A full vehicle simulation of an HEV starter-generator concept with the SmartElectricDrives library

Transcription

1 A full vehicle simulation of an HEV starter-generator concept with the SmartElectricDrives library Johannes V Gragger, Dragan Simic, Christian Kral, Franz Pirker Arsenal Research, Austria Copyright c 2007 SAE International ABSTRACT The presented work proposes a simulation environment using Modelica programming language for full vehicle simulations. A special software library suitable for Modelica simulations is used - the SmartElectricDrives (SED) library. This software is developed to simulate all parts of the electrical drive systems, and furthermore, allows to choose different levels of abstraction in the simulation design. Typical applications for the SED library are simulations of vehicular systems such as full vehicle simulations of hybrid electric vehicles (HEVs) and electric vehicles (EVs), simulations of electrified auxiliary drives, etc. In this work a full vehicle simulation of an HEV concept with a starter-generator is presented and all the parameters of the electrical components are discussed. INTRODUCTION When designing HEVs or EVs, developers have to face a number of challenges. A lot of different drive concepts should be investigated in order to find the most suitable concept for the target application or the target market. These drive concepts include mechanical parts as well as electrical parts. For finding the best drive concept it is very useful to simulate the mechanical components and the electrical components of the different drive concepts in full vehicle simulations so that basic conclusions about the energy consumption and the efficiency can be obtained. Another advantage of full vehicle simulations is that with a small effort the elementary components of the drive can be matched based on their specifications. Modelica language facilitates the simulation of such electromechanical systems. This is why Modelica is used for the development of the proposed SED library[1],[2],[3]. The simulation example presented in this work needs less than 36 seconds CPU-time for calculating the energy consumption of a hybrid vehicle driving a cycle of around 20 minutes. These measurements are taken from a standard personal computer. Due to this very short calculation times simulations such as the one presented have the potential to accelerate the task of matching the drive components in the early concept phase of vehicle development. This work focusses on the description of the used simulation models. In the presented simulation example all electrical parts of the HEV concept are simulated using models from the SED library. The mechanical parts of the simulation are developed with the SmartPowerTrains (SPT) library. Based on this simulation example the meaning of the relevant characteristic electrotechnical quantities as well as the assumptions for the right choice of parameters are discussed. A detailed explanation is given on how to match the parameters of the asynchronous induction machine used as starter-generator with the parameters of the internal combustion engine (ICE) in a well defined drive cycle (e.g. the New European Drive Cycle, NEDC). Also the control strategy of the starter-generator (field oriented control) is explained and the general operation of the starter-generator in boost mode and in recuperation mode is outlined. For the energy buffer of the electromechanical system a battery model (e.g. nickel metal hydride) considering charging/discharging losses, is presented. The parametrization of this battery package corresponding to the starter-generator requirements is also explained. For system security and operational limits some recommendations for current limits and voltage limits are presented. Representative results of the full vehicle simulation, such as the state of charge (SOC) of the energy buffer, the starter-generator power and the fuel consumption of the vehicle validate the proposed simulation environment. THE SED LIBRARY The SED library contains models for all components that are used in a state-of-the-art electric drive system. An example of a typical asynchronous induction machine drive system is shown in Figure 1. The machine is mechanically connected to the load. A battery supplies the drive with electric energy. The converter transforms the dc voltage of the battery to suitable voltage wave forms for the three phase asynchronous induction machine. In the control unit (field oriented control with a combined speed controller) the control signals for the converter get generated based on the reference speed signal and the feedback values obtained from the measurement device. For a proper operation of the drive the electric machine parameters must be chosen with regard to the load specifications. Furthermore the voltage level of the electric source must correspond to the nominal voltage of the machine.

2 equation of a basic ohmic inductive load circuit is: v = R i + L di dt = R i + dλ dt If transient effects are taken into account the current follows the voltage with a lag determined by the time constant τ = L R. Controlling the current in this load would require a feedback controller. In quasi stationary equations only the steady state solutions of the electric system are considered. For a sinusoidal current with the angular frequency Ω, (1) i = 2I rms cos(ωt) (2) Figure 1: An asynchronous induction machine drive system where I rms is the rms values, the time derivative of the current is di dt = 2ΩI rms sin(ωt). (3) Hence, for the voltage equation we derive v = R 2I rms cos(ωt) ΩL 2I rms sin(ωt). (4) The voltage is obviously a sinusoidal quantity with a well determined sine and cosine component. Due to the assumption of sinusoidal currents, the differential equation (1) can be transformed into two algebraic equations obtained from (4), one with respect to the cosine, and the other with respect to the sine coefficient. Such a set of equations is called a quasi stationary set of equations which can be derived for any linear electric circuit with sinusoidal current and voltage waveforms. Figure 2: Screenshot of the SED library in the package browser STRUCTURE OF THE SED LIBRARY In Figure 2 the top level package structure of the SED library is presented. In the Examples package a number of simulations that describe the functionality of the different components of the SED library are included. The QuasiStationaryDrives package The QuasiStationaryDrives package contains very fast ready to use models. These ready to use models represent full torque controlled drives with an included dc supplied converter. These QuasiStationaryDrives models are designed such way that electric transient effects in the machine and switching effects in the converter are neglected. However, mechanical transients caused by the moment of inertia are taken into account. Therefore, QuasiStationaryDrives models cannot be used for investigations of current spikes that are caused by inductances in the electric circuit, for instance. The voltage One of the major applications of QuasiStationaryDrives models are simulations that focus on the energy consumption or the efficiency of electric drives. In such simulations the electric transient effects and switching effects can be neglected for the benefit of decreasing the computation time of the simulation process. The big plus of the QuasiStationaryDrives models is that the negligence of transient effects simplifies control structures for torque control, flux control and current control. These simplified control structures work without feedback signals and standard controller blocks (PI controllers), because only an algebraic system of equations has to be considered. This fact implies two specific advantages. On one hand the user does not have to find applicable controller settings for the different simulation setups: This saves development time. On the other hand the number of algebraic and differential equations in the simulation is decreased, and consequently, the computation time of the simulation is severely reduced [4]. The TransientDrives package In the TransientDrives package all the elementary controller components needed to build a full torque control system for dc and ac machines are included. Furthermore, similar ready to use models such as the QuasiStationaryDrives models are available in the Transient- Drives package. Unlike the QuasiStationaryDrives models, the TransientDrives models consider electric transient effects. In Figure 3 the internal setup of a ready to use model taken from the TransientDrives package is shown. This model represents a torque controlled asynchronous machine with included converter. All the

3 elementary controller components that are used in this model are also separately retrievable in the SED library. The reference voltages for the converter model are generated by the field oriented control (FOC) block. FOC is a commonly used control strategy for ac machines. With FOC flux and current can be controlled independently. The space phasor equation for electric torque production in ac machines can be written as τ Electrical = 3p 2 (i s λ r ) (5) where τ Electrical is the inner torque of the machine, p is the number of pole pairs, i s is the stator current space phasor and λ r is the rotor flux linkage space phasor. If (5) gets transformed to the rotor flux fixed coordinate system in which the y-component of the rotor flux equals zero, λ ry = 0, we obtain Figure 3: Ready to use model of a torque controlled induction machine drive with integrated converter τ Electrical = 3p 2 (i sy λ rx ). (6) This equation shows that the electric torque of the asynchronous induction machine follows to changes of i sy instantly as long as the flux in the machine is kept on a constant level. The transient system equation for the rotor flux linkage is i sx = 1 ( L r dλ rx + λ rx ), (7) L m R r dt where R r is the rotor resistance, L r is the sum of the main field inductance, L m, and the rotor stray inductance, L rσ. According to (7) the rotor flux linkage λ rx follows i sx with a first order delay. Using a PI controller λ rx can be controlled through i sx, commonly called the flux producing current component. All ready to use ac drive models of the SED library have flux weakening implemented so that it is also possible to simulate a machine operating above the nominal speed. A very effective flux weakening method is to control the desired flux level such way that the stator voltage cannot exceed the defined voltage limit of the ac machine [5]. In asynchronous induction machines the flux producing current component, i sx, gets decreased continuously if the voltage limit is reached. In permanent magnet synchronous induction machines a common practice is to apply a flux space phasor via the flux producing current component which opposes the flux of the permanent magnet of the machine [6]. The extra current demand causes a lot of additional losses in the machine. In some automotive applications the permanent magnet synchronous machines are designed such way that the permanent magnet can not provide the full nominal flux. Therefore the permanent magnet field must be supported by a magnetic field produced by the stator current in operation points below nominal speed. The advantage of such a machine design is that flux weakening is more effective compared to conventional strategies [7]. Figure 4: Screenshot of the Converters package in the package browser The Converters package The converters package shown in Figure 4 contains models at two different levels of abstraction. In the PowerBalance package the fast, non switching converter models are modeled. By applying p Supply = p Load (8) these converter models assure power balance between the supply port and the load port of the converter. The power balance models work as ideal power converters. The currents in the load circuit i Load and in the supply circuit i Supply get calculated based on the supply voltage v Supply and the reference voltage of the converter v Ref according to v Supply i Supply = v Ref i Load. (9) More detailed models can be found in the IdealSwitching package. In this package dc/ac converters, ac/dc converters and dc/dc converters are modeled with ideal switches, ideal diodes and ideal thyristors. The computation time of these models depends on the switching frequency of the simulated system, however. The Sources package With the components in the Sources package batteries, supercaps and proton exchange membrane (PEM) fuel cells can be simulated. The battery model used in the example and the supercap model are featured with SOC

4 signal outputs in order to apply control algorithms in the drive system which take the actual battery charge into account and switch between charging and discharging mode based on the available management strategy. Losses are considered as well. In the battery model the charging/discharging efficiency can be parametrized. The total charging/discharging energy efficiency of the battery considers the sum of thermal losses which can be modeled with a serial resistance and general charging discharging losses that are viewed as continuous charge leakage during the charging/discharging process. The maximum charging/discharging energy efficiency is η Q,Max = V Min + V Max 2 R s I Q,Ref V Min + V Max + 2 R s I Q,Ref (10) where V Min is the minimum cell voltage, V Max is the maximum cell voltage, R s is the serial resistance and I Q,Ref is the constant charging and discharging current of the battery. THE SPT LIBRARY Since the SED library contains only electrical components it is necessary to model the mechanical loads connected to the electric machine shafts with components either taken from the Modelica Standard library or from a specialized mechanics library written in Modelica. In the proposed simulation the mechanical components are taken from the SPT library. The SPT library is specially designed for longitudinal simulations of mechanical power trains in vehicular applications. The components in the SPT library are modeled in Modelica language by using Modelica Standard library components. Apart from modeling all the elementary mechanical components in the drive train it is also necessary to implement a virtual driver model. This driver model controls the clutch state, the gear ratio, the brake pedal and the gas pedal. The most important vehicle components included in the SPT library are: Internal combustion engines Clutches Belt drives Mechanical gearboxes Automatic gear boxes Cardan shafts Wheel shafts Chain drives Clutches Belt drives Car bodies Brakes Figure 5: Screenshot of the HEV concept with a startergenerator Axles Environmental effects Driver SIMULATION EXAMPLE OF AN HEV CONCEPT WITH A STARTER-GENERATOR Figure 5 shows the simulation setup of an all wheel drive HEV concept with an asynchronous induction machine starter-generator. The battery supplied startergenerator is permanently connected to the shaft of the internal combustion engine (ICE). With a clutch the manual transmission gearbox can be connected to the shaft of the starter-generator and the ICE. The transmission gearbox is connected to the differential which distributes the torque to the front axle and to the rear axle via cardan shafts. The control bus contains all necessary reference signals and feedback signals that are exchanged by the vehicle components. In the component cycle a new European drive cycle (NEDC) is implemented. Based on this drive cycle signal the component driver operates the overall vehicle model so that the car follows the defined speed characteristic. In the driver component there are parameters which define the points of gear shifts, the response characteristic due to sudden speed changes, the characteristics of the coupling process, etc. In the component Ambient the environmental conditions such as air pressure and temperature are defined. The actual HEV control strategy is implemented in the component Strategy. THE HEV CONTROL STRATEGY The hybrid drive is controlled such way that the startergenerator operation and the ICE operation are dependent on the state of charge (SOC) of the battery and the ICE characteristic. If the driver accelerates the vehicle such way that the ICE would reach an inefficient

5 operation point, then the starter-generator works in motor mode and adds drive torque while the ICE produces a reduced drive torque that corresponds to the most efficient operation point at the instant shaft speed. In cases in which the driver commands a very low drive torque the starter-generator works in generator mode while the ICE produces an increased driving torque that opposes the generating torque of the starter-generator and the load. Beside this operation management the HEV control always assures that the SOC of the battery stays within the defined limits. If the battery reaches the lower SOC limit the starter-generator can not switch to motoring mode and if the battery reaches the upper SOC limit the starter-generator can not switch to generating mode. In this HEV concept energy recuperation is only possible if the clutch is closed. THE PARAMETRIZATION OF THE SIMULATION The starter-generator and the battery of the drive system are parametrized with regard to the specifications of the mechanical components of the vehicle, the control strategy and the defined NEDC cycle. The ICE is a 1.9 liter Diesel Engine with 66 kw and a maximum speed of 4500 rpm. The starter-generator The starter-generator parameters are chosen such way that the electric machine can support the ICE with up to 23 Nm at a speed 4500 rpm. However, in this operation point the machine works in the flux weakening region. In braking mode the machine is capable of electrically braking the vehicle with a power of 11.2 kw and recuperating the saved energy. The machine parameters in this simulation are chosen as follows: Number of pole pairs: p = 4 Nominal stator phase voltage: V N = 115 V Nominal frequency: f N = 200 Hz Nominal stator phase current: I N = 40 A Moment of inertia: J r = 0.15 kgm 2 Stator resistance: R s = Ω Stator stray inductance: L sσ = 2.33e 4 H Main field inductance: L m = H Rotor stray inductance: L rσ = 2.33e 4 H Rotor resistance: R r = Ω The stator is star connected. The following values are calculated from the machine parameters and determine the nominal operation point of the machine: Nominal power: P N = 11.2 kw Nominal torque: τ N = 37.1 Nm The number of pole pairs, p, and the nominal values I N, V N and f N are typical name plate values. Therefore they can be chosen according to available product lists. However, many times the parameters of the single phase equivalent circuit of asynchronous induction machines, such as R s, L sσ, L rσ, L m and R r are not easily available. If these values can not be found in the machine data sheets they have to be measured to reach adequate simulation results. In simulations where the nominal power and speed of electric machines should be varied in order to find the best machine specifications for the electromechanical system it is useful to derive per unit values from a specific machine of which all the single phase equivalent circuit parameters and the name plate values are known. From these per unit values scaled single phase equivalent circuit parameters can be calculated [8]. Usually the base values of the per unit system are I B = I N, V B = V N and f B = f N. The base impedance is Z B = V B I B. (11) By using (11) the per unit resistances are calculated according to r (pu) = R Z B (12) and per unit reactances can be calculated by x (pu) = 2 π f B L Z B. (13) In the SED library a parameter estimation function is implemented that can be used to estimate the parameters of the single phase equivalent circuit from the name plate values of the machine. The battery The battery parameters have to match the specifications of the starter-generator and the HEV control strategy of the vehicle. According to the HEV control strategy the starter-generator must be capable of producing an electric torque of 25Nm at a speed of 4500rpm. However, the nominal speed of the starter-generator is only 2875rpm and therefore the starter-generator works in the flux weakening region if the ICE runs at 4500rpm. In order to assure that the starter-generator produces the required torque at maximum speed the voltage level in the dc-link must not fall below the nominal peak voltage value of the starter-generator. The battery package is parametrized according to V Package,Min V N 2 3 (14) whereas the factor 2 is used to transform the rms voltage value to the corresponding peak value and the factor 3 is used to transform V N from a phase quantity to

6 Serial cell resistance: R s,cell = Ω Overall charging/discharging efficiency: η Q = 0.85 Number of serial connected cells: n s = 228 Number of serial connected cells: n p = 1 The battery parameters are assumptions and correspond to the behavior of a conventional nickel metal hydride accumulator for automotive applications [10]. Figure 6: Linear dependency of SOC and battery voltage a line to line quantity because the stator of the startergenerator is star connected [9]. The minimum voltage of the battery can be calculated by V Package,Min = V Package,Full SOC Min (15) whereas V Package,Full is the voltage of the fully charged battery. In real battery supplied applications the battery does not get charged until SOC = 1. Therefore the used battery model allows the definition of a maximum state of charge SOC Max. By defining the nominal admissible electric charge, the initial SOC, the SOC limits and the voltage limits it is possible to determine the operation of the battery based on the relation described in Figure 6. The battery losses are considered by the serial cell resistances, R s,cell, and the charge/discharge efficiency parameter, η Q. The thermal losses of the battery, P ϑ, are determined by the overall serial resistance of the battery package, R s,package = n s n p R s,cell (16) where n s is the number of serial connected cells and n p is the number of parallel connected cells. Furthermore P ϑ is dependent on the average charge/discharge current I package = n p I Cell. (17) In real applications the serial resistance must be limited to R s,package P ϑ,max (I package ) 2. (18) In the simulation example the battery is modeled with the following parameters: Nominal charge per cell: Q Cell,N = 23400C Minimum voltage per cell: V Cell,Min = 1.24V Maximum voltage per cell: V Cell,Max = 1.32V Minimum state of charge: SOC Min = 0.5 Maximum state of charge: SOC Max = 0.8 Initial state of charge: SOC init = 0.56 The security limits In real applications the electric components of the drive must be protected by specific limitations in order to prevent damage during the operation. These limitations are also implemented in the presented simulation. In the field oriented control algorithm a current limitation is implemented. This current limitation assures that the machine is not overloaded due to thermal losses. If the current limit is reached the machine torque does not follow the desired torque generated by the HEV control strategy anymore. For short term operation the current limit of the machine can be twice the nominal current, I Machine,Max = 2 I N. In the converter model the parameter, I Converter,Max, for monitoring the converter currents is implemented. If this limit is reached the simulation gets aborted and a defined error message is generated. In real applications a fuse protecting the semiconductors would switch off the drive. The value of I Converter,Max may vary a lot among different types of semiconductors. It is useful to use the surge non repetitive forward current, I FS,Max, from the data sheet of the chosen device whereas the maximum RMS forward current, I F,Max, must be considered when analyzing the computed simulation results. In the proposed simulation the converter is assumed to have a current limit of I Converter,Max = 600 A. THE RESULTS OF THE SIMULATION In order to asses the proposed HEV concept the modeled vehicle has to drive the defined NEDC cycle. Figure 7 shows that the vehicle follows the desired speed signal without significant deviations. In Figure 8 the power of the starter-generator, p SG, and the power of the diesel engine, p ICE, are displayed. From time t = 0 s until time t = 800 s the NEDC cycle represents a vehicle operation in a stop and go traffic. In this time the ICE and the starter-generator operate below their rated operation points. From t = 800 s onwards the vehicle drives at speeds that correspond to sub urban traffic characteristics. At time t = 1120 s the vehicle starts to brake at a speed of v vehicle = 33.5 m s and then stops. This is when the starter-generator operates close to the nominal operation point and therefore recuperates the most energy. In Figure 9 the state of charge of the battery is shown. From time t = 900 s the starter-generator boosts additional power to the drive train. Therefore the battery gets discharged. From time t = 1120 s onwards the starter-generator feeds back energy to the battery. This is the periode of the NEDC in which the starter-generator

7 Figure 7: Reference NEDC speed characteristic and real vehicle speed Figure 9: State of charge of the nickel metal hydride battery Figure 8: Power of the starter-generator and power of the ICE saves most of the system energy. In Figure 10 the fuel consumption of the proposed HEV concept is compared with the fuel consumption of a traditional vehicle concept with similiar specifications. In a defined NEDC cycle the implemented starter-generator helps to reduce the fuel consumption of the vehicle by about 10%. In the proposed simulation iron losses and saturation effects of the machine and thermal losses of the converter are neglected. The ongoing research focusses on the development of an ExtendedMachines library. With this ExtendedMachines library iron losses, saturation effects and machine faults causing asymmetries will be simulated. Further objectives are the development of detailed battery models and supercap models considering special electrochemical effects. CONCLUSION A full vehicle simulation built with the SED library and the SPT library is presented. The proposed vehicle concept includes an asynchronous induction machine starter-generator which works in boost mode and recuperation mode. Furthermore, the load point of the ICE gets shifted so that maximum system efficiency is assured. The results of the simulation show that, compared to a conventional vehicle, the proposed starter-generator Figure 10: Comparison of the fuel consumption of the proposed starter-generator concept and the same concept without starter-generator concept saves about 10% of the fuel consumption in an NEDC cycle. The starter-generator is modelled with a quasi stationary ready to use model taken from the SED library. This ready to use model includes physical machine equations, an optimized FOC control system and a power balance converter. Due to the application of quasi stationary equations and the power balance relation the electromechanical efficiency analysis is computed with very small processing effort. Full vehicle simulations such as the one presented have the potential to accelerate EV and HEV development cycles considerably in the early concept phase. REFERENCES [1] H. Giuliani, C. Kral, J. Gragger, and F. Pirker, Modelica simulation of electric drives for vehicular applications the smart drives library, ASIM, [2] D. Simic, H. Giuliani, C. Kral, and J. Gragger, Simulation of hybrid electric vehicles, 5th International Modelica Conference, [3] J. V. Gragger, D. Simic, C. Kral, H. Giuliani, V. Conte, and F. Pirker, A simulation tool for electric

8 auxiliary drives in hevs the SmartElectricDrives library, FISITA World Automotive Congress 2006, Yokohama, Japan, [4] J. Gragger, H. Giuliani, C. Kral, T. Bäuml, H. Kapeller, and F. Pirker, The SmartElectricDrives Library powerful models for fast simulations of electric drives, 5th International Modelica Conference 2006, [5] S.-H. Kim and S.-K. Sul, Voltage control strategy for maximum torque operation of an induction machine in the field-weakening region, IEEE TRANSACTIONS ON INDUSTRIAL ELECTRON- ICS, vol. 44, pp , [6] D. Schroeder, Elektrische Antriebe Regelung von Antriebssystemen. Berlin-Heidelberg: Springer, 2 ed., [7] P. J. Otaduy and J. W. McKeever, Modeling reluctance-assisted PM motors, Tech. Rep. ORNL/TM-2005/185, Oak Ridge National Laboratory, Tennessee, USA, January [8] H. Kleinrath, Stromrichtergespeiste Drehfeldmaschinen. Wien: Springer Verlag, [9] Mohan and Robbins, Power Electronics. New York: J. Wiley Verlag, 2 ed., [10] D. Linden and T. Reddy, Handbook of Batteries. New York: McGraw-Hill Handbooks, Dragan Simic was born in He received the Dipl.-Ing. degree in mechanical engineering from Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture University of Split, Split, Croatia, in Since 2002, he has been with Arsenal Research (Österreichisches Forschungs- und Prüfzentrum Arsenal Ges.m.b.H.), Vienna, Austria. His research activities are focused on the longitudinal dynamics simulation of the conventionaland hybrid vehicles including the simulation of the auxiliaries. Christian Kral AUTHORS Johannes V Gragger phone: +43 (0) received the Dipl.-Ing. and Ph.D. degrees from the Vienna University of Technology, Vienna, Austria, in 1997 and 1999, respectively. From 1997 to 2000, he was a Scientific Assistant at the Institute of Electrical Drives and Machines, Vienna University of Technology. Since 2001, he has been with Arsenal Research (Österreichisches Forschungs- und Prüfzentrum Arsenal Ges.m.b.H.), Vienna, Austria. From January 2002 until April 2003 he was on sabbatical as a Visiting Professor at the Georgia Institute of Technology, Atlanta. His research activities are focused on diagnostics and monitoring techniques, machine models, and the simulation of faulty machine behavior. fax: +43 (0) johannes.gragger@arsenal.ac.at received the Dipl. Ing. (FH) degree in electronic engineering from the University of Applied Sciences, Technikum Wien, Vienna, Austria, in In 2002 and 2003 he worked on research projects at the Engineering College of Copenhagen (IHK), Denmark, and at the University of Zilina, Slovakia. In 2003 and 2004 he worked in IT industry in Metro Manila, Philippines. Since December 2004, he has been a Research Associate at Arsenal Research (Österreichisches Forschungs und Prüfzentrum Arsenal Ges.m.b.H.), Vienna, Austria. His major research interests are drive control, machine monitoring and machine design.

9 Franz Pirker was born in He received the Dipl.-Ing. degree in electrical engineering from Vienna University of Technology, Vienna, Austria, in Since 1999, he has been the Head of Monitoring, Energy, and Drive Technologies at Arsenal Research (Österreichisches Forschungs- und Prüfzentrum Arsenal Ges.m.b.H.), Vienna, Austria. In this area, the main research topics are hybrid electric vehicles (HEV) especially development and design of electric drives. In these fields, Arsenal Research is developing new concepts for HEV s and electric driven auxiliaries like electric water pumps for internal combustion engines.