. T. Ng, „Pressureless Silver Nanopowder Sintered Bonds for Liquid Cooled IGBT Power Modules“, CIPS - International Conference on Integrated Power Electronics Systems, 553-558, 2018 [8] G. Bai, “Low-Temperature Sintering of Nanoscale Silver Paste for Semiconductor Device Interconnection”, Doctoral Thesis, University of Blacksburg, Virginia, 2005 [9] C. Göbl und J. Faltenbacher, "Low temperature sinter technology die attachment for power electronic applications," 2010 6th International Conference on Integrated Power Electronics Systems, Nürnberg, 1-5, 2010 [10] J. Millán, "A review of WBG power semiconductor devices," CAS (International Semiconductor Conference), Sinaia, 57-66, 2012 [11] J. Charbonnier, V. Rat, H. d. Carvalho, D. Bergogne, J. Siegert and K. Pressel, "Towards Wafer Level 3D Power Integration," CIPS - 10th International Conference on Integrated Power Electronics Systems, Stuttgart, Germany, 1-5, 2018 [12] C. Oh, S. Nagao, K. Suganuma, “Diffusional hillock growth in Ag stress migration bonding for power device interconnections”, Proceedings - Electronic Components and Technology Conference, 1128-1133, 2015 [13] J. Hwang, J. Kim, W. Kwon, U. Kang, T. Cho and S. Kang, “Fine pitch chip interconnection technology for 3D integration,” Proceedings 60th Electronic Components and Technology Conference (ECTC), Las Vegas, NV, USA, 1399-1403, 2010 [14] H. A. Mustain, W. D. Brown, and S. S. Ang, “Transient Liquid Phase Die Attach for High-Temperature Silicon Carbide Power Devices,” IEEE Trans. Compon.Packag. Technol., vol. 33, no. 3, 563-570, 2010 [15] K. Suganuma, S. Sakamoto, N. Kagami, D. Wakuda, K.-S. Kim, M. Nogi, “Lowtemperature low-pressure die attach with hybrid silver particle paste”, Microelectron. Reliab., vol. 52, 375-380, 2012 [16] M. Kuramoto, T. Kunimune, S. Ogawa, M. Niwa, K.S. Kim, K. Suganuma, “Lowtemperature and pressureless Ag-Ag direct bonding for light emitting diode dieattachment”, IEEE Trans. Compon. Packag. Manuf. Technol., 548-552, 2012 123 Future Packaging Technologies in Power Electronic Modules [17] C. Oh, S. Nagao, T. Kunimune, K. Suganuma, “Pressureless wafer bonding by turning hillocks into abnormal grain growths in Ag films”, Appl. Phys. Lett. 104, 2014 [18] E. T. Ogawa, J. W. McPherson, J. A. Rosal, K. J. Dickerson, T. - C. Chiu, L. Y. Tsung, M. K. Jain, T. D. Bonifield, J. C. Ondrusek, and W. R. McKee, “Stress- Induced Voiding Under Vias Connected To Wide Cu Metal Leads”, 40th IEEE International Reliability Physics Symposium, Dallas, TX, USA, 312-321, 2002 [19] M. S. Jackson and Che-yu Li, “Stress Relaxation and Hillock Growth in Thin Films”, Acta Metall., vol. 30, 1993-2000, 1982 [20] B. Horváth, B. Illés, T. Shinohara, G. Harsányi, “Whisker growth on annealed and recrystallized tin platings”, Thin Solid Films, vol. 520, 5733-5740, 2012 [21] V. Thompson, C., Secondary Grain Growth in Thin Films of Semiconductors: Theoretical Aspects. Journal of Applied Physics, 58, 763-772, 1985 [22] H. L. Bach, T. M. Endres, D. Dirksen, S. Zischler, C. F. Bayer, A. Schletz, M. März, “Ceramic Embedding as Packaging Solution for Future Power Electronic Applications”, International Power Electronics Conference, 2410-2415, 2018 [23] R. Reiner, B. Weiss, D. Meder, P. Waltereit, T. Gerrer, R. Quay, C. Vockenberger, and O. Ambacher, “PCB-Embedding for GaN-on-Si Power Devices and ICs,” CIPS - 10th International Conference on Integrated Power Electronics Systems, 440-445, Stuttgart, Germany, 2018 [24] Schröder, Dierk. Leistungselektronische Bauelemente. Springer-Verlag, 2008 [25] Matlok, S.; Erlbacher, T.; Krach, F.; Eckardt, B., Switching SiC devices faster and more efficient using a DBC mounted terminal decoupling Si-RC element, Konferenzbeitrag European Conference on Silicon Carbide and Related Materials (ECSCRM), Halkidiki, 2016 [26] Banzhaf, Stefanie; Schwaiger, Stefan; Erlbacher, Tobias; Bauer, Anton J.; Frey, Lothar, Post-trench processing of silicon deep trench capacitors for power electronic applications; International Symposium on Power Semiconductor Devices and ICs (ISPSD), Prague, 2016 [27] F. Krach, N. Thielen, T. Heckel, A.J. Bauer, T. Erlbacher, L. Frey, Silicon integrated RC snubbers for applications up to 900 V with reduced mechanical stress and high manufacturability, Konferenzbeitrag, Device Research Conference (DRC) , Newark/ DE, 2016 [28] F. Krach, T. Heckel, L. Frey, A. J. Bauer, T. Erlbacher, M. März, Innovative Monolithic RC-Snubber for Fast Switching Power Modules, Proceedings of 9th International Conference on Integrated Power Electronics Systems (CIPS), 2016 [29] Gruenler S., Rattmann G., Erlbacher T., Bauer A.J., Frey L., Fabrication and Characterization of High-Voltage Monolithic 3D Capacitors based on Through- Silicon-Via Technology, IEEE Internat. Interconnect Technology and Advanced Metallization Conference, 2015 [30] Erlbacher T., Schwarzmann H., Krach F., Bauer A.J., Berberich S.E., Kasko I, Frey L., „Reliability of monolithic RC-snubbers in MOS-based power modules“, Proceedings of the 5 th Electronics System-Integration Technology Conference, Helsinki, Finland, 2014 124 Future Packaging Technologies in Power Electronic Modules [31] Kraft, S.; Tinter, R.; Schletz, A. Thermal Resistance Simulation and Measurement of a Double Sided Cooled Power Module. In Proceedings of the 8th European Advanced Technology Workshop on Micropackaging and Thermal management, IMAPS, La Rochelle, 2013 [32] Waltrich, U., Bayer, C.F., Zötl, S., Zischler, S., Tokarski, A., Schletz, A., März, M., High Reliable Power Modules by Pressureless Sintering, CIPS - 10th International Conference on Integrated Power Electronics Systems, Stuttgart, 2018 [33] Knörr M., Schletz A., Oertel S., Jank M., Semiconductor Joining through Sintering of Ag-Nanoparticles: Analysis of Suitability of Different Powders Using DSC and TGA Measurements. Proceedings of The World Congress on Particle Technology (WCPT6), Nuremberg, 2010 [34] Kraft S., Zischler S., Schletz A., Properties of a novel silver sintering die attach material for high temperature high lifetime applications. SENSOR, Nuremberg, 2013 [35] Kraft, S.; Schletz, A.; Maerz, M.: Reliability of silver sintering on DBC and DBA substrates for power electronic applications. International Conference on Integrated Power Electronics Systems, Nuremberg, VDE-Verlag, ETG- Fachbericht 133, 439-444, 2012 [36] B. Grummel, R. McClure, Lei Zhou, A. P. Gordon, L. Chow and Z. John Shen, "Design consideration of high temperature SiC power modules," 2008 34th Annual Conference of IEEE Industrial Electronics, Orlando, FL, 2861-2866, 2008 [37] U. Scheuermann, Reliability of Planar Skin Interconnect Technology, CIPS, Nuremberg, 2012 [38] BECKEDAHL, Peter; SPANG, Matthias; TAMM, Oliver. Breakthrough into the third dimension - Sintered multi layer flex for ultra low inductance power modules. In: Integrated Power Systems (CIPS), 8th International Conference on. VDE, 1-5, 2014 [39] Zhong, Yulin, et al. "Design & Analysis of a novel IGBT package with doublesided cooling." Transportation Electrification Asia-Pacific (ITEC Asia-Pacific), IEEE Conference and Expo. IEEE, 2014 [40] Yoon, Sang Won, Koji Shiozaki, and Takehiro Kato. "Double-sided nickel-tin transient liquid phase bonding for double-sided cooling." Applied Power Electronics Conference and Exposition (APEC), Twenty-Ninth Annual IEEE. IEEE, 2014 [41] Yin, Shan, King Jet Tseng, and Jiyun Zhao. "Thermal-mechanical design of sandwich SiC power module with micro-channel cooling." Power Electronics and Drive Systems (PEDS), IEEE 10th International Conference on. IEEE, 2013 [42] Marcinkowski, Jacek. "Dual-sided cooling of power semiconductor modules." PCIM Europe; International Exhibition and Conference for Power Electronics, Intelligent Motion, Renewable Energy and Energy Management; Proceedings of. VDE, 2014 125 Future Packaging Technologies in Power Electronic Modules [43] M. Leicht, et al.: “HEV power modules robust against cyclic thermo-mechanical stress,” in International Conference on Integrated Power Electronic Systems (CIPS), Nuremberg, 2016 [44] Waltrich U., Yu Z., Schletz A., Reichenberger M.: Bonding Copper Terminals onto DBC Substrates of Power Modules by Resistance Projection Welding; IEEE EPDC, Nuremberg, 2015 [45] Fabian, Dresel, Sebastian, Letz, Sigrid, Zischler, Andreas, Schletz, Michael, Novak, (Continental, Division Powertrain), “Selective silver sintering of semiconductor dies on PCB”, PCIM, 2018 [46] J. Müller, M. Novak, F. Dresel, C. F. Bayer, A. Schletz, M. März, Selective Silver Sintering on Organic-Based Circuit Boards, PCIM 2019 126 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle Guoqiang Ao, Lilly Han Abstract With the increasingly stringent on fuel economy and emission control for automotive industry ， the mild hybrid electric system (MHEV), equipped with 48V lithium-ion batteries, providing improved fuel economy at comparatively low cost, are expected to spread rapidly worldwide, especially in Europe and China. However, due to the low voltage of the battery system, it will lead to a large current to satisfy the power demand which challenges the Li-ion battery. Lithium Titanate batteries (LTO) with the characters of high C-rate in charge and discharge, long life, good performance in low temperature, could be the best solution for 48V MHEV. In this paper, an energy management algorithm with Supercharge algorithm is proposed and applied on an SUV. The analysis results show that the mild vehicle with LTO battery can realize extra fuel saving compared to other type of battery chemistry system based on WLTC because of low internal resistance, long cycle life and reliable high C-rate. 1 Introduction More and more countries released their target or legislation to cut down the emission of CO₂. For example, China government will limit the fuel consumption of passage car within 5L/ 100km in 2020. 48V Mild Hybrid Electric Vehicle [1] provide a promising solution to save both CO₂ emission and cost. The 48V battery system play a very important role on optimizing the engine operation point and absorb the recuperated energy which would be waste for the conventional vehicle before. Azureve has involved in many 48V research projects with different MHEV topologies (P0, P2, P3, P4) and gathered much experience in application. We are specialized in deploying Lithium Titanate chemistry (LTO) which has the characters of high C-rate in charge and discharge, long life and good performance in low temperature, thus believe it is the best battery solution for 48V MHEV. The purpose of this paper is to investigate how the MHEV can save the fuel consumption base on WLTC comparing to different chemistry battery based on P2 architectures. The paper is organized as followings. First, a complete HEV system model is presented in order to simulate the fuel consumption including battery models, motor model as well as the power flow model. All the models are validated in laboratory. Additional, the temperature behaviors of the different chemistry of battery are considered. Secondly, a control strategy with a Supercharge algorithm is proposed to minimized the fuel consumption by maximize the usage of the 48V battery system. Finally, 127 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle a conclusion is made that the LTO chemistry battery have more advantages on fuel economy and performance because of the character of high rate of charge and discharge, long life and good performance in low temperature. 2 MHEV System 2.1 MHEV architecture The proposed MHEV includes an Internal Combustion Engine (ICE), a 48V battery system as well as an Electrical Motor (EM). A P2 architecture is selected for the investigations in this paper, as proposed in and presented in Fig.1. The electrical motor is placed between ICE and transmission, which allows a wide range of operating with two clutches controlled separately. The motor can be operated as generator or as starter with C1 connect the ICE while C2 disconnect with transmission. On the other hand, the EM can be used for recuperation and pure electric driving with closed C2 and opened C1. In order to investigate the fuel economy under given cycle (WLTC),A driver model with PI controller is developed to drive the vehicle following the driving cycles. Fig. 1 is the proposed MHEV architecture. Fig. 1: MHEV Powertrain Architecture 2.2 Vehicle model The vehicle model is focus on the power flowing occurring on the power net during drive the vehicle. A velocity profile representing a driving cycle is given as input to the vehicle model. This allows the hybrid control unit (HCU) calculation the power need and how to allocate the power command between ICE and battery to make the whole system working efficiently. The parameters such as weight, drag coefficient and rolling resistance used for energy and power loss calculation are considered to build a unified vehicles system. Thus different vehicle types such as small cars or SUVs can be used for simulation. This paper considers the drive cycle of Worldwide harmonized MCU: Motor Control Unit BMS: Battery Management System HCU: Hybrid Control Unit EMS: Engine Management System TCU: Transmission Control Unit EMS TCU Others Inverter BMS HCU MCU CAN 12V- Loade r 48V-Battery 12V- Battery FD 8AT- Transmission EM C1 C2 128 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle Light duty Test (WLTC) which more close to reality and make more sense for fuel consumptions. The vehicle model [2] is described as equation (1). (1) Where : is the total driving force requirement : is the rolling resistance : is the ascending resistance : is the aerodynamic resistance : is the acceleration resistance 2.3 Battery system The 48V Lithium -ion batteries system is modeled in detail with different cell parameters. The parameters such as energy density, power density, safety and calendar and cycle lifetime, internal resistor and max C rate in wide range of temperature. Usually the LTO battery provides higher C rate in both discharge and charge in wider range of temperature compared with other chemistry material battery. The detail batter model is introduced with internal resistance model [3]. Additional, the LTO battery and other chemistry batter model are validated in laboratory for the contrast of fuel saving contribution. The internal resistance model can be described in Fig. 2 and equation (2). The SOC is calculated as equation (3). (2) Where : open-circuit voltage of battery, : terminal voltage, : required power from electric motor and : current in the circuit. Assume the discharge and charge current is positive and negative value respectively. (3) Fig. 2: Internal resistor model of Battery 129 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle Where : time duration of the sample time : the max capacity for the battery packs ampere hour (Ah). Charge and discharge power is limited here in order to keep the bus voltage in the suitable range : calculated SOC of current sample time : SOC of previous sample time We validated the battery model through test the LTO cell by at different C rate charge and discharge current under different temperature from (-30 ℃ ~60 ℃ ) 。 Fig.3 is 200A and 300A for discharge and and Fig. is 200A and 300A charge data respectively. The proposed battery model is validated well. Fig. 3: Battery model validation for discharge 130 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle Fig. 4: Battery model validation for charge 2.4 Control algorithm The HCU interpreter the driver intention and calculate the required power. And then check the state of all components including ICE, motor, batter state etc to calculate the available power and then split the required power between motor and ICE[4][5][6][7]. The Fig. 5 is shown the power flow of the MHEV. Fig. 5: Power flow of MHEV Where: : Fuel injected into engine cylinder at this sample time : Power required from ICE which depend on come from the fuel combustion of engine at this sample time ： Power from motor at this sample time - : power required by vehicle which interpreted by HCU at this sample time Vehicle dynamic ( ) ( ( ), ( )) req veh P t f v t h t  Engine ( ) ( ( )) ICE ICE fuel t f P t  Motor ( ) ( ( ), ( )) mo mo req P t f soc t P t  Battery max ( ( )) ( ) 1 batt batt f P t SOC t C   ( ) f u e l t ( ) v t ( ), ( ), ( ), ( ) C O t H C t N O x t P M t ( ) IC E P t ( ) req P t ( ) m o P t ( ) batt P t 131 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle : Power from battery pack at this sample time : emissions from ICE Based on the required power, battery state and vehicle speed. The vehicle runs in four modes roughly. These are pure EV, Supercharge, boost and regeneration braking as show in Fig. 6. In pure EV mode, the required power is low and the ICE will shut down to save fuel and emission. The motor drive the vehicle alone power by battery only. In Supercharge mode , the required power is in medium level. The algorithm will allocate the required power between ICE and motor to improve the operation efficiency of the whole powertrain system. The LTO battery will support the motor with large charge or discharge current to make the ICE working in ‘sweet pot’ which will improve fuel economy greatly. In boost mode, the required power is high, the LTO battery will support motor to assist the ICE to work in optimal region with high discharge power. In regeneration mode, the motor work as generator to recover kinetic energy and charge the LTO battery with high charge power to maximize the recuperation which otherwise would be wasted as heat in the break drums during deceleration. In summary, With high charge and discharge power capability, The LTO 48V battery in this hybrid topology will contribute extra 3% fuel saving and less emission for the WLTC drive cycle. Fig. 6: The Operation Mode for MHEV 2.5 The electric motor and ICE model The model of the Internal Combustion Engine (ICE) and electric motor model are relatively simple. The ICE and motor are modeled with and efficiency map of torque Vehicle Torque Demand Engine Start Speed Engine Speed Max powertrain torque Max engine torque Min engine torque 发动机 t 停机区 Boost Supercharge Pure EV 132 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle and speed. A typical mid-size ICE for mid-size car is used[7]. As only regarding fuel consumption, a motor with 25kW peak power is applied in this system. Fig. 7 is the ICE fuel consumption map with the unit of g/ kWh. Fig. 8 is the motor efficiency map. Fig. 7: The fuel consumption map for ICE Fig. 8: The efficiency map of motor 3 Simulation results 3.1 Specification of MHEV powertrain system We co-worked with a Chinese OEM to design a MHEV prototype based on a SUV to calculate the fuel economy. The basic components’ specifications are listed in Table 1. 133 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle The power split under driver cycle of WLTC are conducted. Different type of chemistry of battery are applied to achieve best performance on fuel economy. Table 1: Specification of the target vehicle Item Specification Curb weight (kg) 1835+80 Engine Displacement n(L) 2.0 High of Gravity Center(m) 0.8 Frontal Area (m 2 ) 2.814 Air Resistance Coefficient 0.4079 Rolling Resistance Coefficient 0.0089 8 Automatic Transmission Ratio [5 3.2 2.143 1.72 1.314 1 0.822 0.64] Final Drive Ratio 3.636 Motor Power (kW) 25 Kw with coolant system Battery 20 Ah LTO from Azureve with coolant system 3.2 Torque split between ICE and motor As we know, WLTP is a global harmonized standard for determining the levels of pollutants and CO2 emissions, fuel or energy consumption, and electric range from light-duty vehicles. The WLTC driving cycle, as shown in Fig.9, for a class 3 vehicle is divided in four parts for low, medium, high, and extra-high speed. Based on the WLTC, the proposed MHEV system is simulated and analyzed. Fig. 9: WLTC cycle diagram Low Medium High Extra High Time (s) Speed (km/ h) 134 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle The simulation results show that there is extra 3% fuel saving with the help of LTO high charge and discharge power. In the Low and Medium zone, the Supercharge algorithm raised the ICE load to increase ICE efficiency and charge the battery with high current. So the SOC goes to high. The stored energy will boost the vehicle in high and extra high speed zone which lower the fuel consumption and emissions. So the SOC of battery goes down. The power and torque coordination between motor and ICE are plot out in Fig. 10. The battery voltage and current are shown in Fig.11. Fig. 10: The torque split between motor and engine for the WLTC cycle 135 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle Fig. 11: LTO battery voltage and current for WLTC cycle 3.3 ICE operation points of MHEV The low load operation points are one of the most significant reasons which make the conventional vehicle not environment friendly. But for the MHEV powertrain system, with the help of motor and battery, all those of low load operation points are raised up or canceled which increase the efficiency of the whole system. As shown in Fig. 12. Fig. 12. ICE operation points during WLTC 136 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle 3.4 Accumulated Ah and energy during WLTC Because of the EV, Supercharge, boost and regeneration mode, the battery charges and discharges aggressively. The accumulated Ah and energy is shown on Fig. 13. For the charging Ah, it reaches 33.6 Ah, which means 1.68 cycle times are used and 1.57 kWh energy go into the battery after WLTC. For the discharge Ah, the accumulated discharge Ah is 34.3 with the discharge energy of 1.56 kWh go out of battery. The high throughput of the MHEV system challenges the life cycle of battery. LTO is known as long lift cycle which make it more suitable for the MHEV system. Fig.13: Accumulated charge and discharge Ah and energy 3.5 Regenerative brake Regenerative braking is one of the most important features for the MHEV. It recaptures the mechanical energy by charge the battery which would be wasted as heat for the conventional vehicle. Recapturing braking energy as much as possible is very important. But it is mainly limited by the battery charging capability. LTO have stronger charge capability comparing with other chemistry battery. Fig.14 shows the charge capability comparison between 10 Ah LTO and other chemistry 10 Ah cell for 10 seconds pulse at -30 ℃ ， -20 ℃ ， -10 ℃ and 25 ℃ and different SOC level . 137 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle Fig.14: charge capability between LTO and other cell Fig.15 demonstrates the current charge to the battery during regeneration braking for the end of WLTC cycle. With higher charge current, we get more free energy which benefit the fuel economy. That’s why LTO battery is more suitable for MHEV system. Fig.15: Accumulated charge and discharge Ah and energy 138 The Application of 48V LTO Battery in Mild Hybrid Electric Vehicle 4 Conclusions This paper investigates the fuel saving potential for different chemistry type of battery system. A complete system model is presented to simulate the fuel consumption of MHEVs. The results could serve as support for design and dimensioning of similar recuperation systems. Finally, an extensive comparison of both type of battery is made in fuel saving and life cycle. The LTO battery achieves promising results particularly in fuel savings. At the same time, under real road, rural or city driving cycles, the powertrain system is apt to use more electric power for driving to optimize both the fuel saving and drivability. Reference [1] Lee, S., Cherry, J., Safoutin, M., Neam, A. et al., “Modeling and Controls Development of 48 V Mild Hybrid Electric Vehicles,” SAE Technical Paper 2018-01- 0413, 2018, doi: 10.4271/ 2018-01-0413. [2] Newman, K., Kargul, J., and Barba, D., “Benchmarking and Modeling of a Conventional Mid-Size Car Using ALPHA,” SAE Technical Paper 2015-01-1140, 2015, doi: 10.4271/ 2015- 01-1140. [3] Lee, S., Lee, B., McDonald, J., and Nam, E., “Modeling and validation of Li-ion automotive battery packs,” SAE Technical Paper 2013-01-1539, 2013, doi: 10.4271/ 2013-01- 1539. [4] Johnson, V., Wipke, K., and Rausen, D., “HEV Control Strategy for Real-Time Optimization of Fuel Economy and Emissions,” SAE Technical Paper 2000-01- 1543, 2000, doi: 10.4271/ 2000-01-1543. [5] Kum, D., Peng, H., and Bucknor, N. K., “Supervisory Control of Parallel Hybrid Electric Vehicles for Fuel and Emission Reduction,” ASME Journal of Dynamic Systems, Measurement and Control, Apr 2010 (DS-09-1340). [6] Guoqiang Ao, Hu Zhong, Lin Yang, Jiaxi Qiang, Bin Zhuo. Fuzzy Logic Based Control for ISG Hybrid Electric Vehicle. ISDA, pp. 274-279, Sixth International Conference on Intelligent Systems Design and Applications (ISDA'06), 2006. [7] Lin,C.C., Peng, H., J.W, Kang.J. M. Power Management Strategy for a Parallel Hybrid Electric Truck. IEEE Tran. Control systems technology 11, 6, 839−849. 139 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles Ayman Ayad, Martin Brüll, Andreas Greif, Sebastian Rogge, Matthias Töns Abstract Lifetime is one of the major quality measures in the automotive industry. For optimizing the costs, nearly all the vehicle components are designed to survive a certain time and usage. Over the last decades, this has been optimized perfectly. In the transition to the electric vehicles, many parameters do change now. In this paper we discuss not only the lifetime of the new power electronics, but we also review the vehicle requirements, since both, driving and charging do affect the component’s lifetime - but quite differently. 1 Introduction Reliability is key and thus lifetime requirements are an essential part of the automotive development process. Often, additional lifetime can be associated with additional costs. Just in rare cases, a component exchange can be accepted but requires then a maintenance strategy. For electronics and for power electronic components, this seems not to be acceptable. An electric vehicle (EV) does not only introduce new components but also new use cases, so the topic lifetime must be reevaluated. Power electronic converters in electric vehicles, such as inverters, DC/ DC converters and on-board chargers, have in principle two different kinds of active elements: On the one hand the high voltage path comprising of semiconductors, diodes and passive elements such as capacitors, coils and the contacting layers, being stressed by high currents and thus must overcome the related thermal stress. On the other hand, the low voltage control, e.g. the microcontroller and the other parts on the printed circuit board (PCB) have also to meet the lifetime expectations. Especially the high voltage components depend strongly on the vehicle usage profile which we must consider. In this paper the vehicle usage profiles are discussed. As will be seen, the classical 8 000 h active hours of a vehicle with internal combustion engine (ICE) cannot be simply transposed to an electric vehicle. For instance, the behavior in idle mode is different and the cranking vanished. In addition, during charging, the vehicle is active for many additional hours. In chapter 2 we are going to discuss the vehicle’s lifetime requirements to derive a reasonable specification of the new components. In chapter 3 we dive into the lifetime modelling of a power electronic component and its single elements. Finally, in chapter 4 we will discuss the results in comparison to car manufacturer’s requirements. 140 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles 2 Vehicle lifetime requirements 2.1 Former requirements for ICE vehicles Each car manufacturer has its own set of lifetime requirements, which are specified as several thousand working hours and some hundred thousand kilometers. For passenger cars they vary as shown in Table 1. As a conservative set, we can assume here 8 000 h and 300 000 km. This corresponds to an average vehicle velocity of 37.5 km/ h. Within these profiles, it is accepted to replace tires and brakes, but not belts and chains or even sparks. Moreover, no electronic control unit (ECU) or sensor should be replaced in that lifetime. Table 1: Traction requirements from car manufacturers (bold: worst case) Range [km] Driving time [h] Average speed [km/ h] Lifetime [y] Yearly Mileage [km] 300 000 8 000 37.5 15 20 000 240 000 7 000 34.3 15 16 000 ... … … … … The total number of mileage and working hours is not sufficient to specify the component lifetime consumption. Important is the load collective, i.e. under which condition does the vehicle reach the total working hours and range. Table 2 shows a load collective example of different driving cycles. This is typically a collection of “standard” mission profiles like the WLTC or US cycles as well as OEM specific e.g. measured mission profiles. In addition to meet real vehicle life, hilly driving cycles and acceleration runs are included. Table 2: Typical load collective requirement from car manufactures Mission Profile Duration [s] Average speed [km/ h] Number of repetitions Overall distribution [%] WLTC 1800 46.5 2020 15 Artemis Urban 993 17.65 1815 8 US Highway 775 76.51 200 0.7 OEM city 1 1740 61 419 4 OEM hilly road 1236 36 872 12 OEM cycle… … … … … Acceleration (0...80km/ h) 48 38.49 2000 0.4 Acceleration (80...120km/ h) 27 99.07 1000 0.12 Total 100% For the component life time the distribution of these profiles is very important, i.e. the distribution of city, urban, highway driving and special cases like slopes (e.g. hill-hold) and acceleration driving. 141 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles When an ICE vehicle is used, and the car reaches a red traffic light, the engine runs in idle mode, means being active without need. The modern solution of start-stop systems tries to overcome this, mainly driven by emission targets. But unfortunately, this increased the number of cranking events dramatically, having its issues with the component reliability. So, components like the two-mass flywheel, belts and other mechanical parts influenced by the cranking where challenged by the field test of start-stop. Furthermore, modern features, like keyless entry and a monitoring system of the 12 V battery need to be active all the time when parked, which is by far not covered by 8 000 h of driving. 2.2 Lifetime requirements of an electric vehicle The first approach for the lifetime requirements of an EV should be the same as the ICE vehicle as the EV should be able to substitute the latter. The first guess is then to neglect the red traffic light idle mode, because the e-machine does not rotate. However, for power electronics, this could be an active mode, e.g. in case of the hill-hold function. And even if not, the microcontroller and low-voltage components will be still awake. On top the vehicle is active while being charged. But what does that mean? A very misleading approach would be a gross charging time estimation like: It is charging on every whole night. Assuming here about 11h per night, this would lead to additional 40 000 h in 10 years which is not realistic. A more precise approach would be as follows: Charging with 3.7kW (16A @ 230V) corresponds to a charging speed of 20 km/ h when assuming a driving consumption of ~180Wh/ km. Based on that conversion factor, one could easily derive charging times for typical charging profiles. Table 3 gives a typical European mixed charging profile resulting in 9 313 h for 300 000 km (in [1], even less charging time was calculated), Table 4 gives a 100% Japanese home charger profile resulting in 8 400 h. Means, another 8 000 h to 10 000 h charging time need to be assumed. Table 3: Charging profile 1 (e.g. European Mix Charger) with a total charging time of 9 313 h Charging Power [kW] Occurrence [%] Charging speed [km/ h] Charging time [h] Home 3.7 50 20 7 500 Work 11 25 60 1 250 Public 22 20 120 500 DC 50 5 240 62.5 Table 4: Charging profile 2 (e.g. Japanese Home Charger) with a total charging time of 8 400 h Charging Power [kW] Occurrence [%] Charging speed [km/ h] Charging time [h] Anywhere 6.6 100 35.7 8 400 In case of an architecture with an on-board charger (OBC), either the drivetrain or the OBC needs to be active. In case of a drivetrain with an integrated charging [2] [1], this gives up to 18 000 h maximum for both, driving and charging. But this does not mean 142 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles to double the lifetime requirements of all components, as we will see in the following chapter, especially section 3.2. 3 Lifetime calculation of electronics and power electronics With increasing car manufacturer demand for reliability data in an early design phase, it is necessary to analytically determine the applicable parts for lifetime analysis. This is because the Design Validation (DV) and Product Validation (PV) results or field data are often not yet available at this point of time. The following sections in this chapter show exemplarily how this is done for electronics and power electronic components. Basically, a component failure rate can be described via the Weibull failure rate 𝜆 𝑡 (1) With 𝜂 0, ∞ 𝛾 ∞ and 𝑡 max 0; 𝛾 describing a bathtub curve in three phases: The early phase I with 𝛽 1 describes early failures and show a decreasing failure rate over time , the phase II with 𝛽 1 show a constant failure rate over time, mainly caused by random failures. The phase III with 𝛽 1 show a rising failure rate [3] (see Fig. 1). Figure 1 Weibull bathtub curve In the following we are going to concentrate on the analysis of the phase III of the Weibull failure rate, since it is important to predict the end of life (EOL) failure rate. 3.1 Static qualification of Tier2 supplier parts Looking on the physical effects, there is a huge variety of potential aging effects, different for each part and its environment. Starting with pure temperature dependencies following the law of Servante Arrhenius (1859-1927) to derive a possible failure time to more complex scenarios, wherein also voltage and current plays a role. A good summary of such effects can be found in [4]. A more practical approach for the industrial usage is described in the IEC 61709: 2017 [5], providing a framework to relate given reference conditions with required operation conditions: Basically, the component failure rate 𝜆 can be described as 𝜆 𝜆 𝜋 𝜋 𝜋 𝜋 𝜋 𝜋 (2) 143 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles where 𝜆 is the failure rate under reference conditions, 𝜋 , , is the voltage, current and temperature dependence factor, 𝜋 is the environmental application factor, 𝜋 the switching rate dependence factor and 𝜋 the electrical stress dependence factor [5]. As already stated, failure rates can only be considered constant within phase II of the bathtub curve (see Fig. 1), within the so-called useful lifetime. Since the results of safety analyzes (e.g. Failure Mode, Effects and Diagnostic Analysis: FMEDA and Quantitative Fault Tree Analysis: Quant. FTA) are based on constant failure rates, it must therefore be ensured that the parts used remain within this useful life over the entire predicted operation time. Therefore, the analysis based on wear times makes sense. This failure rate 𝜆 can simply be translated into a failure time giving us following fundamental equation to analyze if the used electrical components stay within their useful lifetime: 𝑡 𝑡 𝛱 (3) It relates operational conditions 𝑡 with the necessary test time 𝑡 using one or more stress factors 𝛱 . The computed test time 𝑡 will then be compared to qualification stress test-time 𝑡 from [6]. If 𝑡 𝑡 holds true this would be a PASS otherwise a FAIL, for the electrical component under investigation. As an example, the temperature dependent stress factor 𝜋 is given via the Arrhenius equation: 𝜋 𝑒 (4) with the activation energy 𝐸 , the Boltzmann constant 𝑘 , the operation temperature 𝑇 and the reference temperature 𝑇 in case, only one activation energy needs to be regarded. The voltage dependent stress factor 𝜋 via following empirical model: 𝜋 𝑒 (5) with the operating voltage 𝑈 , the reference voltage 𝑈 and the rated voltage 𝑈 and two empirical constants 𝐶 and 𝐶 . The activation energy 𝐸 can be interpreted as the minimal energy for the occurrence of a failure. It is different per component and failure mechanism. The same applies to the constants 𝐶 and 𝐶 . Depending on the type of component, different sets of the described stress factors need to be regarded. Looking to the elements used in our power electronics, these parts will be grouped into appropriate categories, in this way not every single component has to be analyzed. From the bill of materials (BOM) two main groups could be figured out: standard and special components. The standard components are further divided into the part categories shown in Table 5. The table further states the stress factors to be regarded following [5]: 144 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles Table 5: Relevant PI Factors of supplier parts [5] Part Categories U I T Capacitors (C) yes yes Resistors (R) - yes Inductors (L) - yes Integ. Semi-Cond. (IC) yes yes Discrete Semi-Cond. (T) yes yes Diodes (D) - yes Optocoupler (OC) - yes As it can be seen from Table 5, for our application, the only relevant stress factors are 𝜋 for temperature and 𝜋 for voltage. The parametrization for the given part classes are listed in the standard. Special components, such as µC, ASICs, Digital Isolators and DC Link Capacitors should be investigated in conjunction with their respective Tier2 suppliers to verify if the IEC61709 parameters can be applied. Often, they have specialized failure rate prediction models based on their field experience. As an example, we give here the calculation for a Multi-Layer Ceramic Capacitor (MLCC): Let’s assume a standard MLCC with the following technical parameters: Dielectric: X7R, Capacity: 100nF, Tolerance: 10%, Maximum Rated Voltage: 50V and Package: 0603. According to Table 5, the stress factors 𝜋 for voltage and 𝜋 for temperature can be used. Referring to Tables 37 and 39 in [5], we get 𝐸 0.35 𝑒𝑉 for equation 4 and 0.5, 𝐶 1 and 𝐶 4 for equation 5. Following Table 2 of AEC-Q200 [6], the test is done at 𝑇 125°𝐶 and lasts 𝑡 = 1 000 h where the capacitor is additionally stressed with full rated voltage of 50V. In this example, we assume a mean operation temperature 𝑇 50°𝐶, the capacitor is used only with 50% of its rated voltage (25V) and an operation time of 𝑡 53 000 ℎ. With all that information we can calculate the necessary test-time 𝑡 to project the qualification test parameters to the operational conditions. 𝑡 𝑡 𝜋 𝜋 53 000ℎ 0,094 0,135 672ℎ (6) In this example, the selected MLCC is qualified for the assumed mission profile, because its qualification time 𝑡 = 1 000 h at the Tier2 supplier was longer than the required test time of 672 h. During the vehicle’s lifetime the mission profiles are not as simple as given in our example above. To meet real use cases, the whole mission profiles from chapter 2 must be transposed into temperature profiles and regarded accordingly. This weighting method will be described in subsection 3.2 for 𝜋 of an IGBT. 145 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles 3.2 Dynamic qualification via system simulation In the following we will discuss the lifetime analysis of the entire system by means of simulation. The knowledge of the individual components described above, as well as the expected mission profiles of the system, play a role here. Figure 2: Cross-section of a power module with Si IGBT semiconductor switches and top side connection Power electronic semiconductors (mainly IGBT, diode, or MOSFET) are considered as the most critical components in EVs. After a certain operation time, aging occurs which affects their reliability. For a precise prediction and potential defect avoidance, lifetime prediction under realistic operation conditions is crucial. Figure 3: Exemplary mission profile applied (top) leads to the temperature stress for IGBT (middle) and Diode (below) Power modules for automotive applications are commonly directly liquid cooled with a very straight thermal coupling between power semiconductor and coolant (see Fig. 2). Since this results in less thermal capacity, their thermal time constants are relatively low (typically below 1s). While driving a typical dynamic mission profile with many temperature swings on the semiconductors and the connections nearby are seen. This stresses the mechanical connections and results in aging of the components (see Fig. 3). 146 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles In contrast to the driving profile effect shown in Fig. 3, while charging the power semiconductors see a constant load and thus stay on a constant temperature. So, the aging of their interconnections is very small in this use case. During working life, the question arises of how much influence a particular mission has on the lifespan of the component or system. Here we follow the laws of Palmgren- Miner, a simple cumulative damage model based on a linear damage hypothesis. Considering 𝑘 different stress levels 𝑆 , then one obtains after 𝑛 cycles at this stress level a partial damage 𝑊 𝑛 ∗ 𝑆 , which one can set in relation to the failure 𝑊 , 𝑁 ∗ 𝑆 after 𝑁 cycles at this stress level. If you sum up the partial damage of all stress levels, you get: ∑ , ∑ 𝐶 (7) 𝐶 is then the proportion of consumed lifetime, which reaches 0 when new and 1 at the end of life [7]. Figure 4: Lifetime analysis algorithm for power electronics in EVs To simulate the lifetime of a power electronic device, certain steps have to be followed. The main steps are summarized in Fig. 4. It starts with the driving cycle, then followed by the electrical machine model that based on the DC-link voltage computes the required rms current, modulation factor, and power factor. These values are then input for the inverter model to calculate the losses of each switch and the total losses based on either datasheet information or experimental values. Using a thermal model (e.g. Foster model [8]) and considering the cooling conditions, the junction temperature of the switches is calculated. Fig. 4 further shows the step “Rainflow Algorithm”, which is skipped for now and will be introduced in section 3.4. The resulting temperature profile of the IGBT is shown in Fig. 5. 147 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles Figure 5: Temperature profile of an IGBT of traction inverter while driving (left) and charging (right) The left temperature profiles in Fig.5 is based on the driving profile shown in Table 2 used to derive a weighted 𝜋 for driving (see Table 6). Based on this, a required test time for driving can be derived as: 𝑡 , 𝑡 𝜋 8 000ℎ 0,0355 284ℎ (8) The right temperature profile in Fig. 5 is based on the charging profile in Table 3 and used to derive a weighted 𝜋 for charging (see Table 7). Based on this, a required test time for charging is: 𝑡 , 𝑡 𝜋 10 000ℎ 0,0021 21ℎ (9) Which is just 7.4% of the driving aging effect and just 2.1% of the component test time of 1000 h. Table 6: Temperature dependent stress factor calculation for driving Weight IGBT Junction Temperature [°C] weighted 𝜋 3.33% -25 6.70085E-09 1.67% 0 6.70302E-0.8 6.67% +25 3.24559E-06 21.25% +50 8.51418E-05 25.83% +75 0.000629358 21.25% +100 0.002471535 10.83% +125 0.004943162 4.58% +150 0.006980948 4.58% +175 0.02037037 Total weighted 𝝅 𝑻 driving 0,0355 Table 7: Temperature dependent stress factor calculation for charging profile 1 (see Table 2) Weight IGBT Junction Temperature [°C] weighted 𝜋 1% +35 1.47326E-06 79% +68 0.001490509 14% +72 0.00034808 6% +80 0.000254241 Total weighted 𝝅 𝑻 charging 0,0021 148 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles These results showed that in case of charging via the traction inverter [2] the effect on the lifetime is quite minor. The first reason is that the charging power is considered as partial load operation for the inverter. Second, the power switches are loaded with a constant power, thus their interconnection aging is uncritical. 3.3 Integration, assembly and connection technology The power electronic modules are manufactured in a multilayer structure (e.g. see Fig. 2). Each material undergoes an expansion depending on the temperature. The coefficients of thermal expansion (CTE) and thus the length expansion of different materials is different, therefore naturally arise at material joints mechanical stresses. These loads are an important cause of component failures and so their lifetimes. For example, the solder and the boundary between Al bonding and Si IGBT chips is very critical. Simplified, such loads can be modeled as a sum of periodic loads, for example sinusoidal oscillations. Small (elastic) stress amplitudes are not critical, but higher amplitudes lead to plastic material deformations. The transition is fluent, even a high number of load cycles with a small amplitude can lead to a long-term failure. These effects affect the time stability, which can be statistically described by a Wöhler curve. In the low-cycle fatigue range (LCF), typically in the range of 10 to10 load cycles, the Wöhler curve can be described by the Coffin-Manson relationship [9]: 𝑁 𝐴 ∗ ∆𝜀 (10) With the failure time 𝑁 (fatigue life), the fatigue coefficient ∆𝜀 (strain amplitude, corresponds to an infinitesimal deformation per cycle), the Coffin-Manson coefficient 𝐴 (a material constant to be determined experimentally) and the material-dependent damage coefficient 𝐵 (to be determined experimentally). Performing lifetime analysis via field tests requires a lot of time. Therefore, one uses a method of standardized temperature cycling tests to shorten the test time. The Coffin & Manson parameters can be determined experimentally by comparing the lifetimes from temperature cycling tests with different test conditions. For chip-based assembly and connection technology, active load change tests (power cycling) are carried out under different test conditions (temperature lift / medium temperature) until failure. Fig. 6 shows an example of the results of two series of measurements with a temperature shock of 100K. The blue dots on the left show a sample of IGBTs soldered at the bottom side and bonded on top. The three straight lines show the mean distribution and the upper and lower confidence limits at 5% and 95%, respectively. The measurements showed a characteristic fatigue life 𝑁 , 94 013 cycles, which we define at 1% failure rate. The red dots on the right show a selection of IGBTs connected via silver sinter technology on both sides. This resulted in 𝑁 , 1 372 267 cycles, which is more than one order of magnitude above the other connection technology. 149 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles Figure 6: Exemplary comparison of soldered/ bonded vs. sintered technology The huge life difference can be explained as follows: Sold and Aluminum on the one hand and silver on the other hand have different CTEs and melting points. If the connection technology turns out to be the weak point, the semiconductor might be oversized in its lifetime budget. To harmonize the subcomponents more economically, therefore, a connection technology is needed that comes closer to the semiconductor’s lifetime. Based on these results one can now vary the technologies and/ or process parameters to achieve target lifetime values in between these results. One obtains a degree of development freedom for the optimization with respect to lifetime (e.g. needed for commercial vehicles) and performance of the same semiconductor area while driving higher ∆𝑇 in the application. In an electric vehicle, one currently assumes maximum coolant temperatures of approximately 65 °C - 70 °C. Due to the presence of the internal combustion engine in a hybrid vehicle, the coolant temperature may be 85 °C, depending on the packaging. The transition from the combustion mode to the EV mode at full performance can therefore have a critical influence on the service life and therefore on the choice of the connection technology used. A lifetime analysis of the component must therefore be carried out in the system context, as we will show in the next section. 3.4 System simulation of the assembly and connection technology Two mechanisms contribute to the inverter temperature: Passive and active temperature changes. Passive temperature changes caused by components, which are connected to the inverter coolant system, e.g. the air conditioner or a passenger compartment heater. These components change the inverter coolant temperature and therefore the IGBT device temperature without any activities of the inverter component itself. However, the passive temperature changes cause lifetime consumption, which will be typically considered as a certain margin in the component design following equation 4. The active temperature changes caused by the inverter (power module) itself through the usage of current. To consider only the effective temperature cycles, the Rain-Flow algorithm [10] is used. Out of this algorithm two important values are provided and used as input for the lifetime prediction in the system simulation: The mean junction temperature 𝑇 and the temperature range ∆𝑇 (see Fig. 4). The following examples 150 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles will show the stress difference between three cycles with a certain vehicle, coolant temperature and battery voltage. The first example is a “standard” urban mission profile (Fig. 7). The left plot shows the velocity profile and the right plot shows the inverter temperature distribution. The second example shows a vehicle full accelerator pedal profile (0…150 km/ h) (Fig. 8). The third example shows the result for the WLTC (Fig. 9). Figure 7: Urban mission profile and inverter temperature distribution Figure 8: Full accelerator pedal mission profile and inverter temperature distribution Figure 9: WLTC mission profile and inverter temperature distribution Although the full accelerator pedal mission profile is much shorter in time than the urban driving profile, it can be seen, that the inverter temperature stress, and therefore the lifetime consumption is much higher in the full accelerator pedal mission profile. The stress of the full accelerator pedal mission profile is higher in consideration of the mean junction temperature as well of the temperature range ∆𝑇. As described above these both parameters finally define the assembly and connection technology and therefore also the component price. In conclusion the selection and distribution of the mission profiles have a huge impact on connection technology and therefore the component costs. 151 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles 4 Discussion The automotive use cases show a huge variety, so a simple lifetime analysis like shown in section 3.1 could be sufficient but could fail also. For certain part categories (e.g. IC, FET), the AEC-Q defined test-times and temperatures sometimes do not cover the operational conditions. This must be handled twofold: Firstly, those Tier2 supplier tests just show a positive test and does not reveal the component’s potential further lifetime. Secondly in real life, not all 𝜋 factors appear on maximum at the same time. As rough indication for test parameters (e.g. for high temperature operational life test (HTOL)) when choosing electrical components, it can be mentioned that 1 000 h at 125 °C is often not sufficient to represent a vehicle lifetime, 1 000 h at 150 °C is borderline, a qualification at 2 000 h at 150 °C and 1 000 h at 175 °C would pass our requirements in most cases. More care needs to be taken for components with high self-heating caused by high power dissipation or a high thermal resistance of the package. A proper and reliable choice can only be taken based on a system simulation of the vehicle’s mission profiles, like shown in section 3.2. When focusing on straight thermal coupled components, such as IGBTs in a power module, the right connection technology needs to be selected properly. Thermal design considerations shall also encompass parts on the PCB, not only the power module, because part temperatures should be as low as possible to gain appropriate acceleration factors. The analysis of sections 3.3 and 3.4 revealed the need of system analysis to choose the right connection technology: In cases the car manufacturer does not expect many situations leading to high junction temperatures and temperature changes, the cheap and established bond technology could be the right choice. On the other hand, the high performant sinter technology gives much higher lifetime and opens the door for different design criteria, such as less thermal coupling or smaller chip sizes. Finally, we’ve seen that charging profiles do affect the lifetime much less than often expected in today’s specifications. Furthermore, since this mode gives a constant load and thus a thermal equilibrium, rather than the thermal cycling while driving, one charging hour counts much less than one driving hour in lifetime. Finally, in case of a drivetrain integrated solution, each AC charging event runs just in partial load of the inverter and thus is even uncritical. This is beneficial for drivetrain integrated charging solutions since this function costs just additional ~2% of the lifetime budget. When looking into the car manufacturer’s requirements of up to 40 000 h additional operating hours of electric vehicles, compared to classical ICE vehicles, we have several potential interpretations: Some may not have much experience with EVs and thus fear or misunderstand the charging part. But maybe some have additional use cases in mind, such as using the car for different purposes and thus require more energy for comfort than just for driving. Probably this is also a hint towards bidirectional charging use cases, such as vehicle-to-load or vehicle-to-grid. If the additional 40 000 h for charging would become truth, one would need to analyze, if one could assume charging-like profiles or some with a bit more dynamic, but far from a driving profile. So, the additional functions of an EV will most likely not be lifetime critical. And this is a promising result for electric vehicles! 152 Lifetime Analysis of Electronics and Power Electronic Components in Electric Vehicles References [1] M. Bruell und P. Brockerhoff, „Avoid the DC charging trap,“ in Electric Vehicle Symposium EVS30, Stuttgart, 2017. [2] M. Bruell, P. Brockerhoff, F. Pfeilschifter, H.-P. Feustel und W. Hackmann, „Bidirectional Chargeand Traction-System,“ in EVS29 Symposium, Montréal, Québec, Canada, 2016. [3] J. Juskowiak, Beanspruchungsgerechte Bestimmung des Weibull- Formparameters für Zuverlässigkeitsprognosen, Stuttgart: Institut für Maschinenelemente der Universität Stuttgart (ISBN 978-3-936100-74-7), 2017. [4] A. Middendorf, Lebensdauerprognostik unter Berücksichtigung realer Belastungen am Beispiel von Bondverbindungen bei thermomechanischen Wechselbeanspruchungen, Dissertation: Technische Universität Berlin, 2010. [5] IEC 61709 , Electric components - Reliability - Reference conditions for failure rates and stress models for conversion, International Electrotechnical Commission (IEC). [6] AEC-Q200: Stress Test Qualification for Passive Components, Automotive Electronics Council, Rev D: 2010. [7] „Miner’s Rule and Cumulative Damage Models,“ ReliaSoft Corporation, [Online]. Available: https: / / www.weibull.com/ hotwire/ issue116/ hottopics116.htm. [Zugriff am 29 01 2019]. [8] Infineon Technologies AG, „Transient Thermal Measurements and thermal equivalent circuit models,“ AN 2015-10, 2015. [9] T. Hannach, Ermittlung von Lebensdauergleichungen vom Coffin-Manson- und Morrowtyp für bleihaltige und bleifreie Weichlote durch Kombination von FE und Experiment, Berlin: Technische Universität Berlin, 2017. [10] G. Glinka und J. Kam, „Rainflow counting algorithm for very long stress histories,“ International Journal of Fatigue, Bd. 9, Nr. 3, pp. 223-228, Feb. 1987. [11] M. Matsuishi und T. Endo, Fatigue of metals subjected to varying stress, Japan Soc. Mech. Engineering, 1968. [12] Manca, J., W. Wondrak, K. Croes, W. De Ceuninck, V. Haeger und L. De Schepper, „The Arrhenius Relation for Electronics in Extreme Temperature,“ IEEE, pp. 29-32, 1999. [13] AEC-Q100: Failure Mechanism Based Stress Test Qualification for Integrated Circuits, Automotive Electronics Council, Rev H: 2014. 153 Voltage Ripple on Electric Vehicle System Test benches Jozsef Gabor Pazmany, Samuel Siegel, Klaus Rechberger, Bernard Bäker Abstract In high-voltage (HV) supply systems of battery electric vehicles (BEV), conducted disturbances occur due to the switching operation of power electronic devices. These oscillations impact power quality and influence the safety and efficiency of the system operation. Increasing the number of power electronic devices connected to the high voltage supply system increases complexity and the probability of undesired mutual effects. Design parameter choice to limit voltage ripple becomes a challenging task for the system integration. Evaluations on test benches and modified prototypes carried out during the development process, raises questions in terms of the transferability and usability voltage and current ripple measurements relative to production vehicles.. This work summarizes the effect of test bench configurations on measureable ripple in comparison to production vehicle measurements. It focuses on the accuracy of events measured on test benches and prototype vehicles prepared for measurements in comparison with vehicles in mass production. It reviews the parameter dependencies and proposes a metric to define the usability of different test bench configurations for ripple measurements. 1 Introduction A vehicle’s high voltage DC power system is influenced by conducted voltage and current disturbances in the frequency range of 0-150kHz due to the switching operation of power electronics [8] [18]. To cope with this issue system designers determine power quality metrics such as maximum voltage ripples by their magnitude in the frequency domain and the maximum peak-to-peak voltage in the time domain. The magnitude of a ripple at certain frequencies depends on system parameters such as filter capacitors C x or the cable inductivity besides the properties of the semiconductor switches or control schemes [5]. System test benches are proposed to mimic the behavior of the vehicle concerning the power distribution and the functionality of its measurement, protection and communication system. However, at higher frequencies the test benches’ system inductivity is different from real vehicle, due to the modified cable length. During the development process of a BEV, besides prototype vehicles, device and system level tests are carried out on test benches for validation. This work focuses on 154 Voltage Ripple on Electric Vehicle System Test benches the performance of system level test benches, which are designed to mimic the functionality and behavior of the vehicle’s HV system. Safety and measurement requirements on test benches often do not allow a design as compact as in the real vehicle. This yields to increased cable lengths with higher cable inductances. The higher cable inductances effect the RLC circuits of the HV system resulting in resonance frequencies that differ between the test bench system and real vehicle system. By this, the dynamic and high frequency events occurring on the test bench are not mimicking the real vehicle behavior and vice versa. This work investigates on how the vehicle HV system designer can evaluate high frequency measurements on system test benches in the frequency range of [0-150kHz]. This paper organized as follows. In section 2. we briefly define the physical origin of the examined transient events and their critical influences. Section 3. gives an analytical calculation method for the peak voltage and peak current with parameter dependencies caused by these transient events, and we define the worst cases. In section 4. we highlight the developed simulation program with its features presenting different use cases. Section 5. shows simulation results validated on real prototype vehicles. 2 System Description In this section, we describe the HV system of the investigated BEV, operating at 800V. We briefly show the functionality of every energy storage, converter and transfer devices. The system provides four-wheel-driving, two-wheel-driving, AC charging and DC charging (with 400V or 800V charging infrastructure) user functions. To satisfy such advanced functional requirements a highly complex power electronics interfaced HVDC power system topology is proposed [8] [9] [10]. Figure 1. HV-system on an 800V battery electric vehicle The system is designed to satisfy the numerous functional and safety requirements of a modern power intensive BEV. Besides the HV lithium-ion battery, the system contains two electric drives and motors, a DC-DC converter to provide lower voltage levels, a heater as comfort load, an onboard charger to provide AC charging and a booster to provide 400V charging. 155 Voltage Ripple on Electric Vehicle System Test benches 2.1 Electric Drives Electric drives are powered by bidirectional six-pulse-inverters that provide the energy for the electric motors. These devices can be built with different semiconductors as silicon IGBT-s or silicon-carbide MOSFET technology. The implemented technology determines the possible operating switching frequency range. 2.2 HV-Battery The HV battery is the main energy storage in the system. The high frequency modelling of the battery is out of the scope of this work. A simplified RL battery model will be used to mimic the high frequency behavior. For this project, we assumed a battery with 20°C temperature and 80% SoC [15]. 2.3 Wiring Harness The wiring harness determine the resonance phenomena and impedance spectrum of the high voltage system. The wiring harness is modelled as a serial RL element. The influence of the skin effect on the size of the conductor cross-section is also computed [16]. The effects of shielding strategies are out of scope of this work thus they are neglected. 2.4 HV Auxiliary loads HV auxiliary loads power numerous function of the system such as charging, supplying lower voltage levels, and comfort functions as heating and cooling. All these devices are connected in parallel to the traction inverter and all of them are power electric components joined to the system via an input filter. Therefore, several power electronic devices are integrated to the HV system, this subsection gives a brief list of these components. HV Heater The HV heater provides heat for the thermal control of the battery and the interior of the vehicle. In the core, it contains electric resistors and silicon IGBT-s, switching at frequencies of around 10 kHz, with a relatively low power drain. DC-DC Converter The DC-DC converter interconnects the lower voltage systems with the 800V system, transfering energy of the HV battery to a lower voltage level. Alternatively, it performance as an infinite load dump for the low voltage systems. OBC The OBC is a power electronic device designed to provide AC charge functions for the customer. It is a rectifier providing 800V DC for the vehicle from the AC charging power. The OBC is connected to the system via a diode in order to protect the AC system from conducted disturbances of the vehicular network. HV Booster The HV booster is a power electronic device which is essential for the 400V DC charging of an 800V battery. The booster is a high power converter using the principal function of a voltage-multiplier (doubler). It operates unidirectionaly with a fixed proportion between the input and output voltage. 156 Voltage Ripple on Electric Vehicle System Test benches Figure 2. Schematic topology of the proposed HV system architecture 3 Analytical calculations In this section, we analytically observe a system test bench with a prototype vehicle’s HV system topology, but modified wiring length. We focus on the frequency range from 20Hz to 150kHz, in which conducted disturbances cause aging effects and disturb measurement systems [2] [3] [4]. The system requirements in this frequency range are considering in voltage domain the maximally allowed emitted voltage amplitude for each frequency and in time domain as well and the maximal voltage stress that the component has to deal with. Maximal current levels are only defined by the functional safety requirements. This work’s research scope is to describe the differences between events on a system test bench compared to a real vehicle between DC and 150kHz. All the power electronic devices have at their HV connection a DC-link capacitor (C x ), in order to limit the disturbances caused by the device and to protect the device from disturbances by the system. Concerning the system level modelling, the power electronic devices behave as constant power loads [3]. In the specified frequency range the wiring behaves as serial RL-elements, the parasitic capacities can be neglected. Serial RL-elements are determined by the damping factor and the natural frequency of RLC circuits. For analytic calculations, one device is replaced by a high frequency disturbance source, which can work as a voltage or as a current source. If it works as a current source, different system impedances result in different voltage amplitudes and current flows in the system. If it works as a voltage source, different impedances result in different current amplitudes and current flows in the system. Concluding we define a metric that helps the system developer to evaluate phenomena measured on different system configurations. This tool is essential to assess the usage of different test bench configurations for examining conducted high frequency disturbances. 157 Voltage Ripple on Electric Vehicle System Test benches 3.1 The inverter as ripple current source in normal operation conditions For analytic calculations, we considered the load dynamics to be much slower than the PWM operation. Thus, we focus on the harmonics of the PWM switching frequency and its sidebands. Note that the sidebands are depending on the number of motor revolutions. Equations (7) - (9) of [1] define the amplitude of the switching harmonics and their sidebands, by Bessel function based approximation. For sinusoidal sampled PWM signals of a three-phase inverter, the emitted ripple as a function of switching harmonics (n) and rotor speed depending side bands (q) is shown in figure 3. Figure 3. High frequency current disturbance emission of PWM modulated inverter, as a function of switching harmonics (n) and their sidebands (q) with the parameters M=0.6 and  0 =0, where M is the modulation index and  0 is the phase angle of the AC load. In [1] and [2], the RMS current calculation in different systems is proposed. In order to examine automotive systems first the RMS of the emitted ripple current is examined. The RMS current on the DC-link capacitor can be calculated as a function of the actual load and the modulation strategy, and it is independent of the impedance of the DC-link capacitor. 158 Voltage Ripple on Electric Vehicle System Test benches Figure 4. Emitted RMS current of a voltage-source inverter in the function of the modulation depth and  0 3.2 Impedance of an automotive high voltage system The impedance of the HV system is to be calculated in order to analyze the system in the frequency range of 0-150kHz. For the analytic calculations, the following assumptions are made, valid for the specified frequency range:  All HV components have a DC-link capacitor  The DC-link capacitors are dominantly larger than the parasitic capacitance  The ESRs of the DC-link capacitors are taken into account  There are no differential mode EMC filters in the system  The battery impedance is assumed as known 159 Voltage Ripple on Electric Vehicle System Test benches The system impedance is calculated from the point of view of the main disturbance source, for instance the front traction inverter. Equation 2 highlights the impedance of an HV system with one auxiliary load (indexing with 1) and with the DC-link capacitor (C x ) of the traction inverter. Where ESR is the equivalent series resistance of a capacitor [5], L is the cable inductivity and R is the sum of the cable resistance. Indexing x refers to the DC-link capacitor of the main disturbance source and indexing 1 refers to the auxiliary load. 3.3 Analytical calculation of ripple current and current distribution The HV system in the specified frequency range is a parallel structure of several LC elements and constant power loads (CPL) since the impedance is frequency dependent. Moreover, the high frequency disturbance source at a certain time contains currents at some discrete frequencies. That implies that the current distribution at a certain frequency is calculated, afterwards extrapolated to a frequency range. The method of current calculation at a time (t0) where the speed and the torque of the electric motor is constant around t0 for a specified auxiliary load or system component contains the following steps:  Calculation of the characteristic of the disturbance source (see section 3.1)  Calculation of the system impedance at every discrete frequency in the spectrum of the switching harmonics (see section 3.2)  Calculation of the current divider equation at every frequency, so the ripple current in frequency domain  Calculation of the RMS current and numerical inverse Laplace transformation of the currents in order to get the time domain ripple current. 3.4 Metric definition This work considers two different high frequency phenomena up to 150kHz, conducted voltage and current disturbance. These phenomena appear by normal operation of every power electronic converter. This section defines an evaluation metric concerning the sensitivity of ripple phenomena for different vehicle architectures and test benches. The proposed metric supports the evaluation of the measured high frequency current distribution and maximal current levels on test benches. It also evaluates the mimicking behavior of the test bench at high frequencies. Evaluation of conducted disturbances in current domain The influence of high frequency current on high voltage devices is a main issue regarding safe and stable operation. Maximum allowed current levels are defined by the system specification and current disturbances are also reducing the effectivity of measurement systems and feedback control loops. The source of the current disturbances are the switching operation of semiconductor devices, as previously highlighted in section 2. For the following examination, a high frequency current source is assumed with 160 Voltage Ripple on Electric Vehicle System Test benches 1A amplitude. The metric definition is based on the difference between the current amplitude at every measurement point in the real vehicle and on the system level test bench. First, the sensitivity of the observed current amplitudes compared to the changed system impedance and cable length has to be determined. Afterwards the sensitivity has to be observed in the whole frequency range up to 150kHz. By defining the error caused by the test bench for every measured current, the elemental metrics has to be overlapped to a closed form definition. The metric definition summarizes resonance shifting and the quality of the RLC elements in the high voltage system. Assumption 1. For the further examination, we consider DC-link capacitors as part of the high voltage system and not only part of a single high voltage device. Definition 1. Assume an automotive HV system with n HV devices, a current source C i where i is part in n with a clean sinusoidal signal at the frequency f inside [0-150kHz], with the magnitude of 1A, where C i replaces one of the high voltage devices. A system test bench is suitable to examine high frequency events if for all f the resulting relative error of the current amplitude at all HV devices is maximum 20%. A system test bench is conditionally suitable to examine high frequency events if for all f the resulting relative error of the current amplitude at all HV devices is between 20- 50%. A system test bench is not suitable to examine high frequency events if for all f the resulting relative error of the current amplitude at all HV devices is more than 50%. Assumption 2. The HV system of a vehicle is a parallel connection of a low pass filter with some band stop filters. There is no discrete resistor in the system just the ESR of the capacitors and the serial resistance of the wiring, and there is no two components with the same DC-link capacitor. Proposition 1. If there are no two resonance frequencies within 1kHz distance from each other in the system and there are no discrete resistances in the system, the HV system can be decoupled to individual RLC elements. The proposed metric declares the difference between the distribution of the high frequency currents in a complex HV system in a real vehicle and in a system test bench configuration. In order to quantify the described metric, three discrete values are defined, where the measure have the following values:  0 (green) if the test bench reproduces sufficiently the real vehicle  1 (yellow) if the test bench reproduces restrictedly sufficient the real vehicle  2 (red) if the test bench reproduces not sufficiently the real vehicle This metric definition allows a fast analysis of the high frequency performance of a system test bench and the fast evaluation and classification of the system behavior after the measurement of certain events. 161 Voltage Ripple on Electric Vehicle System Test benches Current difference in the system Difference evaluation, Functional safety current and voltage limits, the limits in percentage are to define such that under 20% vehicle and test benches are sufficiently similar, 20-50% vehicle validation required due to the restricted sufficiency of the reproduction performance, 50% completely different events, the reproduction performance is not sufficient. 4 Simulation and Validation As part of this project, we have developed a MATLAB/ Simulink/ Simscape simulation model to calculate the proposed metric, and we have validated the results on system test-benches. The simulation model contains the RL model of the wiring harness and the RC model of all DC-Link capacitors and all the high voltage devices besides the disturbance source are modelled as constant power load. The simulations include a high frequency current source as disturbance generator. The scope of the simulation was to mimic the behavior of the test bench and the system in the frequency range of 0-150kHz. The simulation contains the passive model of all HV-devices and the active model of the source devices. The simulations shows the resulted current distribution with different vehicle configurations. As expected, the difference between current distribution in a vehicle and in the system test-bench is critical around the resonance frequencies of the parallel RLC elements built by the HV-devices. We provide simulation that show at critical frequencies how the current distribution differs between prototype vehicle and test bench, with different test bench configurations. The different cable length affects the current distribution and the ripple current at the input of a certain HVdevice. Since maximal current can be critical for a device the robustness of the test bench system can be different from the robustness of the HV-system of the vehicle. Figure shows the simulation results with the defined metric, where all cable length are enlarged with the same factor. Assuming a traction inverter built with IGBT-s its typical switching frequency is between 5-12kHz the dominant frequency of the voltage and current ripple is between 10-24kHz, figure 4 implies that a wiring harness with more than 20% cable length difference has already an observable impact on the ripple current. That means the amplitudes on the system test-bench must be proofed on prototype vehicles as well. 162 Voltage Ripple on Electric Vehicle System Test benches Figure 5. Current distribution between DC-link capacitor and HV-system in a real vehicle, in frequency domain with 1A Ripple current On the other hand, in the scope of the examination stays a test bench with doubled cable length. The effect of such cable length modification is shown on figure 6. Is to be observed that the modification of the properties of the resonance circuits in the test bench configuration. Figure 6. Current distribution between DC-Link capacitor and HV-System in a test bench with double cable length, in frequency domain with 1A Ripple current Figure 5 and Figure 6 shows the difference between the high frequency current distribution in the system test-bench. From the labels on Figure 5 and Figure 6 one can observe that the resonance frequency of a specific auxiliary load is shifted from 10,67kHz to 7,48kHz by rescaling the cables with the factor of 2. Obviously, the reso- 163 Voltage Ripple on Electric Vehicle System Test benches nance frequencies of the RLC elements in the system test bench are the most influencing parameter of current distribution. Furthermore, the longer cable and more cable inductivity implies lower resonance frequencies that means higher skin depth at the resonance frequency. Larger skin depth yields to less cable resistance. On the other hand, due to the longer cable length the resistance of the longer cable is risen. Due to this contradiction between the cable length and skin depth it worth to visualize. For instance, let a DC-link capacitor be 100µF afterwards examined the resonance frequency and skin depth assuming a 50mm 2 straight copper cable in the function of the cable length, from 0.1m to 10m. From the point of view of the other end of the cable. The example shows the effect of cable length on the distribution between the DC-link capacitor of a device and the rest of the HV-system. Figure 7 Current distribution between DC-link capacitor and HV-system at a certain frequency in the function of cable length Figure 7. shows the simulation results of the current distribution between the DC-link capacitor of a power electric device and the other parts of the high voltage system. That device is assumed, to be the source of the current disturbances. The simulation results show the effect, how the longer cables influence the usability of the test bench for the reproduction of a high frequency event. To combine the two dimensions of figure 6 and 7 a complete analysis of the high frequency performance of the examined test bench configuration. 164 Voltage Ripple on Electric Vehicle System Test benches Figure 8. The performance of test benches form 1kHz up to 150 kHz in the function cable length. While “green”= proper reproduction performance “yellow”= bounded reproduction performance “red”=not sufficient reproduction performance. We produced a system level simulation with MATLAB/ Simulink/ Simscape, and we conducted a model in the loop simulation to evaluate the performance of different test bench configuration with different disturbance sources. On figure 8 the results of all different cable length configuration scaled from 0,1 to 10 times the real cable length with current ripple sources from 1kHz to 150kHz are evaluated. It is shown, that if the cable length is inside +/ -20% from the real vehicle (such as prototype vehicles with additional measurement devices) the test bench provides a satisfying reproduction of the current distribution in the HV system. It is also shown that disturbances around the frequency of 10kHz (which is typical for IGBT based devices) the test benches, with larger than 20% cable length difference from the vehicle, are providing bounded or not sufficient reproduction performance. 5 Results We have shown that the different cable length of the system test bench has is a major factor in the evaluation of events in the frequency range of 0-150 kHz. If the system does not contain common mode filters, then the system test bench is not recommended to be used for validation of disturbances caused by the switching operation of IGBT-s, the measurement should be conducted on prototype systems or rather simulated. We have shown, that however additional measurement devices adding sable length to the system, if the additional cable length is smaller than 20% of the original cable length, the current distribution in the system will just be minor disturbed. We found that the traction inverter, in the case of acceleration with maximal torque, causes the most critical voltage and current ripple phenomena. The frequency of the voltage ripple are at the first sidebands of the switching frequency and at the first harmonic of the switching frequency of the inverter. So we examined the possible doubled switching frequencies and the frequency range of the first side bands of the switching frequency. Therefore we examined the switching frequency of 8kHz and we found that 165 Voltage Ripple on Electric Vehicle System Test benches the resonance phenomena between a specific auxiliary load and the front inverter (see figure 1.) is not reproducible on the system test bench so the RMS current stress and the maximal current levels on the test bench are with several Amperes smaller than in the real vehicle. 6 Conclusion In the development process of a BEVs the validation test on system level test-benches are playing a key role. However, the understanding of the differences between the behavior of a test bench and the real vehicle is essential for the efficiency of the development project. Our work shows a method to understand the differences between the behavior of a test bench and the real vehicle. This work provided a contribution to the understanding of the usability of high voltage test benches for the high frequency requirements of the system specification of electric vehicles. This work has proposed a method to evaluate the error caused by the test bench topology against the vehicle topology. The results of this work can be used by the evaluation of the measurement results or errors measured on the system test bench for every hardware configuration during the development. Bibliography [1] M. Zhang, N. Wheeler and D. Grant, ``Switching Harmonics in the DC Link Current in a PWM AC-DC-AC Converter,'' IEEE, 1995. [2] J. W. Kolar and S. D. Round, "Analytical calculation of the RMS current stress on the DC-link capacitor of voltage-PWM converter systems," in IEE Proceedings - Electric Power Applications, vol. 153, no. 4, pp. 535-543, July 2006. [3] H. Wen and W. Xiao and X. Wen and P. Armstrong, ``Analysis and Evaluation of DC-Link Capacitors for High-Power-Density Electric Vehicle Drive Systems'' IEEE Transactions on Vehicular Technology, vol.61 nr.7" 2012, pp. 2950-2964. [4] Marschalko, R., Fodor, D., Teodosescu, P. D., and Bojan, M. (2011). Influence of DC-link capacitor aging on the PWM converters operation. Acta Electrotehnica, 52(4), 197-202. [5] H. Wen, W. Xiao and Xuhui Wen, "Comparative evaluation of DC-link capacitors for electric vehicle application," 2012 IEEE International Symposium on Industrial Electronics, Hangzhou, 2012, pp. 1472-1477. [6] B. A. Welchko, "Analytical calculation of the RMS current stress on the DC link capacitor for a VSI employing reduced common mode voltage PWM," 2007 European Conference on Power Electronics and Applications, Aalborg, 2007, pp. 1-8. [7] Rosa, J. "The harmonic spectrum of DC link currents in inverters." Proc. 4th International Conf. on Power Conversion. 1982. [8] S. Haghbin, A. Karvonen and T. Thiringer, "Harmonic modeling of a vehicle traction circuit towards the DC bus," 2014 International Power Electronics Conference (IPEC-Hiroshima 2014 - ECCE ASIA), Hiroshima, 2014, pp. 1373-1378. 166 Voltage Ripple on Electric Vehicle System Test benches [9] A. Karvonen and T. Thiringer, "Co-Simulation and Harmonic Analysis of a Hybrid Vehicle Traction System," 2015 IEEE Vehicle Power and Propulsion Conference (VPPC), Montreal, QC, 2015, pp. 1-6. [10] A. Karvonen and T. Thiringer, "Parameter analysis of current and voltage ripple in a hybrid vehicle traction system," 2015 IEEE International Electric Machines, Drives Conference (IEMDC), Coeur d'Alene, ID, 2015, pp. 1838-1845. [11] K. Jia, C. Bi and H. Li, "Mitigation of current oscillations in the DC link of electric vehicles," 2017 Asia-Pacific International Symposium on Electromagnetic Compatibility (APEMC), Seoul, 2017, pp. 76-78. [12] A. Henriksson, J. Simonsson, U. Lundgren and P. Ankarson, "Cable Modeling for Accurate Estimation of Current and Voltage Ripple in Electric Vehicles," 2018 IEEE Transportation Electrification Conference and Expo (ITEC), Long Beach, CA, 2018, pp. 714-719. [13] S. Schoerle, E. Hoene, S. Hoffmann, A. Kuczmik and K. D. Lang, "System Simulation of Automotive High Voltage Grids: Modelling of Power Converters and Connecting Cables," 2014 IEEE Vehicle Power and Propulsion Conference (VPPC), Coimbra, 2014, pp. 1-6. [14] A. Mariscotti, "Analysis of the DC-link current spectrum in voltage source inverters," in IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, vol. 49, no. 4, pp. 484-491, April 2002. [15] T. Doersam, S. Schoerle, E. Hoene, K. -. Lang, C. Spieker and T. Waldmann, "High frequency impedance of Li-ion batteries," 2015 IEEE International Symposium on Electromagnetic Compatibility (EMC), Dresden, 2015, pp. 714-719. [16 ] S. Schoerle, E. Hoene, S. Hoffmann, A. Kuczmik and K. D. Lang, "System Simulation of Automotive High Voltage Grids: Modelling of Power Converters and Connecting Cables," 2014 IEEE Vehicle Power and Propulsion Conference (VPPC), Coimbra, 2014, pp. 1-6. [17 ] B. P. McGrath and D. G. Holmes, "A General Analytical Method for Calculating Inverter DC-Link Current Harmonics," in IEEE Transactions on Industry Applications, vol. 45, no. 5, pp. 1851-1859, Sept.-oct. 2009. [18] D. G. Holmes and T. A. Lipo, "Pulse Width Modulation for Power Converters" Piscataway, NJ: IEEE Press, 2003 167 Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles Möglichkeiten und Potenziale der aktiven (EMV-)Störungsunterdrückung zur Bauraumreduktion von DC/ DC-Wandlern im Kfz Andreas Bendicks, Tobias Dörlemann, Stephan Frei, Norbert Hees, Marc Wiegand Abstract The electromagnetic compatibility (EMC) of power electronic systems is a severe issue since the switching of transistors causes high electromagnetic interferences (EMI). To reduce the EMI, usually passive filters are applied that tend to be bulky, heavy and expensive. Active noise cancellation is a promising approach to get rid of these problems. Existing methods, namely active EMI filters, suffer from unavoidable delay times since they inject a cancellation signal that origins from a measured signal. These delay times limit the suppressible frequency range and the achievable EMI reduction. To resolve this issue, synthesized cancellation signals are applied by the authors. Since the signal is artificial, there is no systematic delay and all bothersome effects, like phase-shifts or magnitude responses, can be compensated. It is only required that the cancellation signal can be synchronized with the power electronic device to maintain a destructive interference. This can be realized in most digital controlled systems. The method has already been successfully applied to automotive DC/ DC converters and is currently being extended for the application to power factor correction circuits and motor inverters. In this contribution, the state of the art and the innovation are described, a demonstrator and measurement results are presented, and the potential field of application is evaluated. Kurzfassung Leistungselektronische Systeme können aufgrund der verwendeten schnellschaltenden Transistoren erhebliche Quellen für elektromagnetische Störungen darstellen. Zur Begrenzung dieser Störungen gegenüber der Umwelt werden üblicherweise passive Filter eingesetzt, welche jedoch häufig groß, schwer und teuer sind. Die aktive Störungsunterdrückung ist ein vielversprechender Ansatz zur Lösung dieser Probleme. Bereits existierende Methoden (aktive Filter) leiden unter unvermeidlichen Verzögerungszeiten, da diese das Gegenstörsignal direkt aus einem gemessenen Signal erzeugen. Diese systematische Verzögerungszeit sorgt dafür, dass Störungen und Gegenstörungen niemals gleichzeitig sind. Daher sind die erzielbare Störungsreduktion und der entstörbare Frequenzbereich systematisch eingeschränkt. Zur Lösung dieses 168 Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles Problems können synthetisierte Gegenstörsignale eingesetzt werden. Da diese künstlich erzeugt wurden, besteht keine systematische Verzögerungszeit und alle problematischen Effekte, wie Phasendrehungen oder Betragsgänge, können kompensiert werden. Für die destruktive Interferenz zwischen den Signalen muss ein synchroner Betrieb zwischen Leistungselektronik und dem Entstörsystem gewährleistet werden, was jedoch in den meisten digitalen Systemen realisiert werden kann. Die Methode wurde bereits erfolgreich bei DC/ DC-Wandlern in Automotive-Anwendungen eingesetzt und wird derzeit für weitere wichtige Systeme, wie Leistungsfaktorkorrekturen oder Antriebswechselrichter, erweitert. In diesem Beitrag wird der bisherige Stand der aktiven Filter beschrieben, und die Störungsunterdrückung mithilfe von synthetisierten Signalen wird vorgestellt. Es wird ein Demonstrator präsentiert und Messergebnisse werden diskutiert. Potenzielle Einsatzbereiche werden dargestellt. 1 Introduction Power electronics is a key technology of the 21st century due to the possibility of very efficient energy conversion. Usually, power electronics achieve this efficiency by switching transistors that may be considerable sources of electromagnetic interferences (EMI). These interferences may harm the operation of other systems, such as communication or broadcasting services or susceptible sensors, which steadily gain more significance due to digitization. To ensure the functionality of sensitive systems in proximity to power electronic devices, electromagnetic compatibility (EMC) is a major enabler for electrification with power electronics. 2 State of the Art 2.1 Passive Filters Passive line filters consisting of coils and capacitors are commonly applied to attenuate the conducted EMI of power electronic devices. Since the disturbances as well as the operating voltages and currents of the converters are typically quite high, filters tend to be bulky and heavy. In Figure 1, an automotive on-board charger is depicted. Figure 1: Exemplary on-board charger (3.6 kW) 169 Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles Here, the passive filter takes up approximately one third of the electronics volume. In many applications, this can be a serious problem. In e.g. hybrid vehicles, there is not much volume for the power electronic systems since the combustion engine and its components take up basically all of the available space. In electric vehicles, there is more free space since there is no combustion engine anymore. Nevertheless, there are very high requirements regarding the vehicle’s total mass to improve the driving range. So, there are many reasons to shrink the passive filter component. 2.2 Active EMI Filters To resolve the mentioned issue, active EMI filters have been developed (Figure 2) [1]- [4]. In contrast to passive filters, active EMI filters inject cancellation signals that cause a destructive interference with the disturbances of the power electronics to reduce the EMI of the system. Active EMI filters consist of a small circuit for sensing the disturbances, an amplifier (e.g. operational amplifier) to generate the cancellation signal from the disturbances, and a small circuit for injecting the cancellation signal. The major limitation for active EMI filters is the inevitable delay time introduced by the analog (or recently also digital) circuitry. Due to this delay time, noise and anti-noise can never be exactly simultaneous. This effect systematically limits the achievable EMI reduction and the suppressible frequency range as analyzed exemplary for feedforward-types in [7]. Figure 2: Fundamental topologies of active EMI filters (top: feedforward-type, bottom: feedback-type) Noise- victim Power electronics Injector Sensor Amp. Feedforward active EMI-filter Systematic delay Noise- victim Power electronics Sensor Injector Amp. Feedback active EMI-filter Systematic delay 170 Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles 3 Innovation 3.1 Active Noise Suppression with Synthesized and Synchronized Cancellation Signals To resolve the issue of a systematically delayed signal path, the cancellation signal can be artificially synthesized and applied in synchronicity with the disturbances (Figure 3) [5]-[8]. Remaining magnitude responses, phase-shifts and delays are compensated by the shape of the synthesized signal improving the cancellation’s effectivity widely. For (quasi-)periodic disturbances, the signal can be easily constructed from harmonic sine waves. In the following, an adaptive approach is presented to find the right amplitudes and phases for the cancelling sine waves and, therefore, the overall cancellation signal. Figure 3: Concept of the innovation 3.2 Adapted Harmonics Cancellation The fundamental concept of Adapted Harmonics Cancellation (AHC) is illustrated in Figure 4. The power electronic system is the source of the stationary disturbing harmonics that must be cancelled out. A DC/ DC converter would be a typical example of a system that creates harmonic disturbances due to the periodic switching of the transistors. The cancellation signals can be generated by digital hardware (e.g. by an FPGA with analog-to-digital converters [ADCs] and digital-to-analog converters [DACs], Figure 5). The digital hardware comprises an optimizer and a synthesizer. For cancellation, the synthesizer generates a sine wave for each disturbing harmonic. The optimizer is used to find the right amplitudes and phases for cancellation. The synchronicity of the generated sine waves and the disturbing harmonics is maintained by a suitable synchronization signal. To link the power electronic system and the cancellation system, interfaces are necessary. A sensor consisting of an analog circuit and an ADC is used to measure the disturbances. An injector is applied to couple the cancelling waveforms into the power electronic system. This injector consists of an injecting circuit and a DAC. These components can be very small, especially in comparison to passive EMI filters. Noise- victim Power-- electronics Innovation Trigger Injector Synthe‐ sizer No delayed signal path! 171 Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles Figure 4: Concept of Adapted Harmonics Cancellation Figure 5: FPGA evaluation system with high-speed DACs and ADCs (Red Pitaya STEMlab 125-14) 4 Demonstration For demonstration, the disturbances of an automotive 48 V/ 12 V DC/ DC are suppressed by a self-adapting algorithm that is implemented on the FPGA system Red Pitaya STEMlab 125-14. The Device Under Test (DUT) can be seen in Figure 6. In Figure 7, the disturbances at the artificial network are depicted. Obviously, the original disturbances are much higher than the class 5 limit of the standard CISPR 25 [9]. The resulting disturbances for a feedforward active EMI filter with a reasonable delay time of 10 ns are predicted with the formulas derived in [6] and [7]. Due to the systematic delay, the achievable reduction and the suppressible frequency range are severely limited for the active filter. Since there is no systematic delay between noise and antinoise for the authors’ proposed method, the noise suppression is widely improved. The complete frequency range from 150 kHz to 30 MHz is suppressed successfully and complies now with the standard. The fundamental wave at 300 kHz is reduced by Power- electronic- system Synchronization Sensor ADC DAC Sensing circuit Injector Injecting circuit Synthesizer + … 𝑓 Optimizer 2𝑓 𝐾𝑓 𝐾 172 Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles nearly 60 dB, and even the 100th harmonic at 30 MHz is suppressed by approximately 40 dB. More details and further demonstrations can be found in [5]-[8]. 5 Potential Applications Until now, the presented method of active EMI suppression with synthesized and synchronized cancellation signals has been successfully applied to stationary operating DC/ DC converters. Currently, the method is extended for the application in non-stationary operating systems like power factor corrections or motor inverters which play a key role in power electronics. In these applications, the cancelling sine wave properties have to be adjusted continuously. By injecting exactly synchronous cancellation signals without any uncompensated delay, there is an extraordinary potential for the active EMI suppression in many power electronic systems. Figure 6: Realized DUT Figure 7: Measured disturbances at the artificial network 0.15 0.3 1 2 5 10 20 30 f in MHz 0 20 40 60 80 100 Measurement without cancellation Prediction for active EMI filter with 10 ns delay Measurement for new, proposed method Class 5 limit -60 dB -40 dB -55 dB 173 Possibilities and Potentials of Active EMI Cancellation for the Volume Reduction of DC/ DC Converters in Automobiles 6 Conclusion In our research, a new method has been developed to eliminate disturbing voltages or currents by injecting synthesized and synchronized cancellation signals. In comparison to other known active methods, there is no systematic delay time that limits the achievable performance of the system. The effectivity of the method has already been shown for a DC/ DC converter and is currently being extended for other important power electronic systems, namely power factor corrections and motor inverters. The authors expect a considerable reduction of the necessary passive filter effort that can lead to smaller and lighter power electronic systems. 7 References [1] J. Walker, “Designing practical and effective active EMI filters,” in Powercon 11 Proc., April 1984, Paper I-3. [2] L. E. LaWhite, M. F. Schlecht, “Design of active ripple filters for power circuits operating in the 1-10 MHz range,” IEEE Trans. Power Electron., vol. 3, no. 3, pp. 310-317, Jul. 1988. [3] L. E. LaWhite, M. F. Schlecht, “Active filters for 1 MHz power circuits with strict input/ output requirements,” in 17th Annual IEEE Power Electronics Specialists Conf., Vancouver, Canada, 23-27 Jun. 1986, pp. 255-263. [4] T. Farkas, M. F. Schlecht, “Viability of active EMI filters for utility applications,” IEEE Trans. Power Electron., vol. 9, no. 3, pp. 328-337, May 1994. [5] A. Bendicks, T. Dörlemann, S. Frei, N. Hees, M. Wiegand, “FPGA-basierte aktive Gegenkopplung der Schaltharmonischen von leistungselektronischen Systemen,” in EMV Düsseldorf, Düsseldorf, Germany, 20-22 Feb. 2018, pp. 652-661. [6] A. Bendicks, T. Dörlemann, S. Frei, N. Hees, M. Wiegand, “Development of an Adaptive EMI Cancellation Strategy for Stationary Clocked Systems,” in EMC Europe, Amsterdam, Netherlands, 27 30 Aug. 2018. [7] A. Bendicks, T. Dörlemann, S. Frei, N. Hees, M. Wiegand, “Active EMI Reduction of Stationary Clocked Systems by Adapted Harmonics Cancellation,” IEEE Trans. EMC, to be published, Aug. 2018. [8] A. Bendicks, S. Frei, “Broadband Noise Suppression of Stationary Clocked DC/ DC Converters by Injecting Synthesized and Synchronized Cancellation Signals”, IEEE Trans. Power Electron., to be published, Jan. 2019. [9] CISPR 25 - Vehicles, Boats and Internal Combustion Engines - Radio Disturbance Characteristics - Limits and Methods of Measurement for the Protection of On-Board Receivers, Ed.4.0, 2015. 174 Ring Structures in Automotive Power Nets: Idea and Implementation Laurenz Tippe, Julian Taube, Joachim Fröschl, Hans-Georg Herzog Abstract Due to rapid developments in the field of driver assistance systems and a steady increase of auxiliary loads in vehicles, conventional methods of power net design reach their limits rather quickly. While the increasing power demand of a growing number of electrical loads can be handled by the introduction of a 48 V voltage level, the basic structure of power nets remains mostly untouched. In order to ensure a reliable and fault-tolerant supply of energy, innovative solutions and a rethink in the area of system design is necessary. As a result, the implementation of a ring-structured power net is proposed and its basic considerations are outlined. The presented concept combines a number of multifunctional devices and switching units, which are distributed over the vehicle, as well as a 48 V ring structured architecture, which enables a secured energy supply even in case of an occurring fault. Kurzfassung Bedingt durch die rasante Entwicklung im Bereich der Fahrerassistenzsysteme und einem stetigen Zuwachs an Komfortverbrauchern im Fahrzeug, stoßen konventionelle Methoden der Bordnetzauslegung rasch an ihre Grenzen. Während dem steigenden Leistungsbedarf und der wachsenden Verbraucheranzahl durch die Einführung der 48 V-Spannungsebene entgegengewirkt wird, bleibt die grundlegende Architektur des Bordnetzes weiterhin in großen Teilen unberührt. Um speziell im Bereich des autonomen Fahrens eine ausfallsichere und fehlertolerante Energieversorgung sicherstellen zu können, bedarf es innovativer Lösungsvorschläge und einem Umdenken im Bereich der Systemauslegung. Im Zuge dessen wird die Implementierung eines Ringbordnetzes vorgeschlagen und dieses in den Grundzügen dargelegt. Der Entwurf beinhaltet diverse multifunktionale Einheiten, die über das Fahrzeug verteilt werden. Mittels des vorgeschlagenen Konzepts kann im Fehlerfall über die als Ring ausgeführte 48 V- Hauptversorgungsleitung und mit Hilfe diverser Umschaltvorrichtungen weiterhin eine sichere Energieversorgung garantiert werden. 1 Introduction and State of The Art The ongoing evolution of automotive systems and technologies yields great possibilities for future car generations. While the road to fully autonomous driving is paved by new technologies and innovative thinking, there are also a number of obstacles, which 175 Ring Structures in Automotive Power Nets: Idea and Implementation will have to be overcome in order to ensure a flawless operation of all components. A growing number of energy consumers such as electrical sensors as well as on-board entertainment systems and auxiliary loads puts pressure on conventional power supply systems’ dimensioning and design approaches. At the same time, the conversion from a purely 12 V driven power net to a 48 V power net seems to become more and more visible in production vehicles [1]. In addition, aspirations towards higher system efficiency while maintaining a stable, predictable and reliable system behavior will have to be fulfilled, in order to ensure a flawless operation from a technical point of view. The ongoing migration period from 12 V to 48 V power nets is mostly realized by applying a structure, which adds one complete wiring harness for each voltage level [2]. In case of a failure in the main supply line, a conventional power net layout yields limited possibilities in terms of providing a reliable energy supply (see Fig. 1). Furthermore, the converter design has to follow the peak power of all components on either side, which does not necessarily result in high efficiency for all operating points. Moreover, safety relevant devices have to be identified and, subsequently, need to be adapted to maintain a fail operational behavior. The developments of the recent years already show a trend of redundant energy supply to critical components [4]. Nevertheless, the fail operational behavior of a device still has to be implemented beforehand, which stands mostly separate to the power net design. Regarding the proposed ring layout and its components, the individual electrical loads do not necessarily have to be adapted to be fail operational, but rather the proposed power net layout will fulfill the task to ensure reliable and fault-tolerant behavior of critical components. Fig. 1: Conventional automotive power net (Bat.: Battery; Gen.: Generator) [3] 176 Ring Structures in Automotive Power Nets: Idea and Implementation 2 Ring Structured Power Nets In contrast to the conventional power nets used in the automotive field, a ring topology is proposed as a power distribution structure. Ring structured systems can be regarded as renowned solutions for power grid applications (e.g. see [5]) as well as data network applications (e.g. see [6]). Nevertheless, typical automotive power nets are still based on linear bus structures as shown in Fig. 1. In order to adapt to the rising demand in both reliability as well as in the peak power distributed over a power net in future car generations, the ring structure offers new degrees of freedom to achieve the desired goals. The proposed basic structure (see Fig. 2) includes certain key components, which are: 1. Ring structured main supply line 2. Nodes, which act as hubs 3. Safety-relevant electrical load, which needs to remain functional at all cost Regular non-safety-relevant electrical loads can be found outside the above mentioned ring system. Those are connected in a star formation to the embedded nodes, whilst all safety-relevant electrical loads find themselves embedded in the middle of the ring. The nodes are interconnected using a ring supply line and thus carry a number of both safety-relevant and normal electrical loads. So far, recent developments in power net voltage levels are not adapted to the abstract system. Adjusting this concept to stateof-the-art power nets, which inherit electrical loads on multiple voltage levels, also means adapting the system to provide a reliable energy supply on every desired voltage level. Fig. 2: Ring Structure Including Safety-Relevant Load [3] 177 Ring Structures in Automotive Power Nets: Idea and Implementation Hereinafter, it will be assumed that a main supply on the 48 V level is the standard for automotive power nets, while a second 12 V power net will still be necessary in the forthcoming future. As shown in Fig. 3, the ring structured main supply line connects the generator, as well as four nodes called “Integrated Smart Distribution Devices (ISDDs)” on the 48 V level. For now, we assume, that an ISDD is capable of providing an additional 12 V supply to all components, which are not yet migrated to the 48 V level. Consequently, regular electrical loads are connected to the ISDDs on both the 48 V and the 12 V side. A more specific investigation on all elements inside an ISDD can be found in chapter 3. In addition, safety-relevant loads of both voltage levels can be found embedded in the middle of the ring. This allows for these specific types of loads to be powered from either one ISDD or from both connected ISDDs at the same time. Consequently, it can be followed that each node - or ISDD - has to entail its own energy storage in order to provide the necessary power to all components even in case of a cut from the generator. The technology and sizing of those energy storages does not have to be further specified at this point from a system designer’s point of view. Nevertheless, battery storages as well as capacitors can be regarded as common energy storages inside of automotive power nets. As both the ISDD as well as the ISSU are of critical nature to the overall design, those two components will be examined in further detail in the following sections. 3 Integrated Smart Distribution Device (ISDD) An implementation as it is introduced in chapter 2 requires some further investigations concerning the key elements of automotive power nets. Topics such as energy storage and distribution as well as reliability have to be addressed. In addition to the before mentioned ring structured power net, the nodes (see Fig. 2) to which components will be connected also have a tremendous impact on the overall system behavior. Fig. 3: Ring Power Net Including System Components [3] 178 Ring Structures in Automotive Power Nets: Idea and Implementation In its basic form, the nodes called “Integrated Smart Distribution Devices” (ISDDs) will include the following components: 1. Central 48 V potential 2. Segmentation units 3. DC/ DC converter(s) 4. Energy storage (e.g. a battery or capacitors) 5. Fuses The basic layout is displayed in Fig. 4 and shows all related components. In this case, electrical loads are simplified as resistive loads. In addition to the above mentioned features, additional benefits can be gained by basing the ISDD’s design on a distribution rail (see [7]) rather than on conventional wire structures. This allows for the sizing of each component to be adapted to the individual requirements of each ISDD. In contrast to software, which can easily be upgraded in modern cars, it is not possible yet to easily upgrade hardware once the vehicle is built. Utilizing the power rail as a starting point for construction in future cars’ power nets technically allows for easy upgradability not only in software, but also in hardware, which is due to the plug-and-play features of the distribution line. In order to enable these features, it is helpful to define a standard for plug connections such that hardware components like control units, fuses etc. can easily be repaired or upgraded, no matter of the hardware setup in which a car was first delivered to customers. Furthermore, human errors and wrong cable management in wiring harnesses can be omitted to a great extent. Additionally to wiring faults, EMC design has to be regarded as a key factor while designing a power rail distribution based power net. In order to gain a benefit, the power rails on each voltage level are stacked upon their corresponding ground potentials, which allows for the effects of occurring electromagnetic fields to cancel each other out. Fig. 4: Structure of ISDD utilizing power rails on 48 V and 12 V levels and included components 179 Ring Structures in Automotive Power Nets: Idea and Implementation Moreover, the proposed structure of each ISDD allows for a highly modular and adaptive system in terms of its mounting position throughout the car and the desired voltage levels. Technically, it would also be possible to replace the 12 V level with any other voltage level, e.g. a 5 V level in order to directly power LED illumination or control units inside the vehicles. Taking the idea of a ring structured power net one step further, it would also be possible to use the power rail introduced in [7] as a main distribution system throughout the car. Since the layout of power rails is adaptable to any desired length, different variants can be implemented ranging from short distributed rails mounted behind side panels up to long rails running through the middle of a vehicle. In this case, it is possible to implement a ring structure using the two rails, which then both provide 48 V levels. The ISDD can then be adapted to the before defined mounting standard and simply be added to the rails (see Fig. 5). More details on said power rails can be retrieved from [7]. In this variant, all ISDDs still contain the basic components and function as coupling bridges between shorter versions of the power rails. A concurrent use of ISDDs is also possible, in which ISDDs including power rails on one hand and power rails utilizing mounted ISDDs on the other hand collaborate. Independent of the implemented variants, it is necessary to ensure a high efficiency throughout the entire spectrum of the converter’s power output range. In order to achieve this effect, it is helpful to use a number of parallel converters to achieve a fixed maximum output power (see e.g. [8]). The idea is based on tracking the point of maximum efficiency throughout a broad spectrum of possible power demands and sequentially switching on the desired number of parallel converters. This also yields the benefit of increased reliability since a failure of a single converter can be compensated by the loss of efficiency, while still maintaining the operation of the entire system without any loss of functionality. Defending on the sensitivity of the connected electrical loads, additional capacitors may be used to stabilize the voltage at certain points within the power net. Beyond that, a reliable and robust way for communication within the ring structure has to be found, while keeping the amount of messages sent throughout the network to a minimum. This could either be achieved by the utilization of data over power encoders, or data cables which can then be embedded inside the rail structures. By implication, this would also allow the integration of another level of communication in the manner of a data over power line for the purpose of diagnoses. In addition to this component, additional units sustain the desired behavior of the power net, such as the below mentioned ISSU. Fig. 5: ISDD utilizing power rail on 48 V levels 180 Ring Structures in Automotive Power Nets: Idea and Implementation 4 Integrated Smart Switching Unit (ISSU) In order to ensure a smooth switching behavior between the above mentioned hubs embedded in the ring, another addition has to be made to the proposed network structure. The Integrated Smart Switching Unit (ISSU) is a key component to the behavior of the ring power net introduced in chapter 2. The ISSU itself is connected to at least two ISDDs (see Fig. 6) and also to the most safety-relevant electrical loads (e.g. sensors for autonomous driving or electric power steering supplies). In case of an occurring failure, the ISSU is cut from one of the ISDDs, while switching to the secondary battery located in the remaining ISDD. In any case of a switching event, it has to be ensured, that each connected ISDD is capable of delivering the needed output power to all critical components connected to the ISSU. While the main goal is to ensure a quick and smooth switching behavior between two power supplies represented by the ISDDs, additional features such as a short term, high power density storage (e.g. an additional capacitor) can be added to counteract a possible voltage drop in the event of a demanding load requirement. During the design process, it is the goal to design the system in a way so that the communication between units is kept to the necessary minimum. This rule is regarded as a key rule for the management design guidelines, which are explained in the following paragraphs. Certain variants of control strategies are possible in order to achieve this goal e.g. a completely autonomous switching behavior, whereby all switching commands are automatically generated by an analog circuit without the necessity of a microcontroller. Fig. 6: Connection of Safety-Critical Loads Using ISSU [3] 181 Ring Structures in Automotive Power Nets: Idea and Implementation 5 Management And System Behavior As the basic physical setup of the components as well as the overall structure is laid out in the previous chapters, a suitable and efficient way of managing energy and power flow within the system has to be found. The following thoughts and strategies are adapted from the viable system model (VSM), which was developed by Stafford Beer [9] and is originally based on the cybernetic approaches introduced by Norbert Wiener [10]. As a control strategy, which follows the cybernetic approach, has not yet been developed for ring power nets, the pillars of this approach will be explained briefly. Since mostly undervoltages have been in the focus of considerations throughout the last years (see e.g. [11]), the design also has to be extended to other possible causes of instability in automotive power nets such as overvoltages. The goal is, to design a system which follows the basic rules of cybernetics, primarily the separation of function and management. This directly results in the separation of operational units, such as a fuse or converter and management entities such as their corresponding control units. One can conclude, that a control unit inside a car therefore can act independently, as long as no other higher management entity dictates new guidelines, which have to be followed. A reduction of complexity can be achieved by the introduction of levels of hierarchy and decentralization. This provides the opportunity of simplification inside systems with a number of different participants, allowing every unit to act within prespecified boundaries. The hierarchy levels will have to be deliberated for the system layout proposed in chapter 2. One possible basic hierarchy structure, showing the individual VSMs for different layers, is displayed in Fig. 7. In addition, the general rule of subsidiarity and autonomy keeps each subsystem as independent as possible, while limiting the communication to a necessary minimum and therefore minimizing transfer times and controller delays. Moreover, the design of each hierarchy level should inherit the same structural layout, as it is introduced in [9]. Fig. 7: Possible structure of hierarchy levels for a ring power net 182 Ring Structures in Automotive Power Nets: Idea and Implementation 6 Conclusion A new design proposal for the structural layout of automotive power nets is proposed in this paper. The well-known linear topology of power nets is adapted to a ring structured power net, which yields numerous technical advantages in contrast to the conventional solutions. Since the number of auxiliary as well as safety-critical loads are expected to increase due to recent developments in autonomous driving, ring structured power nets offer certain advantages and help to ensure a stable and reliable energy transmission in case of occurring failures. The key components of the proposed system can be stated as: 1. Ring structured main supply line 2. Hubs (ISDDs) 3. Switching Components (ISSUs) Especially in case of the ISDDs and ISSUs, numerous different manifestations (e.g. different energy storage technologies, different converter topologies etc.) have to be evaluated in terms of efficiency and reliability. Those considerations also heavily depend on the specific circuit layout of the connected components and their characteristics. In terms of voltage stability as well as EMC, approaches for voltage stabilization as well as the influence of a possible neutral point shift has to be examined. Two different versions of ISDDs in particular have been introduced, which both yield a number of advantages and offer easy upgradability and flexibility. This also enables the possibility of hardware changes or upgrades in a later point of development or even after a vehicle is shipped to a customer. Besides, since a redundant energy transmission is guaranteed by the system’s layout itself, costly additions to the electrical loads such as dual power inputs etc. can be unified or neglected for a large number of loads, which as such simplifies development of individual components. Furthermore, a possible approach is discussed to handle the system complexity by the introduction of the viable system model and its adaption to the described system. Since the runtimes of messages increase with the number of system entities, the fundamental idea of extensive autonomy of subsystems provides the basis for a possible energy management system. The next steps include further elaboration based on the cybernetic approach as well as a clear definition of interfaces between each layer of the energy management. The resulting conclusions will be implemented and verified both in simulation and in a second step on a test bench. It can be concluded, that a ring-structured power net offers possible solutions to arising challenges in the automotive field, such as providing a reliable and fault-tolerant energy supply under all circumstances. Nevertheless, an intelligent way of power and energy management structure has to be adapted to the new system design, in order to provide full functionality while ensuring a smooth running behavior in case of occurring failures. 183 Ring Structures in Automotive Power Nets: Idea and Implementation References [1] DeMattia, N.; November 2018; “BMW to debut 48-volt electrical systems by 2020”; https: / / www.bmwblog.com/ 2018/ 11/ 16/ bmw-to-debut-48-volt-electricalsystems-by-2020/ [Accessed: Feb. 15, 2019] [2] Bilo, J. et al.; December 2015; „ZVEI: 48-Volt-Bordnetz - Schlüsseltechnologie auf dem Weg zur Elektromobilität“ [3] Tippe, L. et al.; November 2018; “Introduction of Ring Structures in Future Car Generation’s Electrical Systems”; ESARS-ITEC 2018 [4] Bosch Mobility Solutions; 2017; “Neue Lenksysteme für die Mobilität von morgen - Elektrolenkung Servolectric® Fail-operational“; https: / / www.bosch-mobility-solutions.de/ media/ global/ products-and-services/ passenger-cars-and-light-commercial-vehicles/ steering-systems/ servolectric-steering-systems/ systemmappe_lenksysteme_pkw.pdf [Accessed: Mar. 15, 2019] [5] Wadi, M. et al; 2017; "Comparison between open-ring and closed-ring grids reliability"; 4th International Conference on Electrical and Electronic Engineering (ICEEE) [6] Yue, D. et al; 2010; “A fault-tolerant safety communication model based on dual ring bus”; Proceedings of 2010 IEEE/ ASME International Conference on Mechatronic and Embedded Systems and Applications [7] Taube, J. et al; 2019; “Concept for a 48V / 12V Power Rail with Integrated Power Converter and ECUs”; Konferenzband der EEHE - Elektrik / Elektronik in Hybrid- und Elektrofahrzeugen 2019 [8] Winter, M. et al; 2016; “Using the Viable System Model to control a system of distributed DC/ DC converters”; 2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC) [9] Beer, S.; 1966; „Decision and Control: The Meaning of Operational Research and Management Cybernetics“; ISBN 978-0-471-94838-4 [10] Wiener, N.; 1961; “Cybernetics: Or Control and Communication in the Animal and the Machine“; ISBN 978-0-262-73009-9 [11] Giovanazzi, T.; 2014; “Prädiktives Leistungsmanagement in Fahrzeugbordnetzen“; ISBN 978-3-658-05011-5 184 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Christian Sültrop, Thomas Lang, Franz Rohlfs, Jan Helfrich, Pedro Bossay de Almeida Nogueira, Helga Weber, Uwe Prüfer, Dominik Bergmann, Jürgen Gebert Abstract Automated vehicles must be moved to a safe stop position in the event of a fault which affects longitudinal or lateral control, environment perception or trajectory planning. As of SAE automation level 3 and above, the driver is not available as a fallback. The fallback must therefore be implemented technically. The on-board power supply system (PSS) is therefore required to be fault tolerant, since an adequate power supply of the safety-relevant components must always be ensured. This paper describes how PSS architectures that meet this requirement were developed, evaluated and selected in the HiBord research project based on the concept of intelligent network nodes. The premise was to make use of the implicit redundancies in multi-voltage vehicle electric systems with little additional effort. This creates a PSS that remains fail-operational in the event of a fault until the end of an emergency stop maneuver. Finally, the ongoing implementation of the developed nodes is briefly discussed and an outlook on the demonstration and evaluation environment currently being established is given. Kurzfassung Automatisierte Fahrzeuge müssen bei Fehlern, die die für die Längs- und Querführung, die Umfelderkennung oder die Trajektorienplanung benötigten Komponenten betreffen, in eine sichere Halteposition gesteuert werden. Ab SAEbzw. VDA- Automatisierungsstufe 3 steht der Fahrer nicht als Rückfallebene zur Verfügung. Die Rückfallebene muss daher technisch realisiert werden. Für das Energiebordnetz (EBN) ergibt sich daraus die Anforderung der Fehlertoleranz, da dieses auch im Fehlerfall eine ausreichende Energieversorgung der sicherheitsrelevanten Komponenten sicherstellen muss. Die vorliegende Arbeit beschreibt, wie ausgehend vom Konzept der intelligenten Netzknoten im Forschungsprojekt HiBord EBN-Architekturen entwickelt, bewertet und ausgewählt wurden, die dieser Anforderung gerecht werden. Prämisse war dabei, mit 185 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen geringem Zusatzaufwand die im Mehrspannungsbordnetz implizit vorhandenen Redundanzen nutzbar zu machen. Damit wird ein EBN geschaffen, das im Fehlerfall bis zum Abschluss eines Nothalt-Manövers fail-operational bleibt. Schließlich wird kurz auf die laufende Realisierung der entwickelten Knoten eingegangen und ein Ausblick auf die im Aufbau befindliche Demonstrations- und Evaluationsumgebung gegeben. 1 Einleitung In den Stufen 3 und 4 des automatisierten Fahrens führt das Fahrzeug Längs- und Querführung sowie Navigation und Trajektorienplanung selbstständig aus. Der Fahrer fällt als unabhängige Rückfallebene aus, die daher technisch realisiert werden muss. Die automatisierte Fahrfunktion muss das Fahrzeug im Fehlerfall in eine sichere Halteposition steuern können. Die dazu nötige Fehlertoleranz erfordert die redundante Auslegung von Sensoren, Aktuatoren und Steuergeräten. Dazu gehört zwingend auch, die Versorgung dieser Komponenten durch das Energiebordnetz (EBN) bereitzustellen. Das EBN muss dafür ebenfalls fehlertolerant ausgelegt werden. Fehler im EBN gefährden die Energieversorgung der Komponenten, beispielsweise durch einen Kurzschluss und Spannungseinbruch. Je nach Fehlerart muss die akute Fehlerbehandlung innerhalb einer kurzen Zeitspanne ohne übergeordnete Koordination erfolgen. Weniger zeitkritische Fehler erlauben eine abgestimmte Fehlerbehandlung, die von einer übergeordneten Einheit koordiniert werden kann. In jedem Fall ist nach der Fehlerisolation eine schnelle Stabilisierung des Zustands des Gesamtsystems erforderlich. Die Analyse möglicher Fehler und Risiken im Rahmen von Hazard Analysis und Risk Assessment sowie die Erarbeitung eines Sicherheitskonzepts sind Aufgaben, die zuerst auf der Ebene des Gesamtfahrzeugs durchgeführt werden müssen [1]. Basierend auf einer Gefahren- und Risikoanalyse werden dabei Automotive Safety Integrity Levels (ASIL) für die beteiligten Funktionen erarbeitet. Für das automatisierte Fahren sind insbesondere die Funktionen Lenken und Bremsen sowie Umfelderkennung, Trajektorienplanung und -regelung von zentraler Bedeutung. Da mit einer kurzfristigen Fehlerreaktion des Fahrers nicht gerechnet werden kann, sind insbesondere Fehler in den Lenk- und Bremsfunktionen in vielen Fahrsituationen kritisch, da diese somit durch den Fahrer nicht beherrschbar sind und zu hohen Schäden führen können. Daraus folgt, dass den Lenk- und Bremsfunktionen regelmäßig hohe ASIL zugewiesen werden. An das ASIL ist eine Zielvorgabe für die FIT-Rate geknüpft. Ein entscheidender Schritt diese zu erreichen ist die zweifach redundante Auslegung der zugehörigen Aktuatoren. Das EBN versorgt die sicherheitsrelevanten Aktuatoren mit elektrischer Energie. Anforderungen an Ausfallrate und Verfügbarkeit werden daher an dieses vererbt. Insbesondere dürfen einfache Fehler im EBN nicht zu einem Ausfall der Energieversorgung beider redundanter Aktuatoren führen. Das Projekt HiBord erarbeitet und realisiert geeignete EBN-Topologien auf Basis intelligenter Netzknoten. Diese Arbeit beschäftigt sich mit den Hintergründen und den angewandten Entwurfs- und Auswahlmethoden. 186 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen 2 Referenzszenario In der Norm ISO26262: 2018 [2] ist der Sicherheitsmechanismus mit Notbetrieb (Safety Mechanism with Emergency Operation) eines fehlertoleranten Systems beschrieben. Gemäß der Norm soll die Notbetriebsdauer EOTI (Emergency Operation Time Interval) mindestens so lang bemessen sein, dass das System in einen sicheren Zustand übergehen kann. Dieser Notbetrieb ist für eine variable Zeitspanne zulässig, solange nicht mit einem unangemessenen Risiko (z.B. Totalausfall) zu rechnen ist. Die Norm spricht von einem EOTTI (Emergency Operation Tolerance Time Interval). Aufgrund dieser Randbedingungen muss der Übergang in den sicheren Zustand relativ rasch erfolgen. Der Notbetrieb kann entweder in einem fail-safe Szenario (z.B. Nothalt) enden, welches einen sicheren Zustand darstellt oder in einen fehleroperativen („failoperational“) Betriebszustand überführt werden. Die Zeitspanne, für die das EBN Energie bereitstellen muss, steigt mit der Automatisierungsstufe und dem Wunsch nach Fehleroperabilität an, wie in Abbildung 1 dargestellt. In niedrigen Automatisierungsstufen kann der Fahrer als Rückfallebene innerhalb kurzer Zeitspannen aktiviert werden. In den höheren Automatisierungsstufen können die sicherheitsrelevanten Systeme nicht auf die mechanische Aktivierung durch den Fahrer zurückfallen. Um die Anforderungen fail-operational erfüllen zu können steigt hier die Zeit, in der die automatisierte Fahrfunktion trotz EBN-Fehler garantiert werden muss, bis in den Bereich mehrerer Minuten an. Bild 1: Zeitliche Anforderungen an den Fail-Operational-Betrieb in Abhängigkeit der Automatisierungsstufe. In HiBord wird als Referenzszenario ein Fahrzeug betrachtet, das von einem vollautomatischen Autobahnpiloten (Level 4) gesteuert wird. Für dieses Szenario kann ein Sicherheitsmechanismus mit Notbetrieb wie folgt beschrieben werden: Zunächst bewegt sich das Fahrzeug, wie in Abbildung 2 dargestellt, mit 130 km/ h auf der linken von drei Fahrspuren, als ein Fehler im EBN eintritt. Das Fahrzeug wird nun in einen sicheren Zustand überführt. Als sicherer Zustand wird in diesem Szenario das Anhalten auf dem Standstreifen definiert. Dazu muss das Fahrzeug unter Berücksichtigung des umgebenden Verkehrs abbremsen und über die mittlere und rechte Spur auf den Standstreifen wechseln und dort zum Stehen kommen. 187 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Bild 2: HiBord-Anhaltemanöver auf dem Standstreifen mit drei Spurwechseln in Gegenüberstellung zu den Phasen des Sicherheitsmechanismus mit Notbetrieb nach [2]. Es wird vorausgesetzt, dass die sicherheitsrelevanten Systeme für Lenkung, Bremse, Trajektorienplanung usw. zweifach redundant ausgelegt sind. Weiterhin wird angenommen, dass diese Systeme jeweils fehlersicher ausgeführt sind und somit im Fehlerfall keine negative Rückwirkung auf das Gesamtsystem haben. Um diesen Sicherheitsmechanismus umsetzen zu können, muss das EBN insofern fehlertolerant sein, dass es trotz des aufgetretenen Fehlers zumindest jeweils eines der redundant vorhandenen sicherheitsrelevanten Systeme mit Energie versorgen kann, bis der sichere Zustand des Gesamtfahrzeugs erreicht ist. Der Vorgang als Ganzes stellt also ein fail-safe-Szenario dar, doch zu seiner Umsetzung muss das EBN für eine begrenzte Zeit fail-operational sein. Um energetische Aussagen zum gezeigten Szenario und weiteren Szenarien machen zu können, wurde ein Referenzprofil erarbeitet, das auf Strommessungen im realen Fahrversuch basiert. Die Messdaten stammen aus standardisierten Fahrmanövern. Mittels der in Abbildung 3 dargestellten Arbeitsschritte wurden diese zu einem Referenzlastfall zusammengestellt, der einen Worst-Case des beschriebenen Anhaltemanövers darstellt. Dazu wurden die während der Fahrmanöver gemessenen Lastströme zunächst in einzelne Lastfälle (Grundlast, periodische Lasten, manöverabhängige Lasten wie Lenken und Bremsen) zerlegt und anschließend mittels eines Editors neu zusammengestellt. 188 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Bild 3: Arbeitsschritte zum Erzeugen von Referenz-Lastprofilen aus im Fahrversuch gemessenen Stromverläufen. Die resultierenden Stromverläufe des Referenzlastfalls werden sowohl für Auslegung und Simulation als auch für die Evaluation der realisierten Systeme mittels Emulation eingesetzt. 3 Lösungsansatz Ein naheliegender Ansatz, eine unabhängige Energieversorgung der sicherheitsrelevanten Komponenten zu erreichen besteht darin, das EBN doppelt redundant auszulegen [3]. Dieser Ansatz führt jedoch schnell zu einem EBN, in dem Komponenten wie DC/ DC-Wandler, Batterien, Energieverteiler, Leitungen usw. in deutlich größerer Zahl vorhanden sind, als für die eigentlich Funktion des EBN erforderlich - woraus großer Bauraumbedarf, erhöhtes Gewicht und schließlich erhöhte Kosten folgen. Es ist daher ein ganzheitlicher Ansatz erforderlich, der diese Restriktionen in Betracht zieht. Ein wichtiges Ziel besteht dabei darin, dem EBN möglichst wenige zusätzliche Komponenten hinzuzufügen und dennoch dessen erforderliche Verfügbarkeit sicherzustellen. Aufgrund der Transition zur Elektromobilität, der Hybridisierung konventioneller verbrennungsmotorisch angetriebener Fahrzeuge und der breiten Einführung des 48V- Netzes werden zukünftige Fahrzeuge über zwei (HV-12V bzw. 48V-12V) oder sogar drei (HV-48V-12V) Spannungsebenen verfügen [4]. Die Beweggründe für die Einführung der Spannungsebenen Hochvolt (HV) und 48V zusätzlich zum etablierten 12V- Standard liegen in der effizienten Energiebereitstellung für die elektrischen Antriebsmaschinen sowie für große Energieverbraucher in den Domänen Fahrwerk und Komfort und außerdem in der Möglichkeit zur Energierückgewinnung durch Rekuperation. Hohe Transitionskosten für den Wechsel von der 12V-Ebene auf die 48V-Ebene sorgen dafür, dass dieser Aufwand voraussichtlich nur für diejenigen Komponenten durchgeführt wird, für die sich daraus ein klarer Nutzen ergibt - sei es aus Kostensicht durch Einsparungen im Halbleiterbereich aufgrund nun kleinerer zu schaltende Ströme, aus Sicht Rekuperatiosfähigkeit oder aus Funktionssicht wegen der ca. viermal höheren verfügbaren Leistung des 48V-EBN gegenüber der 12V-Ebene. Wichtige Schlüssel für ein effizient gestaltetes hochverfügbares EBN sind die Wahl einer optimalen Energieverteilungsstruktur, Topologie genannt, und die gezielte Nutzung und Funktionserweiterung der vorhandenen Komponenten. Dabei gilt es, die im 189 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Mehrspannungsbordnetz implizit vorhandenen Redundanzen zu nutzen und möglichst wenige strukturelle Redundanzen zusätzlich hinzuzufügen. Das Design des EBN muss folglich eine Vielzahl von Anforderungen und Randbedingungen berücksichtigen, die aus Entscheidungen auf der Gesamtfahrzeugebene resultieren. Dazu gehören  die Automatisierungslevel und -modi  die Ergebnisse der Gefahren- und Risikoanalyse  das Sicherheitskonzept auf Fahrzeugebene, heruntergebrochen auf das EBN  die Fahrzeugarchitektur, die räumliche Anordnung der Komponenten und die verfügbaren Bauräume  die Anforderungen, die die kritischen Komponenten an die Energieversorgung stellen  der Energie- und Leistungsbedarf in den möglichen Notbetriebsszenarien Das hochverfügbare EBN sollte modular sein, so dass es in ein Plattformkonzept mit unterschiedlichen Automatisierungsgraden, Fahrzeugklassen, Antriebskonfigurationen, Ausstattungsvarianten integriert werden kann. Aus den dargestellten Einflussgrößen wird ersichtlich, dass eine universelle Lösung eine Vielzahl von Einflussgrößen berücksichtigen muss und daher sehr komplex sein kann. Um den skizzierten Ansatz prototypisch realisieren zu können und die Integration eines Demonstrationssystems in ein bestehendes Elektroauto (Battery Electric Vehicle, BEV) zu ermöglichen, wurde daher der Betrachtungsrahmen eingeschränkt: Es wird ein BEV mit einem HV-Antriebsstrang, einem 48V-Netz und einem 12V-Netz betrachtet. Sämtliche sicherheitsrelevante Komponenten werden aus dem 12V-EBN versorgt. 4 Sicherheitsdomänen Sicherheitsdomänen spiegeln die Aufteilung und Strukturierung von Fahrzeugfunktionen wieder, um das Gesamtsystem abzusichern. Dabei gibt es zwei Betrachtungsebenen, die der Anwendungsorganisation, bei der die Funktionen ihrer Aufgabe entsprechend geklustert sind und die der EBN-Struktur. Bordnetze sind historisch gewachsen, durch die Zunahme an Komponenten, wobei die Kommunikation „wild“ zwischen den einzelnen Komponenten über den Bus erfolgt, da keine ausreichende Anwendungsorganisation erfolgte. Durch die Entwicklung über Assistenzsysteme und autonomen Teilfunktionen hin zu vollautonomen Fahrzeugen, steigt die Zahl der Komponenten und die Arbeit für das Kontrollsystem, was eine Anwendungsorganisation unabdingbar macht, bei der der Grad der Vernetzung über zentrale Komponenten erfolgt (siehe Abbildung 4). 190 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Bild 4: Links: Historisch gewachsene Organisations- und Netzstruktur in Fahrzeugen. Rechts: Strukturierte Anwendungsorganisation und Bordnetzstrukturen für redundante System von autonomen Fahrzeugen. Durch die direkte Vernetzung aller Komponenten muss die Absicherung des Systems erhöht werden, da nicht nur eine einzelne Komponente, sondern das komplette System ausfallen kann. Dies wird dadurch erreicht, dass für autonome Systeme die elektronischen Kontrollsysteme zentralisiert werden und die zugeordneten Anwendungen entsprechend organisiert werden. Diese Struktur ist notwendig um der Komplexität, den wirtschaftlichen Aspekten (Gewicht/ Kosten), der Sicherheit und der Flexibilität gerecht zu werden. Es ergeben sich dabei drei Ebenen, die des Gesamtfahrzeuges, eine Funktionsebene in der einzelne Funktionsgruppen ihre Aufgaben erfüllen (Sensoren für die Umgebungserfassung, Kontrollgruppen wie Lenkung, Bremse, Licht, …) und die der einzelnen Komponenten, welche spezifische Aufgaben haben. Um dem Aspekt der Sicherheit gerecht zu werden, müssen die Bordnetzstrukturen entsprechend ausgelegt sein für ein „Fail-Operational“ System, bei dem Fehler in Subnetzen keine Auswirkungen auf den Rest des Systems haben. Ein Bordnetz besteht dafür aus verschiedenen Komponentenklassen:  Energie: Fungiert als Quelle für die Energie aller Systeme, welche in den einzelnen Subnetzen lokalisiert sind.  Transfer: Verbindet die einzelnen Subnetze. Dies ermöglicht die Kontrolle des Gesamtsystems, um Energie zu transferieren und Fehler zu separieren.  Subnetz: Beinhalten alle Komponenten des Fahrzeugs, dabei erfolgt die Lokalisierung in verschiedene Netze um die Redundanz von sicherheitsrelevanten Funktionen zu gewährleisten. Durch geeignete Strukturierung wird damit ein fehlerresistentes System ermöglicht. Energie Transfer Subnetze Kompo‐ nenten Funktion Fahrzeug Konventionelle- gewachs ene-Struktur VDA-3+-Struktur-(Funktions‐-und-Domänenorganisiert) Kamera Lidar Strom SoC ... Lenkung-1 Bremse-1 Lenkung-2 Bremse-2 … Zentrale-CPU Sensoren Kontrolle Rest Speicher Generator Energie Leistung (Rekuperatio n) Wandler Sicherungen Trennelemente Lenkung-1 Bremse-1 Kamera ... Verteiler Lenkung-2 Bremse-2 Lidar ... Verteiler Anwendungsorganisation Bordnetzstruktur Anwendungsorganisation-+ Bordnetzstruktur 191 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen 5 Intelligente Netzknoten Das EBN kann als Graph abstrahiert werden, dessen Kanten die Leitungen und dessen Knoten Stromverteiler, DC/ DC-Wandler, Speicher, Relais, Sicherungen usw. bilden. Diese Komponenten können als abstrakte Verteiler-Knoten betrachtet werden [5], die über die folgenden Subfunktionen verfügen:  Energiefluss schalten  Fehler im Knoten erkennen  Fehler auf Leitungen erkennen  vordefiniertes Verhalten  mit der Umgebung Kommunizieren - In Abbildung 5 wurde dieses Modell um die Subfunktionen Energie wandeln zu einem Wandler-Knoten bzw. Energie speichern zu einem Speicher-Knoten erweitert. Die abstrakt-funktionale Sicht auf die Verteilknoten im EBN ermöglicht eine Betrachtung, die zunächst losgelöst ist von konkreten Implementierungen und den dahinterstehenden Technologien. Bild 5: Energie-Verteilknoten nach [5] (o.) und Erweiterungen zum Energie- Speicherknoten (u.l.) und Energie-Wandlerknoten (u.r.). Der Knoten kann mittels seines vordefinierten Verhaltens auf Ereignisse an seinen Eingängen reagieren. Die Realisierungsmöglichkeit eines Knoten reicht dabei von einer einfachen Schmelzsicherung, deren Reaktion auf hohe Ströme durch die Sicherungskennlinie beschrieben ist, bis hin zu vollelektronischen Systemen mit eigenem Steuergerät, Leistungshalbleiterschaltern und komplexer Sensorik. Ein solcher Energie-speichern-+-verteilen Eingänge- überprüfen Fehler-erkennen Auf-Fehler- reagieren Mit-Umgebung- kommunizieren Energie-schalten Energie- speichern Energie-wandeln-+-verteilen Eingänge- überprüfen Fehler-erkennen Auf-Fehler- reagieren Mit-Umgebung- kommunizieren Energie-schalten Energie- wandeln 192 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen intelligenter EBN-Knoten verfügt über die nötigen Voraussetzungen, um Fehlerzustände an seinen Ein-und Ausgängen zu detektieren und darauf angemessen zu reagieren. Ein intelligenter EBN-Knoten kann sowohl eigenständige Komponente sein, als auch in vorhandenen Komponenten aufgehen. Durch die zunehmende Elektrifizierung der Fahrzeuge zeigt sich auch im EBN ein Trend zur elektronischen Realisierungen von Energiewandlung und Verteilung. Dazu gehören DC/ DC-Wandler, elektronische Stromverteiler und parametrierbar elektronische Sicherungen [6, 7]. Die Funktionen dieser Komponenten erfordern neben elektronische Leistungsschalter als Aktuatoren auch Messwerterfassungseinheiten, Controller zur Datenverarbeitung und Schnittstellen zur Kommunikation. All diese Komponenten können also als Realisierungen des abstrakten intelligenten Energie-Verteilknotens betrachtet werden. Die abstrahierte Sicht ermöglicht darüber hinaus einen neuen Blick auf die intelligenten Netzknoten, ohne auf die bestehenden Komponenten eingeschränkt zu sein. 6 Fehlerreaktion Unterschiedliche Fehler im EBN erfordern unterschiedliche Fehlerreaktionszeiten. In Abbildung 6 sind diese für die wichtigsten Bordnetzfehler dargestellt. Den Fehlerreaktionszeiten werden die Signallaufzeiten der gängigen Automotive- Bussysteme gegenübergestellt. Bild 6: Gegenüberstellung der Fehlertoleranzeiten typischer EBN-Fehler, der Umlaufzeiten typischer Bussysteme und der Zeitkonstanten von EBN-Komponenten. Der Vergleich von Fehlerreaktions- und Signallaufzeiten zeigt, dass ein intelligenter Netzknoten, der einen Fehler wie etwa Kurzschluss oder Lichtbogen detektiert, seine Fehlerreaktion nicht über die gängigen Busschnittstellen mit anderen Knoten oder einer übergeordneten Steuerung (ÜGS) abstimmen kann. Die Fehlerreaktion muss CAN 1-µs 1-ms 1-s 1-ks Ethernet E‐Fuse-(HW) E‐Fuse-(SW) Schmelz sicherung Kurzschluss Lichtb ogen Spannun gs‐ unterbrechu ng Spannun gs‐ einbruch Lastspitze Last‐ Unterbrechu ng Langzeit‐ Unterspannu ng Indu ktiver- Einsch altstrom Langzeit‐Überspann ung Generator DC/ DC Sprungantwort 193 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen daher in diesen Fällen reflexartig, d.h. schnell, autonom und lokal, durch den erkennenden Knoten durchgeführt werden. Die reflexhafte Fehlerreaktion erfordert, dass die Fehlerdetektion und die Fehlerisolation von den Energieverteilknoten selbständig ausgeführt werden können. Diese müssen dazu mit entsprechenden Sensoren und Algorithmen ausgestattet sein. Sowohl die Fehlerreaktionen der EBN-Knoten als auch die Topologie des EBN müssen so gestaltet sein, dass nicht untereinander abgestimmte Reflexe der EBN-Knoten nicht zu einem unbestimmten Zustand des EBN führen. Insbesondere muss sichergestellt sein, dass es aufgrund der Reflexe nicht zur Abschaltung der Energieversorgung einer redundanten sicherheitsrelevanten Komponente kommt, die von dem vorliegenden Fehler im EBN nicht betroffen ist. Für andere Fehlerarten kann die Fehlerreaktionszeit deutlich länger sein, so dass diese Fehler zunächst an die ÜGS gemeldet werden können. Die ÜGS kann dann eine abgewogene Entscheidung auf Systemebene treffen und etwa Abschaltbefehle für einzelne Lasten oder ganze Netzsegmente an die Energieverteilknoten oder an die Lasten selbst senden. In Abbildung 7 sind die beiden möglichen Pfade Reflex und Entscheidung mit ihren unterschiedlichen zeitlichen Abläufen gegenübergestellt. Ziel ist in beiden Fällen eine Wiederhergestellte Energieversorgung, die den Notbetrieb des EBN und damit die Energieversorgung mindestens einer redundanten sicherheitsrelevanten Komponente gewährleistet, bis das Fahrzeug in den sicheren Zustand gebracht wurde. Bild 7: Reflex- und Entscheidungspfad zur wiederhergestellten Energieversorgung. 7 Architekturentwurf und -bewertung Auf der Grundlage des in Abschnitt 2 beschriebenen Szenarios und den in den Abschnitten 3 bis 6 vorgestellten Überlegungen wurden mögliche Lösungen für hochverfügbare EBN generiert, bewertet und zur prototypischen Realisierung in einem Demonstrator ausgewählt, ähnlich wie dies ausführlicher in [8] und [9] beschrieben wird. Das in Abbildung 8 dargestellte Vorgehensmodell zeigt die erforderlichen Schritte von der Gefahren- und Risikoanalyse auf der Fahrzeugebene, der Beschreibung Fehler‐ erkennung Fehler t Fehler‐ isolation Informationsaustaus ch-mit-ÜGS Fehler‐ erkennung Synchroni sation-mit-ÜGS Fehler‐ isolation Wieder‐ herstellung Wieder‐ herstellung Fehler-isoli ert Fehler-isoli ert EBN-stabilisiert Reflex abgestimmte- Entscheidung 194 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen technischer Sicherheitsanforderungen auf EBN-Ebene bis zu Spezifikation und Implementierung auf Komponentenebene. Bild 8: Top-Down-Ableitung der intelligenten EBN-Knoten aus der Gesamtfahrzeugebene. Die in HiBord angewandte Vorgehensweise ist in Abbildung 9 dargestellt. Ausgehend von den abstrakten Beschreibungen der intelligenten EBN-Knoten wurde zunächst eine Diskussion geführt, die von technischen Implementierungsdetails auf der Funktions- und Topologieebene weitgehend losgelöst war. Die in dieser Phase erarbeiteten Topologievorschläge wurden dann in technische Anforderungen überführt. Zur Filterung der für das Projekt am besten geeigneten Topologien wurde eine Nutzwertanalyse [10] durchgeführt, in der die Topologien hinsichtlich projektspezifisch gewichteter quantitativer und qualitativer Kriterien bewertet wurden. Zu den untersuchten Kriterien gehören  Single Points of Failure  Anzahl möglicher Versorgungspfade  Minimalschnitte  Entkopplung der Teil-EBN  Performanz der Redundanz  Komplexität  Umsetzungsmöglichkeiten  Modularisierung (Skalierbarkeit u. Rückwärtskompatibilität)  Anzahl gleicher / unterschiedlicher EBN-Knoten 195 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Bild 9: Angewandte Vorgehensweise zum Auffinden geeigneter Topologien. Zur Unterstützung eines vollständigen, transparenten und nachvollziehbaren Bewertungs- und Auswahlprozesses wurde das Software-Werkzeug HiTool entwickelt. In dessen Datenmodell werden die intelligenten EBN-Knoten und weitere EBN- Komponenten abgebildet und zu den zu untersuchenden Architekturen verknüpft. Das Tool enthält dazu eine Komponentendatenbank, in der FIT-Raten, Leistungen, Speicherkapazitäten, Spannungslagen, Relevanz für das hochautomatisierte Fahren u.a. hinterlegt werden können. Aus einer im Tool erstellten EBN-Skizze können die Anzahl der möglichen Versorgungspfade eines Verbrauchers, die Anzahl der benötigten EBN-Knoten und weitere Metriken berechnet werden. Dabei werden nicht nur externe, sondern auch interne Verbindungen der EBN-Knoten berücksichtigt. 8 Ausgewählte Architekturen Mittels der beschriebenen Methode wurden zwei Architekturen ausgewählt, welche einen hohen Nutzwert oder einen hohen Nutzwert kombiniert mit einem hohen Neuheitsgrad aufweisen. Die Auswahl wurde auf solche Architekturen beschränkt, anhand derer sich das Konzept der intelligenten Knoten in der Praxis demonstrieren und evaluieren lässt. Dazu wird das bestehende EBN eines BEV modifiziert und Demonstratoren der intelligenten EBN-Knoten eingebracht. Die Möglichkeiten der Integration in ein Versuchsfahrzeug sind insofern eingeschränkt, als derzeit weder BEV mit Dreispannungsbordnetz noch mit redundanten X-by-Wire-Brems- und Lenksystemen am Markt verfügbar sind. Beide ausgewählten Architekturen verfügen über ein 48V-Segment, das über einen DC/ DC-Wandler aus der HV-Batterie versorgt und durch eine LiIon-Batterie gestützt wird. In diesem Segment befinden sich keine sicherheitsrelevanten Verbraucher, die eine redundante Energieversorgung erfordern würden. Eine ÜGS ist für die auf globaler Ebene angesiedelten Fehlererkennungs- und Reaktionsmechanismen verantwortlich. Reflexhafte Fehlerreaktionen werden in beiden Architekturen von einem Trennelement (TE), d.h. einem intelligenten Schalter, einerseits und einem DC/ DC-Wandler andererseits umgesetzt. 196 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen 8.1 Architektur A Die erste Architektur, dargestellt in Abbildung 10, wurde in ähnlicher Form schon in anderen Veröffentlichungen vorgestellt, etwa [11, 12]. Das EBN kann ist hier in die drei Spannungsebenen HV, 48V und 12V sowie in die vier Segmente HV, 48V, 12V Primär (1st Branch) und 12V Sekundär (2nd Branch) unterteilt. Bild 10: Ausgewählte Architektur A eines fehlertoleranten EBN erstellt mit dem im Projekt entwickelten Software-Werkzeug HiTool. Die 48V-Ebene wird über einen DC/ DC-Wandler aus der HV-Ebene versorgt. Die 12V-Ebene wird über einen weiteren DC/ DC-Wandler aus der 48V-Ebene versorgt. Die 12V-Ebene ist in ein primäres und ein sekundäres Segment aufgeteilt. Das primäre Segment der 12V-Ebene entspricht weitgehend einem 12V-EBN nach Stand der Technik. Es enthält sowohl Komfort-Verbraucher als auch die sicherheitskritischen Lenk- und Bremssysteme EPS und DSC sowie das sicherheitskritische System V4. Im sekundären Segment befinden sich ein redundantes Lenk- und Bremssystem EPS‘ und DSC‘ sowie ein redundantes System V4‘. Die Platzierung der 12V-Stützbatterie im sekundären 12V-Segment sorgt dafür, dass beide 12V-Segmente über eine eigene Energiequelle verfügen. Für das sekundäre Segment ist dies die 12V-Stützbatterie, während das primäre Segment über den DC/ DC-Wandler mittelbar aus der HV-Traktionsbatterie gespeist wird. An der Grenze zwischen dem primären und dem sekundären Segment befindet sich ein intelligenter Knoten, der die Funktion eines TE realisiert. Im Fehlerfall trennt das TE die elektrische Verbindung zwischen den beiden Netzsegmenten. Damit zerfällt die 12V-Ebene in zwei Segmente. Nur in einem der beiden Segmente ist anschließend der Fehler weiter wirksam. 197 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Dem TE kommt somit die entscheidende Funktion zu, Fehler in EBN zu detektieren und eine Trennung herbeizuführen. Das Auseinanderfallen der 12V-Ebene wird immer dazu führen, dass mindestens eine redundant vorhandene sicherheitsrelevante Komponente mit Energie versorgt wird. Für die Funktion des Energieversorgungssystems kommt es im separierten Betrieb zu Einschränkungen: Die verfügbare Energie im sekundären Segment ist durch den Energieinhalt der 12V-Stützbatterie beschränkt. Dieser muss also so bemessen sein, dass der Notbetrieb und der Übergang in den sicheren Zustand abgedeckt werden. Der State of Function dieser Batterie muss überwacht werden, um die Verfügbarkeit dieser redundanten Energiequelle sicherzustellen. Im primären 12V-Segment ist die verfügbare Energie nicht eingeschränkt, da dieses nach wie vor mittelbar aus der Traktionsbatterie versorgt wird. Bei geöffnetem TE steht allerdings die 12V-Pufferbatterie nicht mehr zur Stabilisierung dynamischer Spannungsschwankungen zur Verfügung. Diese Funktion kann vom DC/ DC-Wandler übernommen werden, der dazu über eine verbesserte Regeldynamik verfügen muss. Alternativ kann ein zusätzlicher Speicher eingebracht werden, der hier nicht im eigentlichen Sinne als Energiespeicher fungieren muss, sondern lediglich Leistungsspitzen abdeckt, die der DC/ DC-Wandler nicht unverzüglich bedienen kann. In Hi- Bord wird hierzu ein DC/ DC-Wandler mit einem Doppelschichtkondensator kombiniert, der Leistungsspitzen abdecken kann, ohne dass eine zweite Bleibatterie erforderlich wird. Zusätzlich können Komfortverbraucher degradiert werden. Insgesamt wird somit durch gezielte Erweiterung eine noch relativ einfache Architektur geschaffen, die mit den heute üblichen Architekturen weitgehend kompatibel ist. 8.2 Architektur B Die zweite Architektur, dargestellt in Abbildung 11, besteht aus zwei 12V-Zweigen, die jeweils als Backbone [13] realisiert und über einen eigenen DC/ DC-Wandler mit dem 48V-Segement verbunden sind. Zwischen den 12V-Segmenten besteht keine direkte Verbindung. Im ersten 12V-Zweig ist ein Segment mit sicherheitsrelevanten, redundanten Komponenten direkt mit dem DC/ DC-Wandler verbunden. Ein Segment mit Komfortverbrauchern ist über ein TE angebunden. Durch Abtrennen der Komfortverbraucher kann verhindert werden, dass sich Fehler in diesem Segment auf die sicherheitsrelevanten Komponenten auswirken. Ein DC/ DC-Wandler in Kombination mit Doppelschichtkondensatoren dient zur Energieversorgung und bietet die gleichen Vorteile bei der Abdeckung transienter Leistungsspitzen wie schon in Architektur A. Im zweiten Zweig befinden sich die redundanten sicherheitsrelevanten Verbraucher, sowie eine 12V-Batterie. Ein TE sichert den Zweig gegen Fehler ab, die vom DC/ DC-Wandler übertragen werden. Dieser Zweig kann aus der 12V-Batterie versorgt werden und ist damit kurzzeitig von der Verfügbarkeit der 48V-Ebene unabhängig. 198 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Bild 11: Ausgewählte Architektur B eines fehlertoleranten EBN erstellt mit dem im Projekt entwickelten Software-Werkzeug HiTool. 9 Realisierung der intelligenten Netzknoten Die beiden ausgewählten EBN-Topologien A und B sollen zur Demonstration und Evaluation im Labor und in einem weiteren Schritt in ein Fahrzeug implementiert werden. Neben einem angepassten Kabelbaum müssen dazu die Schlüsselkomponenten TE, DC/ DC-Wandler und ÜGS realisiert werden. Aus den beiden ausgewählten Topologien wurden daher konkrete Anforderungsprofile an die intelligenten EBN- Knoten abgeleitet. Die Anforderungen umfassen Funktionen, elektrische Kenngrößen sowie Reaktionsmuster und Kommunikationsschnittstellen. Trennelement Die Abbildung 12 zeigt das Blockschaltbild eines TE. Im Normalfall verbindet es die beiden EBN-Segmente, an die es über die Kontakte BN1 und BN2 angeschlossen ist. Für die Realisierung des HiBord-Demonstrators muss es einen maximalen Dauerstrom von 160 A in beide Richtungen übertragen können. Im Fehlerfall muss ein Leistungsschalter die beiden Segmente voneinander isolieren können, so dass sich der Fehler nicht von einem auf das andere Segment auswirkt. 199 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Bild 12: Blockschaltbild des Trennelements. Es wird davon ausgegangen, dass die sicherheitsrelevanten Verbraucher nach LV124 ausgelegt sind und daher kurzzeitige Spannungseinbrüche tolerieren. Falls ein Kurzschluss im EBN auftritt, sinkt die Spannung schnell gegen 0 V. Das TE muss den Kurzschluss dennoch rasch detektieren und die angeschlossen EBN-Segmente innerhalb von 100 µs separieren können. Zur Fehlerdetektion werden die Spannung der beiden EBN-Segmente und der Strom über das TE erfasst und miteinander verknüpft. Die Trennfunktion wird über einen Halbleiterschalter realisiert, der im Fehlerfall geöffnet wird. In Abbildung 13 ist grün der nominelle Spannungsbereich des 12V-EBN dargestellt. Wenn die Spannung in die orange dargestellten Bereiche steigt oder abfällt, wird dies für die angegebenen Zeitspannen toleriert und an die ÜGS gemeldet. Die ÜGS kann darauf gegebenenfalls mit geeigneten Maßnahmen reagieren. Beim Verlassen des orangenen Bereichs trennt das TE die beiden EBN-Segmente reflexhaft, d.h. ohne vorherige Kommunikation mit der ÜGS. Bild 13: Zeit-Spannungsgrenze für das Trennelement. Erkannte Fehler, die das TE nicht reflexhaft behandelt, werden an die ÜGS kommuniziert. Die ÜGS kann somit den Fehler auf Systemebene lokalisieren und isolieren. Hi-Side Treiber Latch µC Spannungsanpassung Spannungsanpassung Temperatursensor CAN Speicher Programming/ Debug LEDs Schutz Energieversorgung Energiepuffer BN1+ CAN H/ L BN1- BN2- BN2+ Schutz Leistungsschalter Quellwahl ODER Schutz Stromsensor Spannung in V 32 27 18 16 9,8 Zeit 6,5 5,5 4,5 0 100 µs 100 ms 200 ms 60 s ∞ 200 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Zu diesen Fehlern gehören etwa kurzbis mittelfristig tolerierbare Verletzungen des nominalen Spannungsbereichs. Neben Fehlerzustandsinformationen werden über die Kommunikationsschnittstelle Sensordaten und Statusinformationen vom TE an die ÜGS übertragen. Die ÜGS kann über die Schnittstelle das TE kommandiert schalten, um Fehlerbehandlungsstrategien auf Systemebene umzusetzen. Die Energieversorgung des TE erfolgt redundant aus beiden angeschlossenen EBN- Segmenten. Die Energieversorgungseinheit muss dabei so gestaltet sein, dass dadurch die Trennfunktion des TE nicht umgangen werden kann. Sie muss außerdem ebenso tolerant gegen Spannungsschwankungen sein wie die sicherheitsrelevanten Komponenten. Speicherwandler Der Speicherwandler (SW) übernimmt im Normalfall die Aufgaben eines herkömmlichen unidirektionalen Bordnetzwandlers zwischen der 48V- und der 12V-Ebene. Falls aufgrund einer Fehlerreaktion des TE (Architektur A) die Verbindung zur spannungsstabilisierenden 12V-Batterie unterbrochen wird, oder das vom Wandler versorgte Segment über keinen eigene Batterie verfügt (Architektur B, erster Zweig), muss der Wandler diese Funktion mit übernehmen. Hierzu wird der DC/ DC-Wandler mit einem ausgangsseitigen Doppelschichtkondensator (Double Layer Capacitor, DLC) kombiniert. Bild 14: Blockschaltbild des Speicherwandlers. Abbildung 14 zeigt das Blockschaltbild des DC/ DC-Wandlers, der in HiBord für die Realisierung der in Abschnitt 8 vorgestellten Architekturen A und B eingesetzt wird. Der Wandler ist aus unterschiedlichen Modulen eines Baukastens für intelligente EBN-Knoten aufgebaut. Vier parallele 48V-12V-DC/ DC-Module werden zu einem DC/ DC-Wandler mit ca. 1650 W Nennleistung kombiniert. Ein DLC-Modul mit einer Kapazität von ca. 17 F dient als Leistungspuffer um Lastspitzen abzufangen und die Dynamik des Gesamtsystems zu erhöhen. Eine Vorladeschaltung begrenzt den Ladestrom des Kondensators. Der Systemausgang wird von einem Schaltermodul gebildet, so dass der Systemausgang elektronisch vom angrenzenden 12V-EBN- Segment abgetrennt werden kann. DC/ DC DC/ DC DC/ DC DC/ DC Vorlade‐ schaltung --DLC Leistungs‐ schalter Spannungs‐ versorgung --Diode Logikmodul CAN 12V‐ 48V‐ 48V+ 12V+ CAN 201 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Weitere Module bilden die CAN-Busschnittstelle und die Hilfsspannungsversorgung für die interne Elektronik. Steuerungslogik, Fehlerdetektions- und Reaktionsmechanismen sowie die Kommunikation mit der ÜGS werden durch das auf einem Logikmodul realisierten internen Steuergerät des Speicherwandlers umgesetzt. Die Module sind hinsichtlich ihrer Abmaße und ihrer Schnittstellen standardisiert, so dass sich neben dem in Abbildung 15 dargestellten DC/ DC-Wandler auch andere intelligente EBN-Knoten aufbauen lassen. Bild 15: v.l.n.r.: Draufsicht CAD-Modell des modularen Speicherwandlers mit Hilfsspannungsversorgung, Logikmodul, DLC-Modul, 2x Wandlermodul (obere Reihe) und Kommunikationsmodul, Schaltermodul, Vorlademodul, 2x Wandlermodul (untere Reihe); Bidirektional trennendes Schaltermodul; 48V-12V-DC/ DC-Wandlermodul. Übergeordnete Steuerung Die ÜGS ist dem TE und dem SW hierarchisch übergeordnet. Sie übernimmt im Normalfall koordinierende Funktionen, wie etwa das Aufstarten des HiBord-Systems. Darüber hinaus erhält die ÜGS über den CAN-Bus ständig Messwerte und Einschätzungen des Systemzustands von den intelligenten EBN-Knoten. Diese Informationen verarbeitet die ÜGS für eigene Fehlererkennungen und Lokalisierungen und kann damit langsam wirkende Fehler erkennen und etwa durch geänderte Soll-Vorgaben an den DC/ DC-Wandler oder durch kommandiertes Abschalten von Lasten behandeln. Falls die intelligenten Knoten reflexhafte Fehlerisolationen vornehmen, reagiert die ÜGS im Nachhinein, indem sie die nötigen weiteren Maßnahmen auf Systemebene - etwa Lastabschaltungen - bestimmt und durch intelligente EBN-Knoten oder andere Komponenten umsetzen lässt. 10 Ausblick Dazu werden derzeit die in Abschnitt 9 beschriebenen intelligenten EBN-Knoten realisiert. In einem anderen Arbeitspaket wird der Kabelbaum eines BEVs so modifiziert, dass die in Abschnitt 8 beschriebenen Topologien A und B in diesem Fahrzeug realisiert werden können. In weiteren Arbeitspaketen werden Laborumgebungen aufgebaut, um die EBN-Knoten und die damit aufgebauten Topologien bzgl. ihrer Eignung und ihre Leistungsfähigkeit für das automatisierte Fahren und bezüglich der Anforderungen an Fehlertoleranz und Fehleroperabilität vergleichen und evaluieren zu können. 202 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen Danksagung Das Verbundvorhaben „HiBord - Hochverfügbare und intelligente Bordnetztopologien für automatisierte Fahrzeuge“ wurde vom Bundesministerium für Bildung und Forschung (BMBF) im Rahmen des Förderprogramms IKT2020 unter dem Förderkennzeichen 16EMO0189K gefördert. Die Autoren bedanken sich bei allen Kollegen, die an der Erarbeitung der vorgestellten Ergebnisse mitgewirkt haben. Abkürzungen ASIL Automotive Safety Integrity Level BEV Battery Electric Vehicle DLC Double Layer Capacitor EBN Energiebordnetz EOTI Emergency Operation Time Interval EOTTI Emergency Operation Tolerance Time Interval FIT Failures in Time, Ausfälle pro 10 9 h HV Hochvolt SAE Society of Automotive Engineers SW Speicherwandler TE Trennelement ÜGS Übergeordnete Steuerung VDA Verband der Automobilindustrie Literatur [1] Road vehicles - Functional safety - Part 3: Concept phase, ISO 26262- 3: 2018(E), 2018. [2] Road vehicles - Functional safety - Part 1: Vocabulary, ISO 26262-1: 2018(E), 2018. [3] V. Usseglio, A. Graf, and D. Gennermann, “Automatisiert fahren mit doppeltem Netz: Smarte Batterieschalter für einen sichere redundante Stromversorgung,” Automobil Elektronik, Nr. 09-10/ 2018, S. 30-33, 2018. [4] ZVEI - Zentralverband Elektrotechnik- und Elektronikindustrie e.V., 48-Volt- Bordnetz: Schlüsseltechnologie auf dem Weg zur Elektromobilität, 2015. [5] J. Barthlott, Modellierung und Parametrierung von Energie-Bordnetz- Architekturen im Kraftfahrzeug. Berlin: Logos-Verl., 2004. [6] L. Gysen, M. Ayeb, and L. Brabetz, “Modellbasierte Sicherungscharakteristik für elektronische Sicherungen in einer Modelica-Bordnetzsimulation,” in Haus der Technik Fachbuch, Band 147, Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement VIII, C. Hoff and O. Sirch, Eds., Renningen: expert Verlag, 2018, S. 118-130. 203 Intelligente Netzknoten als Baustein für fehlertolerante Energiebordnetze in automatisierten Fahrzeugen [7] M. Jaiser, “Intelligentes Bordnetzdesign mit elektronischen anstelle von Schmelzsicherungen,” in Haus der Technik Fachbuch, Band 142, Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement VII, C. Hoff and O. Sirch, Hrsg., Renningen: expert Verlag, 2016, S. 288-298. [8] M. Horn, O. Koller, and S. Kriso, “Development of safe and reliable Powernets for new vehicle functions: using the example Start-Stop Coasting,” in Haus der Technik Fachbuch, vol. 138, Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement VI, C. Hoff and O. Sirch, Hrsg., Renningen: Expert-Verl., 2015, S. 59-73. [9] A. Kilic, C. Große, and T. Shen, “Bordnetz für autonomes Fahren,” in Haus der Technik Fachbuch, Band 147, Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement VIII, C. Hoff and O. Sirch, Eds., Renningen: expert Verlag, 2018, S. 148-159. [10] R. Haberfellner, Ed., Systems Engineering: Grundlagen und Anwendung, 13. Aufl. Zürich: Orell Füssli, 2015. [11] B. Mohrmann, L. Brabetz, G. Hirtz, and S. Preisler, “ToSKa: Power supply Topologies, Stabilization and Communication for future vehicles and automated drive,” in Haus der Technik Fachbuch, Band 147, Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement VIII, C. Hoff and O. Sirch, Hrsg., Renningen: expert Verlag, 2018, S. 160-176. [12] Schipperges, F., Tomanic, T., Ratsch, J., Thele, M., Bäker, B., „Examination of fault-tolerant on-board power supply topologies Elektrik/ Elektronik“, in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement Vlll, expert verlag, Renningen, 2017 [13] A. Sedlmaier-Fuchs, M. Wortberg, S. Lobmeyer, and K. Ring, “Energieverteilung neu gedacht,” ATZ Elektronik, Vol. 9, Nr. 5, S. 52-57, 2014. 204 Towards reliable power supply for highly automated driving Kay Klobedanz, Stefan Grösbrink, Sebastian Kahnt  Abstand / distance ca. 3 cm  Abstract Advanced driver assistance systems (ADAS) more and more evolve from single-purpose functions, e.g. Adaptive Cruise Control with Stop and Go or Lane Keeping Assistance, into integrated partly automated driving features such as Traffic Jam Assistance or Remote Parking Assistance. The next steps are highly automated driving functions such as Highway Pilot Systems or Automated Valet Parking. Highly automated driving systems must ensure functional safety without relying on human supervision and action, resulting in a transition from fail-safe to fail-operational systems. This implies high requirements regarding fault tolerance for involved components to continue a safe and reliable operation even in presence of possible faults. Any electronic component cannot operate without an adequate and reliable energy supply. Therefore, the power supply systems for automated driving features must fulfill equivalent requirements regarding functional safety and reliability. This paper is addressing this challenge of developing adequate concepts for reliable power supply systems. These concepts adapt to existing vehicle power supply topologies to fulfill the specific use-cases and requirements from different drive train platforms. Based on the underlying vehicle architecture, suitable components e.g. intelligent battery sensors, battery management systems, and DC/ DC converters are integrated for a reliable energy management. The combination of these components implements fault tolerance, even by means of redundancy, if required. Different power classes of components are applicable to meet diverse required energy levels with a scalable and cost-optimized topology. Kurzfassung Advanced driver assistance systems (ADAS) entwickeln sich mehr und mehr von Systemen zur teilweisen Unterstützung, beispielsweise einem adaptivem Geschwindigkeitsregelungssystem oder einem Spurhalteassistenten, zu integrierten teilautomatisierten Fahrfunktionen wie einem Stauassistenten oder einem fernbedienbaren Parkassistenten. Der nächste Schritt sind hochautomatisierte Fahrfunktionen wie ein Autobahnpilot oder ein automatisierter Parkservice. Systeme für das hochautomatisierte Fahren müssen funktionale Sicherheit sicherstellen, ohne sich auf menschliche Überwachung oder Fahrereingriff verlassen zu können. Dies geht einher mit einem Übergang von Fail-Safezu Fail-Operational-Systemen. 205 Towards reliable power supply for highly automated driving Impliziert sind hohe Anforderungen an die Fehlertoleranz der beteiligten Komponenten, um einen sicheren und zuverlässigen Betrieb selbst bei Auftreten möglicher Fehler fortzusetzen. Alle elektronischen Komponenten sind auf eine geeignete und zuverlässige Energieversorgung angewiesen. Aus diesem Grund müssen die Spannungsversorgungssysteme für automatisiertes Fahren gleichhohe Anforderungen an funktionaler Sicherheit und Zuverlässigkeit erfüllen. Diese Veröffentlichung adressiert die Herausforderung für geeignete Energieversorgungskonzepte. Diese Konzepte passen sich in bestehende Bordnetztopologien ein und erweitern diese, um den spezifischen Anforderungen der verschiedenen Antriebs- und Fahrzeugplattformen gerecht zu werden. Basierend auf der vorliegenden Fahrzeugarchitektur werden geeignete Komponenten - intelligente Batteriesensoren, Batteriemanagementsysteme, Gleichspannungswandler - zu einem zuverlässigen Energiemanagement integriert. Die Kombination dieser Komponenten implementiert Fehlertoleranz, auch anhand von Redundanz, sofern erforderlich. Komponenten verschiedener Leistungsklassen können eingesetzt werden, um die unterschiedlichen Energielevel-Anforderungen mit einer skalierbaren und kostenoptimierten Topologie zu erfüllen. 1 Evolution: From Advanced Driver Assistance Systems to Highly Automated Driving The evolution of driver assistance systems (DAS) starts with the integration of anti-lock braking systems in the 1970s. In the following decades, motivated by a safety increase (reduction of the frequency and severity of accidents) as well as a comfort increase for the driver, more and more advanced driver assistance systems (ADAS) were developed. These systems can be categorized as information and warning functions such as blind spot warning, driver drowsiness detection or intelligent head beam assistance, passive safety functions such as pre-crash with seat belt preparation, collision mitigation functions such as automatic emergency braking, vehicle control functions such as adaptive cruise control (longitudinal control: accelerating/ deceleration), lane-keeping assistance (lateral control: steering), traffic jam assist or parking assist (longitudinal and lateral control). This paper addresses the evolution of vehicle control systems. The SAE International defines in the standard Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving Systems [1] levels of driving automation, depicted in Figure 1. Figure 2 maps functional examples to the different levels. Level L0 refers to no automation: longitudinal and lateral control are performed completely by the human driver. A L1 system performs either longitudinal or lateral control: specific driving subtasks of driving are shifted from the human driver to the driver assistance system. Examples are adaptive cruise control (ACC, only longitudinal control) and lane-keeping assistance (LKA, only lateral control). 206 Towards reliable power supply for highly automated driving Figure 1: Definition of Autonomous Driving Levels. Level L2 systems perform both longitudinal and lateral driving tasks in certain situations. An example is a traffic jam assist system that combines adaptive cruise control with Stop&Go and lane-keeping, state-of-the-art since 2013. Another example is a parking assist, which performs the entire maneuver with steering, accelerating and breaking. The evolution continues with an extension of driving situations which are covered by automated vehicle control. The traffic jam assist system performs driving and steering in a specific subset of driving situations: highway, congested traffic, speed less than or equal to 60 km/ h, a vehicle ahead of the ego vehicle. The evolution highway pilot extends vehicle control to all driving situations on the highway, covering stop-and-go situations, slow-moving congested traffic and driving up to the speed limit on a clear lane. The levels L2 “Partial automation”, L3 “Conditional automation” and L4 “High automation” differ regarding the expectation of the ability of the human driver to take over control and regarding the available period of time for this take-over. This aspect is discussed in detail in Section 2. Level 5, “Full automation” or widely called “Autonomous driving” refers to a complete vehicle control by the automation system, no driver required from start to finish of a drive as the system can handle all scenarios, road types and conditions. A human driver is obsolete. This level of automation disrupts the present business model of individual mobility and attracts therefore non-automotive technology companies, which do not follow the described evolutionary approach, but aim directly at the ultimate goal. Starting with level L3, the term Highly Automated Driving (HAD) is commonly used. 207 Towards reliable power supply for highly automated driving Figure 2: Customer Functions on different Autonomous Driving Levels. 1.1 Components for Autonomous Driving Powerful sensors are key components for autonomous driving. Radar sensors determine distance, speed and movement direction of objects for ambient recognition. Cameras supplement the information from the radar sensors by identifying the detected objects as well as detecting objects that radar cannot detect, such as car lanes and markings on the ground, traffic signs and traffic lights. LIDAR sensors complement radar sensors for precise ambient recognition at great distance. Ultrasonic sensors determine distances to objects in the short range. These sensor technologies provide vehicles with the ability to see the environment, shake sensors (vibration sensors) add the ability to feel. They cover the immediate vehicle proximity and detect contact and trigger an emergency stop to prevent damage and allow to record the condition of vehicle fleets in real time. Electronic control units, offering computational powers that is orders of magnitude greater compared to previous automotive components, create an environmental model of the area around the vehicle by sensor fusion, recognize the traffic situation, predict other traffic participant’s behavior, derive a driving strategy, plan the ego vehicle’s maneuver and compute a resulting trajectory. These electronic control units provide the necessary input for the actuators. Actuation refers to the actual control of the vehicle, performing actions such as accelerating, changing gears, applying brakes, and steering. Major components are the entire powertrain, brake systems and steering systems [3, 4]. In order to implement complex functions i.e. use cases such as automated parking or a highway pilot that controls from entry to exit sensors, control units and actuators have to meet highly innovative and challenging requirements by themselves, plus must be aligned tightly in order to cooperate effectively. Automotive suppliers react by building collaborations with network partners, complementing technology and integrated it through development partnerships. 208 Towards reliable power supply for highly automated driving 2 Implications for Functional Safety The shift of control from human driver to driver assistance system accompanies with a shift of responsibility and liability. Driver assistance systems that control the vehicle are inherently safety-critical, with the overall safety goal to prevent accidents. A failure or malfunction may result in a loss of life, severe injury or severe environmental damage. The international standard ISO 26262 "Road vehicles - Functional safety" addresses the development of safety-related electrical/ electronic systems in production automobiles, preventing possible hazards caused by the malfunctioning of electrical/ electronic systems [2]. It classifies safety-critical vehicle components into four integrity levels, from lowest Automotive Safety Integrity Level (ASIL) A to highest level ASIL D, depending on the consequences of failures. For an autonomous driving function, typically the highest level ASIL D applies [6]. The safety requirements for components can be reduced by decomposition, if redundancy and independence are ensured. Using a L0 or L1 system, the human driver always has to keep his hands on the steering wheel, as these systems do perform either longitudinal or lateral control, but not both. L2 systems enable the driver to take off his hands temporarily during specific driving situations, which are covered by the system with both longitudinal and lateral control. The human driver must keep his eyes on the road and pay attention to traffic in order to be able to take-over control immediately as soon as the system’s boundaries in terms of manageable driving situation are exceeded. Driver monitoring systems might be integrated to check whether the driver is still able to take-over, for example by tracking the driver’s eyes or by limiting the hands-off time, regularly requesting a driver control input. The human driver remains responsible for monitoring the traffic situation and vehicle state and taking control if necessary. The fundamental difference of level L3 is the option for the driver to take his eyes off. The driver no longer needs to permanently monitor the traffic situation but must be prepared to take-over control within a short period of time, for example 10 to 20 seconds. The driver still acts as the fallback for the driver assistance system, as it is the case for the levels L1 and L2. To allow the take-over time, fundamental changes to the electric/ electronic architecture are required. For driver assistance systems up to level L2, it was safe in case of a hardware or software failure to detect it, inform the driver and deactivate/ degrade the operation, a safety approach called fail safe. The human driver as the fallback in this safety concept takes back the control of the vehicle. Starting with level L3, the safety concept cannot any more fully rely on the driver. Instead, the system must implement a fallback function, which continuously provides vehicle control during the take-over time. Such a safety architecture is called fail operational. The driver is informed by the driver assistance system about the presence of a failure and he is requested to take back the control in the next 10 to 20 seconds, during which the system must provide the fallback. A single failure can no longer imply an unavailability of the function. If the driver did not take over control, the system must perform a minimal risk maneuver that stops the vehicle. At level L4, the driver is no more requested to take-over within a period of time smaller than a minute, but might still require driver intervention from time to time. At level L5 there is not even a driver present, so the vehicle in case of a failure must perform a 209 Towards reliable power supply for highly automated driving safe stop at a non-hazardous place, which depending on the driving situation might request the fallback to perform for several minutes. Summing up: for the levels L3, L4 and L5, the safety concept cannot rely on an immediate human take-over, for which reason a fallback function has to be implemented to react electric/ electronic failures. 2.1 Fail-operational System Architectures The system architecture faces the following challenges: safety requirements up to ASIL D, no human driver as a fallback, fail-operational behavior that guarantees full or degraded operation of vehicle control if a failure occurs. In addition to a prevention/ detection of wrong behavior, non-availability of the function must be prevented. The major architectural pattern to implement a fail-operational system is redundancy: multiple independent channels realize the safety-critical function. Redundancy must be applied to all processing steps of an autonomous vehicle: perception, localization and planning, actuation, plus human machine interface and electrical/ electronic architecture.[3] Multiple and different types of sensors (based on different physical principles) detect objects in the vehicle’s environment, e.g. radar and camera. Multiple electronic control units process the sensor input, localize the vehicle, derive a trajectory and provide input for the actuators. These control units must be diverse to ensure independence: different hardware and software elements might be used to prevent a systematic fault from taking down primary and secondary channel at the same time. Redundant braking and steering are each able to decelerate and steer the vehicle. The human machine interface informs the human on different channels, for example visually and acoustically. If a component fails (single component failure), the function is kept up by the second channel. Regarding the electrical/ electronic system architecture, fault tolerant communication networks are required. To detect failures in one channel, high diagnostic coverage is needed for all components, defined as the effectiveness of the safety mechanisms at detecting faults and determined by the requirements of the specific ASIL. The fail operational mode management with fault detection and reaction has to be implemented in a distributed manner on independent hardware. The function cannot be offered, if an unavailability of the fallback channel was detected. The power supply in a fail-safe architecture monitors overand under-voltage to the components, provides limited output ranges and detects short circuits. The safe state, realized by switching off components and opening protection elements, is entered in case of a failure. Redundant power supplies are required to meet the high level of availability.[5, 6] In case of a breakdown of one supply, the second path has to offer at least enough energy to power sensors, electronic control units and actuators for the period of time required to perform the minimal risk maneuver, which stops the vehicle at a safe place. A valid assumption for this time period is 30 seconds. Failure isolation is crucial to prevent a failure in one path to strike through and affect the other path as well. Common cause failures that take down both paths have to be avoided. In the following, we discuss different power supply solutions for highly-automated driving systems. 210 Towards reliable power supply for highly automated driving 3 Solutions for reliable power supply In this chapter we describe the solutions and products available and under development to enable a reliable power supply for automotive applications. This includes the existing state-of-the-art components for energy management given in section 3.1 as well as trends and upcoming products which are described in section 3.2. 3.1 State-Of-The-Art Components for Energy Management The supplier market comprises solutions for the energy management and power supply of different automotive applications and components. Figure 3 depicts three solutions for drive train platforms. For internal combustion engine (ICE) vehicles the Intelligent Battery Sensor (IBS) allows the implementation of an automatic start/ stop functionality, also called Micro Hybrid level 1. Depending on the level of implementation for energy management in ICE vehicles, voltage stabilizers and DC/ DC 12V converters in combination with suitable battery-technologies are integrated in addition to realize Micro Hybrid level 2, which utilizes braking-energy recuperation and intelligent charging/ discharging strategies for the 12V battery [7]. Currently, numerous automotive OEMs are introducing vehicles based on the so-called Mild Hybrid topology [7]. This topology combines the 12V vehicle power supply with a second vehicle power supply with a voltage level of 48V. By replacing the 12V alternator and starter with an integrated-starter-generator (ISG) unit in the 48V vehicle power supply plus adding a 48V battery, significantly higher energy levels can be generated and stored during recuperation maneuvers. The combustion engine can be boosted by the ISG. Moreover, very dynamic high-power loads such as an electronic active roll stabilization (ARS) can be migrated to the 48V net for energy supply and recuperation. However, also in a Mild Hybrid topology, the majority of electronic loads - in particular the safety relevant driver assistant systems, e.g. electrical power-steering (EPS) actuators - are connected to the 12V vehicle power supply. Thus, an energy transfer from the 48V to the 12V vehicle power supply is required. For this purpose, bi-directional DC/ DC 48V/ 12V converters are utilized as interface for the energy transfer between the vehicle power supplies (ref. Figure 5). Utilizing the described components, the Mild Hybrid topology offers the ability to perform engine-off coasting. In this scenario the engine respectively, generator can be turned off as power supply and the saved energy from the 48V battery can be transferred to the 12V loads. Another class of drive train platform with growing importance and additional requirements regarding energy management are Full Hybrid and Battery Electric Vehicles [7]. In these topologies high voltages and amounts of energy have to be managed and resulting safety constraints must be fulfilled for complex and costly components. Here, as a central control unit, a so-called Battery Management System (BMS) acts as a complete solution for the management of high voltage (HV) Li-Ion batteries. 211 Towards reliable power supply for highly automated driving Figure 3: Drive Train Platforms with integrated Energy Management Solutions. 3.2 Trends & Upcoming Products Beside existing products, there are additional solutions currently under development to fulfil the growing and changing needs of the individual car manufacturers. As mentioned above, the Mild Hybrid topology depicted in Figure 5 comprises a DC/ DC 48V/ 12V converter and a 48V battery. One trend for this topology is the customer demand for a combined solution from one supplier to get harmonized components and to ease the system integration. For this purpose, a so-called PowerPack 48V can offer a cost efficient 48V concept by bundling power electronics with battery and battery management. Figure 4 shows a schematic of such a component, which enables the described mild hybrid topology functions whilst ensuring the reliable supply of the 48V and 12V loads for vehicles with a high-power demand in an optimized package size. Figure 4: Schematics of a PowerPack 48V and a Dual Voltage Battery (2VBM) 212 Towards reliable power supply for highly automated driving For smaller car segments with lower electrical power consumption, the component integration can be driven even further by combining a 12V battery and a 48V battery in one single device. For this purpose, a so-called Dual Voltage Battery Management 12V/ 48V (2VBM) currently is being developed. This dual-voltage battery is composed of a battery module and a battery management system which can handle both voltage levels by putting battery cell blocks into series circuit or parallel circuit depending on the current vehicle situation and power demands. By means of this cell block switching concept, the necessity of an additional DC/ DC converter is avoided. Summed up, the 2VBM component enables mild hybrid applications with a resulting package size fitting to the size of a conventional 12V battery (ref. schematic in Figure 4). As mentioned above, due to the quickly growing importance of electrification, Full Hybrid and Battery Electric Vehicles imply a drive train topology with an integrated E- Engine and HV-components. Currently, also in this topology the safety relevant driver assistant systems are connected to the 12V vehicle power supply and an energy transfer mainly from the HV to the 12V vehicle power supply is required. For this purpose, bi-directional HV/ 12V converters are a common interface for the energy transfer between the vehicle power supplies (ref. Figure 5). Beside this state-of-the-art component, g the development of additional options to offer HV/ 48V converters or even a combined HV/ 48V/ 12V converter are being investigated. These components will allow more modular and flexible topologies by keeping and migrating loads to the 48V vehicle power supply in Full Hybrid and Battery Electric Vehicles. 4 Case studies & solutions for fail-operational architectures In Chapter 3 we presented different applications and components for the power supply and energy management at different drive train concepts based on their underlying state-of-the-art vehicle power supply topologies: ICE incl. Micro Hybrid (12V/ 12V), Mild Hybrid (48V/ 12V), and Full Hybrid and Battery Electric Vehicles (HV/ 12V). In this chapter we provide case studies and solutions for upcoming topologies addressing the requirements resulting from fail-operational system architectures for HAD applications as described in Chapter 2. This comprises the evolvement of state-of-the-art vehicle power supply topologies to more modular and flexible setups described in Section 4.1 as well as applicable components to implement different variants given in Section 4.2. 4.1 Vehicle Power Supply Topologies Figure 5 depicts examples for state-of-the-art vehicle power supply topologies in Mild Hybrid vehicles (48V/ 12V) and Full Hybrid & Battery Electric vehicles (HV/ 12V). Such a topology includes an energy generator, i.e. typically a starter/ alternator which is connected to the 48V respectively HV vehicle power supply to provide the required energy to the high-power loads also attached there. As described above, the corresponding DC/ DC converters are integrated as interface between the different voltage levels to transfer the required energy to the majority of electronic loads including safety-relevant driver assistant systems which are connected 213 Towards reliable power supply for highly automated driving to the 12V vehicle power supply. Additionally, each vehicle power supply also comprises a battery to buffer a certain amount of energy for vehicle situations without energy generation, e.g. a start/ stop maneuver. These topologies ensure a reliable and fail-safe power supply for all integrated components implementing currently established energy management and driver assistant systems up to level L2 e.g. ACC with Stop&Go or Traffic Jamand Park Assist (ref. Chapter 1). Figure 5: State-of-the-art Topologies for underlying Power Train Platforms in Mild Hybrid vehicles (48V/ 12V) and Full Hybrid & Battery Electric vehicles (HV/ 12V). However, as described in Chapter 2, the requirements regarding the availability and reliability of the power supply for the assistant systems are increasing significantly evolving towards highly automated driving. Starting with level L3 a fail-operational system architecture must be implemented. In case of a fault the underlying system must be able to continue vehicle control and perform a minimal risk maneuver that stops the vehicle. Furthermore, upcoming drive train platforms also imply the necessity for flexible and modular topologies to handle and combine several power supplies and provide the required energy management and power supply for all components connected at the different voltage levels. Figure 6 and 7 illustrate two concepts for such modular topologies. The flexibility is realized by combining the different power supplies for HV, 48V, and 12V - marked in the figures as (1), (2), and (3) by means of integrating the corresponding DC/ DC converters for the energy transfer as described in Chapter 4. Depending on the underlying drive train concept and the implemented functionalities, the required topology can be realized as an arbitrary combination of (1), (2), and (3) - e.g. (2) + (3) for Mild Hybrid and (1) + (3) for Full Hybrid & Battery Electric vehicles, like the architectures shown in Figure 5. 214 Towards reliable power supply for highly automated driving - Figure 6: Modular Topology with redundancy at 12V vehicle power supply implemented by DC/ DC 48V/ 12V converters with multiple 12V outputs. - Figure 7: Modular Topology with redundancy at 12V vehicle power supply implemented by additional DC/ DC 12V/ 12V converter or switch. Logically, in a Mild Hybrid system the electric motor depicted by (M) in the illustrated topologies would be replaced by a starter/ alternator in the 48V vehicle power supply also shown in Figure 5. Furthermore, additional energy sources can be integrated in such a topology. For instance, a solar roof connected to the 48V vehicle power supply as depicted in Figure 7. As described in Section 2.1, the major architectural pattern to implement a fail-operational system is redundancy. Thus, with respect to the availability and reliability of power supply for safety-relevant driver assistant systems, both variants implement redundancy at the 12V vehicle power supply to realize a fail operational system architecture as depicted in the topology areas marked with (3) in Figure 6 and 7. Therefore the 12V vehicle power supply consist of two isolated subnets with individual energy storages (batteries) and interfaces (DC/ DC converters) to the other vehicle power supplies. By means of this doubling of the 12V vehicle power supply, the safety-relevant ADAS respectively HAD functions can be designed redundantly by integrating component instances connected to each of the subnets, e.g. as main and backup system. Beside the fail-operational power supply that guarantees full or degraded operation of vehicle control if a failure occurs, this system architecture also allows the fulfillment of the resulting functional safety requirements for HAD applications up to ASIL D. In the next section we describe how existing products can be applied and combined to implement these kinds of vehicle power supply topologies. 215 Towards reliable power supply for highly automated driving 4.2 Applicable Product Combinations Figure 6 presents a fail-operational vehicle power supply topology where two separated 12V power supplies (3) are connected to the 48V vehicle power supply (2) via a DC/ DC converter which acts as power supply for theses subnets and the connected loads. Figure 8 illustrates how such a topology can be implemented for a Mild Hybrid topology with existing state-of-the-art components to ensure redundancy for the ADAS/ HAD functions and fulfil functional safety requirements with an integrity level of ASIL D. In this architecture a DC/ DC 48V/ 12V converter provides a reliable power supply for the main 12V vehicle power supply containing most electronic loads in the vehicle - marked with the dashed blue line. Such a high-power converter must be able to provide a few kilowatts peak-power when required by the loads. Furthermore, as part of a failsafe architecture it must fulfill the functional safety requirements of the 12V main vehicle power supply, e.g., avoidance of overand under-voltage to the components or detection of short circuits in ASIL B. This high-power DC/ DC converter is combined with a dedicated second 48V/ 12V converter which acts as power supply for the redundant backup vehicle power supply - marked with the dashed red line. This redundant vehicle power supply contains a backup battery and more importantly additional backup instances of the components for ADAS/ HAD functions. Also, redundant instances for other loads can be added to this backup to ensure their availability even in case of a failure of the main vehicle power supply. Focusing on the redundancy of the ADAS/ HAD functions a converter with a lower power class e.g. 500W will be sufficient to provide the required amount of energy to the backup loads, resulting in a cost-efficient solution. This topology realizes the functional safety level with an integrity of ASIL D by means of a decomposition to two converters with ASIL B (D): ASIL B(D) + ASIL B(D) = ASIL D [2]. The two converters ensure a fail-operational power supply to the ADAS/ HAD functions by providing energy even in case of a failure in the main 12V vehicle power supply. Figure 8: Redundancy at 12V vehicle power supply implemented by combination of two DC/ DC 48V/ 12V converters. 216 Towards reliable power supply for highly automated driving Figure 9: Redundancy at 12V vehicle power supply implemented by combination of DC/ DC 48V/ 12V converter with DC/ DC 12V/ 12V converter or switch. Another applicable solution is illustrated in Figure 9. This architecture depicts the implementation of the redundancy in a Mild Hybrid topology by means of an additional 12V/ 12V converter or a switch between the 12V subnets. Equal to the topology shown in Figure 8, the main vehicle power supply is supplied by a high-power 48V/ 12V converter which provides the required energy and fulfills the functional safety requirements. In contrast to the solution described above, the backup vehicle power supply in this topology is not also directly supplied by an additional dedicated 48V/ 12V converter. Here, a 12V/ 12V converter or a switch implement an isolating interface between the main vehicle power supply and the backup subnet. Similar to the required energy for the redundant loads mentioned above, for this converter a power class of 500W is sufficient. Depending on the current situation in the vehicle, the integrated interface component can be utilized to establish an energy transfer between the two subnets or to disconnect them. For instance, in case of a shoot-through from the 48V side to the 12V side, by means of this setup the backup instances of the ADAS/ HAD components are protected against the over-voltage which might damage the components in the main vehicle power supply. In this case the ADAS/ HAD backup can use the energy out of the backup battery to perform a minimal risk maneuver as described in Section 2.1. Another possible scenario is an internal error in the 12V/ 12V converter which results in a failure of the energy transfer from the main vehicle power supply to the backup net. In this case the components in the main board will stay operational. Hence, this topology also realizes a fail-operational power supply for ADAS/ HAD systems in the 12V vehicle power supply by means of decomposition as described above. As described above, these two solutions based on state-of-the-art components are applicable for Mild Hybrid topologies. However, similar architectures can be implemented for other drive train platforms based on the modular topologies described in 217 Towards reliable power supply for highly automated driving Section 4.1, e.g. by integrating appropriate combinations of HV/ 12V or HV/ 48V converters. Even though the focus of the presented solutions was on the power electronic components acting as interfaces and power supply for the loads in the 12V vehicle power supplies, also the other relevant components such as the batteries and their control and supervision measures must be integrated in the topologies to ensure a comprehensive energy management and power supply for all application scenarios (ref. Section 3.1). Considering the growing constraints regarding available space and weight for additional components in modern vehicles especially in Mild Hybrid platforms the integration of these components is a logical consequence. Thus, as described in Section 3.2, upcoming and currently being developed solutions, i.e. Power- Pack 48V and 2VBM, will be the next step forward. 5 Summary & Outlook In this paper we gave an overview about the current development towards reliable power supply for highly automated driving. We described how the ongoing evolution of vehicle control applications from simple assistance systems to complex automated driving systems result in rising functional-safety requirements and the necessity of failoperational system architectures based on a reliable power supply implemented by redundancy. As a starting point we presented state-of-the-art components for energy management and power supply of several different automotive applications and drive train platforms. Based on this we introduced trends and upcoming products addressing the customer demands for combined solutions to ease the system integration and meet the rising constraints regarding available space and weight in automotive vehicles. The main contribution of this paper consists of the presented case studies and solutions for fail-operational architectures based on existing components. Therefore, stateof-the-art vehicle power supply topologies were combined and extended to modular and flexible architectures. Depending on the underlying drive train concept and the implemented functionalities, the required topology can be realized as an arbitrary combination for emerging Mild Hybrid and Full Hybrid & Battery Electric vehicles. The presented architectures implement redundancy at the 12V vehicle power supply with respect to the availability and reliability of power supply for safety-relevant driver assistant systems. Utilizing this approach, we described exemplary for a Mild Hybrid topology how such architectures can be implemented with existing state-of-the-art components to ensure fault-operational behavior and fulfil the resulting functional safety requirements for automated driving functions. Similar architectures can be implemented for other drive train platforms based on the presented modular topologies. Considering the above-mentioned constraints regarding available space and weight for additional components in modern vehicles, the presented integrated components will be the next step. 218 Towards reliable power supply for highly automated driving 5.1 Outlook The development towards integrated components described above has already started. Alongside this trend another development will imply additional challenges and requirements regarding the design of fault-operational architectures and future vehicle power supply topologies. Currently, the majority of electronic loads - in particular the safety-relevant systems - is connected to the 12V vehicle power supply. Thus, the described redundancy for a reliable power supply can be limited to this vehicle power supply. However, there are efforts at several automotive OEMs to implement or migrate different safety-relevant systems with high dynamics and power-consumption, e.g. anti-roll stabilization or X-by-wire functions, to the 48V vehicle power supply. Considering this development, the medium-term goal for new topologies must be to also ensure a reliable power supply for the 48V vehicle power supply. In the long-term there will probably be a need for modular and flexible architectures with redundancies across voltage levels. Thus, new solutions and components as the proposed integrated HV/ 48V/ 12V converter will be needed to fulfill the upcoming requirements. Literatur [1] SAE International: Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving Systems, Standard J3016_201401, January 16 th , 2014. [2] ISO 26262-2011(en) Road vehicles — Functional safety, part 1-10. International Standardization Organization, 2011. [3] S. Behere and M. Törngren: "A functional architecture for autonomous driving",Proceedings of the First International Workshop on Automotive Software Architecture. ACM, pp. 3-10, 2015. [4] Daniel Watzenig and Martin Horn (Eds.): “Automated Driving - Safer and More Efficient Future Driving”. Springer, 2017. [5] Ludwig Ramsauer: “Electric power consumption on use case AD (Automated Driving)”. Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement VII, Haus der Technik, pp. 366-386, 2016. [6] Peter Grabs and Udo Hornfeck: “Power Distribution and Functional Safety in the Context of Highly Automated Driving”. Forum Bordnetze von Bayern Innovativ, 2015. [7] Anton Karle: “Elektromobilität: Grundlagen und Praxis”, Hanser, 2017. 219 High Power Density Modular Six Phase Drive Inverter Niklas Langmaack, Günter Tareilus, Markus Henke Abstract In automotive drive systems multiphase electrical machines are used to build drives with maximum powerand torque-density accompanied by advantages in redundant and fail save operation. The increasing number of phases from three to five, six or even more has well-known advantages. In addition the complexity of the drive inverter system rises with an increasing number of power devices, gate driver circuits, sensor elements and signals between power stage and control electronics. This report deals with new design solutions to realize highly efficient and compact power electronic systems for multiphase drive operation. Therefore a holistic design approach is presented, including distributed control via intelligent gate driver units and new highly integrated current sensing technologies. Kurzfassung In Antriebssystemen für Elektrofahrzeuge werden mehrphasige elektrische Maschinen genutzt, um höchste Leistungs- und Drehmomentdichten zu erreichen und Vorteile wie redundanten oder fehlertoleranten Betrieb auszunutzen. Die Steigerung der Phasenzahl von drei auf fünf, sechs oder mehr Phasen hat einige bekannte Vorteile. Im gleichen Zuge steigt jedoch die Komplexität des Antriebsumrichters mit der steigenden Zahl an Leistungshalbleitern, Gate-Treiber-Schaltungen, Sensoren und Signale zwischen Leistungsteil und Steuerungselektronik. Diese Veröffentlichung behandelt neue Entwurfsansätze zur Realisierung effizienter und kompakter Antriebswechselrichter für den Betrieb mehrphasiger Maschinen. Hierzu wird ein ganzheitlicher Designansatz aufgezeigt, der eine verteilte Regelung mittels intelligenter Gate-Treiber-Einheiten sowie eine hochintegrierte Strommessmethode einschließt. 1 Introduction The design process starts with an inverter design for a six phase permanent magnet synchronous machine (PMSM) using six IGBT half bridge modules with individual intelligent gate driver units. Each of them contains a local microcontroller for signal processing and switching signal generation [1]. The gate driver is designed with a serial communication interface to the master control unit of the drive inverter which is highly immune to EMI. It and can easily be implemented multiple times on the master control unit if applied to multiphase systems. The gate driver PCB also incorporates a rogowski coil current sensor for device current sensing. With a sophisticated evaluation algorithm implemented in the microcontroller, this sensor can be used for a load current determination as well [2]. This sens- 220 High Power Density Modular Six Phase Drive Inverter ing technology offers a highly integrative substitute for conventional Hall Effect based current sensors. Furthermore it is a cheap, compact and yet robust solution. The whole drive inverter system containing the power stage, gate driver units with distributed controllers and current sensors, main control unit with digital interfaces for the gate drivers as well as external interfaces for rotor position sensors and vehicle communication has been developed and is operated in a battery electric sports car which represents a multifunctional research environment for automotive traction applications [3]. 2 System Design using Intelligent Gate Drivers In conventional power electronic systems like drive inverters the interface between the control electronics and the gate drivers of the power modules is realized using a set of discrete analog and digital signals. Typically these are the switching signals, fault and reset signals as well as the output signals of current, voltage and temperature sensors. There are two important drawbacks of this conventional approach that will be addressed with the proposed new system design. Especially the analog signals are sensitive to electromagnetic interferences caused by fast switching power devices. With an increasing number of half-bridges the number of discrete signals also increases and makes the whole system more and more complex and less reliable. Figure 1 shows a schematic overview of a conventional three phase drive inverter system with three half-bridges and three conventional half-bridge gate drivers. Figure 1: Conventional system layout With an increasing number of half bridges or power modules the number of discrete signals will also rise. This is the main motivation for completely rethinking the interface between control unit and gate driver unit. The proposed intelligent gate driver circuits include a small microcontroller. This is used for generating the switching signals for one half bridge and evaluating the corresponding sensor signals locally. A robust point-to-point serial interface is used for the communication with the central V DC i U i V i W M conventional control unit Fault PWM conventional gate driver Enable Reset Fault PWM conventional gate driver Enable Reset Fault PWM conventional gate driver Enable Reset Current Current Current Temperature Angle/ Speed Voltage 221 High Power Density Modular Six Phase Drive Inverter control unit (host) instead of the individual digital and analog signals. Figure 2 shows an example. Figure 2: Proposed system design using intelligent gate driver circuits 2.1 Properties, Advantages and Possibilities The design of power electronic systems using the proposed intelligent gate driver circuits with distributed microcontrollers offers a couple of advantages and possibilities that will be discussed in this section. 2.1.1 EMI Robustness Depending on the chosen communication interface the system can be very resistant to EMI problems. In the realized prototype a simple serial interface using a full-duplex RS-485 physical layer with differential signal lines is used. It is fast enough, has just very little overhead and basic fault detection using a parity bit. For future designs a more sophisticated interface using enhanced fault-tolerant encoding technologies like 8b/ 10b encoding could be used. The main benefits regarding the EMI robustness are the very short analog interconnections from the sensor elements to the ADC in the microcontroller of the gate driver board. 2.1.2 Flexibility and Scalability Using intelligent gate driver units results in a highly modular and easily scalable system. Different modules can be equipped with intelligent drivers and may be operated using the same central control unit. Increasing the number of serial interfaces is simple compared to increasing the number of discrete switching signals, fault signals and analog inputs due to the resulting total number of signals. V DC i U i V i W M central control unit for intelligent gate drivers intelligent gate driver intelligent gate driver intelligent gate driver Temperature Angle/ Speed Current Current Current Voltage Sync En/ Flt 222 High Power Density Modular Six Phase Drive Inverter 2.1.3 Safety and Reliability The safety and reliability of power electronic systems can be improved by keeping the system design clear and simple while integrating more features in terms of supervision, fault detection, health monitoring, measurements and tight control. A decentralized approach perfectly meets these requirements. The component diversity and the complexity of the interconnections between power and control electronics can be massively decreased. The intelligent gate driver unit can perform very fast supervision and fault detection tasks for their respective power module and can further be enhanced with more sophisticated monitoring features without the need to change the central control unit. For high power systems a replacement of single power modules can be an interesting option. This is simple using intelligent gate drivers, since the module specific compensation values for analog signals like current offset etc. can be stored in the gate driver’s controller. No parameters need to be tuned in the central control unit’s software. 2.1.4 Galvanic Isolation A typical task of gate driver circuits is to ensure a safe galvanic isolation between the power electronics and the control electronics. This is necessary both for functional reasons and for safety reasons. The proposed digital interface can be implemented with galvanic isolation very easily. Therefore the effort for galvanic isolation in the gate driver circuit itself may be reduced. This allows some savings in size, weight and cost for the intelligent gate driver electronic. Two possible implementations are shown in figure 3. The first variant uses isolated components for each gate driver circuit including individual power supplies and isolated sensors for current and voltage measurements. Therefore the controller on the gate driver board can be directly connected to the host controller. The second variant moves the galvanic isolation to the digital interface. This allows connecting the gate driver controller to the DC link and removing a lot of isolated devices like gate drive optocouplers and isolated analog amplifiers. Figure 3: Possible realizations for the galvanic isolation controller controller Interface Supply Isolated power supply Isolated power supply Isolated current meas. Isolated voltage meas. Interface Supply Non-isolated low-side driver Boot-strap power supply Non-isolated current meas. Non-isolated voltage meas. Galvanic Isolation Isolated high-side driver Isolated low-side driver Level-shifter high-side driver 223 High Power Density Modular Six Phase Drive Inverter 2.1.5 Decentralized Control With increasing switching frequencies thanks to the low switching losses of wideband-gap power semiconductors also the control system of such power electronic converters has to be rethought. Running a control loop for a drive inverter at PWM frequency is always recommended, but can be very challenging, if the PWM frequency reaches e.g. more than 50 kHz. An appropriate way to cope with these challenges is a decentralization of the control algorithms to smaller units. With the supposed system layout using intelligent gate driver units, this way would be consequent. The inner control loop of a cascaded control structure, which is normally a current controller, can be implemented directly in the gate driver unit. Since this controller does not have many other tasks, this control algorithm can also run at a high switching frequency. The outer control loop, which is e.g. a field oriented motor control or speed control, does not necessarily need to run at the switching frequency and can therefore be unaffected by the increased speed. It can be implemented in the central control unit and evaluates the sensors directly connected to the electric machine like an encoder or resolver for position and speed determination and motor temperature sensors. 2.2 Challenges The possible benefits of the proposed concept come along with some challenges that have to be addressed. These can mainly be summed up under the headlines latency and synchronization. 2.2.1 Latency of the Communication Depending on the control structure, the distribution and the operating frequency of the control loop the latency of the serial communication between gate driver controller and central control is more or less critical and has to be taken into account very thoroughly. Figure 4 shows a simplified timing diagram for a conventional control loop structure using intelligent gate drivers. The control loop is calculated on the central microcontroller and can be for example a field oriented motor control. The gate driver units are used as remote ADC and PWM peripherals only. With the control loop operating synchronously with the PWM pattern, two data transfers are necessary within one PWM period. A simple example shows that the maximum calculation time for one control loop iteration is reduced by approximately 4 µs, when the interface is operated at 10 Mbps and the current measurement value and PWM set point value are encoded with 16 bit + 4 bit data overhead each. This corresponds to 3.2 % of the PWM period at 8 kHz switching frequency. 224 High Power Density Modular Six Phase Drive Inverter Figure 4: Timing diagram for a conventional control loop design Figure 5 shows an example for a distributed cascaded control structure. The inner control loop (e.g. phase current) is calculated at high sample rate (PWM frequency) on the intelligent gate driver. An outer control loop (e.g. field oriented machine control) is implemented in the central host controller and operating at a lower frequency. The set point value for the gate driver is changed from duty cycle to reference current. This layout is especially suited for systems using high switching frequencies. Figure 5: Timing diagram for a decentralized cascaded control setup 2.2.2 Synchronization and Shutdown A very important task is the reliable synchronization of the PWM patters that are generated by the individual microcontrollers on the gate drivers. Without synchronization the resulting switching signals would lead to non optimal operation and increased ripple currents in the inverter’s load. A precise synchronization using the serial communication interface is possible but rather complex. In the hardware demonstrator presented in section 4 a simple digital synchronization signal is generated by the central controller and passed to all gate drivers commonly. For an immediate shutdown of all gate drivers in case of a severe fault condition like a short circuit or overvoltage detected by any of the gate drivers a second discrete digital signal is added to the interface. This enable signal is common for all gate drivers and can be used in a bidirectional way as a hardware interlock. Data Transfer S&H Control Loop Calculation Data Transfer Gate Driver Controller Host Controller PWM Signal S&H ADC Data Transfer PWM Period ADC Set PWM S&H Gate Driver Controller Host Controller PWM Signal Data Transfer PWM Period Control Loop 2 Data Transfer Data Transfer S&H S&H Control Loop 2 S&H ADC Control Loop 1 Set PWM ADC Control Loop 1 Set PWM ADC Control Loop 1 Set PWM ADC 225 High Power Density Modular Six Phase Drive Inverter 3 Embedded Current Measurement using Rogowski Coils The rogowski coil is a well known current sensor for AC currents and is widely used for contactless line current measurement or as laboratory equipment. It is lightweight and compact compared to other sensor principles like compensated sensors with ferrite cores and hall elements. In power electronic applications with variable output frequency like drive inverters rogowski coils are not used for current sensing due to the lack of information about the low frequency or DC current components. Using a special sensor arrangement and evaluation method this problem can be eliminated. [2, 4] 3.1 Fundamental Operating Principle of Rogowski Coils The measurement principle of the Rogowski coil was published in 1912 [5]. It can be realized as a toroidal air-core coil, which surrounds the wire carrying the current to be measured. This basic arrangement is shown in figure 6. The function of this inductive sensor can be understood easily with Ampère's Law in mind:  ∮ 𝐻⃗ ∙ 𝑑𝑠⃗ 𝐼  The toroidal coil collects the magnetic field along a closed loop around the current that is to be measured. With the law of induction (2), it can be determined that the induced voltage of the Rogowski coil is proportional to the time derivative of this current. The mutual inductance M is a constant depending on the geometry of the arrangement.  𝑢 𝑀 ∙   An integrator is used to obtain the time domain current signal. The DC component of the output signal has to be filtered using a high pass filter, because the integration constant is unknown. Figure 6: Basic arrangement of a Rogowski coil [6] 226 High Power Density Modular Six Phase Drive Inverter 3.2 New Embedded Current Measurement Concept For measuring the load current of a drive inverter typically a rogowski coil does not suit well because of the lacking DC current capability. This drawback can be eliminated by arranging the sensor at the DC link connections of one half bridge circuit like it is shown in figure 7. The current in the DC link connections (i 1 and i 2 ) can be assumed to be zero whenever the according switching signal for the power semiconductor is zero. This additional information can be used in an evaluation algorithm to determine the integration constant and therefore the DC current component of the half bridge’s output current. For this special application the rogowski coil itself can be implemented using standard PCB technologies like it is shown in figure 8. This makes the sensor element extremely cheap, very lightweight and compact, very precise and highly reproducible. Figure 7: Current measurement in one inverter leg using inductive sensor Figure 8: Hardware implementation of a rogowski coil on the gate driver PCB U DC i L i 1 T gate driver gate driver rogowski coils integrator u R = M ꞏ di 1 / dt + M ꞏ di 2 / dt i 1 = i L , i 2 = 0, when high-side switch turned on i 1 = 0, i 2 = -i L , when low-side switch turned on controller u R i 2 PCB Rogowski Coils 227 High Power Density Modular Six Phase Drive Inverter A very important hardware component of rogowski coil current sensors is the integrator circuit. Figure 9 shows a very simple but proven circuit. The choice of the operational amplifier is very critical for an accurate operation. It needs to have low noise and extremely low offset and bias for a stable output value at DC conditions but at the same time it needs to have high bandwidth and high slew rate for accurate tracking of the high frequency components at the switching events. Figure 9: Proposed analog integrator circuit 3.3 Evaluation Results Figure 10 and 11 show some measurement results of a prototype. The first graph shows a time series of current values measured with the embedded rogowski coil compared to a reference measurement using an oscilloscope current probe. The second graph shows the good linearity of the measurement method in an X-Y representation of the same data. Figure 10: Measurement result in time domain + - ADC 228 High Power Density Modular Six Phase Drive Inverter Figure 11: Measurement result in X-Y representation 4 Hardware Implementation of the Drive Inverter System The presented technologies are implemented in a six phase drive inverter for the allelectric sports car IMAB-Racer, which is a demonstrator and research platform build at IMAB for research and teaching purposes. [7] 4.1 Power Components The power stage of the inverter is based on six identical IGBT power modules in twolevel half bridge topology. They are rated for 1200 V / 450 A to be prepared for DC voltages of up to 800 V. Today, the nominal voltage of the IMAB-Racer’s battery pack is 480 V. The maximum voltage reaches 525 V when it is fully charged. At 300 A rms load current of the inverter this results in an output power of about 320 kVA at the nominal battery voltage. The inverter can be used to drive a six phase electrical machine, which is the actual configuration. Nevertheless this modular topology can also be used to supply two separate three phase machines. The DC link consist of two MKP film capacitors rated 1 mF / 800 V each and a low inductance bus bar connection. The power modules are mounted on a cold plate using a direct liquid cooling similar to the Danfoss ShowerPower technology [8]. This has been found to be a very efficient cooling method with a low thermal resistance compared to conventional liquid cold plates in earlier research work. 229 High Power Density Modular Six Phase Drive Inverter 4.2 Control Electronics and Sensor Integration According to the proposed layout of figure 2, the control electronics is distributed on six identical intelligent gate driver boards and one central control PCB. The gate driver boards each include the gate driver circuits for the high-side switch and the low-side switch of one half-bridge, the rogowski coil current sensor, an isolated voltage measurement, a microcontroller and the RS485 host interface. They are directly mounted onto the IGBT modules like shown in figure 12. Figure 12: Power modules equipped with intelligent gate drivers The main control circuit utilizes a Texas Instruments Tiva C microcontroller which has a total number of eight UART peripherals. The board offers six identical interfaces to the gate drivers and general interfaces for the machine control like encoder interfaces and additional temperature sensor ports. A picture of the control unit is shown in figure 13. Figure 13: Central control unit of the six phase driver inverter PCB Rogowski Coils Microcontroller Conventional current sensor for reference only Host interface Host controller Gate driver interfaces Vehicle and machine interfaces 230 High Power Density Modular Six Phase Drive Inverter 4.3 Inverter and vehicle integration The active components of the drive inverter are supplemented by common mode filter elements and connectors. Everything is integrated into a lightweight rapid prototyping housing (figure 14). Finally the inverter is built into the IMAB-Racer, which is shown in figure 15 and 16. Figure 14: Complete six phase drive inverter 5 Conclusions In this contribution technological approaches to enable modular, scalable and high power density power electronic systems are presented. The intelligent gate driver appears as a sustainable concept opening interesting new possibilities like the application of decentralized and very fast control strategies. Together with the embedded current measurement method using PCB based rogowski coils it creates a design environment perfectly addressing the requirements of fast switching power devices and future power electronic systems for mobile applications. To validate the concepts in practice a six phase drive inverter has been built and set into operation in a real vehicle application. It shows good performance, the expected high power density and impressively demonstrates the advantages of the proposed system design. 231 High Power Density Modular Six Phase Drive Inverter Figure 15: Drive inverter integrated to the research platform IMAB-Racer Figure 16: Demonstrator and research platform “IMAB-Racer” Drive inverter Battery pack Brake system 232 High Power Density Modular Six Phase Drive Inverter Literature [1] N. Langmaack, G. Tareilus, M. Henke, Intelligent Driver Circuit with Robust Serial Interface, IEEE 12th International Conference on Power Electronics and Drive Systems (PEDS 2017), 12.-15.12.2017, Honolulu (HI), USA [2] N. Langmaack, G. Tareilus, M. Henke, Novel Highly Integrated Current Measurement Method for Drive Inverters, IEEE Applied Power Electronics Conference and Exposition (APEC 2016), 20.-24.03.2016, Long Beach (CA), USA [3] M. Henke, G. Tareilus, N. Langmaack, SiC boost converter with high power density for a battery electric sports car, 11th Symposium on Hybrid and Electric Vehicles, 18./ 19.02.2014, Braunschweig [4] N. Langmaack, Technische Universität Braunschweig, DEVICE AND METHOD FOR MEASURING CURRENT, Euro-PCT application EP000003170010A1, 24.05.2017 [5] W. Rogowski, W. Steinhaus, "Die Messung der magnetischen Spannung," Electrical Engineering / Archiv für Elektrotechnik, Volume 1, Issue 4, pp 141- 150, 1912 [6] Luque Alfredo, Rogowski coil.png, Wikimedia Commons, 28.02.2011 [7] T.-H. Dietrich, C. Heister, N. Langmaack, Q. Maurus, C. Uzlu, M. Henke, IMAB- Racer - Research Platform for Efficient Drive Train Components with high Power Density, 16th Symposium on Hybrid and Electric Vehicles, 20./ 21.02.2019, Braunschweig [8] K. Olesen, R. Bredtmann, R. Eisele, ShowerPower - New Cooling Concept for Automotive Applications, Int. Conference Automotive Power Electronics (APE 2006), Paris 233 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Teresa Bertelshofer, Andreas März, Mark-M. Bakran Abstract This paper presents a numerical method combined with a device simulation model used to analyse the parallel connection of several SiC MOSFET dies. Parallel connection is necessary to achieve the desired current carrying capability of main inverters for xEV-drives. With this method, the effect of asymmetries within the chips’ on-state resistance coupled with asymmetries in the threshold voltage is investigated. The investigation result quantifies, to what extent the threshold voltage and the onstate resistance of the chips can vary at the same time, when the output power of the inverter should only be derated by 5 % at most and no single chip in the inverter should be overheated. To achieve this derating limit, the impact of the positive temperature coefficient (PTC) of the output characteristic and the thermal coupling between the paralleled chips is examined. Kurzfassung Diese Arbeit zeigt eine numerische Methode auf, die in Kombination mit einer Bauteilsimulation verwendet wird, um die Parallelschaltung mehrerer SiC MOSFET Dies zu untersuchen. Parallelschaltung ist notwendig, um die gewünschte Stromtragfähigkeit des Hauptinverters eines xEV-Antriebsstrangs zu erreichen. Mit der vorgestellten Methode wird der Einfluss von Asymmetrien innerhalb des Durchlasswiderstandes der Chips und innerhalb der Transferkennlinie untersucht. Das Untersuchungsergebnis ermöglicht eine quantitative Abschätzung, in wie weit Transferkennlinie und Durchlasswiderstand gleichzeitig abweichen dürfen, ohne dass einzelne Chips überhitzen und ohne dass es notwendig wird, die Ausgangsleistung des Inverters um mehr als 5 % zu reduzieren. Um dieses Deratinglimit von 5 % einzuhalten wird der Einfluss der PTC-Charakteristik der Ausgangskennlinie des MOSFETs betrachtet. Als weiteres Einflusskriterium wird die thermische Kopplung innerhalb der Dies untersucht. 234 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation 1 Introduction With the commercial availability of 650 V SiC MOSFETs, their performance in comparison to 650 V Si-inverter modules already used in electric and hybrid vehicles were investigated. It was found that the use of SiC MOSFETs enables a reduction of 70 - 80 % of inverter losses during typical driving cycles compared to Si IGBT inverters [1, 2], which in turn facilitates the use of smaller batteries or an increase of the vehicle range. Therefore, investigating the feasibility of a full-SiC inverter becomes more and more important [3]. The concept of paralleling is fundamental to achieve the desired output power of drive inverters. The challenges of paralleling modules are for example discussed in [4-6]. While many challenges described in these works can be dealt with using carefully designed layouts, there remains a mismatch between modules or chips caused by parameter variations, which are unavoidable during production [6-9]. To ensure that no single chip in the inverter is overheated, a current derating is often applied. The derating, however, should be kept to a minimum, since it directly relates to the necessary chip area for a certain output power. Because SiC is still a costly material, the chip area must be used to full capacity. 2 Differences of Si IGBT and SiC MOSFET regarding paralleling Paralleling Si IGBTs in an inverter module is state of the art and a well investigated topic. When paralleling SiC MOSFETs, one must keep in mind important differences to the IGBT. Challenges are for example:  SiC MOSFET dies are considerably smaller than Si IGBT dies → For the same output power (e.g. 100 kW for drive inverters) one needs to parallel a higher number of SiC MOSFETs (even if they have a higher current carrying capability per area) (>10 SiC dies instead of ≈ 3 IGBT dies) → Higher numbers of paralleled chips bear more risk of differences between these chips’ characteristics → Higher risk of deviations in the static and transient current sharing.  Datasheets of SiC MOSFETs specify high variation limits of the chips characteristics → combined with the high number of paralleled chips this contributes to a high risk of parameter deviation.  The transfer characteristic exhibits a negative temperature coefficient (NTC) (see Fig. 1) → The thermally most stressed chip in parallel connection will turn off later and turn on sooner than the other devices → This chip will overheat even more. On the other hand, paralleling is simplified by the PTC behaviour of the output characteristic (see Fig. 3, case 1). Provided that all chips have the same R on (T J )-dependency, then hotter chips have a higher on-state resistance and consequently bear lower currents than the other devices in parallel during the conduction phase. Hence, they generate less conduction losses, which partially relieves their thermal stress. If one chip has a lower R on (T J )-dependency then the other chips, this chip will bear higher currents and thus produce higher losses during the conduction phase. However, since this chip will be hotter than the rest, the PTC characteristic again mitigates the effect of the described imbalance in the static current sharing. 235 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Fig. 1: Typical transfer characteristic of SiC MOSFET Fig. 2: Output characteristic of SiC MOSFET (case 1) Fig. 3: Temperature dependency of on-state resistance 3 Calculation Basis In this investigation, the number of necessary paralleled 650 V/ 17 mΩ SiC MOSFETs [10] to enable a current carrying ability of I rms,max = 565 A at fsw = 10 kHz are calculated. Important conditions for this calculation, e.g. maximum junction temperature (= 150 °C), specific thermal resistance and module stray inductance (influencing the switching speed and therefore switching losses) are based on an established Si-IGBT inverter module [11]. This reference module was chosen because it represents a challenging case for paralleling chips with switching loss imbalances. Due to its high stray inductance in comparison with low-inductive module prototypes, the switching losses have a high share in the total losses, even at full load. 11 paralleled dies of a size that is commercially available are necessary for the desired current rating of 565 A, resulting in an active area of ≈ 2 cm 2 per switch. The transfer 236 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation and output characteristic of the whole switch is shown in Fig. 1 and Fig. 2, respectively. The temperature dependency of the on-state resistance is shown as case 1 in Fig. 3. In order to quantify to what extent the PTC behaviour of the on-state resistance can mitigate the overheating of one chip caused by switching loss imbalance, two other possibilities are taken into account as well: The theoretical situation where the on-state resistance shows no temperature dependency (case 2 in Fig. 3), and the situation where the MOSFET’s channel resistance is reduced further and consequently shows a more prominent temperature dependency of R on (case 3 in Fig. 3). In order to ensure that the relative increase of R on is the only determining factor in the following analysis, the current carrying ability I rms,max and the ratio P cond / P total at full load must stay the same (at full load, the conduction losses make up 63 % of the total losses). Therefore, R on (150 °C) has to be identical for all three cases. 4 Device simulation and verification The device simulation model that was developed for this investigation is a modified behavioural model. As input the output and transfer characteristic for different dc-link voltages and junction temperatures and the small signal capacities are provided in look-up tables. It should remodel the exact switching transients of a SiC MOSFETs in a specific layout. As a first step, the model was validated with experimental data of the switching behaviour of a single chip. The chip was measured in a double pulse test setup with increased stray inductance, in order to ensure overvoltage peaks identical to the peaks that would occur if the chip was used in the reference module. The test setup also features an increased commonsource inductance, which emulates L cs found in the reference module [12]. The gate drive was adapted accordingly. Only this way the switching losses can be measured correctly [2]. The validation of the device simulation was conducted for different current values, dclink voltages, junction temperatures and gate driving methods. The model offered low deviations of < 3% for predicting the MOSFET’s switching losses. To ensure that the model’s accuracy is also sufficient to evaluate a chip’s switching losses in parallel connection, the current distribution between two discrete MOSFETs was measured. Since this investigation focuses on the effect of device mismatches and not on circuit mismatches, the setup itself was designed as symmetrical as possible, with special attention being paid to ensure identical commutation inductances L D and L S and identical common source inductances L cs of the two paralleled MOSFETs. The switching cell of the DUT (LSS) is shown in Fig. 4. The same driver is connected to the gate and source legs of the paralleled MOSFETs in a TO-247 package. Hence, special arrangements have to be taken to measure the current sharing. In this measurement the drain currents I D,A and I D,B are of interest, since these are the currents that thermally stress each device. High-bandwidth coaxial shunt resistors are used, which must be inserted in the power source path of the MOSFETs, so that device voltage and device current can be measured at the same time. 237 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Now, however, one has to make sure that the current measured at the power source is identical to the drain current of the devices, and no circulating current (marked with a blue dashed line in Fig. 3) is allowed to flow over the shared source connection of the driver. This is achieved by inserting an inductance L S’ (for high frequency decoupling) and a resistance R S’ (for low frequency and DC decoupling) into the source connection of the driver. Only then, the shunt can measure a signal corresponding to I D,A and I D,B . The hardware realisation is shown in Fig. 5. It should be noted, that this setup is used for the validation of the device simulation only. The simulated circuit for the investigation in the next chapters is based on the reference module. The MOSFETs used in this setup feature a shift in the transfer characteristic and therefore show a distinct transient current imbalance. Exchanging the MOSFETs’ position in the test setup and comparing the switching behaviour before and after the exchange guarantees that the current imbalance is mainly caused by the chips’ inherent characteristic and not by the test setup. Fig. 5: Hardware prototype for validation of device simulation. For better display, the driver for the HSS MOSFETs is not connected. Fig. 6 shows both the simulated and measured current imbalance at a turn-on and turn-off event at I D,total = 100 A and U DS = 400 V. The simulation can recreate the measured switching behaviour with a good accuracy. Compared to chip A, whose transfer characteristic is in good accordance with the datasheet, chip B’s transfer characteristic features a shift of ≈ 350 - 500 mV to the left towards lower gate threshold voltages U th . Hence, chip B has to conduct more current during the Miller phase, turns on sooner and turns off later than chip A. Additionally, the combination of the on-state resistance and the position in the test setup leads to a small DC current imbalance. Fig. 7 shows the measured and simulated relative increase of chip B’s switching losses caused both by the difference in the transfer characteristic and the small DC current imbalance. This value will subsequently be called α tot and is defined as: 𝛼 𝐼 𝐸 , 𝐼 𝐸 , , 𝐼 𝐸 , , 𝐼 for 𝐼 , 0 … 100𝐴 Fig. 4: Current measurement of two paralleled SiC MOSFETs 238 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Fig. 6: Simulated and measured current sharing of two paralleled SiC MOSFETs Fig. 7: Simulated and measured relative switching loss increase of chip B E sw,B is chip B’s switching losses, when paralleled with chips that behave differently (in this case chip A), while E sw,sym,B is its switching losses, when paralleled with chips that behave identically to chip B. Again, it can be established that the simulation is well suited for predicting how much a chip’s switching losses increase, when this chip’s transfer characteristic is shifted to the left towards lower threshold voltages. A small deviation between simulation and reality was to be expected, since in the simulation a constant shift was applied, but in reality the discrepancy between chip A’s and the datasheet’s transfer characteristic is slightly different for each current value. The deviation between simulated and measured α tot for low switching currents plays only a minor role in the following derating calculation, as will be shown later. In the next chapter, the parallel connection of 11 dies with arbitrary R on and transfer characteristic shifts is evaluated with the help of the device simulation based on Fig. 8. Compared to Fig. 4 all chips share the same L S , representing part of the dc-link inductance and the inductance of the reference module’s screw contact. The potential after L cs is short-circuited via the Kelvin-Source of the driver. Fig. 8: Equivalent circuit of simulation 239 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation 5 Simulation Results and Derating Calculation The 11 chips are divided into two categories: Category A, where the complete transfer characteristic is shifted to the right towards higher gate threshold voltages U th and category B, where the transfer characteristic is shifted towards lower U th . Similarly, category A’s onstate resistance is shifted towards higher values, category B’s resistance is shifted towards a lower R on . This is done by multiplying R on for all current and temperature values by a certain constant factor (see Fig. 9). Comparing the typical threshold voltage U th,typ for all admissible junction temperatures and the minimum and maximum value for U th given in the datasheet, U th can deviate 700 mV from the typical value in both directions. The same deviation is assumed for the pinch-off voltages for higher currents (equivalent to a shift towards the left or right of the transfer characteristic). Similarly, the datasheet gives a limit for the maximum deviation in the on-state resistance: R on,max on chip level can be up to 17.2 % higher than R on,typ . A smaller deviation towards lower on-state resistances is assumed: R on,min can be up to 8.6 % lower than R on,typ [10]. In total, the maximum deviation of the onstate resistance is represented by the value ΔR on,AB = 25.8 %. Unfortunately, due to the physical properties of the channel resistance, which in 650 V devices make up a large portion of the total on-state resistance, the chip with the lowest threshold voltage also often features the lowest on-state resistance [13]. As a consequence, this chip will be overstressed during switching as well as during conduction. 5.1 Algorithm for derating calculation In a worst-case scenario, 10 of the n total = 11 chips per switch belong to category A and only n B = 1 chip belongs to category B, which means that n B takes up only 9% of the total chip area. As a first step the switching losses for the single chips are simulated for I D,total = 0...800 A without changing the transfer characteristic (= symmetric situation). These reference values are referred to as E sw,sym,A and E sw,sym,B . In the next steps, the transfer characteristic is shifted 700 mV to the right in category A; the same voltage shift to the left is applied to category B. The relative increase of chip B’s switching loss energies was defined in eq. (1). For the maximum U th -variation of 700 mV, the resulting increase of chip B’s switching losses caused only by the shift in the transfer characteristic is displayed as the blue curve in Fig. 10. At low current values, chip B even produces almost twice the switching loss energy E sw,B than it would generate without a U th - Fig. 9: Maximum R on deviation of category A and B Fig. 10: Additional switching loss energies (α) of category/ chip B 240 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation asymmetry (α ΔUth ≈ 100% @ I D,total → 0 A). At I D,total = 800 A E sw,B is 29.3 % higher than E sw,sym,B . The chips of category A produce less switching loss energies compared to E sw,sym,A , so that the sum of all the paralleled chips’ switching losses remains constant and is not influenced by the U th -asymmetry. Fig. 11: Effect of ΔR on on turn-off @ ΔU th,device = ± 700 mV Usually, one would expect the turn-off behaviour of SiC MOSFETs to be independent of their previous conduction phase, since they are unipolar devices (in contrast to IGBTs, where the plasma distribution and thus also the turn-off process is coupled with the current in the device during the previous conduction phase). This is indeed the case, when the MOSFET dies are placed in a setup without a common-source inductance (L cs in Fig. 8). Fig. 11a shows a turn-off process of I D,total = 400 A, where each half of the chips behave identically and feature a threshold voltage shift of 700 mV to the left and right respectively. It must be noted, that this fifty-fifty distribution is only used in this figure for clarity reasons. The solid line shows the turn-off event, when all the chips feature the same on-state resistance and there is no DC current imbalance. The dashed line shows the turn-off event, when the chips of category B have a lower onstate resistance then the chips of category A. Since there is no L cs that impedes the current imbalance before the main di D / dt-phase, the current is distributed immediately according to the transfer characteristic at the beginning of the switching event. Hence, the turn-off losses are independent of the R on mismatch. The situation is different when a distinct common-source inductance is included in the layout. It partially counterbalances the current mismatch before and during the main di D / dt -phase (Fig. 11b). This effect is explained in detail in [3]. However, L cs can only limit a current change. Since I D of the chips of category B already is higher when their on-state resistance is smaller then R on,A , the turn-off current reaches a higher value, although the peak current is still smaller than it would be without the presence of L cs . As a consequence, the R on mismatch influences the turn-off losses. By contrast, the turn-on losses are not affected by ΔR on . When the algorithm for the derating calculation is explained, it will become apparent how important it is that the switching loss mismatch caused exclusively by the threshold voltage variance and the switching loss mismatch caused exclusively by different on-state resistances can be calculated independently of each other. This is paramount, since threshold voltage and on-state resistance are both temperature dependent and the junction temperature of the chip is in turn dependent on the mismatch of switching and conduction losses. 241 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation The blue curve for α ΔUth in Fig. 10 shows the relative switching loss mismatch in a module layout with L cs for the case 1 chip vs. 10 chips and ΔU th = ±700mV, but no ΔR on . The orange curve for α ΔRon in Fig. 10 shows the additional relative switching loss mismatch, when the maximum admissible on-state resistance deviation according to the datasheet occurs. It is only affected by ΔR on , but not by ΔU th . For different deviations in the threshold voltage α ΔUth can be updated via the blue curve in Fig. 12. Fig. 12: Update of α; above: regarding change in ΔU th caused by ΔT AB ; below: regarding change in ΔR on This dependence behaves almost linearly for all current values. Similarly, the additional switching loss mismatch caused by DC current imbalance α ΔRon can be updated via the orange curve in Fig. 12. It, too, is an almost linear function valid for all current values and not dependent on the threshold voltage variance. Chip B’s switching loss energy for all current values I D can be evaluated with eq. (2). 𝐸 , 𝐸 , , 𝛼 𝛼 ∙ 𝐸 , , 1 𝛼 ∙ 𝐸 , , During a sine-halfwave with output current I rms all current values between 0 A and √2 ∙I rms are switched. If α tot is evaluated over a complete sine-halfwave i D (φ), the switching losses behave according to the following equation: 𝑃 , 𝐼 𝑓 2𝜋 1 𝛼 𝑖 𝜑 ∙ 𝐸 , , 𝑖 𝜑 𝑑𝜑 for 𝐼 , 0 … 565𝐴 At full load (I rms = 565 A), the shown example for α ΔUth in Fig. 10 results in 35.5 % higher switching losses P sw,B of category B. This value is mainly influenced by the values of α ΔUth for I D = 600...800 A. So, even for the low deviations of α at low switching currents shown in Fig. 7, P sw,B at full load is still correct. Fig. 13 describes the numerical approach to establish the derating of the inverter output power that is necessary in order to prevent thermally overstressing the single chip of category B. For each I rms value the initial switching loss mismatch for a certain device variation ΔU th and ΔR on are calculated via Fig. 10 and Fig. 12. The initial DC current imbalance and thus conduction loss mismatch are determined also with ΔR on,device . 242 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Since the single chip of category B has a decreased on-resistance, it has to bear higher currents then a single chip of category A. Subsequently, the steady-state junction temperature of category A and B is calculated with the afore established mismatch in P cond and P sw . This step also includes the possibility to consider thermal coupling between the paralleled chips (Fig. 14): It is realised by inserting a coupling layer in the thermal path, which can be moved between fluid and junction. The temperature of the coupling layer is calculated with the sum of all the chips’ losses and the thermal resistance between fluid and coupling layer. T fluid is assumed to be constant. The junction temperatures of the chips of category A and B are calculated with their mismatched losses and the thermal resistance between junction and coupling layer, respectively. In this work no additional thermal asymmetries are examined, meaning all chips feature the same area specific thermal resistance. Fig. 13: Algorithm for calculation of derating 243 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Fig. 14: Effect of thermal coupling Because the single chip in category B is stressed with higher switching and conduction losses (see Fig. 15), it reaches a higher junction temperature than the chips of category A. It also bears a higher current. Both effects raise chip B’s on-state resistance (see Fig. 2 and Fig. 3) and reduce ΔR on . In the next iteration the current distribution is determined with the new values of R on,A (T J,A ,I D,A ) and R on,B (T J,B ,I D,B ). Since ΔR on is now slightly reduced, α ΔRon is also reduced according to Fig. 12. However, since chip B is hotter than the rest of the chips, the discrepancy between chip A and B’s transfer characteristic is further increased according to ΔT AB (because of its NTC behaviour) and the switching loss mismatch becomes even more pronounced. The increase of ΔU th,AB is calculated with ΔT AB and the temperature coefficient of U th given in the datasheet ( ≈ 6.9 mV/ K for relevant junction temperatures of 135 ... 150 °C). Hence, α ΔUth rises according to Fig. 12. Afterwards, chip A and B’s switching losses for the next iteration are updated (see eq. (3)). These steps are repeated by an iterative solver until the current distribution for a certain I rms,total value reaches steady state. Fig. 15: P cond and P sw distribution for single chip of category A and B @ max. allowable device variations of U th and R on The current I B,max , where chip B reaches T J,max , determines the current derating (see Fig. 16): 𝑑𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝐼 , 𝐼 , 𝐼 , 244 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation For this algorithm to work, the junction temperature of a chip must be temporarily constant while conducting the sine half wave of a certain I rms,total value. This is fulfilled, if the output frequency of the inverter is higher than ≈ 50 Hz [14]. In the next chapter, not the derating itself is the desired output parameter of the algorithm, but the information, to what extent the threshold voltage and the on-state resistance of the chips can vary at the same time without exceeding a given derating of e.g. 5 %. To obtain that information, the input for ΔR on,A/ B,device is fixed and ΔU th,device is varied until the desired derating value is achieved. 5.2 Limit of device variation to achieve derating requirements To better understand the derating caused by R on or U th device mismatches, Fig. 17 shows the necessary derating for three situations: The blue curve shows the derating, when ΔR on,device is varied from zero to its maximum value and ΔU th,device is set to zero at the same time. However, there remains a switching loss mismatch, because a temperature difference between the chips still causes a threshold voltage discrepancy and the chip with the lower R on has to turn-off a higher current. Similarly, the orange curve shows the situation, when ΔU th,device is varied from zero to its maximum value and ΔR on,device is set to zero. The purple curve shows the derating, when a variation of ΔU th,device and ΔR on,device is inserted in the algorithm at the same time. Compared to that, the manual sum (black curve) is slightly higher but still a good fit, indicating that the effect of R on and U th device mismatches could also be considered separately. Fig. 17: derating for different ΔR on,device and ΔU th,device Fig. 16: Definition of derating 245 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Fig. 18 shows the maximum allowable deviation of ΔU th,device as a function of ΔR on,device , so that the derating stays within the limit of 5 %. For the PTC-characteristic of R on given in the datasheet (case 1) the maximum device deviation of ΔU th = 700mV is only admissible, if there is no deviation in ΔR on,device at the same time. At ΔR on,AB = 11.3 % only the imbalance in the switching losses caused by the temperature difference of the chips is allowed. ΔU th,device has to be 0. These two reference points can also be extracted from Fig. 17. When there is an increased PTC-behaviour of the on-state resistance (e.g. by decreasing the channel resistance) the curve ΔU th (ΔR on,AB ) is shifted to the right, which enables higher parameter deviations. The theoretical case 2 without any PTC-behaviour shows that even at ΔR on,AB = 0 a current derating of 5 % might not suffice to prevent the overheating of one chip. Fig. 18: max. ΔU th for different ΔR on (T J ) To increase the range of the admissible U th and R on deviation and therefore avoiding the need of a preselection of the chips, another solution is to guarantee a good thermal coupling between the devices. In the previous calculation, it was assumed that the thermal path of category A and B are completely separated until they reach the cooling fluid (see Fig. 14 left). Now the effect of a thermal coupling layer between junction and fluid that merges the loss flow of chips A and B, should be investigated (see Fig. 14 right). The closer the coupling layer is situated to the junction, the higher the loss flow 𝑄 ′ can be for the same temperature difference Δ𝑇 Δ𝑇 . The coefficient for the thermal coupling thc is defined as the ratio of the thermal resistance between coupling and fluid R th,cl-F and the total thermal resistance of all chips between junction and fluid. Fig. 19 shows the admissible ΔU th (ΔR on,AB ) deviation for different coefficients thc. As a reference point the curve for thc = 0 is shown. This is the same curve as in Fig. 18, case 1. The higher the thermal coupling, the higher the loss imbalance between the chips of category A and B can be. For the theoretical case of perfect thermal coupling of the chips (thc → 1) the loss imbalance between chips A and B becomes irrelevant. The new limit is defined only by the total inverter losses and thus by ΔR on,AB : Since the chips of category A make up 91% of the total chip area, the shift to higher R on,A increases the whole inverter losses. The derating limit would only be exceeded when R on of category A was 19 % higher than the typical value. This in turn means, that for perfect thermal coupling ΔU th,device and ΔR on,device can vary arbitrarily, as long as they stay within the datasheet limits. 246 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation Fig. 19: max. ΔU th for different thermal coupling 6 Conclusion This paper presents a numerical approach and device simulation to analyse the parallel connection of 11 SiC MOSFETs. The worst-case scenario regarding paralleling, namely that one chip exhibits a smaller threshold voltage and thus is stressed by higher switching losses while at the same time exhibits a smaller on-state resistance and thus is stressed by higher conduction losses, was investigated. A necessary current derating factor was defined, so that no single chip is thermally overstressed. Since this derating factor directly relates to the overdimensioning of the chip area and therefore SiC material cost, it should be limited to e.g. 5 %. To achieve this goal, the required limits of ΔR on and ΔU th device mismatches were given. In order to expand the admissible range of the parameter deviation and omit the necessity for preselection of the chips, the beneficial effect of the PTC-behaviour of the on-state resistance and thermal coupling between the devices were investigated. References [1] M. Hayes, J. Casady, and J. Palmour, “650V, 7mOhm SiC MOSFET Development for Dual-Side Sintered Power Modules in Electric Drive Vehicles,” PCIM, 2016. [2] T. Bertelshofer, R. Horff, A. Maerz, and M.-M. Bakran, “A performance comparison of a 650 V Si IGBT and SiC MOSFET inverter under automotive conditions,” PCIM, 2016. [3] T. Bertelshofer, A. Maerz, and M.-M. Bakran, “Derating of parallel SiC MOSFETs considering switching imbalances,” PCIM, 2018. [4] M. Wissen, D. Domes, and A. Groove, “Effects of influencing the individual leg inductance in case of paralleling modules on basis of XHPTM 3 and EconoDU- ALTM,” PCIM, 2017. [5] J. Weigel, J. Boehmer, A. Nagel, and R. Kleffel, “Paralleling High Power Dual Modules: A Challenge for Application Engineers and Power Device Manufacturers,” EPE, 2017. 247 Limits of SiC MOSFETs’ Parameter Deviations for Safe Parallel Operation [6] A. Wintrich, J. Nascimento, and M. Leipenat, “Influence of parameter distribution and mechanical construction on switching behaviour of parallel IGBT,” PCIM, 2006. [7] U. Scheuermann, “Statistical Evaluation of Current Imbalance in Parallel Devices,” PCIM, 2016. [8] D. Sadik, J. Colmenares, J. Rabkowski, and H. Nee, “Experimental investigations of static and transient current sharing of parallel-connected Silicon Carbide MOSFETs,” EPE, 2013. [9] Y. Mao, Z. Miao, C. Wang, and K. Ngo, “Balancing of Peak Currents Between Paralleled SiC MOSFETs by Drive-Source Resistors and Coupled Power-Source Inductors,” IEEE Trans. Ind. Electron., vol. 64, no. 10, 2017. [10] Datasheet SiC MOSFET Bare Die S4003, Rohm, 2017. [11] Datasheet FS800R07A2E3_B31, Infineon, 2017. [12] G. Engelmann, S. Quabeck, J. Gottschlich, and R. De Doncker, “Experimental and simulative investigations on stray capacitances and stray inductances of power modules,” EPE, 2017. [13] D. Lu, H. Takubo, S. Takano, and Y. Suzuki, “Paralleling Six 320A 1200V All-SiC Half-Bridge Modules for a Large Capacity Power Stack,” IPEC, 2018. [14] A.Wintrich, U. Nicolai,W. Tursky, and T. Reimann, Application Manual Power Semiconductors. Semikron International GmbH, 2015. 248 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Janis Lehmann, Benjamin Löwer, Mohamed Ayeb, Ludwig Brabetz Abstract Current development trends such as automated driving functions and autonomous vehicles lead to an increasing, average energy demand as well as to a significant higher peak power consumption of the electrical system. In order to ensure high reliability, in the first step, the existing electric components such as batteries, converters or alternators can be upscaled. Within the BMWi-sponsored project ToSKa, predictive power management is being developed as an alternative to the usage of larger or additional components. The information available in the automated vehicle on the condition of the vehicle, the planned driving trajectory and the environmental characteristics are used intelligently to estimate the future power needs of the electrical consumers in the electrical system. After estimating the power demand, it’s compared with the capability of the available electrical sources. Criteria-based counteractions are calculated on the basis of criticality criteria and then, using the communication busses, implemented at the level of consumers and sources. Kurzfassung Aktuelle Entwicklungstendenzen wie automatisierte Fahrfunktionen und autonome Fahrzeuge führen zu einem steigenden, mittleren Energiebedarf sowie zu einer deutlich höheren Spitzenleistung des Bordnetzes. Um eine hohe Zuverlässigkeit und Ausfallsicherheit zu gewährleisten, können im ersten Schritt die vorhandenen Elemente wie Speicher, Wandler oder Generatoren hochskaliert werden. Im Rahmen des BMWi geförderten Projektes ToSKa wird als Alternative zum Einbau größerer oder zusätzlicher Komponenten ein prädiktives Leistungsmanagement entwickelt. Dabei werden die im automatisierten Fahrzeug vorhandenen Informationen über den Zustand des Fahrzeugs, die geplante Fahrtrajektorie und die Umwelteigenschaften intelligent genutzt, um den zukünftigen Leistungsbedarf der elektrischen Verbraucher im Bordnetz abzuschätzen. Nach Abschätzung des Leistungsbedarfes wird dieser mit der Leistungsfähigkeit der verfügbaren Quellen und Wandler abgeglichen. Auf der Basis von Kritikalitätskriterien werden situationsangepasste Gegenmaßnahmen berechnet und mithilfe der Kommunikationsbusse auf Verbraucher- und Erzeugerebene implementiert. 249 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge 1 Prädiktive Vermeidung leistungskritischer Situationen für eine stabile Systemspannung Die Einführung hochautomatisierter Fahrfunktionen führt zu einem steigenden Leistungsbedarf im automobilen Energiebordnetz. Die versorgenden Komponenten werden zusätzlich durch die Implementierung neuer oder elektrifizierter Hochleistungsverbraucher wie aktive Fahrwerksysteme oder elektrische Aufladesysteme belastet. Bei der klassischen Auslegung der Energieversorgung im Fahrzeug wird diese Steigerung durch eine Skalierung der elektrischen Quellen, Wandler oder Generatoren kompensiert. Für den überwiegenden Anteil der Fahrsituationen ist die elektrische Versorgung überdimensioniert und im Hinblick auf Bauraum, Kosten und Fahrzeugmasse nicht optimal. Fortschrittliche Fahrerassistenzfunktionen und (teil-)automatisierte Fahrzeuge erfordern eine sehr genaue Erfassung der Fahrzeugumwelt und Prädiktion des Verhaltens anderer Verkehrsteilnehmer. Die gesammelten Informationen werden bestimmungsgemäß zur Trajektorienplanung des Eigenfahrzeugs verwendet. Durch die intelligente Vernetzung der im automatisierten Fahrzeug verfügbaren Informationen können leistungskritische Situationen im Fahrzeugbordnetz prädiziert werden. Mithilfe einer Zustandserfassung der Quellen kann die Kritikalität der Situation abgeschätzt und bei Bedarf entsprechende Gegenmaßnahmen eingeleitet werden, um den Einbruch der Systemspannung auf ein definiertes Maß zu begrenzen. Ein prädiktives Leistungsmanagement kann die steigenden Leistungsanforderungen damit ohne wesentliche Hardwareerweiterungen ausgleichen. Bekannte Ansätze für ein prädiktives Leistungsmanagement verfolgen einen dezentralen Ansatz, bei dem den Verbrauchern sowie übergeordneten Steuerungsebenen ein Managementalgorithmus zugeordnet wird [1]. Das hier vorgestellte Verfahren basiert auf einer zentralen Steuerungseinheit, womit der Implementierungsaufwand minimiert wird. Das entwickelte Algorithmuskonzept ist in Abbildung 1 dargestellt. Abbildung 1: Algorithmuskonzept des prädiktiven Leistungsmanagements 250 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Der Ausgangspunkt des Konzeptes ist das Fahrzeug mit der Umfeldsensorik sowie den Steuergeräten der Verbraucher, dem Generator, den Batterien und dem DC-DC- Wandler. Im Block Erfassung des aktuellen Zustandes werden die relevanten Informationen über vorrausfahrende Fahrzeuge sowie vor dem Fahrzeug befindliche Objekte mit sämtlichen Eigenschaften aus der Umfeldsensorik extrahiert. Die Rohdaten der verschiedenen Systeme wie Kamera und Radar müssen fusioniert werden, um alle erfassten Umfeld-Informationen für die jeweils erkannten Objekte nutzen zu können. Außerdem werden die aktuellen Aktivierungszustände der Verbraucher sowie allgemeine Informationen zum Fahrzeugzustand erfasst. Zur Bewertung der Leistungsfähigkeit der Quellen werden Zustandsgrößen von Generator, Batterien und DC/ DC- Wandlern abgeleitet. Die gesammelten Daten werden in den Abschnitten der Prädiktion genutzt, um den Strombedarf der Verbraucher zu prognostizieren. Dabei wird zwischen den fahrdynamikrelevanten und nicht fahrdynamikrelevanten Verbrauchern sowie der Grundlast unterschieden. Die entwickelten Prognoseverfahren werden in Abschnitt 2 und 3 näher erläutert. Im Block Leistungsmanagement wird der erwartete Summenstrom der Verbraucher mit der Leistungsfähigkeit der Quellen abgeglichen. Dabei werden Erkenntnisse aus dem laufenden Projekt über fehlertolerante Bordnetztopologien und die dementsprechend verbauten elektrischen Quellen angewendet. Erkannte, kritische Zeitbereiche werden mit definierten Kritikalitätskriterien abgeglichen. Schlussendlich werden für die erkannten, leistungskritischen Situationen angepasste Gegenmaßnahmen auf Verbraucher- und Quellenebene berechnet und an das Fahrzeug übermittelt. 2 Probabilistische Beschreibung des erwarteten Strombedarfs der nicht fahrdynamikrelevanten Verbraucher In dem Energiebordnetz von Fahrzeugen befindet sich eine Vielzahl elektrischer Verbraucher. Für die Gruppe der Verbraucher, die nicht von der Fahrdynamik abhängig sind, wurden die 30 für das prädiktive Leistungsmanagement relevanten Verbraucher ausgewählt. Dabei wurde die Leistungsaufnahme und das elektrische Verhalten der Verbraucher betrachtet. In dem Demonstratorfahrzeug (Ford Mondeo) wird von 27 der 30 Verbraucher der jeweils aktuelle Aktivierungszustand über die Kommunikationsbusse gesendet, der dem prädiktiven Leistungsmanagement in Echtzeit zur Verfügung steht. Die verbleibenden drei Verbraucher sind nicht an die Kommunikationsbusse angebunden. Um den Aktivierungszustand dieser Verbraucher auslesen zu können, wird im laufenden Projekt ein Verfahren entwickelt, das aus dem Summenstrom aller Verbraucher auf die Aktivierung eines Verbrauchers schließt. Dabei werden Verfahren der Mustererkennung mithilfe neuronaler Netze verwendet. Ausgehend von dem aktuellen Aktivierungszustand der Verbraucher wird nun eine Prognose für den erwarteten Strom der nicht fahrdynamikrelevanten Verbraucher im Prognosehorizont entwickelt. Das Verfahren ist in Abbildung 2 dargestellt. 251 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Abbildung 2: Verfahren zur Prognose der nicht fahrdynamikrelevanten Verbraucherströme Zunächst wird mithilfe von Bayes’schen Netzen eine Zustandswahrscheinlichkeit p on bzw. p off des betrachteten Verbrauchers aus definierten Umfeldparametern abgeleitet. Die Umfeldparameter können deterministischer oder probabilistischer Natur sein. Zur Berechnung der Zustandswahrscheinlichkeiten werden für jeden Verbraucher Übergangstabellen definiert. Die Übergangstabellen können durch die Auswertung umfangreicher Messungen zum Verhalten des betrachteten Verbrauchers angelernt oder durch empirische Auswahl bestimmt werden. Die Umfeldparameter werden in diskrete Kategorien eingeteilt und deren bedingte Wahrscheinlichkeiten in der Übergangstabelle den Aktivierungsstufen des jeweiligen Verbrauchers zugeordnet. Interpretiert werden diese Wahrscheinlichkeiten als Zustandswahrscheinlichkeiten für ein verbraucherabhängiges Zeitintervall. In Tabelle 1 ist beispielhaft die empirisch ausgewählte Übergangstabelle für eine Sitzheizung abgebildet. Die eingehenden Umfeldparameter sind die Innenraumtemperatur temp in sowie die Sitzbelegung seat_occupancy. Tabelle 1: Übergangstabelle des Bayes’schen Netzes für eine Sitzheizung P (X | temp in , seat_occupancy) (a 1, b 1) cold, ~occupied (a 1, b 2 ) cold, occupied (a 2, b 1) normal, ~occupied (a 2, b 2) normal, occupied (a 3, b 1) warm, ~occupied (a 3, b 2) warm, occupied x 1 =on 0 0.7 0 0.2 0 0 x 2 =off 1 0.3 1 0.8 1 1 252 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Für die Klassifizierung der Innenraumtemperatur wird eine Fuzzylogik verwendet, die je nach Temperatur eine Wahrscheinlichkeit für die jeweilige Klasse cold, normal und hot liefert (Abbildung 3). Für die Innenraumtemperatur sind diese so ausgelegt, dass eine vom Menschen als angenehm empfundene Temperatur zwischen 20°C und 22°C die Klasse „normal“ beschreibt. Für die Außentemperatur werden wiederrum andere Bereiche verwendet. Abbildung 3: Fuzzy-Logik des Umfeldparameters Innentemperatur Bei einer Innenraumtemperatur von beispielsweise 18.2°C ergeben sich Wahrscheinlichkeiten von P(cold) = 0.9, P(normal) = 0.1 und P(hot) = 0. Aus diesen Wahrscheinlichkeiten, einem belegten Sitz und der Übergangstabelle (Tabelle 1) lässt sich eine Wahrscheinlichkeit für den Verbraucherzustand wie folgt berechnen. 𝑃 𝑥 𝑃 𝑥 𝑎 , 𝑏 ∙ 𝑃 𝑎 ∙ 𝑃 𝑏 (1) 𝑃 𝑋 0 0.7 0 0.2 0 0 1 0.3 1 0.8 1 1 ∙ ⎝ ⎜⎜ ⎛ 0.9 ⋅ 0 0.9 ⋅ 1 0.1 ⋅ 0 0.1 ⋅ 1 0.0 ⋅ 0 0.0 ⋅ 1⎠ ⎟⎟ ⎞ 0.65 0.35 (2) Für den aktuellen Zustand der Umfeldparameter ergeben sich somit Zustandswahrscheinlichkeiten von P on (X) = 0.65 und P off (X) = 0.35, unabhängig vom aktuellen Zustand. Dieser wird erst bei der Berechnung des aktuellen Stroms der Verbraucher berücksichtigt. In Abbildung 4 ist die Zustandswahrscheinlichkeit der Sitzheizung nach dem Motorstart im Fahrzeug dargestellt. Die Zustandswahrscheinlichkeit der Sitzheizung nimmt wie zu erwarten mit steigender Innenraumtemperatur ab. probability 253 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Abbildung 4: Zustandswahrscheinlichkeit bei sich ändernder Innentemperatur Um die Zustandswahrscheinlichkeit der Sitzheizung noch besser zu beschreiben, können weitere Umfeldparameter wie der aktuelle Zustand und die Zustandszeit berücksichtigt werden. Neben einem durch die Prognose vorhergesagtem Einschaltvorgang (großes P on bei inaktivem Verbraucher) sind auch geringe Zustandswahrscheinlichkeiten bei inaktivem Verbraucher für die Leistungsprognose von Wert, da der Einschaltstrom des Verbrauchers somit für die Berechnung der Leistungsreserve innerhalb des Prognosehorizonts nicht berücksichtigt werden muss. Insgesamt werden durch die Netze die Zustandswahrscheinlichkeiten von 30 Verbrauchern berechnet, woraus sich eine komplexere Netzstruktur ergibt, da jeder der Verbraucher von mehreren Umfeldeinflüssen abhängig sein kann. Zur Veranschaulichung der Komplexität ist das bisher verwendete Bayes´sche Netz in Abbildung 5 dargestellt. 0 50 100 150 200 250 300 350 400 450 500 time / [sec] 0 0.5 1 P on 10 20 30 temp in / [°C] P on (seat_heating) temp in 254 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Abbildung 5: Bayes´sches Netz zur Prognose der Zustandswahrscheinlichkeiten der elektrischen Verbraucher Die Zustandswahrscheinlichkeit P X gibt die Wahrscheinlichkeit an, dass der Verbraucher sich aktuell im Zustand X befindet. Vereinfachend kann der erwartete Strom eines Verbrauchers I m durch das Produkt des mittleren Verbraucherstroms I s mit der Zustandswahrscheinlichkeit p on berechnet werden [2]: 𝐼 𝑡 𝐼 𝑡 ⋅ 𝑝 (3) Diese Beschreibung vernachlässigt jedoch die dynamischen Einschaltströme, die bei vielen Verbrauchern auftreten. Die dynamischen Einschaltströme können den statischen Strom der Verbraucher um ein Vielfaches überschreiten und müssen somit im prädiktiven Leistungsmanagement berücksichtigt werden. Um den Schaltzeitpunkt des Verbrauchers abschätzen zu können, wird die Exponentialverteilung zur Beschreibung des Schaltzeitpunktes in Abhängigkeit der Zustandswahrscheinlichkeit gewählt. Die Exponentialverteilung beschreibt die zufällige Zeitdauer zwischen zwei Schaltzeitpunkten, wobei die Schaltevents unabhängig und mit einer konstanten mittleren Schaltrate auftreten [3]. Eine mögliche Exponentialverteilung ist in Abbildung 6 dargestellt. Der Parameter λ der Exponentialverteilung wird mithilfe der Zustandswahrscheinlichkeit aus den Bayes’schen Netzen bestimmt, sodass das Integral unter der Exponentialverteilung im Beobachtungshorizont T watch der Zustandswahrscheinlichkeit in dem Beobachtungshorizont entspricht. In (6) und (7) sind die Terme für λ on und λ off angegeben. Der Beobachtungshorizont T watch wird für jeden Verbraucher festgelegt, wobei die Zeitkonstante der Funktion des Verbrauchers, wie z. B. Aufheizen eines Sitzes oder Beleuchtung, berücksichtigt wird. 255 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge 𝑓 𝑡 𝜆 ⋅ 𝑒 0 𝑓ü𝑟 𝑡 0 𝑠𝑜𝑛𝑠𝑡 (4) 𝑝 𝜆 ⋅ 𝑒 d𝑡 (5) 𝜆 𝑙 𝑛 1 𝑝 𝑇 (6) 𝜆 𝑙 𝑛 1 𝑝 𝑇 (7) Abbildung 6: Parametrisierung der Exponentialverteilung des Schaltmomentes mithilfe der Zustandswahrscheinlichkeit Zur Beschreibung der dynamischen Einschaltströme wird ein generisches, analytisches Verbrauchermodell eingeführt, mit dem die Großzahl der elektrischen Verbraucher im Fahrzeug modelliert werden kann. Die Modellgleichungen für das Einschalt- und das Ausschaltverhalten sind in (8) und (9) dargestellt und sind vom dynamischen Einschaltstrom I d , dem statischen Dauerstrom I s sowie der Zeitkonstante τ abhängig (Abbildung 7). 𝐼 𝑡 𝐼 𝐼 ⋅ 𝑒 𝐼 für 𝑡 0 0 sonst (8) 𝐼 𝑡 0 für 𝑡 0 𝐼 sonst (9) 256 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Abbildung 7: Generisches, analytisches Verbrauchermodell mit Einschaltstrom I on (links) sowie Ausschaltstrom I off (rechts) Zur analytischen Beschreibung des erwarteten Verbraucherstroms I m (t) wird je nach aktuellem Zustand des Verbrauchers der Einbzw. Ausschaltstrom mit der Verteilung des Schaltmoments im Zeitbereich gefaltet [5]. Durch die Parametrisierung mit dem aktuellen λ, das aus p on bzw. p off abgleitet wird, ergibt sich eine statistische Aussage mit Erwartungswert I m (t) sowie Varianz Var(t) für den Ausgangszustand aus: 𝐼 , 𝑡 𝐼 𝐼 ⋅ 𝜆 𝜏 𝜆 𝜏 1 𝑒 𝑒 𝐼 ⋅ 1 𝑒 (10) 𝑉𝑎𝑟 𝑡 𝜆 𝜏 𝐼 𝐼 2 𝜆 𝜏 ⋅ 𝑒 𝑒 2𝑝 𝑡 ⋅ 𝐼 𝐼 𝜆 𝜏 1 𝜆 𝜏 ⋯ ⋅ 𝑒 𝑒 𝑝 𝑡 ⋅ 1 𝑒 𝐼 , 𝑡 ⋅ 𝑒 (11) mit 𝑝 𝑡 𝐼 𝐼 , 𝑡 und für den Ausgangszustand ein: 𝐼 , 𝑡 𝐼 ⋅ 𝑒 (12) 𝑉𝑎𝑟 𝑡 𝐼 ⋅ 𝑒 ⋅ 1 𝑒 (13) 257 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Abbildung 8: Erwarteter Stromverlauf eines Verbrauchers mit Ausgangszustand „aus“ (Beispiel) Mit der statistischen Beschreibung (10) bis (13) kann der Strombedarf der Verbraucher im Prognosehorizont abgeschätzt werden. Dabei wird vorausgesetzt, dass der Verbraucherstrom zum Zeitpunkt t = 0 sec statisch ist. Das gilt in erster Näherung, wenn die Zeit seit dem letzten Zustandswechsel t slc (time since last change) größer als 5τ ist. Für den Zeitraum t slc < 5τ muss der statische Strom I s in (12) wie folgt angepasst werden: 𝐼 , 𝑡 𝐼 ⋅ 𝑒 für 𝑡 5𝜏 𝐼 𝐼 ⋅ 𝑒 𝐼 ⋅ 𝑒 für 𝑡 5𝜏 (14) Bei dem erwarteten Einschaltstrom I m,on (t) kann die Fallunterscheidung vermieden werden, da der Ausschaltstrom des verwendeten Modells keine Dynamik aufweist. Für das prädiktive Leistungsmanagement müssen die je Verbraucher definierten Zufallsvariablen zu einem erwarteten Summenstrom mit Konfidenzintervallen überlagert werden. Der Erwartungswert der Summe von n Zufallsvariablen ist allgemein als die Summe der einzelnen Erwartungswerte definiert. Auf die hier vorliegenden Zufallsvariablen angewendet ergibt sich der erwartete Summenstrom I erw zu: 𝐼 𝑡 𝐼 , 𝑡 (15) Bei der Varianz einer Summe von Zufallszahlen muss die Kovarianz Cov in die Berechnung mit einbezogen werden. Die Kovarianz gibt die Korrelation zwischen zwei Zufallszahlen an. Im vorliegenden Fall werden die Korrelationen in den Bayes’schen I m,on (t) / [A] 258 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Netzen zur Bestimmung der Zustandswahrscheinlichkeiten betrachtet. Somit ergeben sich an dieser Stelle alle Kovarianzen zu null und die Berechnung der Varianz vereinfacht sich zu: 𝑉𝑎𝑟 𝑡 𝑉𝑎𝑟 𝑡 (16) Zur weiteren Betrachtung werden I erw (t) und Var sum (t) an das Leistungsmanagement weitergegeben. Durch die Wahl des Konfidenzintervalls ist das System auf das gewünschte Vertrauensniveau parametrisierbar, wodurch sich die Sensitivität des Leistungsmanagements beeinflussen lässt. Mit dem in Abbildung 8 dargestellten, analytischen Verbrauchermodell wird ein Großteil der nicht fahrdynamikrelevanten Verbraucher im Fahrzeugbordnetz modelliert. Für einige Verbraucher müssen jedoch weitere Modelltypen definiert werden, um sie adäquat abbilden zu können. Im Rahmen des laufenden Projektes wurden zusätzliche Verbrauchermodelle für Verbraucher mit periodischem Stromprofil (z. B. Scheibenwischermotor), Verbraucher mit mehr als zwei Schaltlevel (z. B. Innenraumlüfter) und Verbraucher mit zeitlich verzögerten Schaltstufen und konstanter Verzögerung (z. B. Frontscheibenheizung) definiert. Auf dieser Basis werden die erwarteten Verbraucherströme sowie die Varianzen analytisch bestimmt. 3 Aktivierungsprädiktion der fahrdynamikrelevanten Verbraucher In diesem Abschnitt wird die Prädiktion der fahrdynamikrelevanten Verbraucher betrachtet. Diese Verbrauchergruppe wird gesondert betrachtet, weil die Verbraucher in der Regel einen sehr hohen Leistungsbedarf mit großen Stromgradienten aufweisen. Bei einer statistischen Beschreibung der Verbraucherströme wie in Kapitel 2 kann der situationsselektive Leistungsbedarf kaum abgebildet werden. Zu den fahrdynamikrelevanten Verbrauchern gehören in jedem Fall die elektrische Lenkung sowie die Aktuatoren der Bremssysteme. Je nach Ausstattungsgrad des Fahrzeugs können weitere Verbraucher wie z. B. aktive Fahrwerksysteme zu dieser Kategorie hinzukommen. Im Folgenden werden ausschließlich die elektrische Lenkung sowie das Bremssystem betrachtet. In Abbildung 9 ist die Prädiktion der Ströme der fahrdynamikrelevanten Verbraucher konzeptionell dargestellt. Das Konzept basiert auf der Erkennung der aktuellen Fahrsituation, welche wesentlich auf den Daten der Umfeldsensorik beruht. Zur Erreichung eines hohen Automatisierungsgrades (SAE L3 und höher) werden mindestens zwei unabhängige Umfeldsensoren (RADAR, LIDAR, Kamera, …) benötigt. Die von den Umfeldsensoren erkannten Objektdaten werden miteinander synchronisiert und zu fusionierten Umfeldobjekten vereint. Aus den unterschiedlichen Objekttypen wie Fahrzeug, Fußgänger, Zweiradfahrer oder feste Bebauung kann die Fahrzeugumwelt modelliert werden. Das Abbild der Fahrzeugumwelt kann durch Informationen aus der Car2X-Kommunikation weiter verbessert werden. Zur Situationserkennung werden außerdem Daten über die Fahrzeugsituation wie aktueller Lenkwinkel, Fahrzeuggeschwindigkeit und der Aktivierungszustand des Abstandstempomaten genutzt. 259 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Abbildung 9: Modellbasierte Prädiktion der Ströme der fahrdynamikrelevanten Verbraucher Im Rahmen des vorliegenden Beitrags wird die Situationserkennung der drei in Abbildung 9 dargestellten Situationen ACC Bremsung, Kurvenfahrt sowie Notbremsung implementiert. Für die durch das ACC-System (Abstandstempomat) ausgelöste Bremsung werden die Objektdaten der Kamera- und Radar-Systeme positionsgetreu miteinander fusioniert, um die genauen Abstandsinformationen des Radar-Systems mit den visuellen Informationen wie Blinker- und Bremslicht oder aktuelle Fahrspur aus dem Kamera-System zu verbinden. Ist der Abstandstempomat aktiviert, kann die berechnete Sollbeschleunigung anhand vorausfahrender Fahrzeuge in der eigenen Fahrspur abgeschätzt werden. Um hohe, negative Beschleunigungen zu erreichen, wird die Rückförderpumpe des Bremssystems aktiviert. Durch die Prognose der Sollbeschleunigung kann der Aktivierungszeitraum des Bremssystems prognostiziert werden. Der Strombedarf des Bremssystems teilt sich im Wesentlichen auf die Rückförderpumpe und die Magnetventile auf (Abbildung 10). Die genaue, zeitliche Prädiktion der Strompeaks ist auf dem gewählten Abstraktionslevel nicht möglich. Deshalb wird im prognostizierten Aktivierungszeitraum des Bremssystems mit einem konstanter Strom I brake = 80 A gerechnet. Abbildung 10: Gemessene Ströme des ABS-Systems bei einer Bremsung des Abstandstempomaten (Links: Rückförderpumpe. Rechts: Magnetventile) I ABS-pump / [A] 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 time / [sec] 0 0.5 1 1.5 2 2.5 3 3.5 4 260 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Für die Prognose des Strombedarfs der elektrischen Lenkung bei einer Kurvenfahrt werden zunächst elektrisch unterstützende Lenksysteme betrachtet, die das Lenkmoment des Fahrers geschwindigkeitsabhängig unterstützen. Zur Prognose des Strombedarfs werden Informationen zum Straßenverlauf aus der Car2X-Kommunikation verwendet, um den erwarteten Lenkwinkel zu berechnen. In Kreuzungsbereichen können zusätzlich Navigationsdaten und Fahrereingaben wie Blinker oder Blickrichtung genutzt werden. Aus der prognostizierten Fahrzeuggeschwindigkeit und dem Lenkwinkel kann mithilfe eines datengetriebenen Modells der Strombedarf des Lenksystems im Prognosehorizont T Hor abgeschätzt werden. Für die Situationserkennung der Notbremsung werden, wie bei der ACC-Bremsung, die Objektdaten der Umfeldsensorik genutzt. Bei Gegenständen oder Hindernissen, die ausreichend lange erfasst werden, kann die Situation rechtzeitig erfasst und die Aktivierung des Bremssystems mit hoher zeitlicher Genauigkeit prognostiziert werden. Bei plötzlich auftretenden Hindernissen verringert sich die Zeit zwischen der sicheren Erkennung des Objektes und der Auslösung der Notbremsung dermaßen, dass das prädiktive Leistungsmanagement unter Umständen keine rechtzeitigen Gegenmaßnahmen einleiten kann. Zur sicheren Erkennung eines Objektes muss das Objekt mehrere Zyklen nacheinander erkannt werden, um Geisterobjekte auszuschließen. Die Zeitspanne zwischen dem ersten Erkennen eines Objektes und der Ausgabe als sicheres Objekt muss vom prädiktiven Leistungsmanagement genutzt werden, um eventuelle Gegenmaßnahmen einzuleiten. Die unter Umständen falschen Entscheidungen aufgrund von Geisterobjekten sind hier der verspäteten Reaktion des Leistungsmanagements vorzuziehen. Bei deaktiviertem Notbremsassistent bleibt aufgrund der menschlichen Reaktionszeit genug Zeit zwischen der sicheren Erkennung der Objekte und der Einleitung der Notbremsung durch den Fahrer. Somit muss in diesem Fällen nicht auf das erste erkannte Objekt reagiert werden. Abbildung 11: Mögliche Prognose des Strombedarfs im Fahrzeug für den Prognosehorizont T Hor = 5 sec In allen Fällen der Notbremsung ist zu überprüfen, ob der Fahrer ein Ausweichen dem Bremsen vorziehen könnte. Falls das Ausweichen nicht ausgeschlossen werden kann, muss zusätzlich zum Strombedarf der reinen Notbremsung mit dem Strombedarf des 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 time / [sec] 0 20 40 60 80 100 120 140 160 180 200 Baseload non drivingdynamic consumers drivingdynamic consumers 261 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Ausweichens gerechnet werden. Die Strombedarfe werden aus gemessenen Stromprofilen abgeleitet, die mit den aktuellen Situationsparametern skaliert werden. In Abbildung 11 ist eine beispielhafte Prognose des Summenstroms dargestellt, die sich aus der Grundlast (baseload), dem prognostizierten Strom der nicht fahrdynamikrelevanten Verbraucher (non drivingdynamic consumer) und dem erwarteten Strom der fahrdynamikrelevanten Verbraucher (drivingdynamic consumers) zusammensetzt. 4 Einleitung von situationsangepassten Maßnahmen Zur frühzeitigen Erkennung leistungskritischer Situationen muss zunächst der von der Batterie bereitgestellte Strom abgeleitet werden, weil in erster Näherung die Batterie das spannungsdefinierende Element ist. Der in Abschnitt 2 und 3 hergeleitete, erwartete Strombedarf des Fahrzeuges muss von den verfügbaren Quellen und Speichern bereitgestellt werden, wobei die Aufteilung des Laststroms die erwartete Systemspannung wesentlich beeinflusst. Als Quelle wird im Fahrzeug je nach Bordnetztopologie üblicherweise ein Gleichspannungswandler (DC-DC-Wandler) oder ein Drehstromgenerator mit Gleichrichter (Lichtmaschine) verwendet. Zur Vereinheitlichung der Beschreibung der dynamischen Leistungsfähigkeit von Generatoren und Wandlersystemen wird die Leistungsspezifikation verwendet, welche sich aus den drei Kenngrößen  Maximaler Ausgangsstrom I max ,  Minimaler Ausgangsstrom I min (z. B. für rückspeisefähige Wandlersysteme),  Maximaler Betrag des Ausgangsstromgradienten dI max / dt zusammensetzt [4]. Die Leistungsspezifikation wird im Echtzeitsystem anhand des Zustandes der Quelle berechnet. Der Quellenstrom folgt dem Laststrom entsprechend der aktuellen Leistungsspezifikation. Die verbleibende Differenz muss von der Batterie geliefert werden, was zu Spannungsschwankungen führt. Anhand eines definierten Sollspannungsbereiches wird die so prädizierte Systemspannung auf kritische Zeitbereiche untersucht. Wenn eine ausreichende Kritikalität vorliegt, werden mögliche Gegenmaßnahmen auf Verbraucher- oder Quellenebene und deren Auswirkungen gegeneinander gewichtet. Schließlich werden die notwendigen Maßnahmen mit der geringsten Wahrnehmbarkeit für den Nutzer an das Fahrzeug zurückgemeldet, wobei die verbraucherspezifischen Latenzen bei der Wahl des Maßnahmenzeitpunktes berücksichtigt werden. 5 Fazit Es wird ein Ansatz für ein prädiktives Leistungsmanagement präsentiert, der auf einer zentralen Steuerungsebene implementiert werden kann. Dabei wird der Strombedarf der fahrdynamikrelevanten Verbraucher durch eine implementierte Situationserkennung für die Situationen ACC Bremsung, Kurvenfahrt sowie Notbremsung und anschließender Anwendung verschiedener Strommodelle prognostiziert. Zur Prädiktion der Ströme der nicht fahrdynamikrelevanten Verbraucher werden zunächst die Zustandswahrscheinlichkeiten der Verbraucher mithilfe von Bayes’schen Netzen abgeleitet. In Verbindung mit der Verteilung des Schaltmoments über dem Prognosehorizont kann auf den erwarteten Summenstrom sowie die zugehörige Varianz der nicht fahrdynamikrelevanten Verbraucher geschlossen werden. 262 Prädiktives Leistungsmanagement für automatisierte Fahrzeuge Der Summenstrom aller Verbraucher inklusive der Grundlast wird mit der Leistungsfähigkeit der verbauten Quellen und Speicher abgeglichen und anhand von Kritikalitätskriterien bewertet. Bei Bedarf werden situationsangepasste Gegenmaßnahmen ausgewählt und ausgeführt. Literatur [1] Tom Kohler: Prädiktives Leistungsmanagement in Fahrzeugbordnetzen. München, Technische Universität München, Dissertation, 2013 [2] Prof. Dr. rer. nat. Ludwig Brabetz. Statistische Berechnung des erwarteten Summenstroms von Konstantstrom-Verbrauchern. Interner Bericht. Universität Kassel, Okt 2018. [3] Mohamed Ayeb, Patrick Graebel, Ludwig Brabetz, and Giscard Jilwan. Electrical Power System Assessment Method Based on Bayesian Networks. In SAE 2013 World Congress + Exhibition. SAE International, Apr 2013. [4] Janis Lehmann, Benjamin Löwer, Björn Mohrmann, Rainer Knorr: Prädiktives Leistungsmanagement für zukünftige Fahrzeugbordnetze, ATZextra 48 V, April 2019, Springer Verlag [5] Dr.-Ing. Mohamed Ayeb: Faltung des Stromverlaufs eines Verbrauchers mit der Exponentialverteilung zur Ermittlung des statistisch zu „erwartenden“ Stromverlaufs. Interner Bericht. Universität Kassel, Okt. 2018 263 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen Stefan Schwimmbeck, Quirin Buchner, Hans-Georg Herzog Abstract Safety relevant functions in vehicles require a minimal occurrence of short circuits in the power supply system, which might cause dangerous undervoltages. This contribution presents a methodology optimizing the voltage stability with respect to the power of the DC/ DC converter, the capacity of the battery and the quantity of consumers with degradation capabilities. The proposed approach considers the gravimetric and volumetric impact as well as costs of individual components. The results show that an intelligent approach in the topology design can significantly increase the robustness against short circuits. Kurzfassung Um sicherheitsrelevante Funktionen im Fahrzeug zuverlässig gewährleisten zu können, dürfen potentielle Kurzschlüsse nur in möglichst wenigen Situationen die Energieversorgung gefährden. Im Folgenden wird eine Methode zur Optimierung der Kurzschlussresistenz des Energiebordnetzes hinsichtlich des maximalen Stroms des DCDC-Wandlers, der Batteriekapazität und des Anteils an Verbrauchern mit Degradationsfähigkeiten dargestellt. Dabei wird Bezug auf die Auswirkungen auf Masse, Volumen und Bauteilkosten der einzelnen Komponenten genommen. Die Ergebnisse zeigen neben den optimalen Parameterkombinationen auch, dass ein geeignetes Design der Energiebordnetztopologie die Kurzschlussresistenz deutlich erhöhen kann. 1 Motivation Eine zuverlässige Energieversorgung der sicherheitsrelevanten Verbraucher wird vor allem im Hinblick auf autonomes Fahren gegenwärtig diskutiert. Gemäß der Bewertung der funktionalen Sicherheitsanforderungen mittels der Gefahren- und Risikoanalyse in der ISO26262 [1] müssen autonome Fahrfunktionen die Anforderungen bis ASIL D erfüllen. Darüber hinaus benötigen auch im konventionellen Fahrzeug sicherheitsrelevante Verbraucher wie Licht, Scheibenwischer, Lenkung oder Bremse eine zuverlässige Energieversorgung mit ASIL A bis C. Im Sicherheitskonzept werden sowohl funktionale als auch technische Umsetzungen von Sicherheitszielen beschrieben. 264 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 1.1 Problematik bei der Erstellung eines Sicherheitskonzepts für die sichere Energieversorgung im Automobilbordnetz Die Erstellung des Sicherheitskonzepts für das Energiebordnetz (ENBN) stellt aus einer Vielzahl von Gründen eine Herausforderung dar. Durch Unterscheidungen wie Rechts- und Linkslenker oder Elektro-, Hybrid und Verbrenner-Fahrzeuge herrscht eine hohe Variantenvielfalt, die sich in verschiedenen Energiebordnetztopologien widerspiegeln. Dies hat auch unterschiedliche Quellenkonzepte bestehend aus Generator oder DCDC-Wandler zur Folge. Darüber hinaus stellt jede dieser Varianten ein komplexes System aus Quelle, Energiespeicher und einer Vielzahl an Verbrauchern dar. Zwar kann zur Beherrschung einiger Fehlerfälle eine redundante Versorgung, bestehend aus Quelle und Batterie, angenommen werden, wenn diese dementsprechend ausgelegt werden. Allerdings haben sowohl Generator [2] als auch DCDC-Wandler [3] mit dem nachgeschalteten 48 V- oder Hochvoltsystem [4] tendenziell niedrige Verfügbarkeitswerte. Zudem muss das drehzahlabhängige Verhalten [5] des Generators berücksichtigt werden. Als 12 V-Energiespeicher werden meistens AGM-Batterien verwendet. Die Batterie stellt einen Sonderfall im Sicherheitskonzept dar, da ihre Brauchbarkeitsdauer gewöhnlich kürzer als die Lebensdauer des Fahrzeuges ist. Informationen über ihre Zuverlässigkeit werden beispielsweise in [6] und [7] beschrieben. Vervollständigt wird das komplexe System des Energiebordnetzes mit zahlreichen Verbrauchern, die häufig nicht gemäß eines ASIL-Prozesses entwickelt wurden (wird mit „QM“ bezeichnet) und somit eine Fehlerquelle (beispielsweise in Form eines Kurzschlusses oder eines erhöhten Stromverbrauchs) darstellen können. Durch die hohe Komplexität des Systems kann nur schwer identifiziert werden, welche Ereignisse darin zu einer relevanten Unterspannung führen können und welche keine Gefährdung ausstrahlen. Sicherheitsmechanismen wie Trennelemente helfen hier, da sie funktional so konzipiert werden, dass sie Fehlerfälle wie Kurzschlüsse beheben können. Solche Trennelemente werden meistens in Form von Halbleiterschaltern ausgeführt. Konzepte werden beispielsweise in [8] und in ähnlicher Weise für hochautomatisiertes Fahren in [9] diskutiert. Hierbei ist zu unterscheiden, ob eine Trennung der Energiequelle von der Batterie durch das Schaltelement erfolgt (Konzept 1) oder ob beide weiterhin bei den sicherheitsrelevanten Lasten (SRL) verbleiben (Konzept 2). Vorteilhaft im ersten Konzept ist, dass der gesamte Versorgungsstrom der Verbraucher nicht über das Schaltelement fließt, wodurch dieses klein und kostengünstig gestaltet werden kann. Außerdem wird der Schalter hauptsächlich auf den Batterieladestrom und den Verbrauch der sicherheitsrelevanten Lasten ausgelegt, also üblicherweise unabhängig von der Konfiguration der Sonderausstattung. Nachteilig wirken sich hierbei die hohen Anforderungen an die Güte der Batteriezustandsdiagnose aus, die hohen Einfluss auf die Kundenverfügbarkeit haben kann. Bei der Umsetzung von Konzept 2 hingegen wird viel Halbleiterfläche benötigt, da sämtliche nicht sicherheitsrelevanten Verbraucher von den Schaltelementen getrennt werden müssen. Dies kann entweder per Abschaltung von Verbrauchergruppen oder aber individuell in Form von „smart fuses“ (wie bspw. in [10] gezeigt wird) erfolgen. 265 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 1.2 Zielsetzung: Optimierung der Kurzschlussresistenz von Energiebordnetzarchitekturen ohne Trennelement Auch in Bordnetzen ohne einem Trennelement (Abb. 1) kann die Energieversorgung mit einer ASIL-Einstufung gewährleistet werden. Dafür muss, neben zahlreichen anderen Aspekten, das Schaltungsdesign der Verbraucher so konzipiert sein, dass ein Kurzschluss nur mit einer maximalen Ausfallrate auftreten kann. Dies muss für jeden Verbraucher umgesetzt werden, der die Spannungsstabilität des Bordnetzes durch einen Kurzschluss gefährden kann. Grundsätzlich gilt: Je spannungsstabiler das Bordnetzes konzipiert wird, desto weniger Verbraucher können einen kritischen Kurzschluss verursachen, der die definierten Spannungsgrenzen verletzt. Darüber hinaus muss berücksichtigt werden, wie häufig sich das System in Betriebszuständen befindet, in denen Kurzschlüsse der Verbraucher zum Ausfall der kompletten Energieversorgung führen können. In der hier vorgestellten Methode werden folgende drei Parameter verwendet, um die Spannungsstabilität, aber auch die Systemkosten im Energiebordnetz während eines Kurzschlusses zu optimieren: - Maximaler Strom des DCDC-Wandlers, - Nennkapazität der Batterie, - Anteil an Verbrauchern mit Degradationsfähigkeiten. Das Konzept für die spannungsgesteuerte Degradation ist aus [11] bekannt: Sinkt die Spannung am Verbraucher unter eine bestimmte Schwelle, degradiert er selbstständig seinen Strombedarf, sodass sich die Spannung im System erholen kann. Ziel der Optimierung ist eine ideale Auslegung der drei genannten Parameter, mit der die maximale Kurzschlussresistenz bei gegebenen Systemkosten erreicht wird. Dies wird anhand einer Bordnetztopologie gezeigt, die im folgenden Kapitel kurz vorgestellt wird. Abbildung 1: ENBN-Architektur mit gemäß nach ASIL- Anforderungen entwickelten (ASIL), sicherheitsrelevanten (SRL) und ohne ASIL entwickelten (QM) Verbrauchern. R ASIL R SRL R ASIL R QM,2 R QM ,n ... 266 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 2 Bordnetztopologie Die Bewertung von Kurzschlüssen erfordert eine detaillierte Betrachtung sämtlicher Leitungs- und Übergangswiderstände im BN. Dazu wird gemäß der ENBN-Architektur aus Abb. 1 auch die zugehörige BN-Topologie erläutert. In der zu optimierenden ENBN-Topologie sind vier Stromverteiler (SV) im Fahrzeug angeordnet. Über den SV im Vorderwagen sind der DCDC-Wandler und ein weiterer SV (links) angeschlossen. Am Stromverteiler im Heck befinden sich ein zusätzlicher Stromverteiler (rechts) und die Batterie, wie auch Abbildung 2 zeigt. Die Stromverteiler versorgen die Verbraucher, die nach ihrer Sicherungsgröße den jeweiligen SV gemäß Abb. 2 zugeteilt sind. Während die Leitungsquerschnitte und damit auch die Kontaktgrößen auf die Sicherungsgrößen angepasst sind, wird für sämtliche Verbraucher eine konstante Leitungslänge von insgesamt 2 m angenommen. Die Nachbildung der Alterung der Batterie erfolgt, über ihren internen, ohmschen Widerstand, der 150 % seines Wertes bei optimalen Bedingungen (hoher SOC und Raumtemperatur) entspricht. Damit wird die Batteriealterung abhängig von der Batteriekapazität modelliert. Weiterführende, detailliertere Angaben über das Simulationsmodell des Energiebordnetzes befinden sich in [12], worin auch dessen Validierung diskutiert wird und Bewertungen bzgl. der Kurzschlussresistenz der Energiebordnetzarchitektur gezeigt werden. Im nächsten Kapitel wird eine Einführung in die Pareto-Optimierung gegeben und es werden die definierten Optimierungszielfunktionen und -algorithmen, ihre technische Umsetzung sowie die Ergebnisse der Optimierung diskutiert. Abbildung 2: BN-Topologie mit sicherheitsrelevanten Verbrauchern, Kurzschlussstellen (durch Sicherungsbezeichnung charakterisiert), DCDC-Wandler im Vorderwagen und Batterie im Heck. 267 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 3 Rechnergestützte Pareto-Optimierung Das Konzept der rechnergestützten Optimierung wird in [13] genutzt, um mit Hilfe einer Variation des Partikelschwarm-Optimierungsalgorithmus das Bordnetz gewichtsoptimal und spannungsstabil auszulegen. In dem hier vorgestellten Ansatz soll nicht nur das Gewicht der gesamten Komponenten, sondern auch eine Kennzahl für ihren Bedarf an Bauraum und Bauteilkosten optimiert werden. Diese drei Größen sind durch die Zielfunktion 𝑍 charakterisiert. Die Minimierung der relevanten Ausfallrate von Verbraucherkurzschlüssen 𝜆 stellt eine weitere Zielfunktion 𝑍 dar, sodass sich ein Problem der multikriteriellen Optimierung bzw. eine Pareto-Optimierung ergibt [14]: min 𝑍 𝒙 , 𝑍 𝒙 . - (1)- Eine Pareto-Menge beschreibt die Menge an Punkten bzw. Vektoren 𝒙′, die dadurch charakterisiert ist, dass es keine weiteren Punkte 𝒙 gibt, für die gültig ist: 𝑍 𝒙 𝑍 𝒙 -für- 𝑖 1 -und- 2 -- 𝑍 𝒙 𝑍 𝒙 -für-j 1 --oder- 2 .- (2)- Abb. 3 zeigt diesen Zusammenhang, wobei 𝑍 alle möglichen Punkte und die rote Hervorhebung die Pareto-Menge beschreibt. Der Suchraum wird hier durch die drei kontinuierlichen Parameter maximaler Strom des DCDC-Wandlers 𝐼 , Nennkapazität der Batterie 𝐶 und Anteil an Verbrauchern, die einen spannungsgesteuerten Degradationsmechanismus implementiert haben, 𝑎 beschrieben. Dabei sind folgende Definitionsbereiche möglich: 𝐷 𝑥 𝑥 ∈ ℝ|150 A 𝑥 375 A - 𝐷 𝑥 𝑥 ∈ ℝ|50 Ah 𝑥 90 Ah - 𝐷 𝑥 𝑥 ∈ ℝ|0 𝑥 0,75 . - - (3)- Abbildung 3: Mögliche Punkte 𝑍 über die Zielfunktionen 𝑍 und 𝑍 mit deren Pareto-Menge in Rot nach [14]. 𝑍 𝑍 𝑍 268 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen Eine besondere Herausforderung bei der hier gezeigten Optimierung ist die Tatsache, dass die relevante Ausfallrate von Kurzschlüssen, welche durch die in [12] vorgestellte Bewertungsmethode ausgewertet wird, selbst eine Vielzahl an Simulationen benötigt. Die Bewertungsmethode wird im folgenden Abschnitt kurz erklärt. 3.1 Relevante Ausfallrate von Verbraucherkurzschlüssen Inwiefern sich ein Kurzschluss und der damit einhergehende Einbruch der Energiebordnetzspannung kritisch auf das System auswirken, hängt stark von der Bordnetztopologie und den Umgebungsbedingungen wie Außentemperatur, Ladezustand der Batterie oder dem kumulierten Verbraucherstrom ab. Es treten große Potentialunterschiede innerhalb des BN während eines Kurzschlusses auf. Die niedrige Spannung niedervoltseitig am DCDC-Wandler kann dazu führen, dass er in einen Degradationsmechanismus wechselt oder sich sogar abschaltet. Die Höhe des Kurzschlussstroms wird hauptsächlich durch seinen Fehlerwiderstand, die Leitungs- und Kontaktwiderstände sowie den Ort des Kurzschlusses bestimmt. Die Dauer des Kurzschlusses ist definiert durch die Zeit-Strom-Charakteristik der Schmelzsicherung. Abb. 4 zeigt einen Kurzschluss nahe des Stromverteilers rechts, dessen Spannung unten in Schwarz aufgetragen ist. Der Kurzschlussstrom (oben in Rot) und der Batteriestrom (oben, schwarz) betragen bei dieser Messung über 1000 A. Wie in [15] gezeigt wurde, reagiert das System sehr sensitiv auf den transienten Spannungseinbruch, der in Abb. 4 an der Spannung am Stromverteiler im Heck bei 0 s zu Abbildung 4: Messung des Batterie-, DCDC-Wandler- und Kurzschlussstroms (KS) sowie der Spannungen an der Batterie und am Stromverteiler rechts (SV,R). 269 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen sehen ist. Er wird bei schnell auftretenden Kurzschlüssen durch die parasitären Induktivitäten der Leitungen hervorgerufen. Da die zu optimierenden Parameter kaum Einfluss auf ihn haben, muss der Effekt anderweitig, z.B. mit Designmaßnahmen, beherrscht werden. In dieser Arbeit werden sie allerdings vernachlässigt, sodass durch die reduzierte Betrachtung die Übertragbarkeit der Optimierung erhöht wird. Die Optimierung benutzt die Bewertungsmethode aus [12], um eine Aussage zu erhalten, wie viele Verbraucher einen kritischen Kurzschluss verursachen können und mit welcher Wahrscheinlichkeit sich das System in einem Zustand befindet, in dem der Kurzschluss kritisch ist. Die Methode bewertet die drei Betriebszustände Temperatur 𝑇, State-of-Charge der Batterie (𝑆𝑜𝐶) und die Summe der Ströme sämtlicher Verbraucher 𝐼 inklusive der Wahrscheinlichkeiten ihres Auftretens. Dafür werden die Betriebszustände in die in Gleichung (4) dargestellten Definitionswerte abgerundet bzw. aufgerundet und jede der Kombinationen simuliert. Daraus ergibt sich die Wahrscheinlichkeit, mit der ein Verbraucher einen Kurzschluss in einem relevanten Zustand verursacht. Da teilweise höhere Fehlerwiderstände einen schwerwiegenderen Fall darstellen, werden zusätzlich - wenn dies notwendig ist - bis zu vier verschiedene Fehlerwiderstände in der Simulation bewertet, sodass der schlimmste Fall bewertet wird. - 𝐷 20; 10; 0; 10; 20; 30 °C - 𝐷 0,45; 0,55; 0,65; 0,75; 0,85; 0,95 - 𝐷 40; 80; 120; 160; 200 A - - (4)- - - Als Annahme wird gesetzt, dass jeder der Verbraucher eine Fehlerrate von 100 FIT (eng. failure in time, entspricht der Anzahl an Fehlern in 10 Stunden) für den Fehlermodus Kurzschluss hat. Durch die Bewertungsmethode reduziert sich die maximale Fehlerrate der 12 bewerteten Verbrauchern 𝜆 von 1.200 FIT auf einen niedrigeren Wert: die relevante Fehlerrate 𝜆 , die gleich der Zielfunktion 𝑍 ist. 𝑍 𝜆 𝑘 ⋅ 𝜆 , mit- 𝑘 1 - (5)- 3.2 Diskussion der Kennzahlen der Parameter Wie bereits beschrieben soll die Zielfunktion 𝑍 eine Funktion von Masse (m), Volumen (V) und der Bauteilkosten (Btk) der Komponenten DCDC-Wandler (DCDC), Batterie (Batt) und der Verbraucher mit Degradationsfähigkeiten (VD) darstellen. Die Gleichung (6) gibt diesen Zusammenhang zusammengefasst an: 𝑍 𝑘 𝑘 𝑘 𝑘 .- (6)- - Jede Funktion für eine Kennzahl bildet wiederum die Addition der Kennzahlen für die Masse, das Volumen und die Bauteilkosten ab: 𝑘 𝑘 , 𝑘 , 𝑘 , . - (7)- 270 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen Sämtliche Massen werden mit 3/ kg, die Volumina mit 3/ l und die Btk mit 1/ € in die jeweilige Kennzahl umgerechnet. Als Annahme wird vorausgesetzt, dass stets identische Technologien ohne Berücksichtigung von Sprungfunktionen diskutiert werden. Somit sind die Kennzahlen der Komponenten nur auf AGM-Batterien und einer festgelegten DCDC-Wandler-Variante mit Wasserkühlung gültig. 3.2.1 Kennzahl für die Batteriekapazität Eine Erhöhung der Batterienennkapazität führt dazu, dass der Spannungsabfall über die inneren Widerstände der Batterie während des Kurzschlusses geringer ist. Eine Auswirkung der Kapazitätsänderung auf die SoC-Zustandsverteilung wird vernachlässigt. Das Volumen und die Masse der Batterie werden anhand bekannter Fahrzeugbatterien abgeleitet und auf die Nennkapazität der Batterien bezogen. Für die Massekennzahl ist die dafür durchgeführte Regression in Abb. 5 dargestellt. In der Abbildung ist das Gewicht verschiedener Batterien über ihre Nennkapazität und die Regressionskurve eingetragen. 𝑘 , 3 kg 1 kg 3,5 Ah ⋅ 𝐶 , - (8)- - Abbildung 5: Gewichts- und Kapazitätsdaten verschiedener 12 V-AGM-Starterbatterien und deren Regressionskurve. 271 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen Analog zu oben gezeigtem Vorgehen wird auch das Volumen bestimmt und durch folgende Funktion ausgedrückt: 𝑘 , 3 l ⋅ 0,123 l Ah ⋅ 𝐶 , . - (9)- Die Kennzahl für die Bauteilkosten der Batterie wird mit 1/ Ah angenommen. Wie beschrieben ergibt sich die Funktion für die Kennzahl der Batterie 𝑘 aus der Addition der einzelnen Funktionen (Abb. 6). 3.2.2 Kennzahl für die maximale DCDC-Wandler-Leistung Während eines Kurzschlusses wird der DCDC-Wandler maximal belastet. Wird seine kurzzeitig mögliche, maximale Leistung vergrößert, wird die Spannung im System besser gestützt. Aktuelle Veröffentlichungen mit Informationen über eine mögliche Auslegung von 48 V-12 V-DCDC-Wandlern, wie bspw. [16], sehen stationäre Leistungen von 3,5 kW vor. Dementsprechend ist der Definitionsbereich für diesen Parameter in einem Bereich um 3,5 kW angesetzt. Da die Spannung während des Kurzschlusses deutlich unter dem stationären Betriebsbereich von ca. 14 V liegt, wird als relevanter Parameter der niederspannungsseitige Nennstrom 𝐼 , 250 A verwendet. Für den für Kurzschlüsse relevanten Zeitbereich kleiner 150 ms ist es sinnvoll, eine Überlastungsfunktion zu implementieren. Deswegen wird davon ausgegangen, dass der Gleichspannungswandler kurzzeitig einen 25 % höheren Strom 𝐼 , liefern kann. Abbildung 6: Kennzahlen für das Gewicht 𝑘 , , das Volumen 𝑘 , und die Bauteilkosten 𝑘 , der Batterie, sowie deren Addition, der Kennzahl 𝑘 . 272 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 𝐼 , 125 % ⋅ 𝐼 , - (10)- Die Annahme, dass der DCDC-Wandler modular aus verschiedenen Phasen aufgebaut ist, welche beliebig erweiterbar bzw. reduzierbar sind und deren Stufen vereinfacht vernachlässigt werden, begründet die lineare Abhängigkeit der Bauteilkosten-, Masse- und Volumenkennzahlen vom maximalen Strom 𝐼 , . Anhand einer Studie aus [17] kann abgeschätzt werden, dass die Nennleistung des Wandlers einen deutlich höheren Einfluss auf die Bauteilkosten als auf das Volumen hat. Dementsprechend wird angenommen, dass 50 % der Bauteilkosten direkt proportional zu seiner Nennleistung sind, aber nur 30 % des Volumens und 20% der Masse des Wandlers. Somit ergeben sich mit dem in [18] angegebenem Volumen von 2,7 l und den Schätzungen der Bauteilkosten zu 80 € und dem Gewicht mit 2,5 kg folgende Funktionen für die Kennzahlen (Abb. 7): 𝑘 , 3 kg ⋅ 0,2 ⋅ 2,5 kg 300 A ⋅ 𝐼 , 300 A 2,5 kg - 𝑘 , 3 l ⋅ 0,3 ⋅ 2,7 l 300 A ⋅ 𝐼 , 300 A 2,7 l - 𝑘 , 1 € ⋅ 0,5 ⋅ 80 € 300 A ⋅ 𝐼 , 300 A 80 € . - - (11)- Abbildung 7: Kennzahlen für das Gewicht 𝑘 , , das Volumen 𝑘 , und die Bauteilkosten 𝑘 , des DCDC-Wandlers, sowie deren Addition, der Kennzahl 𝑘 . 273 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 3.2.3 Kennzahl für den Anteil an degradationsfähigen Verbrauchern Die spannungsabhängige Verbraucherdegradierung wird selbstständig mit ASIL B in jedem Verbraucher umgesetzt. Abhängig von der Eingangsspannung der jeweiligen Komponente reduziert sie ihren Strombedarf nach einer spezifizierten Funktion. Dafür ist oftmals eine zusätzliche Schaltung bestehend aus MOSFET, Gate-Treiber, ASILzertifizierter Recheneinheit und evtl. einer oder mehrerer Spannungsmessstellen notwendig. Schätzungsweise ist eine solche Schaltung in 4 cm³, mit 20 g und für 1 € umsetzbar. Als Prämisse wird gesetzt, dass die Größe, die Masse und die Kosten der Schaltung unabhängig von der Leistung des Verbrauchers sind. Somit verteilt sich der abschaltbare Strom gemäß der Verbraucherverteilung im System, die durch den Kurvenverlauf in Abb. 8 gezeigt und in der Gleichung (12) als 𝑓 𝑎 bezeichnet wird. Idealerweise werden zunächst leistungsstarke Verbraucher befähigt, sodass bereits mit wenigen Schaltungen viele Degradationsmöglichkeiten entstehen. Um den maximal möglichen Anteil an degradationsfähigen Verbrauchern (hier angenommen mit 75 %) zu erreichen, müssen viele 5 A Verbraucher befähigt werden, die allerdings nur einen geringen Anteil am Stromverbrauch im BN haben. Dieser Zusammenhang macht sich in einem starken Anstieg auf der rechten Seite der Kurve bemerkbar. 𝑘 , 𝑘 ⋅ 𝑓 𝑎 mit 𝑘 ⎩ ⎪⎨ ⎪⎧ ⋅ 20 g für Gewicht ⋅ 4 cm für Volumen € ⋅ 1 € für Btk (12)- Abbildung 8: Kennzahlen für das Gewicht 𝑘 , , das Volumen 𝑘 , und die Bauteilkosten 𝑘 , über den Anteil an Verbrauchern mit Degradationsfähigkeiten, sowie deren Addition, der Kennzahl 𝑘 . 274 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 3.3 Optimierungsalgorithmus Paretosearch Für die multikriterielle Optimierung mit einem 3-dimensionalen Suchraum eignet sich der Optimierungsalgorithmus Paretosearch, der einen Patternsearch-Ansatz verwendet. Paretosearch ist ein in Matlab implementierter Optimierungsalgorithmus [19]. Dabei wird der einkriterielle Optimierungsalgorithmus Patternsearch mit mehreren Punkten parallel oder sequentiell angewendet, um Punkte nahe der Pareto-Front zu finden. Ein Punkt besteht dabei aus einer Kombination der Parameterwerte (hier 3-dimensional) und den Ergebnissen der Zielfunktionen. Patternsearch Beim Optimierungsalgorithmus Patternsearch wird ausgehend von einem Startpunkt sequentiell ein Schritt in beide Richtungen jeder Dimension gemacht und geprüft, ob die Zielfunktion einen geringeren Wert als die Zielfunktion im Startwert hat. Dies wird als Umfrage (engl. poll) bezeichnet. Sobald ein besserer Funktionswert gefunden wird, ist die Umfrage erfolgreich und der betrachtete Punkt wird als Startpunkt für den nächsten Schritt verwendet. Im nächsten Schritt wird die Schrittbzw. Maschenweite des Betrachtungsrasters verdoppelt und die Umfrage erneut vollzogen. Dies wird solange durchgeführt, bis kein geringerer Funktionswert als der des Startwerts gefunden werden kann. Tritt dies ein, ist die Umfrage nicht erfolgreich und das Vorgehen wird in der nächsten Iteration mit halber Schrittweite wiederholt. Paretosearch Paretosearch nutzt das Vorgehen des oben beschriebenen Patternsearch-Algorithmus, um Punkte zu suchen, die den Pareto-Zustand erfüllen. Dabei sollen die Punkte möglichst gleichmäßig auf der Pareto-Front verteilt sein. Im Unterschied zur einkriteriellen Optimierung müssen beim Paretosearch-Algorithmus alle Werte der verschiedenen Zielfunktionen mindestens einen ebenso guten Wert und mindestens bei einer Zielfunktion einen besseren Wert ergeben, damit der neue Punkt den vorherigen dominiert (siehe (2)). Wird ein Punkt nicht dominiert, ist sein Rang eins. Alle Punkte mit Rang zwei werden jeweils von einem Punkt mit Rang eins dominiert, etc. Nach der Initialisierung wird die erste Iteration mit den Startwerten (entspricht der Anzahl der Punkte auf der Pareto-Front) durchgeführt. Die Startwerte werden sequentiell abgearbeitet, indem immer eine Umfrage des Patternsearch-Algorithmus pro Punkt durchgeführt wird. Wird der betrachtete Punkt von einem neuen Punkt dominiert, ist die Umfrage erfolgreich. Es werden noch in derselben Iteration sämtliche Schritte in die gefundene Richtung durchgeführt, indem die Schrittweite solange verdoppelt wird, bis ein Punkt mit mindestens einem höheren Funktionswert gefunden wird. Der vorletzte ausgewertete Punkt hat die niedrigsten Funktionswerte und somit Rang eins. Er wird gemeinsam mit der verwendeten Schrittweite gespeichert. War die komplette Umfrage bei einem Punkt nicht erfolgreich, wird die Schrittweite für die nächste Iteration halbiert. Sämtliche neu gefundenen Punkte und auch die Punkte aus der letzten Iteration werden für den kommenden Turnus in Betracht gezogen. Überschreitet die Anzahl der gefundenen Punkte die der gewünschten Punkte auf der Pareto-Front, werden bereits 275 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen dominierte Punkte entfernt. Dafür dient das Volumen der Punkte als Entscheidungskriterium. Das Volumen ist ein Maß für die Gleichmäßigkeit der Punkteverteilung auf der Pareto- Front. Die Änderung des summierten Volumens einer Iteration im Vergleich zur vorherigen ist eines der Abbruchkriterien des Paretosearch-Algorithmus. 3.4 Anwendung des Optimierungsalgorithmus Abb. 9 zeigt die Anwendung des Algorithmus auf die beschriebenen Zielfunktionen für eine Pareto-Front mit sieben Punkten. Nach der ersten Iteration ergeben sich fünf nicht dominierte Punkte. Diese dienen als Startwerte für die zweite Iteration, nach der bereits sieben Punkte mit einer breiteren Verteilung gefunden wurden. Die dritte Iteration führt zu zwei neuen Punkten. Im rechten Bild der Abb. 9 wurden zusätzlich die Punkte der Iteration zwei in grau eingezeichnet. Der rechte, graue Punkt wird vom neuen Wert dominiert, da sowohl die 𝑘 als auch 𝜆 einen niedrigeren Wert aufweisen. Der linke, graue Punkt wurde durch einen neuen Punkt ersetzt (grauer Pfeil), weil dieser ein größeres Volumen besitzt. 3.5 Effizienz der technischen Umsetzung Tabelle 1 zeigt die maximale Anzahl an Auswertungen, die für eine Pareto-Front mit 20 Punkten benötigt wird. Die Abbruchkriterien werden so gewählt, dass Paretosearch für die Erstellung der Pareto-Front ca. zehn Iterationen mit insgesamt ca. 300 Auswertungen durchführt. Eine Bewertung eines Punktes benötigt 25 s bis 150 s. Hierbei wurde hoher Anspruch an die Effizienz gestellt, die durch die aufgeführten drei wesentlichen Umsetzungen gelang: 1) Minimierung der Zeit einer Kurzschlusssimulation 2) Parallelisierung der Auswertung auf mehrere Rechenkernen 3) Heuristiken zur Auswahl wichtiger Simulationen Abbildung 9: Die ersten drei Iterationen des Paretosearch-Algorithmus für eine Pareto-Front mit sieben Punkten. Im rechten Bild sind zusätzlich die beiden Punkte aus Iteration 2 eingezeichnet, die in der Iteration 3 ersetzt werden. 276 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen Eine Minimierung der Simulationszeit gelingt vor allem dann, wenn die Modelle in sämtlichen Arbeitspunkten reibungslos funktionieren. Gerade hinsichtlich der Batterie und des Wandlers ist dies sehr aufwendig umzusetzen, da die Komponenten sowohl durch die Parameter der Optimierung, als auch durch die verschiedenen BN-Zustände verändert werden. Ein zweiter wichtiger Schritt ist, nur die notwendigsten Informationen als Ergebnis auszugeben, sodass die Simulation komplett auf dem Arbeitsspeicher und nie auf die deutlich langsamere Festplatte des PCs zugreifen muss. Grundsätzlich können schnelle Zielfunktionen effizient mit Hilfe der Umsetzung innerhalb des Optimierungsalgorithmus parallelisiert werden. Dies ist hier nicht der Fall, sodass die Parallelisierung bei der Betrachtung der Verbraucher implementiert ist. Auf zwölf verschiedenen Kernen werden die zwölf Verbraucher bewertet. Ein Nachteil bei dieser Umsetzung ist, dass die Verbraucher unterschiedliche Simulationszeiten aufweisen und auf den langsamsten gewartet werden muss. Bei der Bewertung der BN-Zustände Temperatur, SoC und Verbraucherstrom dienen heuristische Verfahren dazu, die Anzahl der zu betrachtenden Fehlerfälle zu minimieren. Unter anderem müssen BN-Zustände nicht simuliert werden, wenn bereits ein KS bei einem Zustand kritisch war, bei dem der SOC höher oder der BN-Strom niedriger war. Auch der Fehlerwiderstand muss nur in seltenen Fällen variiert werden. Der Nutzen dieser Vorgehensweise kann aus Tabelle 2 abgelesen werden. Ohne die heuristischen Verfahren würde die Simulationszeit für sämtliche BN-Zustände und deren Fehlerwiderstände ca. 864 s betragen. Tatsächlich werden maximal nur 150 s benötigt, oft auch deutlich weniger. Tabelle 1: Maximale Anzahl und die dafür benötigte Zeit von Simulationen bzw. Auswertungen für einen Teilbetrachtungsumfang. Teilbetrachtung Anzahl Tats. Zeit Pareto-Front mit 20 Punkten und 10 Iterationen ca. 300 2.592.000 ca. 50.000 s Verbraucher 12 8.640 25 s - 150 s BN-Zustände max.180 720 KS-Fehlerwiderstand max. 4 4 Kurzschlusssimulation 1 1 ca. 1,2 s 4 Ergebnisdiskussion Abb. 10 zeigt die Pareto-Front mit der jeweiligen Parameterkombination. Dabei stellen Punkte, die in den vier Bildern auf derselben vertikalen Linie liegen, eine Parameterkombination und die Funktionswerte der beiden Zielfunktionen dar. Beispielsweise werden die Systemkennzahl 260 und 75 FIT erreicht, wenn 𝐼 , 375 A, 𝑎 9 % und 𝐶 , 72 Ah betragen. Für niedrige Kennzahlen, d.h. alle Parameter werden minimal gewählt, ergibt sich eine relevante Ausfallrate von ca. 360 FIT. Durch die größere 277 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen Dimensionierung des Wandlers, der Batterie oder durch einen höheren Anteil an degradationsfähigen Verbrauchern kann eine relevante Ausfallrate von unter 30 FIT erreicht werden. Dazu wird zunächst der kurzzeitige, maximale Strom des Wandlers erhöht. Dadurch kann die relevante Ausfallrate bereits auf 200 FIT reduziert werden. Anschließend wird die Nennkapazität der Batterie vergrößert, bevor der Anteil an Verbrauchern mit Degradationsfähigkeiten zunimmt. Die sichtbaren Schwankungen bei den Parametern sind der Tatsache geschuldet, dass die Methode zur Bewertung der relevanten Ausfallrate als Ergebnisse diskret verteilte Ausfallraten ausgibt. Dieser Effekt ist beispielsweise beim Wandler bei 𝑘 270 zu sehen. Abbildung 10: Pareto-Optimum von Systemkennzahl 𝑘 und relevanter Ausfallrate 𝜆 sowie die dazugehörenden Parameterwerte 𝐼 , , 𝑎 und 𝐶 , . 278 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 4.1 Validierung durch den Partikelschwarmalgorithmus Durch die Überführung der multikriteriellen in eine einkriterielle Optimierung aus [20] min 𝑍 𝒙 , 𝑍 𝒙 → min 𝑎 ⋅ Z 𝒙 1 𝑎 ⋅ 𝑍 𝒙 --- - (13)- - kann auch der singuläre, globale Optimierungsalgorithmus Partikelschwarm verwendet werden. Die Schrittweite der sukzessiven Änderung von 𝑎 im Bereich 0 bis 1 bestimmt die Anzahl von Punkten, die die Kurve des Pareto-Optimums darstellen. Der Partikelschwarm-Algorithmus wird in [13, 21] beschrieben. Er gliedert sich in die Algorithmen der Schwarmintelligenz ein, die dadurch charakterisiert sind, durch die Interaktion untereinander und mit der Umwelt intelligentes Verhalten aufweisen zu können. Mehrere Partikel wissen dabei ihren besten Punkt und den der gesamten Population. Sie können sich somit in Richtung des globalen Optimums bewegen. [13] Abbildung 11: Vergleich der durch Paretosearch gefundenen Pareto- Menge mit der des Partikelschwarm-Algorithmus. 279 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen Der Partikelschwarm-Algorithmus wird hier zur Validierung der Pareto-Front verwendet. Den Unterschied zeigt Abb. 11 auf. Die Ergebnisse des in dieser Anwendung erheblich performanteren Optimierungsalgorithmus Paretosearch bilden die Pareto- Front gut ab. Auch der Partikelschwarm-Algorithmus findet die niedrigen Stromwerte für den DCDC-Wandler bei 𝑘 270. Der Anteil der degradationsfähigen Verbraucher ist beim Partikelschwarm-Algorithmus für 𝑘 kleiner 290 stets 0. In der Regel findet der Partikelschwarm-Algorithmus globale Minima besser im Vergleich zu Paretosearch. Für den Vergleich von Pareto-Fronten eignet sich der Paretosearch-Algorithmus besser, da er deutlich effizienter arbeitet und die gefundenen Punkte gleichmäßiger verteilt sind. 4.2 Einfluss verschiedener Klimazonen auf das Pareto-Optimum Die Temperatur wirkt sich wesentlich auf die Widerstände der Leitungen, das Batterieverhalten und die Auslösezeit der Sicherung aus. Abb. 12 zeigt die Sensitivität der Pareto-Mengen aufgegliedert nach der Temperatur. Deutlich erkennbar ist der Unterschied zwischen Alaska und Deutschland. Als Grundlage dienen 10-Grad genaue Häufigkeitsverteilungen der Temperatur für die jeweilige Klimazone. Dabei war die durchschnittliche Temperatur -2,4 °C in Alaska, 9,8 °C in Deutschland und 21,3 °C in Texas. 4.3 Optimierungsergebnisse verschiedener ENBN-Topologien Die Pareto-Fronten bieten eine gute Vergleichsmöglichkeit für ENBN-Topologien hinsichtlich ihrer Robustheit gegenüber Kurzschlüssen. Abb. 13 vergleicht die Pareto- Fronten der bereits diskutierten Topologie (siehe Abb. 2) mit der Batterie im Heck (Batt,H) und dem DCDC-Wandler vorne angebunden mit einer Topologie, in der sowohl der Speicher als auch die Quelle vorne angebunden sind (Batt,V). Es zeigt sich, dass bei der Topologie „Batt,V, SRL,V“ eine deutlich bessere Pareto-Front erreicht Abbildung 12: Pareto-Mengen bzgl. der relevanten Ausfallrate von Kurzschlüssen 𝜆 und der Systemkennzahl 𝑘 für verschiedene Klimazonen. 280 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen wird als in der bisher diskutierten Topologie (Batt,H, SRL,V). Sämtliche sicherheitsrelevanten Verbraucher sind nahe an den Quellen angebunden, wodurch ihre Spannung während des Fehlerfalls gestützt wird. Hinsichtlich autonomen Fahrens werden zahlreiche Komponenten, wie Sensoren oder Rechner, sicherheitsrelevant sein. Die Verteilung dieser sicherheitsrelevanten Lasten im Fahrzeugbordnetz führt dazu, dass die Spannungen an jedem Stromverteiler betrachtet werden müssen. Dies wird in Abb. 13 mit „SRL,vert“ bezeichnet. Sind DCDC-Wandler und Batterie verteilt (grün) hat dies kaum Auswirkungen auf die Ausfallrate. Daraus kann rückgeschlossen werden, dass bei kritischen Zuständen stets die Spannung am vorderen Stromverteiler die niedrigste im System ist. Dies ist immer der Fall, wenn sich der Kurzschluss entweder vorne befindet oder der KS hinten ist und der Wandler spannungsbedingt degradiert. Bemerkenswert ist die Abweichung der Topologie „Batt,V, SRL,vert“ (rot) im Vergleich zur blauen Kurve. Hier sind beide Quellen im Vorderwagen angebunden, aber jeder Stromverteiler versorgt sicherheitsrelevante Lasten. Dies hat eine erhebliche Erhöhung der Ausfallrate des Systems zur Folge. Abb. 13 zeigt auch, wie vorteilhaft sich die Auslegung der Bordnetztopologie auf die Kurzschlussresistenz auswirken kann. Um beispielsweise eine relevante Ausfallrate von 100 FIT zu erreichen, wäre für die normale BN-Topologie eine Kennzahl von ca. 250 notwendig. Ist die Batterie hingegen vorne angebunden, kann dies durch eine Parameterkombination mit einer Kennzahl von ca. 190 bereits erreicht werden. Abbildung 13: Pareto-Mengen bzgl. der relevanten Ausfallrate von Kurzschlüssen 𝜆 und der Systemkennzahl 𝑘 der ENBN-Topologien Wandler im Vorderwagen, Batterie hinten (Batt,H), Batterie vorne (Batt,V), wobei unterschieden wird, ob nur am SV im Vorderwagen sich sicherheitsrelevante Lasten befinden (SRL,V) oder sie auf jeden SV verteilt sind (SRL,vert). 281 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen 5 Zusammenfassung In dem Artikel wird eine Methode zur Bestimmung von optimalen Komponentenparametern gezeigt, die eine Energieversorgung mit möglichst hoher Robustheit gegenüber Kurzschlüssen ermöglicht. Dafür werden Funktionen für Kennzahlen diskutiert, die einen Bezug des maximalen Stroms des DCDC-Wandlers, der Batterienennkapazität und des Anteils an Verbrauchern mit Degradationsfähigkeiten zu deren Massen, Volumina und Bauteilkosten herstellt. Die Ergebnisse der multikriteriellen Optimierung in Form von Pareto-Mengen werden verwendet, um zum einen die Sensitivität der Optimierung hinsichtlich der Temperatur aufzuzeigen, aber auch um verschiedene ENBN- Topologien vergleichen zu können. Es wird deutlich, dass eine optimale Entwicklung eines zuverlässigen Bordnetzes bereits beim Design der Topologie beginnt. Literatur [1] ISO: ISO/ FDIS 26262 Road vehicles - Functional Safety - Part 1-10; 2011. [2] M. Abele, Modellierung und Bewertung hochzuverlässiger Energiebordnetz-Architekturen für sicherheitsrelevante Verbraucher in Kraftfahrzeugen, Kassel, Univ., Diss. 2008. [3] M. Mürken und P. Gratzfeld, Reliability Comparison of Bidirectional Automotive DC/ DC Converters, 2017 IEEE 86th Vehicular Technology Conference (VTC- Fall), Toronto, ON, 2017. [4] Y. Song and B. Wang, Evaluation Methodology and Control Strategies for Improving Reliability of HEV Power Electronic System, in IEEE Transactions on Vehicular Technology, vol. 63, no. 8, pp. 3661-3676, Oct. 2014. [5] M. Hohmann, Ein synthetischer Ansatz zur Auslegung von Kfz-Bordnetzen unter Berücksichtigung dynamischer Belastungsvorgänge, Ilmenau: Dissertation, 2009. [6] M. Mürken, D. Kübel, A. Thanheiser und P. Gratzfeld, Analysis of automotive lead-acid batteries exchange rate on the base of field data acquisition, IEEE International Conference on Electrical Systems for Aircraft, Railway, Ship Propulsion and Road Vehicles & International Transportation Electrification Conference (ESARS-ITEC), Nottingham, 2018. [7] J. Albers und I. Koch, Reliability of Lead-Acid Batteries,Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement, 2016. [8] K. Morgenroth und B. Schneider, Functional Safety within the wiring system, 6. Internationaler Fachkongress Bordnetze im Automobil, Ludwigsburg, 2018. [9] S. Schumi, Energy and Supply Concepts for Automated Driving, 2018 IEEE International Conference on Electrical Systems for Aircraft, Railway, Ship Propulsion and Road Vehicles & International Transportation Electrification Conference (ESARS-ITEC), Nottingham, 2018. [10] S. Frei, M. Kiffmeier und S. Önal, Support of Emergency Functions through Combination of Smart Fusing and Model-Based Diagnosis, 6. Internationaler Fachkongress Bordnetze im Automobil, Ludwigsburg, 2018. 282 Optimierung der Zuverlässigkeit von Energiebordnetzarchitekturen hinsichtlich Kurzschlüssen [11] F. Ruf, A. Barthels, G. Walla, M. Winter, H.-U. Michel, J. Fröschl und H.-G. Herzog, Autonomous Load Shutdown Mechanism as a Voltage Stabilization Method in Automotive Power Nets, 2012 IEEE Vehicle Power and Propulsion Conference, Seoul, 2012. [12] S. Schwimmbeck, Q. Buchner und H.-G. Herzog, Bewertung von Kurzschlüssen in Energiebordnetzarchitekturen, 8. Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement, Würzburg, 2018. [13] F. Ruf, Auslegung und Topologieoptimierung von spannungsstabilen Energiebordnetzen, München: Dissertation, 2015. [14] Branke, Jürgen, Multiobjective optimization. Interactive and evolutionary approaches, Berlin, Heidelberg, New York, NY: Springer, 2008. [15] S. Schwimmbeck, Q. Buchner und H.-G. Herzog, Evaluation of Short-Circuits in Automotive Power Nets with Different Wire Inductances, IEEE International Conference on Electrical Systems for Aircraft, Railway, Ship Propulsion and Road Vehicles & International Transportation Electrification Conference (ESARS- ITEC), Nottingham, 2018. [16] A. Leon-Masich, R. Molina-Llorente, M. Tena-Gil und R. Jiménez-Pino, High-Efficiency GaN-based 48V/ 12V DC-DC Converter for Automotive Applications, 8. Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement, Würzburg, 2018. [17] USDRIVE, Electrical and Electronics Technical Team Roadmap, Driving research and innovation for vehicle efficiency and energy sustainability, 2013. [18] B. Köppl, M. Scheffer, R. Vuletic und T. Röwe, 48V DC/ DC converter and 48V motor drive IFX solutions and test results, 6. Tagung Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Energiemanagement, 2015. [19] Mathworks paretosearch Algorithm, https: / / de.mathworks.com/ help/ gads/ paretosearch-algorithm.html, abgerufen am 12.03.2019. [20] K. Littiger, Optimierung. Eine Einführung in rechnergestützte Methoden, Berlin, Heidelberg: Springer Berlin Heidelberg, 1992. [21] K.-L. Du und M .Swamy, Search and Optimization by Metaheuristics. Techniques and Algorithms Inspired by Nature. Cham: Birkhäuser. Online verfügbar unter http: / / dx.doi.org/ 10.1007/ 978-3-319-41192-7, 2016. 283 Model-Based Analysis of Transient Processes in Highly Available Automotive Energy Systems Martin Baumann, Christoph Weissinger, Hans-Georg Herzog Abstract Dynamic and safety-relevant electrical components in automotive energy systems increase the functional requirements to the energy and power supply. A reduced availability of energy or power within energy systems of self-driving vehicles is particularly safety-critical. Elevated power peaks could damage electrical components, reduce their performance, lead to malfunctions or impair the energy systems stability. Such raised power peaks are triggered by rapidly changing processes or overlaying power demands. These transient phenomena are caused by discontinuous operations of nonlinear components or switching of inductive and capacitive loads. Transient processes in automotive energy systems up to the microsecond range are not sufficiently studied. This contribution shows transient disturbances within automotive energy systems as well as retroactive effects of system integrated components. Disturbances are analysed by means of quantifiable stability parameters. Typical automotive components are compared regarding their transient behaviour. The final sensitivity analysis demonstrates an approach for the reduction of transient disturbances. Obtained simulation results show that transient retroactive effects can be reduced by up to 30 % through the customization of the component-integrated input circuitry. Kurzfassung Dynamische sowie sicherheitsrelevante elektrische Komponenten in Energiebordnetzen von Kraftfahrzeugen erhöhen die funktionalen Anforderungen an die elektrische Energie- und Leistungsversorgung. Eine eingeschränkte Energie- oder Leistungsverfügbarkeit ist beim automatisierten Fahren besonders sicherheitskritisch. Hohe Leistungsspitzen können elektrische Komponenten beschädigen, deren Leistungsfähigkeit reduzieren, zu Fehlfunktionen führen oder die Bordnetzstabilität beeinträchtigen. Ausgelöst werden diese Leistungsspitzen durch schnell veränderliche Vorgänge oder durch gleichzeitige Überlagerung mehrerer Hochleistungsverbraucher. Mögliche Ursachen für transiente Vorgänge sind die diskontinuierliche Arbeitsweise nichtlinearer Betriebsmittel oder das Schalten von induktiven und kapazitiven Lasten. Transiente Vorgänge im Bordnetz bis in den Mikrosekundenbereich wurden bisher noch nicht ausreichend analysiert. Dieser Beitrag zeigt transiente Störungen im Energiebordnetz sowie mögliche Netzrückwirkungen von Komponenten. Störungen werden anhand von quantifizierbaren Stabilitätsparametern analysiert. Typische Fahrzeugkomponenten werden bezüglich ihres transienten Verhaltens verglichen. Die abschließende Sensitivitätsanalyse bietet einen Ansatz für die Reduzierung von transienten Störungen. Die Simulationsergebnisse zeigen, dass transiente Netzrückwirkungen durch Anpassung der Eingangsschaltung um bis zu 30 % verringert werden können. 284 Model-Based Analysis of Transient Processes in Highly Available Energy Systems 1 Disturbances within Electrical Energy Systems Automotive energy systems of self-driving vehicles require a particularly high safety level with respect to energy supply of loads and storage units. The simultaneous operation of safety, comfort and chassis control loads could exceed the power offer of the automotive energy system and lead to voltage drops threatening the functionality of safety-relevant systems. Moreover, the increased use of both, power electronics and switchable loads, leads to high dynamic power fluctuations and particularly small time constants within the energy system. Avalur [1], Bai [2], Diez [3], Frazier [5] and Nasim [10] published measurements of transient processes within automotive energy systems. These rapidly changing processes are standardized by the International Organization for Standardization (ISO) in ISO7637 [7]. There are similar conditions in the Alternating Current (AC) energy system interconnecting substations, power plants and consumers such as industries or smaller-sized households. Disturbances as shown in Table 1 can have serious consequences. Unsatisfactory protection concepts can lead to a division of the energy system into asynchronous decoupled micro-grids leading to a system blackout at worst. Standardized quality criteria for voltage and current form the prerequisite for the localization and elimination of trouble spots. The quality of the energy system is assessed on basis of amplitude, Route Mean Square (RMS) value, frequency, phase symmetry and Total Harmonic Distortion (THD). [4] Table 1: General categorization of electrical disturbances. Disturbance Causes Effects underand overvoltage charging and discharging of bus capacitors sudden load decrease, increase disruption of inductive currents malfunction, breakdown or destruction of components destructive impact on the energy supply supply interruption defective cable routing overstraining of energy supply dynamics flickers load fluctuations harmonics discontinuous and switching operation of nonlinear components periodic switching and balancing processes commutation dips when switching power converters mechanical load disturbances of electrical machines additional losses overstraining of component and risk of malfunction electromagnetic disturbance of neighbouring systems voltage imbalances interharmonics periodic operations asynchronous to the line frequency 285 Model-Based Analysis of Transient Processes in Highly Available Energy Systems In contrast to large-scaled energy systems, automotive Direct Current (DC) systems do not have standardized quality criteria. Each Original Equipment Manufacturer (OEM) defines quality parameters itself. Klötzl [9] suggests the calculation of the quantifiable parameters degree G st and reserve of stability R st according to Equations (1), (2). These parameters are based on the relative frequency of voltages f(U k ) at the component terminals in comparison to stable limits U min , U max and the nominal voltage U N as shown in Figure 1. U degr corresponds to the degradation voltage. If the terminal voltage falls below U degr , the power of the component is brought down in order to unburden the energy system. Both reserve and degree of stability reach their maximum 1 if the terminal voltage equals U N for the entire period considered. As soon as the terminal voltage exceeds or undershoots the nominal voltage U N , both stability parameters G st and R st decrease. 𝐺 ∑ 𝑓 𝑈 1 ∑ 𝑓 𝑈 1 (1) 𝑅 min 1 , 1 (2) DIN ISO16750 [8] defines requirements to component manufactures in respect to minimum U min and maximum voltage U max in different classes. For further considerations U min is assumed to be 10.5 V and U max to be 16 V under consideration of the specified limits. The requested dynamics of a component in form of the maximum temporal current or voltage change is not mapped with the introduced quality criteria G st , R st . A deeper investigation of the automotive energy system is used in order to evaluate the influence of electrical components on the overall system stability in Chapter 3. Figure 1: Frequency of terminal voltages and characteristic component values. 286 Model-Based Analysis of Transient Processes in Highly Available Energy Systems 2 Model-Based Investigation of System Disturbances Electrical components affect the stability of the attached energy system through their dynamic power demand or supply due to recuperation. The strength of influence is analysed using the simulation setup of Figure 2 implemented in Dymola Modelica. The energy supply is solely provided by a 90 Ah Absorbent Glass Mat (AGM) lead battery. This model consists of an Open Circuit Voltage (OCV) U OCV of 13 V corresponding to a State-of-Charge (SOC) of 100 %, an internal resistance R i and three RCstages. Resulting time constants model chemical and electrical conduction processes such as diffusion, migration and charge transfer in a simplified form. The electrical line is modelled as series circuit of ohmic resistance R L1 and inductance L 1 . R L1 is assumed to be 25 mΩ and L 1 = 7 µH. This corresponds to average values of the whole supply path of a component taking into account cables, fuses, distributors and ground return. This simple two-component model is sufficient for frequencies f below 10 kHz (ω < 63 kHz) according to Gehring [6]. The following investigations are focused on a frequency range (ω < 20 kHz). For frequencies above the accuracy of the transmission line model needs to be rechecked. Figure 2: Simulation setup for investigating network disturbances. The component to be examined is modelled via its input circuitry and a characteristic electrical load. The generic input circuitry consists of overvoltage, feedback and reverse polarity protection, Electromagnetic Compatibility (EMC) filter and capacities. Three different loads are applied to the input circuitry in order to model the dynamic behaviour of the component as shown in Figure 3. As a first option, a typical constant load current in dependence of the components power class is connected. Another load is a switched RL-series circuit representing a periodic discontinuous load (e.g. electrical heating). The third equivalent circuit represents a 6-phase DC/ AC inverter and a connected Permanent Magnet Synchronous Machine (PMSM). The mechanical load can be adjusted or changed instantaneously. Figure 3: Examined loads connected to the secondary side of the input circuitry. 287 Model-Based Analysis of Transient Processes in Highly Available Energy Systems 2.1 Input Circuitries as Two-Ports As shown in Figure 2, the input circuitry can be modelled as a two-port and is fully described with four impedances Z 11 , Z 12 , Z 21 , Z 22 according to Equation (3). 𝑈 s 𝑈 𝑠 𝑍 𝑍 𝑍 𝑍 𝐼 𝑠 𝐼 𝑠 (3) Figure 4 visualizes the equivalent input circuitry of the Electric Power Steering (EPS). U 1 is the input voltage which directly links to the automotive 12V energy system (connectors Kl. 30 and Kl. 31). An electric load is applied on the secondary side of the twoport with the input voltage U 2 . The introduced input circuitry consists of common-modethe coupling choke L CMC with the Direct Current Resistances (DCR) R M1 , R M2 . Moreover, the circuit includes an inductance L 1 with resistance R L1 and two parallel capacities C 1 , C 2 with the Equivalent Series Resistances (ESR) R C1 , R C2 . Figure 4: Exemplary input circuitry of the EPS. For R C =R C1 =R C2 , C=C 1 =C 2 and the chokes coupling coefficient k, the two-port can be simplified and fully described with impedances Z 11 and Z 22 of Equation (4). Impedances Z 12 , Z 21 are equal to impedance Z 22 for the proposed input circuitry. 𝑍 2 𝑠 𝐿 𝑘𝐿 2𝑅 𝑅 𝑠𝐿 𝑅 𝑍 𝑅 (4) 2.2 Open Loop Response When switching on the input circuitry without connected load, bus capacitors are charged and the whole system oscillates until steady state is reached. Thus, the energy supply is discharged with an elevated current peak I max . At the same time, the voltage U Bat collapses, which could lead to malfunctions of individual components according to voltage limits given in Figure 1. Figure 5 presents battery voltage U Bat and current I Bat for the interconnected EPS being switched on. Both U Bat and I Bat oscillate with the systems resonance frequency ω res. The magnitude of the voltage dip can be quantified using degrees of stability G st , R st introduced in Chapter 1. Thus, further stability considerations are carried out based on the current profile. 288 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Figure 5: Simulated switching on behaviour of an open loop EPS input circuitry. Charging the circuitry integrated capacities leads to the peak current I max . The current amplitudes decrease, following an exponential decay. This decay can be explained with Equation (5) using the resonance peak current I res and attenuation coefficient D. The resonance current is defined as the amplitude current which would occur when the system is stimulated with battery voltage U Bat oscillating at ω res . Initially, the current increases with a slope m i requesting dynamics of the system. This current slope strongly depends on I res and ω res as shown in Equations (6), (7). i 𝑡 𝐼 𝑒 sin 𝜔 𝑡 (5) i 𝑡 𝐼 𝜔 𝑒 cos 𝜔 𝑡 Dsin 𝜔 𝑡 (6) 𝑚 max i 𝑡 i 𝑡 0 𝐼 𝜔 (7) Both values I res and ω res can be extracted from the magnitude of relevant conductivities in the frequency domain. Figure 6 visualizes the bode plot of the two-ports input conductivity I 1 / U 1 (reciprocal of Z 11 ) and the systems overall conductivity I 1 / U OCV . ω res and I res / U res correspond to the resonance point of the major systems conductivity. Figure 6: Bode plot of the energy system (solid) and EPS input circuitry (dashed). 289 Model-Based Analysis of Transient Processes in Highly Available Energy Systems 2.3 Load Response An attached load to the input circuitry changes the total impedance of the system. For example, the input circuitry is no longer completely described with the input impedance Z 11 , rather impedances Z 12 , Z 22 influence the systems behaviour expressing the electrical connection between the primary and secondary side of the two-port. Figure 7 shows the voltage and current profile of the battery when switching on a constant load current I ss at the secondary side of the EPS’s input circuitry. In analogy to the balancing process of the open loop response, both U Bat and I Bat oscillate with a frequency ω res,load . This frequency differs from the open loop resonance frequency ω res . Moreover, there is a current overshoot I max when switching on the load. Figure 7: Simulated switching on secondary load with EPS input circuitry. Input circuitries are typically based on a π-two-port architecture. As a result, two-port impedances can be transformed into a star equivalent circuit using the delta/ star transformation as shown in Figure 8. Figure 8: Delta/ star transformation of a π-two-port architecture. 290 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Equation (7) shows four characteristic two-port impedances Z 11 , Z 12 , Z 21 and Z 22 resulting from Z a , Z b , Z c . Impedances Z A , Z B , Z C characterizing the star circuit can be expressed with the two-port impedances as shown in Equation (8). 𝑍 𝑍 𝑍 𝑍 (7) 𝑍 𝑧 𝑧 𝑍 𝑧 𝑧 𝑍 𝑧 (8) In order to calculate the resonance frequency and a measure for load-dependent perturbations on the energy system, the transfer function has to be adapted. Equation (9) expresses the correlation of primary current I 1 , secondary current I 2 and primary voltage U 1 according to Figure 2. The resulting transfer functions include star impedances Z A , Z B , Z C being derived in Equation (8) , the battery impedance Z Bat as well as the line impedance Z Line . 𝐼 𝑠 𝐼 𝑠 𝑈 𝑠 (9) Neglecting the disturbance caused by U 1 , the correlation of I 1 / I 2 from Equation (9) can be analysed regarding magnitude and resonance frequency. Figure (9) presents the bode plot of I 1 / I 2 of the EPS (compare circuit in Figure 4). The maximum highlights the resonance frequency ω res,load at which a secondary load change has a major influence on the primary perturbation. This also corresponds to the transient frequency occurring at load change as shown in Figure 7. The maximum magnitude at this frequency is a measure of current overshoot of I 1 due to secondary side load changes. Figure 9: Bode plot of I 1 / I 2 of the EPS. 291 Model-Based Analysis of Transient Processes in Highly Available Energy Systems 2.4 Switched Resistive-Inductive Loads Another typical load profile is triggered by periodically switching systems. Similar profiles can be found at electric heating systems that are located in the rear window or the seat. They are driven by Pulse Width Modulation (PWM) to adjust the average heating power. Heating systems have particularly high time constants. Hence, users cannot perceive any difference in comparison to a constant power with reduced current flow. Figure 10: Simulated switched load as a replica of a PWM heater. Figure 10 illustrates typical disturbances on the energy system triggered by a switched heating system within the armrest of the driver-side door. This component requires a current of about 20 A in an exemplary switching cycle T sw . Similar to the load response in Section 2.3, there is a current overshoot when switching on the load e.g. at t=1ms and a transient at a certain resonance frequency. When switching off the load at t=1.5ms the battery still charges the bus capacitors of the input circuit because of its comparatively sluggish time constants. Ultimately, the capacitors discharge that leads to a current undershoot and a voltage overshoot at the battery terminals. 2.5 Mechanical Load Disturbance Several components of the automotive energy system contain electronically controlled machines. An EPS, for example, is built up of a 6-phase inverter and a PMSM. A sudden mechanical load disturbance on the drive shaft of the machine also leads to disturbances within the current supply. Figure 11 displays battery voltage and current for an EPS when being put into operation at t=0 s. The machine requests an excessive current. As a result, the battery voltage collapses. The three-pole machine drives up to f=50 Hz at a constant ratio of frequency and voltage. At t=0.7 s, the PMSM reaches its nominal number of revolutions and its steady state. 292 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Figure 11: Simulated load disturbance on a PMSM with 6-phase inverter. A sudden load disturbance in height of the nominal moment 6 Nm occurs at t=1 s. This disturbance appears sluggish within the voltage trace U Bat . Additionally, the machine requests a recognizable high current and thus, burdens the dynamics of the energy system, respectively. The connected 1 kW machine represents a typical steering machine used in the automotive industry. 3 Stability Impact of Automotive Components Automotive energy systems contain generators, storages, converters and loads. In general, loads can be divided into electronically controlled machines, Electronic Control Units (ECUs) and switched inductive-ohmic impedances. Six typical components L1 - L6 of each group are compared in order to investigate system disturbances due to their load characteristics. L1, L2, L3 are high-power controlled machines such as electric fan (L1), EPS (L2) and rear axle steering (L3). L4, L5, L6 need significantly less power. L4 summarizes all loads within a door such as electric armrest heating or window lifter. L5 is a high-power ECU and L6 represents roof-integrated loads. Components are compared regarding their influence on energy system disturbances based on previously defined quantifiable criteria. Figure 12 shows both resonance current I res and maximum peak current I max for the consideration of the open loop input circuitries. Electronically controlled machines L1, L2, L3 lead to remarkable elevated peak currents while characteristic current values of other loads are comparatively small. The difference is justified by significantly greater bus capacitors of components L1, L2, L3. The ratio of I res and I max differs for each component due to dissimilar attenuation coefficients D as visualized in Figure 12. 293 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Figure 12: Simulated resonance current I res , amplitude current I max and attenuation D. Table 2 contrasts the resonance frequencies ω res for open loop and load conditions. Components L4, L5 and L3 lead to comparatively high-frequency transient oscillations. These resonance frequencies influence damping and slope of component triggered transients. Table 2: System-inherent resonance frequencies. Component Exemplary Application 𝜔 𝐼 0 𝜔 𝐼 𝐼 L1 electric fan 3.62 kHz 2.48 kHz L2 EPS 5.54 kHz 4.33 kHz L3 rear axle steering 7.32 kHz 7.04 kHz L4 PWM heating 14.40 kHz 12.11 kHz L5 ECU 9.03 kHz 7.85 kHz L6 interior lighting 2.58 kHz 1.23 kHz Figure 13 presents the maximum current slope m i and the magnitude of I 1 / I 2 . In contrast to the previous comparison, components L4, L5 show a strong influence on transient processes within the interconnected energy system. In particular, components with high resonance frequencies ω res lead to elevated dynamic requirements in the form of sharp current slopes m i . Furthermore, components L4, L5 generate relatively high current overshoots when requesting secondary load currents. Figure 13: Simulated current slope m i and load perturbations. 294 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Figure 14 compares stability parameters degree G st and reserve stability R st for the sequence of turn on and off cycles of the input circuitries themselves and typical load currents. Because of greater load currents of controlled machines L1, L2, L3, their stability degree is significantly lower. Also, the reserve stability to the minimum voltage limit R stis especially small with these components. An overshoot of the voltage depends largely on the size of the bus capacitors. Thus, the positive reserve stability as a distance measure to the maximum voltage limit is more crucial with the same components L1, L2, L3. Component L6 contains diodes to ensure a unidirectional power flow. This excludes voltage overshoots due to discharge of bus capacitors. Figure 14: Simulated degree G st and reserve of voltage stability R st . 4 Generic Sensitivity Analysis Various input circuitries of automotive energy components are based on a delta/ πarchitecture. This chapter investigates the influence of the circuitries parameters on the disturbance within the energy system due to the load. The parameters of a highpower audio amplifier shown in Table 3 serve as the starting point of the π-input circuitry visualized in Figure 15. Table 3: Adjustable circuit parameters. System Component Value Line L 1 7 µH R L1 25 mΩ Input Circuitry C 1 560 µF C 2 1680 µF R C1 40 mΩ R C2 13 mΩ L 2 470 nH R L2 3 mΩ 295 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Figure 15: Exemplary π-input circuitry. The first investigation is the influence of the capacities C 1 , C 2 on the key parameters resonance current I r , maximum current I m , resonance frequency ω res , current slope m i and load coupling I 1 / I 2 . Figure 16 summarises the results of the sensitivity analysis of both capacities. Each capacity is changed by ΔC while the other capacity is left at the starting value of Table 3. One remarkeable finding is that an increase in the capacity does not elevate the resonance current inevitably. The overall system is changed and thus, the resonance frequency. An increase of the capacity leads to a drecrease of ω res for both C 1 and C 2 . At the same time, the maximum current I m increases. C 1 affects the resonance current I r . C 2 does not affect the resonance frequency essentially. Figure 16: Simulated sensitivity of the system in dependence of capacities C 1 and C 2 . 296 Model-Based Analysis of Transient Processes in Highly Available Energy Systems The current slope m i decreases for higher capacities as pointed out in Figure 15. This is due to smaller resonance frequencies. Furthermore, the transmission of disturbances in the current profile of the secondary current I 2 to the primary current I 1 can be strongly influenced by the parametrization of the input circuitries capacities. An increased capacitor C 1 reduces the influence of I 2 on I 1 . C 2, however, can only slightly reduce this influence. The second investigation is the impact of capacities ESRs on the same parameters as analyzed in the sensitivity of capacities C 1 and C 2 . Figure 17 indicates the influence of the capacities ESRs R C1 , R C2 . Each resistance is changed by ΔR C while the remainder keeps at the starting value of Table 3. Especially R C2 can reduce I r and I m . Moreover, this resistance increases ω res . In opposite, R C1 does not strongly affect I r , I m , ω res or m i . Current slope m i has a minimum value at a certain R C2 for the given energy system and the input circuitries parametrization. Additionally, the current overshoot is reduced at the same R C2 by about 30 %. Despite this effect, an increased ESR elevates the current overshoot according to R C1 . A power demand cannot be satisfied fast enough from the bus capacity. The resistance hinders the current flow through the capacity and thus increases disturbances within the energy system due to secondary power fluctuations. Figure 17: Simulated sensitivity of the system in dependence of ESRs R C1 and R C2 . 297 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Within the third investigation, the inductance L 2 of the input circuitry visualized in Figure 15 is varied. The inductance is changed by ΔL 2 for the sensitivity analysis as displayed in Figure 18. Apparently, the inductance does not strongly influence the resonance current I res , but it cuts both peak current I max and resonance frequency ω res . Due to the reduction of ω res and an almost constant I res , the current slope m i is reduced as well. Thus, the overall stress within the energy system can be reduced with a larger dimensioned input circuitry integrated inductance. The result differs for the consideration of the closed loop response of the varied input circuitry with a connected load. Objectively, L 2 increases the overshooting of the energy systems current when power fluctuations at the secondary side of the input circuitry occur. Figure 18 shows a rising ratio of input current I 1 and output current I 2 at the load dependend resonance frequency ω res . A greater induced voltage due to the raised inductance appears in the current profile of the energy system. Figure 18: Simulated sensitivity of the system in dependence of inductance L 2 . 298 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Finally, the influence of the electrical connection between energy supply and component should be investigated. The electrical parameters of the line R L1 and L 1 are adjusted relative to their initial values by ΔR L1 ,L 1 . This adaption corresponds to a line extension neglecting the constancy of fuse and distributor values. All critical values I res , I max, ω res and m i are reduced with a longer line according to Figure 19. The elevated inductance and resistance lead to a higher attenuation of transient currents. This influence is clearly visible in the decreasing ratio of I res and I max . The downsized resonance frequency ω res affects the current slope m i accordingly. In contrast to the inductance adaption of L 2 , a line extension positively influences power fluctuation trigged current overshoots I 1 / I 2 . Large-sized resistors and inductors of the line hinder the line current flow. Thus, the line forces the component to be supplied by the charged capacities of the input circuitry at first. This power demand compensation reduces component-side fluctuation triggered disturbances within the energy system. Figure 19: Simulated sensitivity of the system in dependence of the line R L1 , L 1 . 299 Model-Based Analysis of Transient Processes in Highly Available Energy Systems 5 Conclusion This paper introduced disturbances within automotive electrical energy systems in analogy to large-scaled AC energy systems. Possible causes were identified and compared to their effects. Beyond, the stability of automotive DC energy systems was defined using the quantifiable degree and reserve of stability. This definition found in the literature has been extended by dynamic current characteristics such as slope, transient frequency and amplitude. The input circuitry of components strongly influences the stability parameters and thus disturbances within the energy system caused by power fluctuations of the load. Input circuitries can be abstracted as two-ports with four characteristic impedances to examine power transients. Mathematical investigations in frequency and time domain support the characterization of components within an energy system. Typical automotive energy system components have been categorized and considered in respect to their transient behaviour and influence on the energy systems stability. High-power machines controlled by power electronics or switched loads were identified as particularly strong stressors. An insight into the sensitivity analysis of a generic delta input circuitry completes the stability considerations. The transient behaviour can be strongly influenced by the selection of bus capacitors and their ESRs. Depending on the energy system, the influence can even be minimized. Results of Chapter 4 show that system disturbances of an audio amplifier can be reduced by up to 30 % when increasing the input circuit integrated capacity by 1 mF or its series resistance by 50 mΩ. However, a general statement is challenging because the overall energy system and the shift of the resonance strongly depend on precise values. In principle, an extended line reduces critical current peak and disturbances by accepting raised losses. The work done in this paper forms the basis for pervasive investigations of transient processes within automotive energy systems. To ensure a safe energy supply of safety-relevant components, identified stressors need to be modelled physically. In contrast to batteries, the dynamics for the bidirectional power transfer of DC/ DC converters are limited based on their integrated control. Thus, a sole consideration of the impedance-dependent resonances is not sufficient for dynamic processes supplied by controlled energy sources. In addition, the accuracy of the line model for frequencies above 10 kHz needs to be rechecked in order to apply the analysis on the whole energy system consisting of interconnected energy suppliers, storages and loads. Bibliography [1] Avalur, K. K. G. and Azeemuddin, S. ‘Power management IC architecture inautomotive environment: Case study of rear view camera’. In: TENCON 2017 -2017 IEEE Region 10 Conference. Nov. 2017. [2] Bai, H. et al. ‘The Short-Time-Scale Transient Processes in High-Voltage and High-Power Isolated Bidirectional DC-DC Converters’. In: IEEE Transactions on 300 Model-Based Analysis of Transient Processes in Highly Available Energy Systems Power Electronics 23.6 (Nov. 2008), pp. 2648-2656. issn: 0885-8993. doi: 10.1109/ TPEL.2008.2005106. [3] Diez, T. P. et al. ‘Transient voltage characterization for automotive 42 volt power systems’. In: IEEE International Symposium on Electromagnetic Compatibility. Symposium Record (Cat. No.00CH37016). Vol. 2. Aug. 2000, 921-926 vol.2. doi: 10.1109/ ISEMC.2000.874746. [4] Dorner, H.: Wissenswertes über Netzrückwirkungen. Grundlagen - Anlagen- Gesamtbetrachtung - Simulation - Normgrenzwerte - Maßnahmen zur Netzverbesserung. VDE Schriftenreihe. 2013. [5] Frazier, R. K. and Alles, S. ‘Comparison of ISO 7637 transient waveforms to real world automotive transient phenomena’. In: 2005 International Symposium on Electromagnetic Compatibility, 2005. EMC 2005. Vol. 3. Aug. 2005, 949-954 Vol. 3. doi: 10.1109/ ISEMC.2005.1513662. [6] Gehring, R. ‘Beitrag zur Untersuchung und Erhöhung der Spannungsstabilität des elektrischen Energiebordnetzes im Kraftfahrzeug’. PhD thesis. Technische Universität München, 2011. [7] ISO Central Secretary. Road vehicles - Electrical disturbances from conduction and coupling - Part 2: Electrical transient conduction along supply lines only. en. Standard ISO 7637-2: 2011. Geneva, CH: International Organization for Standardization, 2011. [8] ISO Central Secretary. Road vehicles - Environmental conditions and electrical testing for electrical and electronic equipment. Standard ISO 16750-1: 2018. Geneva, CH: International Organization for Standardization, 2018. [9] Klötzl, J.: Stabilität automobiler Leistungsbordnetze. Universität der Bundeswehr München. 2012. [10] Nasim, S. et al. ‘Cost-effective means of protecting electronics from voltage transients and overvoltages in an automotive network’. In: 2009 IEEE Student Conference on Research and Development (SCOReD). Nov. 2009, pp. 357-359. doi: 10.1109/ SCORED.2009.5443000. [11] Schipperges, F. et al. ‘Untersuchung fehlertoleranter Energiebordnetze’. In: Internationale Fachtagung EEHE (Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen). 2017. 301 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications Fabian Schipperges, Stefan Schumi, Stefan Hörtling, Bernard Bäker Abstract Redundant and fault-tolerant on-board power supplies will be necessary for automated vehicles. Safety switches enable fault-tolerance by switching-off overcurrents caused by a short circuit or due to other impermissible voltage conditions. Since harsh and varying conditions challenge these safety relevant devices in the case of a short-circuit current, the avalanche effect is examined. Exceeding the rated avalanche energy destroys the devices. Hence, the design must be proven. In this work, we investigate switching effects according to varying line impedances and different periods to switch-off. For this purpose, a test series is carried out to determine avalanche energies and to prove simulation results. A bidirectional Mosfet-based safety switch is used for the test series. 1 Introduction Future automated driving requires safe and reliable power supply of the safety relevant ECUs. A supply blackout that causes malfunctions of these functions must be avoided. Therefore, a wide variety of redundant and fault-tolerant power net architectures is under consideration. These architectures provide a basis for establishing an appropriate safety concept to fulfill functional safety requirements according to ISO 26262. Since we expect parallel active redundancies in a vehicle, coupling elements like safety switches or DC/ DC converters are required. In this work we focus on Mosfet-based switches. These switches guarantee fault-tolerance in the case of a fault by separating faulty branches. However, switching effects due to line inductances and the technological properties of the semiconductor must be taken into account. We focus on the occurrence of a short-circuit current and consequently the switch-off with a bidirectional safety switch. In section 2, we introduce the principles of redundant and fault-tolerant power nets and discuss the application of unidirectional and bidirectional switches. Since switching-off loads and high currents driving the devices into avalanche, this effect is explained. Section 3 provides a brief overview of literature dealing with the phenomenon of short circuit. We deploy an analytical description for the short-circuit current profile and the avalanche operation mode, considering the mentioned literature. Section 4 presents a test setup and the results of experimental investigations. The avalanche energies under varying line impedances and short circuit durations are calculated. The results are compared to simulation results in section 5. Finally, we discuss how to improve the design of the switch. 302 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications 2 Fault-tolerant On-board Power Supplies Using Mosfet-based Safety Switches 2.1 Redundant Power Net - Fault-tolerant and redundant architecture approaches have the purpose to fulfill requirements of functional safety according to ISO 26262. Redundant structures are based on the method of decomposition allowing safety goals to be partitioned [1]. This means partly reduced requirements related to the fault metrics, though, at the price of at least doubling of all safety-relevant components. Fail-operational emergency operation for the transition of the vehicle to a safe state is available. This is a requirement in the case of a relevant failure of one of the subsystems during automated driving. - -- Figure 1: Example of a redundant on-board power supply providing a 1-out-of-2 redundancy of safety-relevant functions (SR1, SR2) with two independent batteries. To achieve a trade-off within the requirements, in a vehicle, active parallel redundancies are to be expected. Hence, we expect 1-out-of-2 redundancy concepts for safety-relevant ECUs as well as for their power supply, taking functional safety and legal regulations into account. Figure 1 illustrates a possible concept containing two independent power nets ASIL compliant Power Net 1/ 2. Each subsystem consists of an independent battery supply B1/ B2 and a set of (safety-relevant) ECUs SR1/ SR2. They allow transiting the vehicle to a safe state if during automated operation a failure occurs. A comparison of different architectures is not the purpose of this work, but besides the presented topology, there is a wide variety of suitable concepts like ring, tree and backbone architectures. Nevertheless, the properties and application of Mosfet-based switches discussed in this paper, is relevant for all these architectures, since Mosfet-based switches are suitable devices for fault-tolerant operations in general. 2.2 Voltage Requirements Standards (VW 80000 [2], ISO 16750-2 [3] and others) define static and dynamic voltages and correlated functional status classifications. Within these voltage ranges, the ECUs have to work properly as defined. Likewise, the power supply of the safetyrelevant ECUs has to fulfill reliability and availability requirements considering these DC/ DC S0 12V HV/ 48V ASIL compliant Power Net 1 SR1 ... S1 S2 S3 ... SR2 ... ASIL compliant Power Net 2 B2 B1 303 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications voltage range requirements. We find recommended static ranges of 6V to 16V with different functional status classifications as well as dynamic threshold values downward to 0V. Dynamic voltages depend strictly on correlated time intervals. Within these boundaries, all safety-relevant ECUs have to perform as designed. 2.3 Mosfet Safety Switches, Avalanche Effect Figure 1 indicates different types of switches. All of them are assumed to be solidstate Mosfet-based switches. We distinguish between unidirectional and bidirectional current interruption. Typically, bidirectional devices are located between (sub-) power nets with independent voltages sources (DC/ DC converter or battery). These switches are represented by S0, S2 and S3. The switch S1 in Figure 1 depicts a unidirectional switch as a proper device to switch-off loads. A unidirectional switch consists of only one Mosfet device. Since the device structure of a Mosfet include an intrinsic and parasitic diode (body diode), a bidirectional switch consists of at least two Mosfet devices. These two Mosfets are connected in serial. They share a common source or a common drain connection, so we name it back-to-back (B2B) switch. Figure 2 illustrates the B2B switch that is examined in this work. Each row of Mosfets include four Mosfets connected in parallel. Figure 2: Considered and examined (back-to-back) safety switch. The switch consists of 2x 4 Mosfets (parallel) of type IPLU300N04S4 [4] and driver of type AUIR3241S [5]. For the design of the back-to-back safety switch, the maximum current amplitude and continuous current has to be taken into account as well as the dissipation of the energy released at the moment of switch-off. As explained and examined in section 3 and section 4, this energy depends on current amplitude and line inductance ratios. With switching-off a short-circuit current, inductive voltage spikes can occur. If the specific threshold of the breaking voltage V BR,DSS is exceeded, the previously blocking device will be in avalanche operation mode. In this mode, high electric fields in reversed biased p-njunctions are generated and exceed the critical electric field E Crit defined by technological properties. Due to a strong electric field, energetic free carriers provoke electron-hole pairs by impact ionization and lead to a multiplication effect. This effect leads to increased current and consequently to a steep temperature rise inside the device [6; 7]. Since the avalanche effect and potential failure mecha- B1 B2 Driver back-to-back switch (B2B) 4x 4x D S D G G 304 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications nisms are well-known, improved devices provide the effect as a feature for specific applications. Manufactures specify permissible avalanche ratings in the data sheets, depending on current, duration and temperature for a single avalanche pulse. However, if the device is avalanche rated, the avalanche energy dissipating inside the device must not be exceeded to avoid destruction. This requirement has to be fulfilled for the entire typical automotive temperature range. Figure 3: Simulation of a short circuit (current rise until 662µs) and avalanche operation mode due to switch-off. While the Mosfet is facing avalanche, the voltage V av =V DRAIN -V SOURCE across the device is clamping. During this period, the energy previously stored in line inductance is dissipated into the device. Figure 3 shows the simulation results of a short circuit and the clamping characteristics of the B2B switch, consequently operating in avalanche mode (dashed rectangle) due to current switch-off. The avalanche energy E AV can be calculated as follows: 𝐸 𝑉 𝑡 𝐼 𝑡 𝑑𝑡 ∆ (1) 3 Modelling Approach: Short Circuit and Avalanche Effect - It is important to capture the main influencing characteristics of the wiring, the battery and the switching device. The influence of wiring and battery condition is illustrated in Figure 4. - Figure 4: Shorting of a 60Ah LiFePO 4 battery under various conditions. V SOURCE → V DRAIN → avalanche event 305 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications In this experiment, we shorted a 60Ah Lithium-Iron-Phosphate battery (LiFePO 4 ). Two tests were carried out with a total wire length of approximately nine meters and battery temperatures of 22.5°C and -20°C, respectively. The third test was carried out with a wire length of approximately 2.5 meters and a battery temperature of 22.5°C. Without any deeper analysis, the strong impact of the battery impedance and the line impedance on the current profile is easily recognizable. These tests clarify the need for a deeper system knowledge. In the following, the impact of system impedance and reaction time (the time segment between detection and switch-off) is investigated. We deduce how to model a short circuit event and short circuit switch-off. The phenomenom of short circuit has been investigated in literature in different manner and with different foci. Kriston et al. focuses on external short circuit of differnt lithium-ion cells. They examine the thermal behaviour, the current rate and voltage response under varying conditions with an experimental approach. They derive a normalized ratio of the external and internal resistances. The influence of this ratio on the maximum current is pointed out [8]. Okazaki et al. compare observed and predicted external initial short-circuit currents for lead-acid batteries. They consider two different methods for determining the internal ohmic resistance. The prediction of the initial short-circuit current is based on Ohm’s law [9]. IEC 61660-1 presents a method for calculating a short-circuit current [10]. Even though this standard focuses on d.c. auxiliary installations in power plants, it provides an applicable standard approximation function. Figure 5 shows this standard approximation and a diagram of a typical short-circuit current. 𝑖 𝑡 𝑖 , 0 𝑡 𝑡 (2) 𝑖 𝑡 𝑖 1 𝑝 𝑒 𝑝 , 𝑡 𝑡 (3) Figure 5: - Standard approximation function (left) and diagram of a typical short-circuit current of a battery (right). The standard approximation can be described by formula (1) and (2). Illustrations from [10]. Schwimmbeck et al. discuss short circuit in automotive power net and takes the current power net conditions into account [11]. A method is presented that considers t p 0 0  2 i  1 i p i 1 (t) i 2 (t) I k T k t IEC 683/ 97 i B i pB t pB I kB t IEC 680/ 97 306 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications probability distributions and the variation of influencing parameters on short circuit behavior. These investigations are important to understand short circuit mechanisms for determining maximum currents and voltage limits. The above presented literature helps to understand short circuit with respect to varying conditions and maximum and minimum values. In this work, we focus on active switch-off of short-circuit currents. Especially, we examine switching effects provoking avalanche and clamping of the Mosfets. Infineon’s application note Switching Inductive Loads [12] explains how to evaluate the energy to be dissipated during avalanche operation mode. Since we examine the switching-off of short-circuit currents with a high side device, the objective of the mentioned application note differs. Nevertheless, the physics and theory is valid in both cases. 3.1 Wiring and Switch We consider equivalent circuit diagrams relating to the experimental setup as presented in section 4. For the modelling approach the following assumptions are made: (i) Line capacitances are neglected. (ii) The short circuit occurrence and the short-circuit current interruption are assumed to be ideal switching processes. (iii) Since a short circuit resistance could be an arbitrary value, worst case scenario is examined. In the experimental setup, the short circuit impedance, besides the line and battery impedance, is represented by the R DS,on of a Mosfet low side switch as described in section 4. (iv) The battery voltage is assumed to be constant as described in section 3.2. For an analytical description of the system, we distinguish between two periods. the occurrence of a short circuit is the beginning of period A. During this period, current increases rapidly. It ends with current interruption due to switch-off. Now, period B describes avalanche operation mode of the switch and it lasts until the current is decreased to zero and the entire energy is dissipated. - Figure 6: Timeline of a short circuit and switch-off event. For analytical description, it is divided into two periods (A and B). For case A, we assume an equivalent circuit diagram as shown in Figure 7. The line impedances of the wires are modeled by RL-circuits. The B2B switch in on-phase is modeled as ohmic resistance representing R DS,on of the B2B Mosfets. During the period of short circuit, the same assumptions can be made for low side switch as for the B2B. A constant voltage source and an ohmic resistance model the battery properties. Initial short circuit trigger switch-off current interrupted latency 0 ≤ t ≤ t on t on < t ≤ t off avalanche current increasing A B 307 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications Case A (0 ≤ t ≤ t on ) With second Kirchhoff’s law and di/ dt≠0 we define: 𝑉 𝑅 , 𝑅 𝑅 𝑅 𝑅 𝑖 𝑡 𝐿 𝐿 𝑖 𝑡 𝑑𝑡 (4) To determine the short-circuit current i sc (t) we solve the differential equation in (4) with assumption i(t=0)=0: 𝑖 𝑡 , 1 exp 𝑡 𝜏 ; 𝜏 , (5) - Figure 7: Case A. The model includes battery and wire impedance, a short circuit resistance R sc =R DS,on and switch resistance R SW =R DS,on (for conducting mode). Case B (t on ≤ t ≤ t off ) In this period, the Mosfets of one row of the B2B are clamping. A constant voltage source emulate this behavior and a diode allows the direction of current flow. Figure 8 illustrates the equivalent circuit diagram. Since the B2B provides two rows of Mosfets, the diagram shows the complete equivalent circuit for the sake of completeness. With second Kirchhoff’s law we define: 𝑉 𝑉 𝑅 , 𝑅 𝑅 𝑅 𝑖 𝑡 𝐿 𝐿 𝑖 𝑡 𝑑𝑡 (6) With I on,max =i sc (t on ) we solve equation (4) and describe the decreasing current during avalanche: 𝑖 𝑡 , 𝐼 , , 𝑒𝑥𝑝 𝑡 𝜏 ; 𝜏 𝐿 𝐿 𝑅 , 𝑅 𝑅 𝑅 (7) 𝑖 𝑡 0; for 𝑡 𝑡 (8) With equation (1) and equation (5) the avalanche energy can be calculated. V B att i sc (t) Z B att R W 1 L W 1 R SW R W 2 L W 2 R sc Battery Wire 1 B2B Switch Wire 2 308 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications Figure 8: Case B. During avalanche mode, the switch is modeled by a diode and an equivalent voltage source that emulate the clamping voltage. 3.2 Battery In this work, for experimental investigation and simulation, a 40 Ah LiFePO 4 battery is used. The impedance of a battery varies as a function of state of charge, temperature, lifetime, application history and other. It is strongly non-linear. Since the battery impedance influences the behavior of a short circuit fundamentally, appropriate assumptions must be made. For this purpose, a series of tests with batteries and cells was carried out. Afterwards, electrical equivalent circuit models were developed and the measurements were compared to simulation results. A comprehensive description of the results is not the scope of this work. Since the series of tests of this work was carried out at room temperature and the short circuit duration was limited to 500µs, it is sufficient to assume a simple battery model. For the period between 0µs and maximum 500µs, the inductance and battery ohmic resistance dominate. Initially, the load carriers are provided mainly by discharging the electrochemical double layer that is built between electrode and electrolyte, so we neglect the influence of electrochemical charge transfer reaction. Therefore, we assume the state of charge to be constant, too. The inductance of the battery is small in comparison to line inductance, so it is neglected as well. The considered battery model consists quiete simply of a constant voltage source and a series resistance while the resistance includes battery terminal connectors and cell connectors. Please note that the assumptions made are not valid for a short circuit duration t>500µs. 4 Experimental Investigations This section presents the results of experimental investigations. The purpose of these experiments is a better understanding of shorting and switch-off short-circuit currents. The insights gained in experiments help to evaluate the avalanche energy E AV as a function of short circuit duration and line impedances. In addition, the ratio of line inductances is examined. Thus, we compare simulation results to measurements. V cl2 D cl2 V B att i av (t) Z B att R W 1 L W 1 D cl1 V cl1 R W 2 L W 2 R sc Battery Wire 1 B2B Switch Wire 2 309 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications A microcontroller is used to trigger a low side switch for provoking the short circuit and the B2B for switch-off and current interruption. The controller allows the variation of the time period Δt on between shorting and switching-off. Figure 6 illustrates a brief description of the timeline of the test procedure. A schematic representation of the test setup is shown in Figure 9. The equipment, conditions and experimental design are summarized in Table 1. Figure 9: Test setup. The µC allows the variation of the time between initial short circuit and interruption of the current. The line impedances were varied during test procedures. Table 1: Test equipment, experimental design and specified test conditions. µC: Arduino Uno Rev.3 battery: 40Ah LiFePO 4 , T=22.5°C, 80%SoC, V OCV =13.35V wire 1: 1m, 2m, 5m; 35mm 2 ; copper wire 2: 2m, 3m, 6m; 35mm 2 ; copper Each combination of wire lengths was tested with short circuit duration t on (µs): 10, 50, 100, 200, 350, 500 measuring system: ECAT/ EtherCAT CSM GmbH current transducers: LEM LLF 1010-S, bandwidth 200kHz ADC: AD4 OG1000, CSM GmbH, sampling rate 500kHz Impedance analyzer for determination of line impedance: Keysight E4990A. For all test episodes illustrated in Figure 10, the time to switch-off was specified as t ON ~100µs. The upper part of the figure presents the voltages measured at drain and source junction of the switch. For t ≤ t ON , V SOURCE and V DRAIN share approximately the same potential. During avalanche operation mode, we determined a potential difference of approximately 47.5V for all test episodes. This clamping voltage is defined by technological properties. Avalanche operation mode starts at the moment when the voltage spikes exceed the breakdown voltage V BR(DSS) of Mosfets. For the considered Mosfets, it is rated as V BR(DSS) = 40V and increases with temperature [4]. The influence of absolute wire impedance clarifes the measurements in Figure 10 on the left side. The time constant τ=L/ R define the current slew rate. The current ampli- Batt. Z W ir e1 Z W ir e2 uC ∆ t 310 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications tudes differ significantly while the voltage amplitudes and proportion are equal to each other. On the right side of Figure 10, the line impedances equal each other. Consequently, the current slew rates and amplitudes are equal to each other, too. The proportions of line impedances on both sides of the switch are oppositional, so the voltages at Drain and Source follows these distributions. This context should be kept in mind when designing the power net, since it could reduce the bill of materials. If the switch is located close to the battery, in the case of a switch-off, the feedback on intact power net is reduced by design. Figure 10: - Short circuit and avalanche operation mode with t ON~ 100µs and varying wire length. V av (t)=V SOURCE -V DRAIN remains the same value of approx. 47.5V. - The maximum current amplitude at t=t ON can be calculated by equation (5). Figure 11 presents the maximum values at t=t ON of all test episodes as a function of short circuit duration and line inductance. The maximum value of I sc,max = 938.2A was measured with L~4.2µH and t ON ~366µs. The avalanche energy for each test episode was calculated according to equation (1), taking the values of Figure 11 into account. The results for time Δt AV = t OFF -t ON are illustrated in Figure 12 (blue) as a function of line inductance and related to associated t ON . For comparison purposes, additionally, the diagram shows the permissible avalanche engery limits (red). We calculated these limits according to data sheet [4] in consideration of drain current I D . The junction temperature T j,start was assumed to be 22.5°C and to make sure this temperature assumption, we hold on for one hour between test episodes for heat dissipation. The illustrations of the avalanche energy in Figure 12 clarify the dominant influence of the current amplitude on avalanche energy while line impedance mainly influence the current slew rate and finally the current amplitude. We found the first off-limit condition with the parameters t ON ~366µs and L~4.2µH, when E AV ~1.963Ws was exceeded. We want to point out to the reader, that a real application reduces the limits of permissible avalanche energy due to a strong temperature dependence of the Mosfets. voltage V/ V current I sc / A voltage V/ V current I sc / A V DRAIN → V SOURCE → V DRAIN → V SOURCE → A B 311 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications To improve avalanche energy constraints due to the influence of temperature, the ambient temperature and continuous current should be reduced. Additionally, components that reduce the avalanche current are recommended, as we discuss in the next section. - Figure 11: Current I max,on =i sc (t on ) as a function of short circuit duration t on and line inductance L. - Figure 12: Permissible Avalanche Energy E AV (red), calculated under consideration of data sheet characteristics for room temperature. Energy calculated based on current and voltage measurements (blue/ green). 312 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications 5 Discussion and Outlook 5.1 Simulation Results Figure 13: Simulation results (red) in comparison to measurements (blue). Test conditions: 22.5°C, 80% SoC, L total ~8.33µH, R total ~3.7mΩ Simulations were carried out using Spice. Figure 13 exemplarily shows the results of one test episode. Generally, there is a good agreement between simulations and measurements. Deviations are particularly observed in avalanche operation mode. Deviations result from:  Assumptions and simplifications as described in section 3.  Neglect of the influence of temperature change. With using the manufacturer’s Level 1 Spice model, the device temperature is assumed to be constant during the test period.  Production distribution for V BR(DSS) lead to different clamping voltages [7].  Neglect of power net capacity and simplifications influence dynamic behavior.  Finite accuracy of measurement equipment. For a deeper insight we also refer to [13]. In this work, a comprehensive discussion about simulating Mosfet switches in fault-tolerant power nets is given. Additionally, the use of a free-wheeling diode is examined. The simulation results and experimental investigations as described in this work help to understand the avalanche effects and how the energy can be calculated. In future work, a wide temperature range will be investigated. 5.2 Improvement and Reduction of Transients In general, the period 0≤t≤t ON should be kept as short as possible by a fast and reliable overcurrent detection. Furthermore, the results of experimental investigations and simulations point out two challenges. First, the Mosfets of a safety switch provide only a finite avalanche energy capability that varies strongly due to temperature and current rate. For improvement, additional Mosfets in parallel would raise the avalanche energy capability. Likewise, the costs increase. Instead of raising the energy capability of Mosfets, it is advisable to distribute the avalanche current over additional diodes in order to reduce the avalanche energy dissipating into the Mosfets. 313 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications The second challenge, we want to point out, concerns the propagation of voltage spikes and clamping voltage into the intact (sub-) power net. In accordance with the applied standards we mentioned in section two, transient undervoltage or overvoltage must be within the defined limits. To counteract this propagation, again, diodes could be used. - Figure 14: a) voltage clamping with suppressor diode, b) free wheeling path via a diode in case of undervoltage, c) voltage clamping and energy absorption The illustration above depicts three options. In (a), a suppressor diode is connected to ground. The diode limits overvoltage peaks according to its breakdown voltage V BR . If this voltage is exceeded, the diode conducts reversely and the avalanche current of the Mosfets is being reduced. Alternatively, these diodes can be connected in parallel to the Mosfets. Figure 14 (c) shows this configuration considering a bidirectional suppressor diode. To counteract transient undervoltage and also reduction of avalanche current, diodes can be applied as presented in Figure 14 (b). The diode starts to conduct if the absolute value of (under-)voltage exceeds the forward bias of the diode. In this case, the voltage is limited to the diode’s forward voltage. In addition to the passive devices mentioned, active circuits like surge stoppers are also conceivable. However, those solutions require an increased technical effort. Since the selection of appropriate diodes and the positioning on Drain or Source depends on overall system design, only the principal characteristics are mentioned above. Finally, we want to point out to the reader that the connection of diodes to ground increases the probability of faults and influences relevant fault metrics. A further revision to guarantee fault metric requirements may increase the bill of materials again. 6 Conclusion In this paper, we derived how to model a short circuit and the avalanche effect of a Mosfet safety switch caused by overcurrent switch-off. Analytical and experimental investigation as well as simulation results clarify that the current amplitude dominates the energy dissipating during avalanche mode. A high line impedance is less critical than a small one, since the time constant limits current rise. The clamping voltage, as a specific technological property, arises during avalanche operation mode. The ratios of line inductances on both sides of the switch determine the voltage level on each side. This context should be considered when designing the power net in order to minimize transient feedbacks on intact power net per design. V + V + c) parallel energy absorption GN D GN D a) suppresor diode V + b) free-wheeling diode V + 314 Solid-state Safety Switch for Fault-tolerant Automotive Power Net Applications The avalanche energy limit of the examined bidirectional switch was reached with short wires and short circuit durations > 360µs. The employed switch provides a high avalanche energy capability since it consists of four Mosfets in parallel. However, in a real application, the limits of acceptable avalanche energy decrease rapidly with rising current rates and rising temperature. In general, the short circuit duration should be kept as short as possible in order to reduce current amplitudes and to save the bill of materials. Diodes help to improve the energy capability and to reduce transients. In Future work, an improved design and an overcurrent detection will be built and tested. Literature 1] International Organization for Standardization, 2018. ISO 26262: 2018 Road vehicles - Functional safety. [2] Volkswagen AG, 2017. VW 80000: Elektrische und elektronische Komponenten in Kraftfahrzugen bis 3,5t. [3] International Organization for Standardization, 2012. ISO 16750-2: Road vehicles - Environmental conditions and testing for electrical and electronic equipment - Part 2: Electrical loads [4] Infineon Technologies AG, 2015. Product Data Sheet: OptiMOS Power- Transistor IPLU300N04S4-R8. www.infineon.com. [5] Infineon Technologies AG, 2017. Product Data Sheet: Low quiescent current back to back Mosfet driver AUIR3241S. www.infineon.com. [6] Vishay, 2011. Application Note AN-1005: Power Mosfet Avalanche Design Guidelines. www.vishay.com. [7] Infineon Technologies AG, 2017. Application Note AN_201611_PL11_002: Some key facts about avalanche. www.infineon.com. [8] Kriston, A. et al., 2017. External short circuit performance of Graphite- LiNi1/ 3Co1/ 3Mn1/ 3O2 and Graphite-LiNi0.8Co0.15Al0.05O2 cells at different external resistances. In: Journal of Power Sources 361: 170-181. DOI: 10.1016/ j.jpowsour.2017.06.056 [9] Okazaki et al.,1986. Predicted and observed initial short circuit current for leadacid batteries. In: Journal of Applied Electrochemistry, 16(5): 631-635. DOI: 10.1007/ BF01006846 [10] International Electrotechnical Commission, 2000: IEC 61660-1 Short-circuit currents in d.c. auxiliary installations in power plants and substations - Part 1: Calculation of short-circuit currents. [11] Schwimmbeck, S. et al., 2018. Bewertung von Kurzschlüssen in Energiebordnetzarchitekturen. In: Elektrik/ Elektronik in Hybrid- und Elektrofahrzeugen und elektrisches Enegeriemanagement VIII. Würzburg, 12.-13.06.2018. Renningen: Expert Verlag. [12] Infineon Technologies AG, 2011. Application Note: Multichannel Low-Side Switches - Switching Inductive Loads. www.infineon.com. [13] Schumi, S., Schipperges, F.: Short Circuit and Avalanche Effects in 12V Power Distribution for Automated Driving. In: IEEE International Conference on Mechatronics 2019. Ilmenau, 18.-20.03.2019 - 315 Anwendungsmöglichkeiten von all-solid-state Zellen in Niedervolt-Autobatterien Ines Miller, Varvara Sharova, Robert Stanek Abstract The movement towards the use of electric vehicles is leading to the inevitable changes in the development of mobility. Hybrid electric vehicles, especially with a 48V onboard power supply is gaining an increased attention from the automotive OEMs. The main reasons include the improvements in the fuel efficiency and the ability to support a number of electrified functions, enhancing the driving experience. Electrification is enabled through the state-of-the-art lithium-ion batteries with liquid electrolytes. However, the continuously increasing demands on power and safety, push towards the development of new battery types, among which all-solid-state battery is one of the most promising ones. Kurzfassung Die Mobilität erfährt zurzeit eine disruptive Veränderung dahingehend, dass sich der Trend stark in Richtung des Einsatzes von elektrisch betriebenen Fahrzeugen entwickelt. Mild Hybrid Electric Vehicles, insbesondere mit einem 48V Bordnetz, gewinnen dabei zunehmend die Aufmerksamkeit der Automobilhersteller. Gründe hierfür sind ihre verbesserte Kraftstoffeffizienz sowie ihre Fähigkeit, eine Vielzahl an elektrischen Verbrauchern bedienen zu können. Die Elektrifizierung wird durch die Verwendung von wiederaufladbaren Lithium-Ionen-Batterien ermöglicht. Dennoch besteht weiterhin der Bedarf an leistungsfähigeren und sichereren Technologien, weshalb die all-solidstate Batterie zunehmend an Bedeutung gewinnt. 1. Einleitung und Motivation Die Veränderungen in der Welt der Mobilität stehen bisher erst am Anfang ihrer Entwicklung. Dabei gilt der elektrische Antrieb als wichtiger Faktor und Stellschraube für die Realisierung von nachhaltigen Mobilitätskonzepten. 1 Jeder Automobilhersteller muss eine Entscheidung treffen, ob und wie stark er sich im Bereich der Elektromobilität einbringen will. Es scheint jedoch unumgänglich, Elektromobilität zu berücksichtigen, um am Markt weiterhin wettbewerbsfähig zu bleiben. Des Weiteren ist die Elektromobilität ein Thema, das stark politisch getrieben wird. 316 Anwendungsmöglichkeiten-von-all‐solid‐state-Zellen-in-Niedervolt‐Autobatterien- Außerdem werden durch immer restriktivere gesetzliche Regulationen die Automobilhersteller weltweit vor die Herausforderung gestellt, kontinuierlich sinkende CO 2 -Emissionsgrenzwerte sowie strengere Flottenverbrauchswerte einzuhalten. Eine Übersicht der zukünftigen CO 2 -Emmissionsziele für Europa, USA und China ist in Abbildung 1 dargestellt. Diese werden für Normfahrzeuge in Europa auf 95 g CO 2 / km bis 2020 für neu zugelassene Pkws und auf 147 g CO 2 / km für leichte Nutzfahrzeuge herabgesetzt. 2 Folglich dürfen alle Pkw Normfahrzeuge bis dahin nur noch einen Durchschnittsverbrauch von 4,1 L Benzin bzw. 3,6 L Diesel aufweisen. 3 Bei der Überschreitung der Grenzwerte fällt nun ab 2019 der volle Satz an Strafzahlungen in Höhe von 95 €/ g pro gCO 2 / km an. 4 Ausgehend von dem definierten Emissionsgrenzwert für 2020 sollen die Grenzwerte in Europa bis 2025 um weitere 15% und bis 2030 sogar um bis zu 37,5% reduziert werden. 5 - Abbildung 1. Vergleich der zukünftigen CO 2 -Emmissionsziele für Europa, USA und China. 6 Die oben dargestellten Anforderungen sind mittlerweile mit Verbrennungsmotoren allein nicht mehr zu erfüllen. Die Lösung der Industrie ist daher, stets Elektrofahrzeuge weiter zu entwickeln, um einerseits die Emissionen zu reduzieren und gleichzeitig die Wirtschaftlichkeit sowie die Leistungsfähigkeit der Fahrzeuge zumindest konstant zu halten. 7 Dem aktuellen Entwicklungsstand zufolge gibt es noch starkes Verbesserungspotential in Bezug auf die Reichweite von Elektroautos sowie die zugehörige Infrastruktur, um den Kundenwünschen gerecht zu werden. Als gute Übergangslösung bis hin zum voll elektrischen Fahrzeug (EV oder BEV), bei dem der Verbrennungsmotor vollkommen entfällt, scheinen Hybrid-Fahrzeuge (HEV) eine zufriedenstellende Alternative zu bieten. Hierbei sind die Fahrzeuge bereits teilweise elektrifiziert, wobei der klassische Verbrennungsmotor von mindestens einer elektrischen Maschine auf Hochvolt- (HV) oder Niedervoltbasis (NV) unterstützt wird. Je nach Anordnung und Leistung können sie in weitere Klassen wie Mikro-, Mild- (MHEV), Voll- (HEV), oder Plug-In Hybride (PHEV) unterteilt werden. 8 Letztere werden über ein externes Stromnetz geladen, wohingegen die anderen Klassen nur über den Verbrennungsmotor mithilfe des Generators geladen werden und sich anhand ihres elektrischen Leistungsanteils unterscheiden. Die bisher gängigste Spannungsebene in Fahrzeugen ist die 12 Volt Ebene, die für den Start des Fahrzeuges sorgt und der elektrischen Versorgung der Basisfunktionen dient, wie beispielsweise Scheibenwischer oder Heizung. 9 Jedoch treiben die strenger werdenden Emissionsgrenzen und die stetig steigenden Kundenanforderungen an elektrische Funktionalität und Leistungsfähigkeit der Fahrzeuge das klassische 12V- Bordnetz an seine Leistungsgrenzen. 10 Um den Ansprüchen dennoch gerecht werden zu können, wird als Lösungsansatz das 12V-Bordnetz durch ein zusätzliches 48V- 317 Anwendungsmöglichkeiten-von-all‐solid‐state-Zellen-in-Niedervolt‐Autobatterien- Bordnetz in MHEVs ergänzt. Dieses gilt als attraktiver Mittelweg zwischen den 12V und HV Systemen, da zusätzliche Funktionen, wie z.B. Segeln, Boosten, Rekuperation und elektrisches Rangieren, ermöglicht und dennoch Emissionen und Verbrauch eingespart werden können. 11 Hinzu kommt, dass diese Technologie momentan aufgrund ihrer vielversprechenden Funktionen einen großen Entwicklungsaufschwung erfährt, weshalb auch OEMs, wie Daimler, VW oder BMW, mittlerweile einen großen Fokus auf das 48V Bordnetz legen. 12 Eine zentrale Rolle spielt hierbei die 48V Batterie, welche als Energiequelle zur Versorgung der oben genannten Funktionen dient. Diese speichert die Energie als chemische Energie und wandelt diese bei Bedarf in elektrische Energie für die Verbraucher um. Dabei ist die Lithium(Li)-Ionen Batterie (LIB) marktführend, die aber mittlerweile Schwierigkeiten hat, den hohen Kundenanforderungen standzuhalten. In diesem Zusammenhang wird einerseits stark an der Verbesserung der konventionellen Batterietechnologie gearbeitet, andererseits wird weiterhin an der Entwicklung neuer Technologien geforscht. 13 - 2. Einführung in die zu bewertenden Technologien 2.1. Wiederaufladbare Batterien - - 2.1.1. Li-Ionen Batterien Li-Ionen Batterien gehören zu der Klasse der Sekundärbatterien, wobei die erste wiederaufladbare Li-Ionen Batterie 1991 von Sony kommerzialisiert und auf dem mobilen Konsumermarkt eingesetzt wurde. 14 Angefangen mit kleinen Zellformaten für Konsumeranwendungen wurden die Zellen immer größer sowie leistungsstärker und haben sich inzwischen sogar in der automobilen Anwendung in Hybrid- oder Elektrofahrzeugen etabliert. 15 Der Aufbau und die Funktionsweise der Zelle werden im Nachfolgenden beschrieben. In Abbildung 2 ist der klassische Aufbau und das Funktionsprinzip einer Li-Ionen Zelle beim Entladevorgang dargestellt. Eine Zelle besteht aus zwei unterschiedlichen Elektroden, einem flüssigen, ionenleitfähigen Elektrolyten und einem Separator, der aus einer porösen Membran besteht und als elektronischer Isolator zwischen den Elektroden dient. 318 Anwendungsmöglichkeiten-von-all‐solid‐state-Zellen-in-Niedervolt‐Autobatterien- - - Abbildung 2. Aufbau und Funktionsprinzip der Li-Ionen Zelle beim Entladevorgang. 16 Literatur [1] ZF Friedrichshafen AG, 2017, S. 4. [2] Bundesministerium für Umwelt, Naturschutz und nukleare Sicherheit, 2009; European Commission, 2018. [3] VCD - Verkehrsclub Deutschland e.V., 2018. [4] Bundesministerium für Umwelt, Naturschutz und nukleare Sicherheit, 2009, S. 3. [5] European Commission, 2018; VCD - Verkehrsclub Deutschland e.V., 2018. [6] Eigene Darstellung auf Basis von unternehmensinternen Daten. [7] ZF Friedrichshafen AG, 2017, S. 5. [8] Ernst & Heuermann, 2014, S. 20. [9] Ozawa, 2009 [10] Leuthner, 2013, S. 13 f. [11] Eigene Darstellung in Anlehnung an Xu 2004. - 319 The Authors Dipl.-Ing. (Univ.) Ottmar Sirch BMW Group München Dr.-Ing. Carsten Hoff CLAAS E-Systems KGaA mbH & Co KG Dissen a.T.W Guoqiang Ao, Ph.D. Shanghai Azureve Technology Co.,Ltd. Shanghai, China Dr.-Ing. Ayman Ayad Continental, Powertrain Division Technology & Innovation Regensburg Dr.-Ing. Mohamed Ayeb Universität Kassel Fachgebiet Fahrzeugsysteme und Grundlagen der Elektrotechnik Kassel Dipl.-Ing. Hoang Linh Bach Fraunhofer-Institut für Integrierte Systeme und Bauelementetechnologie IISB und Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen Prof. Dr. Bernard Bäker TU Dresden Dresden Prof. Mark-M. Bakran Universität Bayreuth Bayreuth Martin Baumann, M.Sc. BMW Group München Dr.-Ing. Christoph Friedrich Bayer Fraunhofer-Institut für Integrierte Systeme und Bauelementetechnologie IISB Erlangen Andreas Bendicks, M.Sc. TU Dortmund, Arbeitsgebiet Bordsysteme Dortmund Dominik Bergmann Siemens AG München Dipl.-Ing. Teresa Bertelshofer Universität Bayreuth Bayreuth Dipl.-Ing. Pedro Bossay de Almeida Nogueira Kromberg & Schubert KG Renningen Prof. Dr. rer. nat. Ludwig Brabetz Universität Kassel Fachgebiet Fahrzeugsysteme und Grundlagen der Elektrotechnik Kassel Dr. Martin Brüll Continental, Powertrain Division, Technology & Innovation Regensburg Quirin Buchner, B.Sc. BMW Group München Dexin Chen IHS Markit München 320 The Authors Andrea Colognese, PhD Univ. of Padua ON Semiconductor München Dr. Louis Costa AB Mikroelektronik GmbH Salzburg Dr. Richard Dixon IHS Markit München Tobias Dörlemann, M.Sc. TU Dortmund, Arbeitsgebiet Bordsysteme Dortmund Detlev Endner, Dipl.-Ing. Elektrotechnik Opel Automobile GmbH Rüsselsheim Prof. Dr.-Ing. Stephan Frei TU Dortmund, Arbeitsgebiet Bordsysteme Dortmund Dipl.-Ing. Joachim Fröschl BMW Group München Prof. Dr.-Ing. Jürgen Gebert BMW Group München John Grabowski, BSEE Univ. of Mich. ON Semiconductor Ann Arbor, Mi. USA Dr.-Ing. Andreas Greif Continental, Powertrain Division, Business Unit HEV Nürnberg Dr. Stefan Grösbrink Hella KGaA Hueck & Co. Lippstadt Lilly Han, Bachelor of Engineering, MBA Shanghai Azureve Technology Co., Ltd. Shanghai, China Uwe Hauck TE Connectivity Germany GmbH Berlin Dipl.-Ing. Norbert Hees Leopold Kostal GmbH & Co. KG Lüdenscheid Dr. rer. nat. Jan Helfrich Kromberg & Schubert KG Renningen Prof. Dr.-Ing. Markus Henke Technische Universität Braunschweig Institut für Elektrische Maschinen, Antriebe und Bahnen Braunschweig Prof. Dr.-Ing. Hans-Georg Herzog Technische Universität München Fachgebiet für Energiewandlungstechnik München Stefan Hörtling HTWG Konstanz Konstanz Michael Kahlstatt, Dipl.-Ing. Elektrotechnik Opel Automobile GmbH Rüsselsheim Sebastian Kahnt Intedis GmbH & Co. KG Würzburg Dr. Kay Klobedanz Hella KGaA Hueck & Co. Lippstadt Dr. rer. nat. Thomas Lang Robert Bosch GmbH Reutlingen 321 The Authors Dipl.-Ing. Niklas Langmaack Technische Universität Braunschweig Institut für Elektrische Maschinen, Antriebe und Bahnen Braunschweig Janis Lehmann, M.Sc. Universität Kassel Fachgebiet Fahrzeugsysteme und Grundlagen der Elektrotechnik Kassel Dr. Michael Leidner TE Connectivity Germany GmbH Speyer Benjamin Löwer, M.Sc. Universität Kassel Fachgebiet Fahrzeugsysteme und Grundlagen der Elektrotechnik Kassel Dr. Michael Ludwig TE Connectivity Germany GmbH Speyer Roland Matthe, Diplom Ingenieur Elektrotechnik Opel Automobile GmbH Rüsselsheim Ines Miller, Masterandin P3 automotive GMbH München Dipl.-Ing. Jonas Müller Fraunhofer-Institut für Integrierte Systeme und Bauelementetechnologie IISB und Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen Jozsef Gabor Pazmany, M.Sc. Dr. Ing. h.c. F. Porsche AG Weissach Dipl.-Inf. Uwe Prüfer SmartCable GmbH Erlangen Dr. Klaus Rechberger Dr. Ing. h.c. F. Porsche AG Weissach Wirt.-Ing. Sebastian Rickert, M.Sc., P3 group Stuttgart Dipl.-Ing. Sebastian Rogge Continental, Powertrain Division, Business Unit HEV Nürnberg Franz Rohlfs, M.Sc. Fraunhofer IZM Berlin Fabian Schipperges Porsche AG Weissach Dipl.-Ing. Andreas Schletz Fraunhofer-Institut für Integrierte Systeme und Bauelementetechnologie IISB und Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen Dr. Helge Schmidt TE Connectivity Germany GmbH Bensheim Stefan Schumi Infineon AG Neubiberg Stefan Schwimmbeck M.Sc. BMW Group München Dr. Varvara Sharova P3 automotive GmbH München Samuel Siegel, M.Sc. Dr. Ing. h.c. F. Porsche AG Weissach 322 The Authors Andreas März, M.Sc. Siemens Mobility GmbH Nürnberg Robert Stanek P3 automotive Stuttgart Dipl.-Ing. Christian Sültrop Fraunhofer IISB Erlangen Dr.-Ing. Günter Tareilus Technische Universität Braunschweig Institut für Elektrische Maschinen, Antriebe und Bahnen Braunschweig Julian Taube, M.Sc. Technische Universität München Fachgebiet Energiewandlungstechnik München Laurenz Tippe, M.Sc. Technische Universität München Fachgebiet Energiewandlungstechnik München Dipl.-Ing. Matthias Töns Continental, Powertrain Division, Business Unit HEV Regensburg Dr.-Ing. Helga Weber SmartCable GmbH Erlangen Dr.-Ing. Christoph Weissinger BMW Group München Dipl.-Ing. Marc Wiegand Leopold Kostal GmbH & Co. KG Lüdenscheid Marco Wolf TE Connectivity Germany GmbH Speyer Dipl.-Ing. Zechun Yu Fraunhofer-Institut für Integrierte Systeme und Bauelementetechnologie IISB und Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen 323