Jedinger Alexander 2015

Alexander Jedinger, BSc
Development of a real-time
simulation model of hybrid powertrains
in motorsports
MASTER’S THESIS
to achieve the university degree of
Diplom-Ingenieur
Individual Master’s Degree Program

Electrical and Automotive Engineering
submitted to
Graz University of Technology
Supervisors:
Dipl.-Ing. Martin Ackerl

Dipl.-Ing. Dr.techn. Jürgen Fabian
Institute of Automotive Engineering
Dipl.-Ing. Christoph Leichtfried

AVL List GmbH
Graz, June 2015

AFFIDAVIT
I declare that I have authored this thesis independently, that I have not used other than the
declared sources/resources, and that I have explicitly indicated all material which has
been quoted either literally or by content from the sources used. The text document up-
loaded to TUGRAZonline is identical to the present master‘s thesis dissertation.
Date Signature
II
ACKNOWLEDGMENT
I am using this opportunity to express my gratitude to everyone who supported me
throughout the course of this thesis. I am thankful for their aspiring guidance, invaluable
and constructive criticism and friendly advice during the project work.
I would like to thank AVL List GmbH, for giving me the opportunity and the resources
to work on this project. A special thanks to my supervisor, Mr. Christoph Leichtfried,
who was always there to support me.
To my supervisors at the university, Mr. Jürgen Fabian and Mr. Martin Ackerl, I would
like to express my gratitude for their guidance. Through critical thinking and constructive
criticism, they helped to author a sophisticated and well-structured thesis. Thank you very
much for the time and effort.
Many thanks to the Electric Drives and Machines Institute, especially Mr. Klaus Krischan
and Mr. Heinrich Eickhoff, for their support and expertise. I am grateful for their excellent
help on difficult technical issues, often provided on very short notice.
Finally, I would like to express my sincere gratitude to my family. My parents, for giving
me the freedom and opportunity to pursue this degree, and the continuous support
throughout my education. And my siblings, for the emotional support and trust. In addi-
tion, I want to thank my girlfriend, Ariel, for the countless hours of listening and proof-
reading.
III
KURZFASSUNG
Die Verwendung von Elektro- und Hybridantrieben im Motorsport nimmt stetig zu und
stellt die Fahrzeugentwicklung und -simulation vor neue Herausforderungen. Die Ent-
wicklung von Betriebsstrategien für diese neuen Komponenten erfordert detaillierte Mo-
delle für alle Teile des Antriebsstranges, um maximale Leistung bei minimalem Kraft-
stoffverbrauch zu erreichen.
Ziel der Arbeit war die Entwicklung eines MATLAB® Simulink® Modells für Elektro-
motoren und Frequenzumrichter in Hybridantrieben für Anwendungen im Motorsport. In
Zusammenarbeit mit der Rennsportabteilung der AVL List GmbH wurde ein Modell ent-
wickelt, das anhand von Messungen am Prüfstand und allgemeinen Motordaten parame-
triert werden kann. Dieser Ansatz ermöglicht die Abbildung aller gängiger Radialfluss-
maschinen, ohne genaue Kenntnis über deren innere Strukturen. Neben dem elektrome-
chanischen Verhalten wurden auch die thermischen Eigenschaften der Komponenten ab-
gebildet. Das Modell ist echtzeitfähig, und kann für verschiedene Antriebsstrangkonfigu-
rationen verwendet werden. Das entwickelte Modell wurde mit einem realen Antriebs-
system verifiziert, allerdings konnten aufgrund fehlender Messungen nicht alle Eigen-
schaften überprüft werden.
Schließlich wurde das Modell in das bestehende Gesamtfahrzeugsimulationsmodell von

AVL (Vehicle Simulation Model – Powertrain Model VSM-PTM) integriert. Dies wird
sowohl als reine Software-Simulation als auch als Teil von Hardware-in-the-Loop Prüf-
ständen eingesetzt. Die Simulationsergebnisse können als plausibel angesehen werden,
müssen aber erst mit Messungen eines realen Antriebsstranges verifiziert werden.
IV
ABSTRACT
The use of hybrid and electric propulsion systems in motorsports is rising continuously,
posing new challenges for vehicle models and simulations. This requires detailed models
of all parts of the powertrain. The development of operating strategies to incorporate these
new hybrid and electric components is necessary in order to achieve maximum perfor-
mance at the lowest fuel consumption.
The main objective of this project, done in close collaboration with the racing department
of AVL List GmbH, was to create a MATLAB® Simulink® model of electric motors and
inverters in hybrid electric powertrains used in motorsports. A model was developed that
can be parameterized with a defined set of measurements from a test bench and basic
specifications of an electric drive. This approach allows for its use for all major types of
radial flux machines without requiring detailed knowledge of the drive’s inner structure.
Both the electromechanical and thermal behavior of the motor and inverter were de-
scribed. The model is capable of operation in real time and can be adjusted to different
powertrain layouts. It was verified with an existing electric drive, however, some features
could not be tested due to a lack of measurements.
The model was integrated into AVL’s existing simulation software, Vehicle Simulation
Model - Powertrain Model (VSM-PTM). In VSM-PTM, the model operates as a
standalone software model and on real time platforms as part of Hardware-in-the-Loop
tests. The simulation results seem plausible and are awaiting verification with measure-
ments from a real powertrain.
V
ACRONYMS AND ABBREVIATIONS
ACO Automobile Club de l’Ouest
AVL Anstalt für Verbrennungskraftmaschinen List GmbH
DC Direct Current
ERS Energy Recovery System
FEA Finite Element Analysis
FIA Fédération Internationale de l’Automobile
HEV Hybrid Electric Vehicle
HiL Hardware in the Loop
HV Hybrid Vehicle
ICE Internal Combustion Engine
IM Induction Motor
IPM Interior Permanent Magnet Motor
KERS Kinetic Energy Recovery System
MGU-H Motor Generator Unit Heat
MGU-K Motor Generator Unit Kinetic
RMS Root Mean Square
SPM Surface Mounted Permanent Magnet Motor
VSM-PTM Vehicle Simulation Model – Powertrain Model
VI
TABLE OF SIGNS AND SYMBOLS
Symbol Unit Description
a m²/s Thermal diffusivity
Acu,th m² Contact area between winding and stator teeth
Acu,yk m² Contact area between winding and stator yoke
Aslot m² Contact area between winding and stator iron
B T Magnetic flux density
cd - Aerodynamic drag coefficient
cosφ - Power factor
Cp J/(kg⋅K) Specific heat capacity
Cpn J/(kg⋅K) Specific heat capacity
crr - Per wheel rolling resistance coefficient
dc m Width of cooling jacket
deq m Equivalent thickness of air and insulation material in the
stator slot
f Hz Frequency
Gcu,ir W/K Thermal admittance between copper and stator iron
Gth W/K Thermal admittance
H A/m Magnetic field strength
hairgap W/(m²⋅K) Heat transfer coefficient of the airgap
hch W/(m²⋅K) Heat transfer coefficient coolant to housing
hyc W/(m²⋅K) Heat transfer coefficient yoke to coolant
Î A Peak value of current
ia, ib, ic A Phase currents for phase a, b, c
id A Current in direct axis
IDC A Current from dc supply
Idc,cell A DC current for each battery cell
Iph A Phase current (RMS)
iq A Current in quadrature axis
IS A Stator phase current (RMS)
kair W/(m⋅K) Thermal conduction coefficient of air
kcu,ir W/(m⋅K) Equivalent thermal conduction coefficient of air and insu-
lation material in the stator slot
kir W/(m⋅K) Thermal conduction coefficient of the stator core
lC m Axial length of cooling jacket
Ld H Inductance in direct axis
Lq H Inductance in quadrature axis
VII
Table of Signs and Symbols

ls m Length (height) of stator
lsb m Stator slot perimeter
lshf m Length of shaft
m kg Thermal mass
m kg Vehicle mass
mn kg Thermal mass of a specific thermal node
NNu - Nusselt number
Nph - Number of phases
NPr - Prandtl number
NTa - Taylor number
p - Number of pole pairs
Pcu W Motor copper losses
PDC W Heat flux from housing to environment
Pel W Electrical power
PFe W Motor iron losses
Pfr W Motor friction losses
pir - Ratio of tooth iron volume to total tooth plus slot volume
PL W Power dissipation
PL,cell W Power dissipation of each battery cell
PL,inverter W Overall inverter losses
PL,motor W Overall motor losses
PL,other W Motor iron, friction and ventilation losses
PLn W Power dissipation in a specific thermal node
Pmech W Mechanical power
PPU kW Overall power unit power
Pv W Motor ventilation losses
q̇ W Rate of heat transfer
q̇coolant2housing W Rate of heat transfer between coolant and housing
q̇n W Rate of heat transfer to/from a specific thermal node
q̇yoke2coolant W Rate of heat transfer between yoke and coolant
R(T) Ω Temperature dependent resistance
Rcu,ir K/W Thermal Resistance between copper and stator iron
Ri Ω Equivalent series resistance for each battery cell
rir m Inner radius of rotor iron core
ris m Inner radius of stator teeth
riy m Inner radius of stator yoke
rm m Mean radius of stator yoke
rmr m Mean radius of rotor
VIII

ror m Outer radius of rotor
roy m Outer radius of stator yoke
RS Ω Mean stator winding resistance
Rth K/W Thermal resistance
Rth1 K/W Thermal resistance between shaft and rotor
Rth1 K/W Thermal resistance between junction and heatsink
Rth2 K/W Thermal resistance between rotor and teeth
Rth2 K/W Thermal resistance between heatsink and coolant
Rth3 K/W Thermal resistance between teeth and yoke
Rth3 K/W Thermal resistance between coolant and housing
Rth4 K/W Thermal resistance between yoke and coolant
Rth4 K/W Thermal resistance between housing and environment
Rth5 K/W Thermal resistance between coolant and housing
Rth6 K/W Thermal resistance between housing and environment
Rth7 K/W Thermal resistance between winding and teeth
Rth8 K/W Thermal resistance between winding and yoke
RU-V mΩ Ohmic resistance between phase u and v
RV-W mΩ Ohmic resistance between phase v and w
RW-U mΩ Ohmic resistance between phase w and u
Scu m² Cross-sectional area of copper in the stator slot
Sslot m² Cross-sectional area of stator slot
t s Time
T Nm Torque
T K Temperature
T0 K Initial temperature
T0n K Initial temperature of a specific thermal node
Tair K Ambient temperature during dc test
Tcase K Temperature of case during dc test
Tcoolant K Mean temperature of coolant
Tdem Nm Demand torque
Tn K Temperature of a specific thermal node
Twi K Inlet temperature of coolant
Two K Outlet temperature of coolant
Ua,b,c V Phase voltages for phase a, b, c
US V Stator phase voltage (RMS)
VDC V Voltage from dc supply
w kg Mass of liquid in cooling jacket
ẇ kg/s Coolant mass flow
IX

-1
αcu K Temperature coefficient of copper
ΔT K Temperature difference
η - Efficiency
ηI - Inverter efficiency
ηM - Motor efficiency
ηPU % Overall power unit efficiency
ν m²/s Kinematic viscosity
π - Ratio of circle’s circumference to diameter
τ s-1 Time constant
ΨPM Wb Flux linkage from permanent magnets
ω s-1 Angular velocity
X
TABLE OF CONTENTS
Affidavit....................................................................................................... II
Acknowledgment .......................................................................................III
Kurzfassung ............................................................................................... IV
Abstract ....................................................................................................... V
Acronyms and Abbreviations .................................................................. VI
Table of Signs and Symbols.................................................................... VII
Table Of Contents ..................................................................................... XI
1 Introduction........................................................................................... 1
1.1 Motivation ................................................................................................ 1
1.2 Objective ................................................................................................... 1
2 Fundamentals ........................................................................................ 4
2.1 Hybrid Powertrains in Vehicles ............................................................. 4
2.1.1 Topologies of Hybrid Electric Powertrains ............................................... 5
2.1.2 Hybrid Powertrains in Motorsports ........................................................... 7
2.1.3 Requirements for Hybrid Powertrains in Motorsports .............................. 8
2.2 Modeling of Electric Drives .................................................................. 13
2.2.1 Basics of Electric Drives ......................................................................... 13
2.2.2 Electromechanical Behavior of Electrical Machines .............................. 15
2.2.3 Determination of Losses in Electrical Machines ..................................... 20
2.2.4 Thermal Lumped Parameter Model......................................................... 23
3 Methodology ........................................................................................ 25
3.1 Design Of The Electric Drive Model .................................................... 25
3.1.1 Electromechanical Behavior.................................................................... 25
3.1.2 Motor Losses ........................................................................................... 27
3.1.3 Inverter Losses......................................................................................... 29
3.1.4 DC Power Calculation ............................................................................. 29
3.1.5 Motor Thermal Model ............................................................................. 30
3.1.6 Inverter Thermal Model .......................................................................... 33
XI
Table Of Contents
3.2 Parameterization ................................................................................... 34

3.2.1 Electromechanical Parameters................................................................. 34
3.2.2 Motor Loss Parameters ............................................................................ 35
3.2.3 Inverter Loss Parameters ......................................................................... 37
3.2.4 Motor Thermal Parameters ...................................................................... 38
3.2.5 Inverter Thermal Parameters ................................................................... 44
3.3 Model Verification ................................................................................. 44
3.3.1 Electric Drive Test System ...................................................................... 45
3.3.2 Parameter Generation .............................................................................. 46
3.3.3 Comparison of Model and Electric Drive Measurements ....................... 50
3.4 Integration in the Simulation Environment ........................................ 53
4 Results .................................................................................................. 58
4.1 Model Capabilities ................................................................................. 58
4.2 Simulation in VSM-PTM ...................................................................... 59
5 Discussion ............................................................................................ 64
6 References ............................................................................................ 65
7 Appendix .............................................................................................. 68
APPENDIX A Matlab Code: Parameter Generation ...................................... 68
APPENDIX B Motor Thermal Model in Simulink.......................................... 75
APPENDIX C Inverter Thermal Model in Simulink ...................................... 78
XII
1 INTRODUCTION
This thesis is about the design and development of a real-time model of hybrid power-
trains for motorsports applications. The project was done in close collaboration with the
racing department of AVL List GmbH (AVL).
1.1 MOTIVATION
The number of racing classes using hybrid electric propulsion systems is constantly ris-
ing, which creates new challenges for vehicle simulation. Not only are additional compo-
nents, such as electric motors and batteries, increasing the complexity of the powertrain
layout, they are also offering new ways of operation and control of the vehicle.
The Vehicle Simulation Model- Powertrain Model (VSM-PTM) contains detailed models
for most parts of racing vehicles, including a sophisticated model for internal combustion
engines (ICEs). However, the existing models for electric drives were very basic, based
solely on a maximum torque over speed curve without accounting for efficiency or losses
of the electric drive or the battery.
In order to maximize power output at a given fuel consumption and ultimately reduce lap
time, all parts of the hybrid powertrain need to be modelled with great detail. For racing
leagues with open regulations regarding the use of hybrid powertrain components (i.e. the
24 Hours of Le Mans), a flexible vehicle simulation model which covers a wide range of
powertrain layouts is desirable. To fulfill these requirements, detailed models of electric
motor, inverter, and energy storage were necessary.
1.2 OBJECTIVE
The main objective of the project was the creation of a MATLAB® Simulink® model of
the electric components of a hybrid electric powertrain as it is being used in motorsports.
The model had to be capable of operation in real time and also be flexible enough to be
used for different powertrain configurations. In addition, the model had to be imple-
mented in the existing simulation software VSM-PTM. In VSM-PTM, the model operates
as a standalone software model and as a part of Hardware-in-the-Loop (HiL) simulations.
Alexander Jedinger 1
1 Introduction
For this thesis, the focus was on a model for the electric motor and inverter. The result of
a prior project within AVL was used as a model for energy storage in a lithium-ion bat-
tery.
There are many different modeling approaches with large variations in complexity, com-
putation time, and parameterization effort (Section 2.2). In order for the project to be
successful and the result to be useful in the future, it was important to define the appro-
priate level of complexity of the model. AVL’s existing vehicle simulation model was
analyzed and the following requirements for the electric drive model were defined.
Since the model was integrated in VSM-PTM, a simulation platform for HiL-simulations,
it needed the same real-time capabilities as VSM-PTM. This means the model had to run
in Matlab® Simulink® at a frequency of 2 kHz. Additionally, the model needed to be easy
to parameterize while maintaining a high level of accuracy. Motorsport applications typ-
ically push components to their limits, requiring that the model is accurate in extreme
operating conditions. Because manufacturers tend to be secretive about design and control
algorithms for the electric drive, the model could not require this information. For the
thermal layout of motor and inverter, liquid cooling could be assumed.
Another essential requirement was the representation of the actual physical parameters of
the components (i.e. voltages and currents). The electric drive model was later connected
to a battery model and therefore needed to have the proper physical interface. Because of
the simulation frequency of 2 kHz and the requirement for real time operation, high fre-
quency phenomena were not modeled.
In order to fully integrate into VSM-PTM, the desired model required all necessary inter-
faces to the surrounding model. The interface components are comprised of coolant tem-
peratures and flow rates for the connection to the vehicle cooling circuit, the battery in-
terface, and the mechanical connection to the drive shaft (Figure 1). Moreover, the model
parameters had to be integrated in the PTM parameter structure.
1 Introduction
Demand Torque Active Torque

Speed
Electric Drive
Battery Voltage Battery Current
Model
Coolant Temperature In
Coolant Mass Flow In Coolant Temperature Out
Figure 1: Required model interfaces
To attain a detailed thermal model, the losses of the drive were separated (Figure 2).
While the inverter losses are represented in one value, the motor losses are split up into
two parts, according to their origin. These three portions of losses are then mapped onto
their corresponding nodes in the thermal model (Section 3.1.5 and 3.1.6).
Figure 2: Flow of energy for the electric drive model
2 FUNDAMENTALS
2.1 HYBRID POWERTRAINS IN VEHICLES

Hybrid powertrains in vehicles are one of the most promising concepts to reduce the use
of fossil fuels and emissions in traffic. Currently many automotive manufacturers are de-
veloping hybrid vehicles (HVs), particularly hybrid electric vehicles (HEVs). The success
of established hybrid concepts, like the Toyota Prius, proves not only the ability to de-
crease fuel consumption and emissions to meet current and future regulations, but also
that there is a high demand from consumers.
HVs use more than one source of energy to power the vehicle. Typically, one of those
sources works unidirectional (i.e. the energy can only be converted in one direction)
(Figure 3, Powertrain (1)). In most cases, this part represents a gasoline or diesel powered
ICE. The second source of energy acts not only as a source of power, but also as an energy
storage system, harnessing and releasing energy on demand (Figure 3, Powertrain (2)).
The most common way of doing that is with electric motors as the energy converter and
batteries or super capacitors for the energy storage. Another concept is using a flywheel
to store energy, and connect it to the drivetrain through two electric machines or a con-
tinuously variable transmission.
Figure 3: Principal topology of hybrid powertrains [1]
2 Fundamentals
2.1.1 TOPOLOGIES OF HYBRID ELECTRIC POWERTRAINS

As previously mentioned, the vast majority of HVs in series production and development
are HEVs, powered by an ICE in combination with electrical energy storage. The most
common topologies of such powertrains are introduced in the following sections.
2.1.1.1 PARALLEL HYBRID ELECTRIC POWERTRAIN

In a parallel hybrid electric powertrain configuration, the ICE and the electric motor are
both mechanically connected to the wheels and to each other (Figure 4). Several modes
of operation are possible:
• Only the ICE delivers power to the transmission

• Only the electric motor delivers power to the transmission
• Both ICE and electric motor deliver power to the transmission
• The electric motor receives power from the transmission (recuperation)
• The electric motor receives power from the ICE
• The electric motor receives power from the transmission and the ICE
• The ICE delivers power to the transmission and the electric motor.
Figure 4: Parallel hybrid electric powertrain [1]
A big advantage of this concept is the ability to deliver torque from both power sources
directly to the wheels without any energy conversion, resulting in high transmission effi-
ciency for both the ICE and the electric drive. Another advantage is the scalability of the
electric drive. It can be used with small starter-generator units for a start-stop system, as
2 Fundamentals
well as with large electric machines and batteries that offer a substantial range of fully
electric driving. The major disadvantage is the inability to drive the ICE in only the most
fuel efficient operating point due to the mechanical coupling with the drivetrain [2].
2.1.1.2 SERIES HYBRID ELECTRIC POWERTRAIN

A series hybrid electric powertrain can be seen as a fully electric powertrain with an ad-
ditional source of electric energy in form of an ICE coupled to a generator (Figure 5).
This is the typical layout of electric vehicles with range extenders, where the vehicle is
usually powered fully electrically and the ICE is used to extend the range by converting
fossil fuels to electricity when the battery is depleted.
Advantages of the series hybrid electric layout include the ability to operate the ICE in a
narrow speed range, enabling it to be smaller and more fuel efficient. Also, there is only
one source of torque, which reduces the number of mechanical components compared to
a parallel layout. Because of the favorable torque-speed characteristics of suitable electric
machines, a multi-gear transmission becomes unnecessary. Moreover, packaging is more
flexible since most parts of the powertrain are solely connected by electric wires. This
concept also bears disadvantages, such as the low overall efficiency of the ICE, due to
conversion of mechanical to electrical power and back to mechanical power. The concept
requires not only a powerful electric motor for propulsion since it is the only source of
mechanical power, but also requires an electric generator to be attached to the ICE [2].
Figure 5: Series hybrid electric powertrain [1]
2 Fundamentals
2.1.1.3 OTHER HYBRID ELECTRIC POWERTRAIN TOPOLOGIES

Besides the discrete use of series and parallel hybrid electric powertrains, there are also
other layouts. One example is a combination of a series and parallel hybrid powertrain,
which combines advantages of both layouts. In most operation modes it acts as a regular
series hybrid powertrain, allowing fully electric driving with a range extender. However,
when maximum power is needed, the ICE can be mechanically connected to the drive
shaft using a planetary gearbox or a set of clutches resulting in a higher efficiency at high
speeds [1].
2.1.2 HYBRID POWERTRAINS IN MOTORSPORTS

Saving energy is becoming increasingly important not only for passenger and commercial
vehicles but also in motorsports. The following sections give an overview of racing series
with hybrid powertrains and their topologies.
2.1.2.1 FORMULA ONE

In 2009, Formula One introduced their first hybrid system by adding a kinetic energy
recovery system (KERS) to their powertrain. It was a parallel hybrid topology, where an
electric or kinetic energy storage was attached to the crank shaft, enabling the KERS to
recuperate energy when braking and propelling the car with 60kW for up to 6.6 seconds
per lap [3].
In 2014, as a part of major changes in regulations, the Fédération Internationale de l’Au-

tomobile (FIA) introduced the Energy Recovery System (ERS), a much more advanced
and integrated hybrid concept. It comprises two motor generator units (MGU), one at-
tached to the crankshaft (MGU-K) and one attached to the shaft of the turbo-charger
(MGU-H). According to the FIA [4], the MGU-K may operate with an electric power of
up to 126kW (~120kW mechanical power) at no more than 50krpm. The MGU-H may
run at a maximum of 125krpm but is not limited in power. For energy storage, there is a
battery that can provide a maximum of 4MJ of boost energy per lap while being charged
at a maximum of 2MJ per lap. Figure 6 illustrates the components of the 2014 F1 power
unit.
2 Fundamentals
Figure 6: Formula One power unit in 2014 [5]
The 2014 ERS enables a more sustained use of the hybrid components, with twice the
power and ten times the storage capacity compared to the preceding KERS. Although
more powerful, it also requires more sophisticated controls to coordinate MGU-H, MGU-
K, battery and ICE.
2.1.2.2 24 HOURS OF LE MANS

The Automobile Club de l’Ouest (ACO), host of the 24 Hours of Le Mans, changed the
technical regulations in 2009, and again in 2012 and 2014, allowing HVs to enter the race
and giving participating teams more freedom in the design of the powertrain. Since then,
many variations of hybrid powertrains were used and have proved to be more successful
than their ICE powered counterparts. In 2012, the Audi R18 e-tron, equipped with a fly-
wheel energy storage system, was the first hybrid car to win the 24 Hours of Le Mans.
2.1.3 REQUIREMENTS FOR HYBRID POWERTRAINS IN MOTORSPORTS

Like most parts in a motorsports vehicle, the requirements for hybrid powertrains are
different than those in passenger cars. Important factors for passenger vehicles such as
cost, smooth operation, and low noise typically have lower priority in motorsports.
The main goal in racing is to achieve the lowest lap time, which is affected by a large
number of vehicle properties. Those most relevant to the powertrain are discussed in the
2 Fundamentals
following sections and are illustrated with results of VSM simulations. The simulations
were done with a typical setup for open wheeled single-seater racecars, derived from an
existing dataset at AVL Racing. The results show the sensitivity of the system with re-
spect to various parameters. Table 1 gives an overview of important parameters for the
simulation, where 𝑐𝑐𝑑𝑑 is the aerodynamic drag coefficient, 𝑐𝑐𝑟𝑟𝑟𝑟 the per wheel rolling re-
sistance coefficient, 𝑚𝑚 the vehicle mass, 𝑃𝑃𝑃𝑃𝑃𝑃 the overall power of the power unit and 𝜂𝜂𝑃𝑃𝑃𝑃
the overall efficiency of the power unit.
Table 1: Vehicle parameters for simulations
cd crr m PPU ηPU

- - kg kW %
0.952 0.018 695 445 35
The track used in the simulations was the Autodromo Nazionale Monza in Italy
(Figure 53, page 59) with a length of 5,754 meters. Figure 7 shows the vehicle speed
profile of the simulation with the parameters from Table 1, resulting in a lap time of
92.87s.
400
300
Speed / km/h
200
100
0
0 1000 2000 3000 4000 5000 6000
Distance / m
Figure 7: Vehicle speed at the Autodromo Nazionale Monza, simulation result
2.1.3.1 POWER
The most obvious factor for a quick lap is the overall power of the hybrid power unit. The
higher the propulsion power, the faster the car and therefore the lower the lap time. The
result of VSM simulations with different ICE power is shown in Figure 8. Although the
lap time decreases with higher power, the graph also indicates that the improvement is
not proportional to the power gain. This is because of the increasing aerodynamic drag at
higher vehicle speeds. In addition, the duration of full power during a lap decreases with
increased maximum power since the tires are operated at the slip limit for a longer time.
2 Fundamentals
Consequences of a change in ICE power (e.g. from increased vehicle weight, higher fuel
consumption) have been ignored in these simulations.
96
95
94
Lap Time / s
93
92
91
90
360 380 400 420 440 460 480 500 520
ICE Power / kW
Figure 8: Influence of propulsion power on lap time
94
92
Lap Time / s
90
88
360
400 60
80
440 100
120
480 140
160
Power ICE / kW 520 180
Power MGU-K / kW
Figure 9: Influence of ICE power and MGU-K power on lap time
In a parallel hybrid configuration like in Formula One, it is worth considering how pow-
erful combustion engines and electric motors should be. Figure 9 illustrates the effect of
ICE power and MGU-K power on lap time. In this example, the MGU-K always operates
2 Fundamentals
when the ICE is at full load, without considering the MGU-K’s energy limits. It shows
that an increase of the MGU-K’s power by 100kW gains 1.5s in lap time, while the same
increase of power for the combustion engine gains about 2.5s.
Since the power of the MGU-K is limited to a maximum of 120kW, the ICE is the only
device that can be worked on with the current regulations. There is also a limitation of
fuel consumption to 100kg per race and fuel mass flow to no more than 100kg/h at any
time [4]. As a result, the efficiency of the combustion engine is a very crucial part for
success, and a lot of engineering effort is put into its optimization.
2.1.3.2 VEHICLE WEIGHT

To illustrate the effects of vehicle weight, Figure 10 shows VSM simulation results at
different vehicle weights. All other parameters remained constant. Since mass affects the
acceleration of the vehicle, higher mass at a constant engine power leads to a linear in-
crease in lap time.
96
95
94
Lap Time / s
93
92
91
90
640 650 660 670 680 690 700 710 720 730 740 750
Vehicle Mass / kg
Figure 10: Influence of vehicle mass on lap time
2.1.3.3 AERODYNAMIC PERFORMANCE

Another very important contributor to lower lap times in modern motorsports is the aer-
odynamic behavior of the car. While higher downforce improves the car’s traction limit,
allowing for higher cornering speeds, it has negative effects on the drag coefficient. Con-
sequently, the maximum speed is reduced. Figure 11 shows simulation results of different
2 Fundamentals
aerodynamic drag on the car’s lap time. All other parameters, including downforce, were
kept constant.
96
95
94
Lap Time / s
93
92
91
90
0,7 0,75 0,8 0,85 0,9 0,95 1 1,05 1,1 1,15 1,2
Aerodynamic Drag / p.u.
Figure 11: Influence of aerodynamic drag on lap time
2.1.3.4 EFFICIENCY
The efficiency of the power unit gained in importance particularly after some of the big-
gest race classes passed regulations to limit fuel consumption. When fuel consumption is
limited, every improvement in efficiency directly affects the lap time. Figure 12 shows
the lap time over different power unit efficiencies. Similar to the effect of propulsion
power (Figure 8), the lap time gain decreases at higher efficiencies since the additional
power can only be used when the wheels are below the slip limit.
96
95
94
Lap Time / s
93
92
91
90
30 31 32 33 34 35 36 37 38 39 40
Power Unit Efficiency / %
Figure 12: Influence of power unit efficiency on lap time
2 Fundamentals
Another aspect of efficiency is when to use the available energy over the course of a lap.
Figure 13 shows the relationship of energy and MGU-K power and their influence on lap
time. It turns out that with the same amount of energy per lap, different lap times can be
obtained. The more concentrated the boost of the MGU-K at the beginning of each
straight (i.e. the higher the MGU-K power), the lower the overall lap time.
93
92
Lap Time / s
91
90
60
80 0
100 1000
120 2000
140 3000
160 4000
5000
Power MGU-K / kW 180 6000
Used Energy / kJ
Figure 13: Influence of MGU-K power and used energy on lap time
2.2 MODELING OF ELECTRIC DRIVES

Describing the behavior of electric drives has been a topic of interest for decades, and is
still an ongoing field of research. The following sections will discuss some of the most
popular ways of modeling electric drives, both for the electromagnetic and the thermal
domain.
2.2.1 BASICS OF ELECTRIC DRIVES

Traditionally, electric motors run uncontrolled, directly fed from the electric grid. Thus,
there is typically only one rotational speed at which they can continuously operate. It is
determined by the winding configuration of the motor and therefore cannot be changed
without major modification of the motor. By using a power processing unit (PPU), also
2 Fundamentals
referred to as inverter, and appropriate controls, the electric motor can be decoupled from
the electric power source. This enables it to run within a range of speeds rather than at a
single speed. An electric drive system typically consists of a motor, the PPU, a sensor for
the motor’s rotor position and the controller [6]. Some drives, particularly induction mo-
tor (IM) drives, do not always require a sensor for the rotor position or speed. Figure 14
illustrates a typical drive system including the source of electricity and the mechanical
load.
Figure 14: Block Diagram of a typical Electric Drive System [7]
The main part of the PPU is a set of semiconductors that converts the fixed form of elec-
tricity (Figure 14) to the desired form for driving the motor. In traction applications, the
fixed form is typically direct current (DC) voltage from a battery. The most crucial part
of an electric drive is the controller, which generates the demanded waveform for the
inverter by evaluating the input command and the rotor position.
Another significant advantage allowing for the use of electric motors in traction applica-
tions is the ability to not only drive a motor at positive and negative rotational speeds, but
also apply positive and negative torques. This results in the so-called four-quadrant oper-
ation shown in Figure 15.
2 Fundamentals
torque
(+) speed (+) speed

(−) torque (+) torque
Backward Acceleration Forward Acceleration
speed
Backward Braking Forward Braking
(−) speed (+) speed
(−) torque (−) torque
Figure 15: Four-quadrant operation of electric drives in vehicles
Inverters can be used to drive most types of electric motors. The most common types used
in vehicles for propulsion purposes are IMs, interior permanent magnet synchronous mo-
tors (IPM), and surface mounted permanent magnet synchronous motors (SPM)
[8, 9, 10]. Details about the functionality of these machines can be found in [6, 11] and
other literature about fundamentals of electric machinery.
2.2.2 ELECTROMECHANICAL BEHAVIOR OF ELECTRICAL MACHINES

There is a variety of approaches to modeling electric machines which greatly differ in
complexity and accuracy. The choices for a given model have to be made according to
the desired level of detail and available data for parameterization of the model. For the
following models discussed in this section, permanent magnet synchronous motors were
used. However, the introduced models can be used for most common types of electric
machines.
Independent of the chosen model, there are several basic relations for symmetric three
phase electric machines that are used in this thesis [6]. The active power fed into a three
phase electric machine can be calculated as
𝑃𝑃𝑒𝑒𝑒𝑒 = 3 ⋅ 𝐼𝐼𝑆𝑆 ⋅ 𝑈𝑈𝑆𝑆 ⋅ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐, (2.1)
2 Fundamentals
with 𝐼𝐼𝑆𝑆 and 𝑈𝑈𝑆𝑆 as the stator current and voltage, respectively, and 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 as the power factor
of the machine. The mechanical power on the rotor shaft of the machine is calculated as
𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚ℎ = 𝑇𝑇 ⋅ 𝜔𝜔, (2.2)
where 𝑇𝑇 is the mechanical torque and 𝜔𝜔 is the angular velocity of the rotor. By combining
Equation (2.1) and (2.2) the overall efficiency 𝜂𝜂 of the machine can be obtained as fol-
lows:
𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚ℎ
𝜂𝜂 = . (2.3)
𝑃𝑃𝑒𝑒𝑒𝑒
2.2.2.1 EQUIVALENT CIRCUIT

A simple way of describing the behavior of electric machines is to derive equivalent elec-
tric circuits, which represent more complex physical processes taking place inside the
machine. Figure 16 shows the single phase equivalent circuit of a permanent magnet syn-
chronous machine, with R1 as the stator resistance, L1 as the stator inductance, UT as the
terminal voltage, I1 as the stator current and UPM as the induced voltage due to the per-
manent magnets in the rotor.
R1 L1
I1
UT UPM
Figure 16: Single-phase equivalent circuit of a SPM
Single phase equivalent circuits are a good base for modeling and understanding the func-
tionality of electric machines and are rapidly simulated. However, they only describe the
basic behavior of the machine and do not account for magnetic saturation, geometric
2 Fundamentals
asymmetries or frequencies of higher order. Furthermore, there is no simple mathematical

connection between the electrical input and the mechanical output of the machine.
2.2.2.2 SPACE VECTOR MODEL

For three-phase-systems, the space-vector representation is the most commonly used ap-
proach today. It is based on the fact that the sum of the three phase voltages, currents,
etc., in a symmetric three-phase-system without a neutral conductor is always zero. Equa-
tion (2.4) clarifies this relationship for the stator phase voltages as follows,
𝑈𝑈𝑎𝑎 + 𝑈𝑈𝑏𝑏 + 𝑈𝑈𝑐𝑐 = 0. (2.4)
One phase value can be calculated from the two other phases, so it is sufficient to know
the values of two phases to define the system. The relationship can also be described with
a system of two vectors using a Cartesian coordinate system. This is called dq-, or Park-
Transformation, and will be explained in the following paragraphs using the stator cur-
rents as an example. In general, the phase currents of a three-phase system can be de-
scribed as
𝑖𝑖𝑎𝑎 = 𝐼𝐼̂ ⋅ cos(𝜔𝜔𝜔𝜔 + 𝜑𝜑𝑎𝑎 ) = 𝐼𝐼̂ ⋅ 𝑒𝑒 𝑗𝑗𝜑𝜑𝑎𝑎 ,

𝑖𝑖𝑏𝑏 = 𝐼𝐼̂ ⋅ cos(𝜔𝜔𝜔𝜔 + 𝜑𝜑𝑏𝑏 ) = 𝐼𝐼̂ ⋅ 𝑒𝑒 𝑗𝑗𝜑𝜑𝑏𝑏 , (2.5)
𝑖𝑖𝑐𝑐 = 𝐼𝐼̂ ⋅ cos(𝜔𝜔𝜔𝜔 + 𝜑𝜑𝑐𝑐 ) = 𝐼𝐼̂ ⋅ 𝑒𝑒 𝑗𝑗𝜑𝜑𝑐𝑐 ,
2𝜋𝜋 4𝜋𝜋
where 𝜑𝜑𝑎𝑎 = 0, 𝜑𝜑𝑏𝑏 = 3
, and 𝜑𝜑𝑐𝑐 = 3
correspond to the 120° temporal shift between the
phases. In a three-phase electric motor, the phases also have a 120° physical gap with
respect to each other (see Figure 17a), which, when adding them geometrically, results in
a constant magnitude vector rotating with the angular frequency 𝜔𝜔. The idea of the Park-
Transformation is to define a coordinate system that rotates with the same frequency 𝜔𝜔,
creating a domain where the physical three-phase system is represented through non-ro-
tating DC-vectors. Equation (2.6) shows the transformation proposed by Park [12].
2𝜋𝜋 2𝜋𝜋
2 cos(𝜃𝜃) cos �𝜃𝜃 − � cos �𝜃𝜃 + � 𝑖𝑖𝑎𝑎
𝑖𝑖𝑑𝑑 3 3 � ⋅ �𝑖𝑖 �
�𝑖𝑖 � = � 𝑏𝑏 (2.6)
𝑞𝑞 3 2𝜋𝜋 2𝜋𝜋
−sin(𝜃𝜃) − sin �𝜃𝜃 − � − sin �𝜃𝜃 + � 𝑖𝑖 𝑐𝑐
3 3
2 Fundamentals
For most types of electric machines (except induction machines), 𝜃𝜃 is the angular position
of the rotor. Figure 17 illustrates the transition from the physical three-phase system to
the imaginary two-phase Cartesian system.
d
θ θ
stator system
rotor system
(a) three-phase system (b) two-phase Cartesian system
Figure 17: Transition from three-phase to two-phase system [6]
A major advantage of this approach is that the connection between the electrical and me-
chanical behavior of the machine can be written in a simple way [13]. The air gap torque
for SPM machines is given by
3
𝑇𝑇 = 𝑝𝑝 ⋅ Ψ𝑃𝑃𝑃𝑃 ⋅ 𝑖𝑖𝑞𝑞 , (2.7)
2
where 𝑝𝑝 is the number of pole pairs, Ψ𝑃𝑃𝑃𝑃 is the flux linkage due to the permanent magnets
in the rotor, and 𝑖𝑖𝑞𝑞 is the current in the q-axis. Considering the reluctance torque of IPM
machines, the air gap torque expands Equation (2.7) to
3
𝑇𝑇 = 𝑝𝑝 �Ψ𝑃𝑃𝑃𝑃 ⋅ 𝑖𝑖𝑞𝑞 + �𝐿𝐿𝑑𝑑 − 𝐿𝐿𝑞𝑞 � ⋅ 𝑖𝑖𝑑𝑑 ⋅ 𝑖𝑖𝑞𝑞 �, (2.8)
2
where 𝐿𝐿𝑑𝑑 and 𝐿𝐿𝑞𝑞 are the inductances in d- and q-axis. Equations (2.7) and (2.8) are the
basis of the field oriented control method used in many modern electric drive applications.
2 Fundamentals
2.2.2.3 FINITE ELEMENT ANALYSIS (FEA)
The finite element method has become more popular in recent decades, partly due to the
increasing computational power of computers. In contrast to the approaches discussed in
Sections 2.2.2.1 and 2.2.2.2, it focuses on the spatial electromagnetic field distribution in
the machine (i.e. it stays close to the actual physical phenomena). The finite element
method is used in several fields of engineering and is not limited to electromagnetic fields.
Because the electromagnetic FEA is the only relevant application for this thesis, the others
are not discussed in detail.
The basic idea of electromagnetic FEA is to solve the underlying set of partial differential
equations (Maxwell’s equations) for a particular problem. The Maxwell’s equations are
as follows:
𝜌𝜌
∇ ⋅ 𝑬𝑬 = ,
𝜀𝜀0
∇ ⋅ 𝑩𝑩 = 0,
𝜕𝜕𝑩𝑩 (2.9)
∇ × 𝑬𝑬 = − ,
𝜕𝜕𝜕𝜕
𝜕𝜕𝑬𝑬
and ∇ × 𝑩𝑩 = 𝜇𝜇0 �𝑱𝑱 + 𝜀𝜀0 �.
𝜕𝜕𝜕𝜕
Analytical solutions of these equations can only be obtained for simple problems. In order
to solve them for more complex problems, the problem is divided into a large number of
simpler elements. These elements are solved numerically and are reassembled to form the
solution of the initial problem. Figure 18 shows the discretization of a 2D SPM problem
into triangular elements, done with the FEA software JMAG.
Since this is a numeric approach, the solution will never be exact, but the accuracy in-
creases as the number of elements grows. This also increases the computation time, so
there is a trade-off between accuracy and necessary computational power or time. There-
fore, many FEA models for electrical machines are two-dimensional, simulating only one
radial plane of the machine. If there are symmetries in that radial plane, the FEA model
may also simulate only parts of that radial plane. Consequently, the model size of FEA
problems varies greatly, from some thousands of elements (e.g. 2D quarter models) to
2 Fundamentals
several millions of elements (e.g. full 3D models). Because the finite element method is
based on the spatial distribution of fields, the geometry of the machine as well as material
parameters, like magnetic saturation, are taken into account. FEA models are not limited
to symmetric electrical machines, but able to solve various electromagnetic problems.
Figure 18: Division into finite elements for a 2D SPM problem
2.2.3 DETERMINATION OF LOSSES IN ELECTRICAL MACHINES

Particularly for electrical drives in vehicles, high efficiency is of great importance. In
order to calculate the efficiency in a model, the losses that occur in various parts of the
machine need to be determined. Those can be divided into four main contributors as fol-
lows:
𝑃𝑃𝐿𝐿 = 𝑃𝑃𝑐𝑐𝑐𝑐 + 𝑃𝑃𝑓𝑓𝑓𝑓 + 𝑃𝑃𝑓𝑓𝑓𝑓 + 𝑃𝑃𝑣𝑣 . (2.10)
The majority of losses occur in the winding (𝑃𝑃𝑐𝑐𝑐𝑐 ) and the iron core (𝑃𝑃𝑓𝑓𝑓𝑓 ). Additional losses
can be defined as friction losses (𝑃𝑃𝑓𝑓𝑓𝑓 ) and ventilation losses (𝑃𝑃𝑣𝑣 ). In the following sec-
tions, the causes and various modeling approaches for these types of losses will be intro-
duced.
2.2.3.1 COPPER LOSSES

Copper losses occur in the stator winding and are also referred to as ohmic or winding
losses. They are caused by the ohmic resistance of the stator windings, typically made of
copper, and the current flowing through these windings. For three phase machines, the
basic calculation is done by
2 Fundamentals
𝑃𝑃𝑐𝑐𝑐𝑐 = 3 ⋅ 𝐼𝐼𝑆𝑆2 ⋅ 𝑅𝑅𝑆𝑆 , (2.11)
where 𝐼𝐼𝑆𝑆 is the root mean square (RMS) value of the current through one phase, and 𝑅𝑅𝑆𝑆
is the mean resistance of one phase winding. Since the resistance of copper varies with
temperature, 𝑅𝑅𝑆𝑆 is not constant. The thermal behavior of the resistance of copper can be
approximated by
𝑅𝑅(𝑇𝑇) = 𝑅𝑅(𝑇𝑇0 ) ⋅ �1 + 𝛼𝛼𝑐𝑐𝑐𝑐 ⋅ (𝑇𝑇 − 𝑇𝑇0 )�, (2.12)
with 𝑅𝑅(𝑇𝑇0 ) as the resistance at a reference temperature 𝑇𝑇0 and 𝛼𝛼𝑐𝑐𝑐𝑐 as the temperature
coefficient of copper. For commercial copper used in electric machines, the temperature
coefficient is 𝛼𝛼𝑐𝑐𝑐𝑐 = 3,81 ⋅ 10−3 𝐾𝐾 −1 [14].
2.2.3.2 IRON LOSSES

Iron losses occur in the magnetic parts of rotor and stator and are also referred to as core
losses or magnetic losses. The losses are caused by the alternating magnetic field and can
be divided into two separate phenomena: hysteresis losses and eddy current losses.
Figure 19: B-H hysteresis curve of a ferromagnetic material at 10Hz (red) and 200Hz (blue)
[15]
Hysteresis losses occur due to the nonlinear behavior of ferromagnetic materials. If a fer-
romagnetic material is exposed to a magnetic field 𝐻𝐻 in one direction, which then declines
and changes to the opposite direction, the resulting magnetic field density 𝐵𝐵 does not
2 Fundamentals
follow linearly. Due to the reorientation of the Weiss domains in the material [6], the
transition from a positive to negative magnetic field and vice versa follows a hysteresis
loop (Figure 19). The area enclosed by the hysteresis loop represents the specific energy
for each cycle, which is proportional to the iron losses. The specific iron losses at a fre-
quency 𝑓𝑓 of the magnetic field 𝐻𝐻 can be calculated with
𝑝𝑝𝑓𝑓𝑓𝑓 = 𝑓𝑓 � 𝐻𝐻 𝑑𝑑𝑑𝑑. (2.13)
Eddy current losses, or joule losses, are caused by the eddy currents in the conductive
regions of the machine due to the alternating magnetic field and, consequently, the in-
duced voltage. They are divided into classic and anomalous eddy current losses. Classic
eddy current losses originate from the induced voltage, causing circular currents in the
stack of iron sheets, which lead to losses due to the ohmic resistance of the iron. The
anomalous losses refer to losses from microscopic eddy currents caused by the displace-
ment of the Bloch walls [16]. There are numerous ways of mathematically describing iron
losses in electrical machines. Figure 20 gives an overview of common approaches.
Figure 20: Different approaches for modeling the iron losses in electric machines [15]
2.2.3.3 MECHANICAL LOSSES

Mechanical losses occur due to the rotation of the motor’s shaft. The majority of those
losses are friction losses in the rotor bearings and ventilation losses in the air gap. In
electric machines with permanent magnets, the mechanical losses cannot be measured
separately because of the ever present iron losses. The permanent magnets would have to
be removed to do the measurements.
2 Fundamentals
2.2.4 THERMAL LUMPED PARAMETER MODEL

In order to account for the influence of temperature on the operation of an electric drive
system and to design an adequate cooling circuit, the thermal behavior of the machine
needs to be modelled. Since other thermal modelling approaches such as thermal finite
element modelling or computational fluid dynamics would not meet the model require-
ment for real time capability, the concept of lumped parameters is the only approach dis-
cussed in detail.
A lumped parameter model, also called lumped element model or lumped capacitance
model, reduces a thermal system to a number of interconnected discrete points. This sim-
plifies the partial differential equations of the continuous time and space model of the
physical system to a finite set of ordinary differential equations. It is analogous to the
equivalent circuit approach used for electrical networks (Figure 21).
T1 Rth 𝑞𝑞𝑞 T2
Cp1 Cp2
Figure 21: Example lumped parameter network
According to Fourier’s law of thermal conduction, the local heat flux density can be cal-
culated as the product of thermal conductivity 𝜆𝜆 and the negative local temperature gra-
dient −∇𝑇𝑇 as follows:
𝑞𝑞⃗ = −𝜆𝜆 ∇𝑇𝑇. (2.14)
Making the assumptions used for lumped element models, this can be simplified to
Δ𝑇𝑇 𝑇𝑇1 − 𝑇𝑇2

𝑞𝑞̇ = = , (2.15)
𝑅𝑅𝑡𝑡ℎ 𝑅𝑅𝑡𝑡ℎ
2 Fundamentals
where Δ𝑇𝑇 is the temperature difference between two nodes, and 𝑅𝑅𝑡𝑡ℎ is the thermal re-
sistance between them. This is analogous to Ohm’s Law for electrical circuits. The tem-
perature of each mass point can be obtained by
1
𝑇𝑇 = ∫ (𝑃𝑃𝐿𝐿 − 𝑞𝑞̇ )𝑑𝑑𝑑𝑑 + 𝑇𝑇0 , (2.16)
𝑚𝑚 ⋅ 𝐶𝐶𝑝𝑝
with 𝑚𝑚 being the mass of the particular lump, 𝐶𝐶𝑝𝑝 the specific heat, 𝑃𝑃𝐿𝐿 the rate of heat
transfer to the lump, 𝑞𝑞̇ the rate of heat transfer from the lump, and 𝑇𝑇0 the initial tempera-
ture. The size of lumped element models can range from one or two lumps to massive,
auto-generated models with a large number of elements.
3 METHODOLOGY
After covering the technical background for the thesis, the following sections explain the
model that was designed and the assumptions it is based on.
3.1 DESIGN OF THE ELECTRIC DRIVE MODEL

According to the model requirements stated in Section 1.2, a modelling approach for the
physical parts of the model was proposed. Figure 22 gives an overview of the organization
of the different parts.
Figure 22: Electric drive model layout
The model consists of a basic representation of the electromechanical behavior, loss cal-
culations for both inverter and motor, and the thermal models for inverter and motor.
3.1.1 ELECTROMECHANICAL BEHAVIOR

As mentioned in Section 1.2, the model could not require knowledge about the controls
of the electric motor, and thus a space vector based approach was not feasible. For the
same reason and the required real-time capabilities, any finite element modelling would
not have met the requirements either. Consequently, the model for the electromechanical
behavior simply comprises a time delay to account for the delay in torque production and
3 Methodology
a limitation of maximum torque according to the motor speed and DC supply voltage.
Figure 23 illustrates the modelling approach.
Figure 23: Modelling approach for the electromechanical behavior
The torque limitation block ensures that the produced torque never exceeds the motors
maximum rating at a certain speed, and also reduces the maximum rating if the supply
voltage drops below a critical value. Figure 24 illustrates the derating of motor torque
when the supply voltage drops under a critical value 𝑉𝑉𝐷𝐷𝐷𝐷 , which results in a lower base
speed (i.e. an earlier start of field weakening operation).
160
0.2 VDC
140
0.4 VDC
120 0.6 VDC
0.8 VDC
100
|Torque| / Nm
1 VDC
80
60
40
20
0
0 1000 2000 3000 4000 5000 6000 7000 8000
Rotor-Speed / rpm
Figure 24: Motor torque derating due to low supply voltage
The motor’s torque response can be represented with a first order lag element. The ac-
cording mathematical form is
3 Methodology
𝑡𝑡
𝑇𝑇 = 𝑇𝑇𝑑𝑑𝑑𝑑𝑑𝑑 ⋅ �1 − 𝑒𝑒 −𝜏𝜏 �, (3.1)
where 𝑇𝑇𝑑𝑑𝑑𝑑𝑑𝑑 is the demanded torque and 𝜏𝜏 is the torque delay time constant. Figure 25
shows the resulting model layout as it was implemented in Simulink. The go-to flags for
speed and torque were used as inputs for the following loss determination.
Figure 25: Simulink model for electromechanical behavior
3.1.2 MOTOR LOSSES

In order to model the thermal behavior of the electric drivetrain, the losses needed to be
determined. In accordance with the model requirements (Section 1.2), the losses were
estimated based on measurements of the motor efficiency over the whole operating range,
including the temperature of the winding. The model distinguishes between copper losses
and other losses, including iron, friction and ventilation losses. Since the major applica-
tion for this model will be permanent magnet motors, where iron, friction, and ventilation
losses are hard to measure separately, no further separation was made.
The copper losses are determined by
2
𝑃𝑃𝑐𝑐𝑐𝑐 = 𝑁𝑁𝑝𝑝ℎ ⋅ 𝐼𝐼𝑝𝑝ℎ ⋅ 𝑅𝑅𝑆𝑆 , (3.2)
where 𝐼𝐼𝑝𝑝ℎ is the measured mean RMS phase current at each operating point, 𝑅𝑅𝑆𝑆 the mean
stator winding resistance at the current winding temperature, and 𝑁𝑁𝑃𝑃ℎ is the number of
phases. The stator winding resistance was calculated with the following equation based
on a measurement of the mean stator winding resistance at a known temperature, and the
current winding temperature,
𝑅𝑅𝑆𝑆 (𝑇𝑇) = 𝑅𝑅𝑆𝑆 (𝑇𝑇0 ) ⋅ �1 + 𝛼𝛼𝑐𝑐𝑐𝑐 ⋅ (𝑇𝑇 − 𝑇𝑇0 )�. (3.3)
3 Methodology
Here, 𝑇𝑇0 is the temperature where the mean stator winding phase resistance 𝑅𝑅𝑆𝑆 (𝑇𝑇0 ) was
measured and 𝛼𝛼𝑐𝑐𝑐𝑐 is the temperature coefficient of copper. Figure 26 illustrates the pro-
posed model layout for the determination of copper losses.
Figure 26: Modelling approach for the copper losses
Since the iron, friction and ventilation losses are not separated, there is no way to analyt-
ically describe the influence of temperature. Consequently, the temperature dependence
of these losses was implemented as an additional parameter for the losses lookup table
based on efficiency measurements at different motor temperatures. Considering the
model requirements, this was a very accurate way of determining these losses. Figure 27
illustrates this model layout.
Figure 27: Modelling approach for iron and other losses
The determination of all motor losses implemented in Simulink is shown in Figure 28.
3 Methodology
Figure 28: Simulink model for determination of motor losses
3.1.3 INVERTER LOSSES

Since there is little to no knowledge about the inverter layout and operation modes, the
loss determination was based solely on prior measurements. Figure 29 illustrates the sim-
ple implementation of the inverter losses.
Figure 29: Simulink model for determination of inverter losses
3.1.4 DC POWER CALCULATION

The interface to the DC power supply is the demanded DC current, which is calculated
with
𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚ℎ + 𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑃𝑃𝐿𝐿,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖

𝐼𝐼𝐷𝐷𝐷𝐷 = , (3.4)
𝑉𝑉𝐷𝐷𝐷𝐷
where 𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 and 𝑃𝑃𝐿𝐿,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 are the always positive losses of motor and inverter, 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚ℎ
is the mechanical power, and 𝑉𝑉𝐷𝐷𝐷𝐷 is the current battery voltage. Figure 30 shows the im-
plementation of Equation (3.4) in Simulink.
3 Methodology
Figure 30: Calculation of DC current
3.1.5 MOTOR THERMAL MODEL

The approach for the thermal modeling was a lumped element model. This can be easily
parameterized or reduced to fit the available parameters and is sufficiently fast for real
time operation. The proposed model assumed constant temperature in the axial direction
of the machine (i.e. all lumped elements are in one radial plane), except the thermal re-
sistance to the shaft, which also contains an axial component. Therefore, the application
of the model is limited to radial flux machines. Figure 31 gives an overview of the pro-
posed network, including the mapping of the dissipated losses. The blue nodes represent
the mass points and discrete temperatures of the model, and the arrows show the flow of
energy.
Figure 31: Overview of the motor thermal model
3 Methodology
The model consists of six concentrated mass nodes, two constant temperature nodes and
eight thermal resistances between these nodes. Figure 32 shows the thermal network with
the mentioned parameters. The two nodes Tshaft and Tenvironment are boundaries to the sur-
rounding model and were assumed constant.
Tenvironment
Thousing
Tcoolant
Tteeth
Trotor
Tyoke
Tshaft
Rth1 Rth2 Rth3 Rth4 Rth5 Rth6
PL,yoke
PL,rotor PL,teeth
Rth7 Rth8
Twinding
PL,winding
Figure 32: Lumped element model for motor
The mathematical representation for thermal conduction discussed in Section 2.2.4 is

based on Equations (2.15) and (2.16). Therefore, the individual temperatures at the nodes
are calculated by
1
𝑇𝑇𝑛𝑛 = ∫ (𝑃𝑃𝐿𝐿𝐿𝐿 − 𝑞𝑞̇ 𝑛𝑛 )𝑑𝑑𝑑𝑑 + 𝑇𝑇0𝑛𝑛 , (3.5)
𝑚𝑚𝑛𝑛 ⋅ 𝐶𝐶𝑝𝑝𝑝𝑝
with 𝑛𝑛 indicating the individual nodes, 𝑇𝑇0𝑛𝑛 the initial temperature, 𝑚𝑚𝑛𝑛 the active thermal
mass, and 𝐶𝐶𝑝𝑝𝑝𝑝 the specific heat capacity for each node. The rate of energy coming from
the losses 𝑃𝑃𝐿𝐿𝐿𝐿 was zero for all elements except rotor, teeth, yoke and winding. According
to the network in Figure 32, the conductive heat flux 𝑞𝑞̇ 𝑛𝑛 is the sum of two or more heat
flux components.
The iron and other losses in each operating point, determined in Section 3.1.3, are distrib-
uted to three mass points (Figure 32) for the rotor, stator teeth and stator yoke. Due to the
simplicity of the proposed lumped parameter model, the iron, ventilation, and friction
losses are mapped onto those three nodes. Since the distribution is highly affected by the
rotational speed of the motor, the model requires a table with the proportions of losses
occurring in the rotor, stator teeth and yoke over the machine’s operating speed range. As
3 Methodology
an example, Figure 33 shows the effect of motor speed on different sources of iron loss
[17].
Figure 33: Sources of iron losses over rotor speed in an IPM machine [17]
The model of the water cooling was designed following the approach in [18]. There, spa-
tially constant temperatures on the motor and housing surface, and constant liquid density
and heat capacity are assumed. The average temperature of the liquid in the motor was
assumed as
𝑇𝑇𝑤𝑤𝑤𝑤 + 𝑇𝑇𝑤𝑤𝑤𝑤
𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = , (3.6)
2
where 𝑇𝑇𝑤𝑤𝑤𝑤 and 𝑇𝑇𝑤𝑤𝑤𝑤 are the liquid temperatures at the inlet and outlet of the cooling jacket,
respectively. The resulting heat balance equation was
𝑑𝑑𝑇𝑇𝑤𝑤𝑤𝑤
𝑤𝑤𝐶𝐶𝑝𝑝 = 𝑤𝑤̇ 𝐶𝐶𝑝𝑝 (𝑇𝑇𝑤𝑤𝑤𝑤 − 𝑇𝑇𝑤𝑤𝑤𝑤 ) + 𝑞𝑞̇ 𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦2𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 − 𝑞𝑞̇ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐2ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 , (3.7)
𝑑𝑑𝑑𝑑
where 𝑤𝑤 is the mass of liquid in the cooling jacket, 𝑤𝑤̇ the liquid mass flow, 𝐶𝐶𝑝𝑝 the liquid
specific heat capacity, and 𝑞𝑞̇ the heat flux from the adjacent nodes, stator yoke and hous-
ing. These were calculated with
1
𝑞𝑞̇ 𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦2𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = �𝑇𝑇 − 𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 �
𝑅𝑅𝑡𝑡ℎ4 𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦
(3.8)
1
𝑎𝑎𝑎𝑎𝑎𝑎 𝑞𝑞̇ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐2ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 = �𝑇𝑇 − 𝑇𝑇ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 �
𝑅𝑅𝑡𝑡ℎ5 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑡𝑡
3 Methodology
with temperatures and thermal resistances according to Figure 32. The full motor thermal
model that was implemented in Simulink can be found in APPENDIX B.
3.1.6 INVERTER THERMAL MODEL

Similar to the motor thermal model, the inverter thermal model is a lumped parameter
type. Figure 34 illustrates the model layout that consists of four discrete mass points (blue
nodes), where one of them represents the liquid cooling, and the constant environment
temperature (black node). The dissipated heat is mapped onto the node for the semicon-
ductor junction. The yellow arrows show the flow of energy between the nodes. Most of
the energy leaves the system through the coolant outlet.
Figure 34: Overview of the inverter thermal model
The mathematical definition is similar to the motor thermal model, and based on Equa-
tions (3.5)-(3.8). Figure 35 shows the thermal lumped parameter network for the inverter.
Tenvironment
Theatsink
Tjunction
Thousing
Tcoolant
Rth1 Rth2 Rth3 Rth4
PL,inverter
Figure 35: Lumped element model for inverter
The full inverter thermal model that was implemented in Simulink can be found in AP-
PENDIX C.
3 Methodology
3.2 PARAMETERIZATION
An essential part of the project was deriving the parameters for the model from real data
and measurements. The parameters listed below are general parameters, which can be
obtained from the motor datasheet or by conducting simple measurements.
• Rated Torque
maximum motor torque in constant torque range
• Rated Speed
maximum motor speed at full torque (start of field weakening)
• Maximum Speed
absolute maximum motor speed
• Number of Phases
number of stator phases
• Stator Winding Resistance
mean stator winding resistance and winding temperature at measurement
Further necessary parameters and their determination are explained in the following sec-
tions.
3.2.1 ELECTROMECHANICAL PARAMETERS

Parameterization of the electromechanical part is done primarily with the general param-
eters listed above. Below are additional parameters for this particular part.
• Minimum DC voltage for normal operation

minimum supply voltage where motor can run at full power
• Torque delay
time constant of delay between torque demand and actual motor torque
Similar to the general parameters, the minimum supply voltage for normal operation
should be given by the motor or inverter manufacturer. The motor torque delay can be
measured by using a torque demand step as the controller input and measuring the result-
ing motor torque response. The time constant is the time from the input step until the
output response reaches 63% of the demand value.
3 Methodology
3.2.2 MOTOR LOSS PARAMETERS

For calculations of the motor losses, the motor efficiency, stator current and winding tem-
perature needed to be measured for a number of operating points covering the full oper-
ating range of the machine. Figure 36 shows an example distribution of measurements.
120
80
Motor Torque / Nm
40
-40
-80
-120
0 5 10 15 20 25 30 35 40 45 50
Rotational Speed / krpm
Figure 36: Example distribution of measurements over motor operating range
From this set of measurements, the motor losses 𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 were calculated with
(1 − 𝜂𝜂𝑀𝑀 )
𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 = 𝑇𝑇 ⋅ 𝜔𝜔 ⋅ 𝑓𝑓𝑓𝑓𝑓𝑓 𝑇𝑇 > 0 𝑜𝑜𝑜𝑜
𝜂𝜂𝑀𝑀 (3.9)
𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 = −𝑇𝑇 ⋅ 𝜔𝜔 ⋅ (1 − 𝜂𝜂𝑀𝑀 ) 𝑓𝑓𝑓𝑓𝑓𝑓 𝑇𝑇 < 0,
depending on the flow of energy in each operating point. 𝑇𝑇 is the motor torque, 𝜔𝜔 is the
angular frequency, and 𝜂𝜂𝑀𝑀 is the measured efficiency. The motor losses could then be
separated into copper losses and other losses which include iron, ventilation and friction
losses. Using Equation (3.2) to obtain the copper losses 𝑃𝑃𝑐𝑐𝑐𝑐 for each operating point, the
other losses 𝑃𝑃𝐿𝐿,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒 were calculated with
𝑃𝑃𝐿𝐿,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒 = 𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 − 𝑃𝑃𝑐𝑐𝑐𝑐 . (3.10)
The acquired values for stator current and iron and other losses at each point of measure-
ment had to be converted into a continuous plane across the operating range and then
discretized to be used in the Simulink model (Figure 37 and Figure 38). The Matlab code
3 Methodology
for the generation of the necessary maps from measurement data can be found in APPEN-
DIX A.
160
160 160 16/0 A
Stator Current
120 140 140 14 0
120 120 12 0
100 100
80 10 0
14 0
80 80
60 60
40
Torque / Nm
40 40 60
12
0
20 20
80
0
20 20 40
12 10 0
-40 40 40 60
0
60 60
80
80 80
-80 10 0
100 100
120 120 12 0
-120 140 140 140
160 160 16 0
-160
0 1000 2000 3000 4000 5000 6000 7000 8000
Rotor-Speed / rpm
Figure 37: Example stator current map generated from measurements
160
Iron and Other Losses / W
120
22 50
10 00
20 0
12 50
25 00
15 00
50 0
17 5
75 0
0
0
30 0
0
80 27 50
40
Torque / Nm
0
20 00
1000
1250
15 00
22 50
17 50
25 00
750
500
0
27 5
-40
-80 00
30
-120
27 50
17 50
1250
1500
1000
25 00
20 00
22 50
750
50 0
-160
0 1000 2000 3000 4000 5000 6000 7000 8000
Rotor Speed / rpm
Figure 38: Example iron loss map generated from measurements
3 Methodology
3.2.3 INVERTER LOSS PARAMETERS

Similar to the motor loss parameters, inverter efficiency measurements across the full
operating range were required. The inverter losses were calculated and then interpolated
and discretized for use in the Simulink model using
(1 − 𝜂𝜂𝐼𝐼 )
𝑃𝑃𝐿𝐿,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = �𝑇𝑇 ⋅ 𝜔𝜔 + 𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 � ⋅ 𝑓𝑓𝑓𝑓𝑓𝑓 𝑇𝑇 > 0
𝜂𝜂𝐼𝐼 (3.11)
𝑎𝑎𝑎𝑎𝑎𝑎 𝑃𝑃𝐿𝐿,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = �−𝑇𝑇 ⋅ 𝜔𝜔 + 𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 � ⋅ (1 − 𝜂𝜂𝐼𝐼 ) 𝑓𝑓𝑓𝑓𝑓𝑓 𝑇𝑇 < 0
where 𝜂𝜂𝐼𝐼 is the measured inverter efficiency at each operating point and 𝑃𝑃𝐿𝐿,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 is the
calculated motor losses. Figure 39 shows an example map of iron losses for the full op-
erating range.
160
1600
1400
1600 Inverter
14 00 Losses /0W
0
1400 12 00 10
120 1200 12 00 80 0
10 00 10 00
80 0 60 0
80 80 0
600 40 0
600
40 400 400
60
Torque / Nm
0
20 0 20 0
20 0
0 40 0
200 200
200
-40 400 400
40 0
60 0
600 600
-80 800 60 0
80 0
1000 10 00 80 0
-120 1200 1200
12 00 10 00
1400 1400
14 00
1600 1600
-160
0 1000 2000 3000 4000 5000 6000 7000 8000
Rotor-Speed / rpm
Figure 39: Example inverter loss map from measurements
3 Methodology
3.2.4 MOTOR THERMAL PARAMETERS

Parameters for the motor lumped element model include thermal resistances, heat capac-
ities and masses (Figure 32). In addition, the distribution of losses to rotor, teeth and stator
yoke and the initial temperature for each node are required to fully parameterize the
model. All necessary parameters are listed below.
• Thermal resistances
 Shaft – Rotor (𝑅𝑅𝑡𝑡ℎ1 )
 Rotor – Stator teeth (𝑅𝑅𝑡𝑡ℎ2 )
 Stator teeth – Stator yoke (𝑅𝑅𝑡𝑡ℎ3 )
 Stator yoke – Cooling (𝑅𝑅𝑡𝑡ℎ4 )
 Cooling – Housing (𝑅𝑅𝑡𝑡ℎ5 )
 Housing – Environment (𝑅𝑅𝑡𝑡ℎ6 )
 Winding – Stator teeth (𝑅𝑅𝑡𝑡ℎ7 )
 Winding – Stator yoke (𝑅𝑅𝑡𝑡ℎ8 )
• Specific heat capacities, masses and initial temperatures
 Rotor (𝐶𝐶𝑝𝑝 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 , 𝑚𝑚𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 , 𝑇𝑇0 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 )
 Stator teeth (𝐶𝐶𝑝𝑝 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡ℎ , 𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡ℎ , 𝑇𝑇0 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡ℎ )
 Stator yoke (𝐶𝐶𝑝𝑝 𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦 , 𝑚𝑚𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦 , 𝑇𝑇0 𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦 )
 Stator winding (𝐶𝐶𝑝𝑝 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 , 𝑚𝑚𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 , 𝑇𝑇0 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 )
 Cooling liquid (𝐶𝐶𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 , 𝑚𝑚𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 , 𝑇𝑇0 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 )
 Housing (𝐶𝐶𝑝𝑝 ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 , 𝑚𝑚ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 , 𝑇𝑇0 ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 )
• Iron loss distribution between rotor, stator teeth and stator yoke
Since the thermal conductivities and iron loss distribution are not easy to measure, ways
to analytically calculate these parameters based physical data are explained in the follow-
ing sections. The number of nodes in the model can also be reduced to the desired size by
simply setting the unsuitable thermal resistances and masses to zero.
3.2.4.1 THERMAL RESISTANCES

The thermal resistances used in Figure 32 can be calculated using the formulas proposed
in [19]. The necessary geometric dimensions are illustrated in Figure 40.
3 Methodology
Figure 40: Definition of motor dimensions for thermal parameter calculation according to [19]
Rth1 – Shaft to Rotor

𝑅𝑅𝑡𝑡ℎ1 represents the thermal resistance between the rotor and the shaft in radial direction
as well as the resistance of the shaft for heat conduction outside of the engine. It is calcu-
lated by
1 0.5�𝑙𝑙𝑠𝑠ℎ𝑓𝑓 − 𝑙𝑙𝑆𝑆 � 1 0.5𝑙𝑙𝑆𝑆 1 𝑟𝑟𝑚𝑚𝑚𝑚

𝑅𝑅𝑡𝑡ℎ1 = � 2 �+ � 2 �+ ln � � (3.12)
2 𝑘𝑘𝑖𝑖𝑖𝑖 𝜋𝜋𝑟𝑟𝑖𝑖𝑟𝑟 4 𝑘𝑘𝑖𝑖𝑖𝑖 𝜋𝜋𝑟𝑟𝑖𝑖𝑖𝑖 2𝜋𝜋𝑘𝑘𝑖𝑖𝑖𝑖 𝑙𝑙𝑆𝑆 𝑟𝑟𝑖𝑖𝑖𝑖
with 𝑙𝑙𝑠𝑠ℎ𝑓𝑓 as the length of the shaft, 𝑙𝑙𝑆𝑆 the length of the stator and 𝑘𝑘𝑖𝑖𝑖𝑖 the iron thermal
conduction coefficient. The first, second, and third term in the equation represents the
resistance of the external part of the shaft, the shaft inside the rotor, and the resistance of
the inner part of the rotor, respectively.
Rth2 – Rotor to Stator Teeth (air gap)

The convective heat transfer in the air gap follows the Taylor-Couette Flow for fluids
between two concentric rotating cylinders. The following describes the calculation
method according to [19] and [20].
The most common definition for the Taylor number is
3 Methodology
𝑟𝑟𝑖𝑖𝑖𝑖 −𝑟𝑟𝑜𝑜𝑜𝑜 3
𝜔𝜔2 𝑟𝑟𝑜𝑜𝑜𝑜 � � (3.13)
2
𝑁𝑁𝑇𝑇𝑇𝑇 = ,
𝜈𝜈 2
Where 𝜔𝜔 is the angular velocity of the rotor, 𝑟𝑟𝑜𝑜𝑜𝑜 the outer radius of the rotor, 𝑟𝑟𝑖𝑖𝑖𝑖 the inner
radius of the stator, and 𝜈𝜈 the kinematic viscosity of the fluid in the air gap. At low speeds
(𝑁𝑁𝑇𝑇𝑇𝑇 < 1700), the Nusselt number 𝑁𝑁𝑁𝑁𝑁𝑁 is set constant to 2.2. For higher speeds (𝑁𝑁𝑇𝑇𝑇𝑇 >
1700), the Nusselt number is defined as
𝑁𝑁𝑁𝑁𝑁𝑁 = 0.23 𝑁𝑁0.63 0.27

𝑇𝑇𝑇𝑇 ⋅ 𝑁𝑁𝑃𝑃𝑃𝑃 , (3.14)
with the Prandtl number 𝑁𝑁𝑃𝑃𝑃𝑃 = 𝜈𝜈/𝑎𝑎, where 𝑎𝑎 is the thermal diffusivity of the fluid. Once
the Nusselt number is obtained, the convective heat transfer coefficient can be calculated
by
𝑁𝑁𝑁𝑁𝑁𝑁 ⋅ 𝑘𝑘𝑎𝑎𝑎𝑎𝑎𝑎
ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 = , (3.15)
𝑟𝑟𝑖𝑖𝑖𝑖 − 𝑟𝑟𝑜𝑜𝑜𝑜
where 𝑘𝑘𝑎𝑎𝑎𝑎𝑎𝑎 is the air thermal conductivity. Ultimately, the thermal resistance can be cal-
culated with
𝑟𝑟𝑖𝑖𝑖𝑖 +𝑟𝑟𝑖𝑖𝑖𝑖
1 𝑟𝑟𝑜𝑜𝑜𝑜 1 1
𝑅𝑅𝑡𝑡ℎ2 = ln � � + 𝑟𝑟𝑜𝑜𝑜𝑜 +𝑟𝑟𝑖𝑖𝑖𝑖 + ln � 2 �. (3.16)
2𝜋𝜋𝑘𝑘𝑖𝑖𝑖𝑖 𝑙𝑙𝑆𝑆 𝑟𝑟𝑚𝑚𝑚𝑚 2𝜋𝜋ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑙𝑙𝑆𝑆 2𝜋𝜋𝑘𝑘 𝑙𝑙 𝑝𝑝
𝑖𝑖𝑖𝑖 𝑆𝑆 𝑖𝑖𝑖𝑖 𝑟𝑟𝑖𝑖𝑖𝑖
2
The equation is the sum of three terms. The first term is the conductive thermal resistance
of the proportional part of the rotor between mass point and air gap. This has to be adapted
according to the actual geometry of the rotor, including additional air gaps and permanent
magnets. The second term represents the convective thermal resistance of the air gap. The
third term is the conductive thermal resistance of the proportional part of the stator teeth
between the mass point for teeth and air gap, where 𝑝𝑝𝑖𝑖𝑖𝑖 is the percentage of the stator
teeth volume with respect to the total teeth and slots volume.
Rth3 –Stator Teeth to Stator Yoke

Assuming the node for the teeth is in the center of the teeth, and the node for the yoke is
in the radial center of the yoke, the thermal resistance between these nodes can be ob-
tained by
3 Methodology
1 𝑟𝑟𝑖𝑖𝑖𝑖 1 𝑟𝑟𝑚𝑚
𝑅𝑅𝑡𝑡ℎ3 = ln � 𝑟𝑟𝑖𝑖𝑖𝑖 +𝑟𝑟𝑖𝑖𝑖𝑖 � + ln � �. (3.17)
2𝜋𝜋𝑘𝑘𝑖𝑖𝑖𝑖 𝑙𝑙𝑆𝑆 𝑝𝑝𝑖𝑖𝑖𝑖 2𝜋𝜋𝑘𝑘𝑖𝑖𝑖𝑖 𝑙𝑙𝑆𝑆 𝑟𝑟𝑖𝑖𝑖𝑖
2
The first term of the equation is half of the thermal resistance of the teeth, and the second
term is the thermal resistance of the inner half of the stator yoke.
Rth4 –Stator Yoke to Coolant

The thermal resistance between the stator yoke node and the coolant can be calculated
with
1 𝑟𝑟𝑜𝑜𝑜𝑜 1
𝑅𝑅𝑡𝑡ℎ4 = ln � � + . (3.18)
2𝜋𝜋𝑘𝑘𝑖𝑖𝑖𝑖 𝑙𝑙𝑆𝑆 𝑟𝑟𝑚𝑚 2𝜋𝜋ℎ𝑦𝑦𝑦𝑦 𝑟𝑟𝑜𝑜𝑜𝑜 𝑙𝑙𝐶𝐶
The first term of the equation is the outer half of the stator yoke thermal resistance. The
second part is the thermal resistance between stator yoke and coolant, where ℎ𝑦𝑦𝑦𝑦 is the
conductive heat transfer coefficient between stator yoke and cooling liquid and 𝑙𝑙𝐶𝐶 is the
axial length of the cooling jacket. Since ℎ𝑦𝑦𝑦𝑦 is largely dependent on the fluid dynamics
in the cooling jacket, there is no easy way to calculate it analytically. The DC supply test
(see page 43) is a way to obtain it by measurement.
Rth5 – Coolant to Housing

The thermal resistance between the cooling liquid and the motor housing is calculated
with
1
𝑅𝑅𝑡𝑡ℎ5 = , (3.19)
2𝜋𝜋ℎ𝑐𝑐ℎ (𝑟𝑟𝑜𝑜𝑜𝑜 + 𝑑𝑑𝑐𝑐 )𝑙𝑙𝐶𝐶
where ℎ𝑐𝑐ℎ is the conductive heat transfer coefficient from coolant to housing and 𝑑𝑑𝑐𝑐 the
width of the cooling jacket. Like ℎ𝑦𝑦𝑦𝑦 , ℎ𝑐𝑐ℎ is best obtained by the DC supply test.
Rth6 – Housing to Environment

Assuming natural convection outside of the motor, the thermal resistance between the
motor housing and the environment can be obtained by the DC supply test with
3 Methodology
𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑒𝑒 − 𝑇𝑇𝑎𝑎𝑎𝑎𝑎𝑎
𝑅𝑅𝑡𝑡ℎ6 = , (3.20)
𝑃𝑃𝐷𝐷𝐷𝐷
where 𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 is the temperature of the case, 𝑇𝑇𝑎𝑎𝑎𝑎𝑎𝑎 is the ambient temperature and 𝑃𝑃𝐷𝐷𝐷𝐷 is the
heat flux from housing to environment, all obtained from the DC supply test.
Rth7, Rth8 – Stator Winding to Stator Teeth and Yoke

The thermal resistance between stator winding and stator core depends both on the fill
coefficient of the winding in the slot and the insulation technique. The overall conductive
resistance from winding to iron can be obtained by
𝑑𝑑𝑒𝑒𝑒𝑒
𝑅𝑅𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖 = , (3.21)
𝑘𝑘𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖 𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
with 𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 as the contact area between winding and stator iron (𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = 𝑙𝑙𝑠𝑠𝑠𝑠 𝑙𝑙𝑆𝑆 ), 𝑘𝑘𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖 the
equivalent conductivity coefficient of the air and insulation material in the stator slots
(evaluated by the DC supply test), and 𝑑𝑑𝑒𝑒𝑒𝑒 as the equivalent thickness of the air and in-
sulation material of the stator slots, which is calculated with
𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 − 𝑆𝑆𝑐𝑐𝑐𝑐
𝑑𝑑𝑒𝑒𝑒𝑒 = , (3.22)
𝑙𝑙𝑠𝑠𝑠𝑠
where 𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 is the cross-sectional area of the stator slot, 𝑆𝑆𝑐𝑐𝑐𝑐 is the cross-sectional area of
copper in the stator slot, and 𝑙𝑙𝑠𝑠𝑠𝑠 is the perimeter of the stator slot. The obtained overall
heat resistance can then be split up into the teeth and yoke parts considering the thermal
admittance, 𝐺𝐺𝑡𝑡ℎ = 1/𝑅𝑅𝑡𝑡ℎ , with
𝐺𝐺𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖 = 𝐺𝐺𝑡𝑡ℎ7 + 𝐺𝐺𝑡𝑡ℎ8 . (3.23)
The teeth and yoke parts are proportional to the cross-sectional area of the heat transfer,
and therefore have the relationship
𝐴𝐴𝑐𝑐𝑐𝑐,𝑡𝑡ℎ
𝐺𝐺𝑡𝑡ℎ7 = 𝐺𝐺
𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖
(3.24)
𝐴𝐴𝑐𝑐𝑐𝑐,𝑦𝑦𝑦𝑦
𝑎𝑎𝑎𝑎𝑎𝑎 𝐺𝐺𝑡𝑡ℎ8 = 𝐺𝐺
𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖
3 Methodology
where 𝐴𝐴𝑐𝑐𝑐𝑐,𝑡𝑡ℎ is the contact area between winding and stator teeth, and 𝐴𝐴𝑐𝑐𝑐𝑐,𝑦𝑦𝑦𝑦 is the contact
area between winding and stator yoke. Ultimately the thermal resistances can be calcu-
lated as
𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
𝑅𝑅𝑡𝑡ℎ7 = 𝑅𝑅
𝐴𝐴𝑐𝑐𝑐𝑐,𝑡𝑡ℎ 𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖
(3.25)
𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
𝑎𝑎𝑎𝑎𝑎𝑎 𝑅𝑅𝑡𝑡ℎ8 = 𝑅𝑅 .
𝐴𝐴𝑐𝑐𝑐𝑐,𝑦𝑦𝑦𝑦 𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖
DC Supply Test
As mentioned previously, the parameters ℎ𝑦𝑦𝑦𝑦 , ℎ𝑐𝑐ℎ , 𝑅𝑅𝑡𝑡ℎ6 and 𝑘𝑘𝑐𝑐𝑐𝑐,𝑖𝑖𝑖𝑖 can be obtained with
a DC supply test, where the rotor is at a standstill and the only source of losses are copper
losses defined by 𝑃𝑃 = 𝐼𝐼 2 𝑅𝑅. The DC supply measurement proposed in [19] with adapta-
tions to comply with the thermal model used in this thesis is explained in the following
sections.
The motor, at zero speed and with series or parallel connected windings, is supplied with
a DC current (50% to 70% of the rated current). In thermal steady-state conditions, the
temperatures of stator windings, stator teeth, stator yoke, coolant, housing and ambient
have to be measured. By solving the proposed thermal network (Figure 32, page 31), the
desired parameters can be calculated.
3.2.4.2 IRON AND OTHER LOSS DISTRIBUTION

The distribution of iron losses within an electric machine is highly affected by the type of
motor, the motor’s geometry and the materials used. Ways for determining these losses
are using finite element analysis or taking thermal measurements (requires a motor with
several temperature probes in the laminations). These approaches are reported in [17, 21]
for an IPM machine. The model requires data for the proportional iron losses in rotor,
stator teeth and stator yoke over the whole operating speed range. If neither FEA nor
thermal measurements are feasible, the data must be provided by the motor manufacturer.
3 Methodology
3.2.5 INVERTER THERMAL PARAMETERS

Parameters for the inverter lumped element model include thermal resistances, heat ca-
pacities and masses (Figure 35). The necessary parameters are listed below.
• Thermal resistances
 Junction – Heat sink (𝑅𝑅𝑡𝑡ℎ1 )
 Heat sink – Coolant (𝑅𝑅𝑡𝑡ℎ2 )
 Coolant – Housing (𝑅𝑅𝑡𝑡ℎ3 )
 Housing – Environment (𝑅𝑅𝑡𝑡ℎ4 )
• Specific heat capacities, masses and initial temperatures
 Junction (𝐶𝐶𝑝𝑝 𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗 , 𝑚𝑚𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗 , 𝑇𝑇0𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗 )
 Heat sink (𝐶𝐶𝑝𝑝 ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 , 𝑚𝑚ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 , 𝑇𝑇0 ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 )
 Coolant (𝐶𝐶𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 , 𝑚𝑚𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 , 𝑇𝑇0 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 )
 Housing (𝐶𝐶𝑝𝑝 ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 , 𝑚𝑚ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 , 𝑇𝑇0 ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 )
Since the layout and components used in inverters greatly affect these thermal parameters,
detailed knowledge about the inverter’s structure is necessary to obtain them. Per the
model requirements, the user of this model will not have this information. The listed pa-
rameters need to be provided by the inverter manufacturer.
Another way to obtain the thermal resistances is by conducting measurements at a thermal

steady state and calculating them with
Δ𝑇𝑇
𝑅𝑅𝑡𝑡ℎ,𝑛𝑛 = , (3.26)
𝑞𝑞̇
where Δ𝑇𝑇 is the temperature difference and 𝑞𝑞̇ the heat flux between two nodes. This re-
quires temperature sensors for each node of the lumped parameter network.
3.3 MODEL VERIFICATION

To prove the operation of the designed model, an existing electric drive was modelled
and the simulation results were compared with the measurements. The following sections
describe the properties of the drive, the parameterization process, and the comparison of
simulations and measurements.
3 Methodology
3.3.1 ELECTRIC DRIVE TEST SYSTEM

The electric drive system used as a reference was the UQM PowerPhase 75 (Figure 41),
a surface mounted permanent magnet synchronous motor with concentrated windings,
supplied by a three phase half bridge inverter with insulated gate bipolar transistors
(IGBTs).
Figure 41: UQM PowerPhase 75 drive system [22]
Table 2 gives an overview of the main drive characteristics. It shows that the motor is
operated in field weakening mode from 2875rpm to 8000rpm (i.e. can provide the maxi-
mum power in more than 50% of its operating range).
Table 2: UQM PowerPhase 75 motor characteristics
Base Speed 2875rpm

Maximum Speed 8000rpm
Base Torque 150Nm
Base Power 45kW
Peak Torque 240Nm
Peak Power 75kW
Maximum Efficiency 94%
Minimum DC Voltage 240V
Cooling Liquid 50/50 Glycol-Water
Number of Phases 3
3 Methodology
3.3.2 PARAMETER GENERATION

Following the guidelines in Section 3.2, the parameters for the model were obtained. Most
of the general parameters are listed in Table 2 and can be found in the motor datasheet
[22]. Unfortunately, several of the measurements required in Section 3.2 could not be
obtained, since the test system was destroyed during the tests. The following text explains
how all parameters were obtained, either by following the guidelines from Section 3.2 or
making appropriate assumptions in order to obtain a value.
To obtain the mean stator winding resistance, the resistances between the three phases
(𝑅𝑅𝑈𝑈−𝑉𝑉 , 𝑅𝑅𝑉𝑉−𝑈𝑈 , 𝑅𝑅𝑊𝑊−𝑈𝑈 ) were measured at a constant temperature of 25°C (Table 3) and then
the mean resistance was calculated using Equation (3.27).
Table 3: Measured stator winding resistances
RU-V RV-W RW-U

mΩ mΩ mΩ
27.02 27.38 26.48
𝑅𝑅𝑈𝑈−𝑉𝑉 + 𝑅𝑅𝑉𝑉−𝑊𝑊 + 𝑅𝑅𝑊𝑊−𝑈𝑈

𝑅𝑅𝑆𝑆 = = 13,48𝑚𝑚Ω @ 25°C (3.27)
6
3.3.2.1 ELECTROMECHANICAL PARAMETERS

The minimum DC voltage for normal operation was obtained from the motor datasheet
and is 240 volts. From this base value, a power derating table was generated (Figure 24,
page 26).
For the motor’s torque response, a minimum of one step response needs to be measured
on a test bench. This was not possible due to damage on the test motor. Consequently, an
average value based on prior measurements done by the author in laboratory classes at
the Electric Drives and Machines Institute at Graz University of Technology was as-
sumed. These measurements do not perfectly apply to the UQM drive system, but are
sufficiently accurate to prove the validity of this model. Table 4 shows the used electro-
mechanical parameters.
3 Methodology
Table 4: Electromechanical parameters for the UQM PowerPhase 75
VDC_min τ
V ms
240 5.0
3.3.2.2 MOTOR AND INVERTER LOSS PARAMETERS

The motor loss parameters were obtained from measurements of steady state operating
points across the whole operating range. The number of measured operating points was
123, evenly distributed to the motoring and generating mode. Figure 42 shows the meas-
ured operating points.
160
120
80
Torque / Nm
40
0
-40
-80
-120
-160
0 1000 2000 3000 4000 5000 6000 7000 8000
Rotational Speed / rpm
Figure 42: Measured operating points for UQM PowerPhase 75
For each operating point, the following components were measured:

• Motor speed
• Motor torque
• Motor efficiency
• Inverter efficiency
• Average stator phase current (RMS)
• Winding temperature.
With these values, the model parameters for the motor and inverter loss calculation were
generated following Sections 3.2.2 and 3.2.3. The stator current, iron loss, and inverter
3 Methodology
loss maps generated from these measurements are shown in Figure 37, Figure 38 and
Figure 39 (pp. 36-37), respectively.
3.3.2.3 MOTOR THERMAL PARAMETERS

For the thermal parameters, the guidelines from Section 3.2.4 could not be followed be-
cause neither the required geometric data nor the thermal measurements were available.
Therefore, a rough estimation of the geometry of the motor was used to calculate a base
set of thermal resistances and masses. The specific heat capacities were based on typical
material used in this type of machine. These parameters were then adapted and changed
to match the available measurements. Table 5, Table 6 and Table 7 list the resulting pa-
rameters for the motor thermal model.
Table 5: Thermal resistances for UQM PowerPhase 75 Motor
Rth1 Rth2 Rth3 Rth4 Rth5 Rth6 Rth7 Rth8

K/W K/W K/W K/W K/W K/W K/W K/W
0.2 0.392 0.03 0.005 0.0145 0.3 0.023 0.053
Table 6: Specific heat capacities for UQM PowerPhase 75 Motor
Cp rotor Cp teeth Cp yoke Cp winding Cp coolant Cp housing

J/(kgK) J/(kgK) J/(kgK) J/(kgK) J/(kgK) J/(kgK)
500 500 500 385 4181 897
Table 7: Masses for UQM PowerPhase 75 Motor
mrotor mteeth myoke mwinding mcoolant mhousing

kg kg kg kg kg kg
7 5 12 11 2 2
3.3.2.4 INVERTER THERMAL PARAMETERS

Similar to the motor thermal parameters, there was insufficient data available to obtain
the inverter thermal parameters according to Section 3.2.5. Again, values were assumed
and adapted to fit the measurements. Table 8, Table 9 and Table 10 show the resulting
parameters.
3 Methodology
Table 8: Thermal resistances for UQM PowerPhase 75 Inverter
Rth1 Rth2 Rth3 Rth4

K/W K/W K/W K/W
0.013 0.015 0.08 0.3
Table 9: Specific heat capacities for UQM PowerPhase 75 Inverter
Cp junction Cp heatsink Cp coolant Cp housing

J/(kgK) J/(kgK) J/(kgK) J/(kgK)
80 897 4181 897
Table 10: Masses for UQM PowerPhase 75 Inverter
mjunction mheatsink mcoolant mhousing

kg kg kg kg
0.05 1.5 1.8 0.5
3 Methodology
3.3.3 COMPARISON OF MODEL AND ELECTRIC DRIVE MEASUREMENTS

After the model was fully parameterized, the model’s behavior was compared to the real
drive system. The measurements from the UQM PowerPhase 75 used for this comparison
are from an automated test run covering a large part of its operating range. Because this
was the only test run available for this motor, not all features of the model could be eval-
uated. The test run is illustrated in Figure 43.
9000
Engine Speed
6000
[rpm]
3000
0 200
100
Engine Torque
[Nm]
0
100 -100
Mechanical Power
50 -200
[kW]
-50
-100
0 500 1000 1500 2000 2500 3000 3500 4000
Time [s]
Figure 43: Test run of UQM PowerPhase 75 drive
The measurement was done at a frequency of 10Hz (i.e. a time step of 100ms), which is
sufficient for showing the capabilities of the thermal model. However, the transient elec-
tromechanical behavior could not be evaluated since the time constant for the torque delay
is smaller than the time between two measurements.
The temperatures of the motor coolant at inlet and outlet are illustrated in Figure 44, com-
paring the measurements with the simulation results. The measured inlet temperature was
used as an input for the simulation. The figure shows that the simulated outlet temperature
matches very well with the measurement. The largest difference between measurements
and model can be found in the beginning, where the thermal model takes additional time
to equilibrate. There is also a discrepancy between 2000s and 2500s because around
3 Methodology
2100s the motor was operated in overload condition. Overload operation has not been
modelled for this test due to the lack of measurements outside the continuous operating
range. Besides the coolant temperatures, only the stator winding temperature was meas-
ured accurately and could be compared with the simulation results. Figure 45 shows the
measured and simulated stator winding temperatures. The simulation matches the meas-
urements well with exception to the range around 2500s. A possible explanation for this
deviation could be insufficient accuracy when generating the lookup table for the copper
losses.
58
Tcoolant in
56 Tcoolant out measurement
Temperature / °C
Tcoolant out model

54
52
50
48
0 500 1000 1500 2000 2500 3000 3500 4000
Time / s
Figure 44: Comparison of motor coolant temperatures
110
100
Temperature / °C
90
80
70 Twinding measurement
Twinding model
60
0 500 1000 1500 2000 2500 3000 3500 4000
Time / s
Figure 45: Comparison of stator winding temperatures
The coolant temperatures at the inverter inlet and outlet are illustrated in Figure 46. Sim-
ilar to the motor simulation, the measured inlet coolant temperature was used as an input
3 Methodology
for the model. The simulated outlet coolant temperature matches well with the measure-
ment. The simulation deviates from the measurement in the range between 2000s and
2300s, which is probably the result of a measurement error. After reviewing the test setup
and talking to the test rig operators, a possible cause could be additional air convection
or vibration at high rotational speeds. Figure 47 compares the DC current from measure-
ment with the simulation result, which matches well. It also shows the overload operation
between 1800s and 2300s.
57
Tcoolant in
56
Tcoolant out measurement
55
Tcoolant out model
Temperature / °C
54
53
52
51
50
49
0 500 1000 1500 2000 2500 3000 3500 4000
Time / s
Figure 46: Comparison of inverter coolant temperatures
200
IDC measurement
150
IDC model
100
Current / A
50
-50
-100
-150
0 500 1000 1500 2000 2500 3000 3500 4000
Time / s
Figure 47: Comparison of DC currents
3 Methodology
The mean deviations between the measurement and simulation were calculated, and are
listed in Table 11. The simulation results match well with the measurements. Considering
the temperature range of the coolant and the winding temperature, it makes sense that the
mean deviation for the winding temperature is significantly higher than for the coolant
temperatures.
Table 11: Mean deviations between simulation and measurement
ΔTCoolant Out Motor ΔTCoolant Out Inverter ΔTStator Winding

°C °C °C
0.167 0.188 1.864
3.4 INTEGRATION IN THE SIMULATION ENVIRONMENT

The final goal of this work was the integration of the designed model into VSM-PTM,
the transient engine model for VSM. The desired hybrid architecture comprises two elec-
tric drives and one turbocharged 6 cylinder gasoline internal combustion engine. One
electric drive is connected to the ICE crankshaft through a spur gear unit (MGU-K) and
the second one is directly attached to the shaft of the turbocharger (MGU-H).
In the following text, the integration of the model into VSM-PTM is explained and the
corresponding Simulink models are discussed. Parts of the main layout of the PTM power
unit are shown in Figure 48, with the additional blocks for the hybrid parts highlighted
with solid red lines and the modified blocks highlighted with dotted red lines. The new
ERS block contains the electromechanical behavior and loss calculation for both MGU-
K and MGU-H, and also a battery model for the energy storage system. To link the MGU-
K to the crankshaft, the MGU-K Gbx block (see Figure 48) was added, representing the
spur gear unit. In the Cooling ICE Electro block, the cooling circuit for the electric drives
was added, which includes the thermal models for MGU-K and MGU-H and a radiator
(Figure 51, page 57).
3 Methodology
Figure 48: VSM-PTM power unit layout
3 Methodology
The ERS block in Figure 50 consists of two electric drives and a battery model. Both of
the electric drives comprise a model for the electromechanical behavior, loss calculation
for motor and inverter, and the DC power calculation described in Section 3.1. The input
signals are demand torque, speed, DC voltage and the temperatures from the thermal
model. Outputs are the current torque, DC current and the motor and inverter losses. The
calculation to the right of the two MGU blocks converts the output data into the VSM-
PTM architecture.
The battery model was taken from an existing project within AVL and adapted to fit this
application. It is based on a single cell model, including a constant internal resistance and
a voltage source (Figure 49). The open circuit voltage (OCV) is a function of the state of
charge (SOC) and was parameterized from voltage curves at different discharge rates.
Ri IDC
OCV
VDC
= f(SOC)
Figure 49: Battery cell model used in VSM-PTM
The input DC current to the battery pack is divided by the number of cells in parallel
before being fed into the cell model. The resulting cell voltage is then multiplied by the
2
number of cells in series. The power loss for each cell 𝑃𝑃𝐿𝐿,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = 𝐼𝐼𝐷𝐷𝐷𝐷,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑅𝑅𝑖𝑖 are multiplied
by the overall number of cells in the battery pack to calculate the battery losses.
3 Methodology
Figure 50: ERS block layout
3 Methodology
The thermal models for MGU-H and MGU-K, as well as their common cooling circuit,
were implemented in the Cooling ICE Electro block and are shown in Figure 51. On the
left, the losses coming from the loss calculation in the ERS block are fed into the thermal
models for motors and inverters. The single thick connecting-line in the model is the flow
of cooling liquid. After exiting the radiator, the liquid first flows through the MGU-H
motor and inverter, and then MGU-K motor and inverter before flowing back into the
radiator.
Figure 51: Additional ERS cooling circuit in the Cooling ICE Electro block
The Water Circuit Electro block (Figure 52) models the cooling liquid in the hoses be-
tween the different components and the coolant pump. As illustrated, the water pump is
attached to the ICE crankshaft.
Figure 52: Water circuit electro block
4 RESULTS
The result of this thesis is a flexible and fast model for electric powertrains and the electric
parts of hybrid powertrains. The following section first discusses the model capabilities
for a single electric drive, and then presents the model behavior in VSM-PTM.
4.1 MODEL CAPABILITIES

The designed model for an electric drive is versatile and meets all the requirements de-
fined in Section 1.2. Due to its simplicity, it can be used for most types of radial flux
machines, such as permanent magnet motors, induction motors, or synchronous reluc-
tance motors. In addition, the computation times of the model are short, making it per-
fectly suitable for real time operation.
As the test with the UQM PowerPhase 75 drive system indicates, the model accuracy
seems promising. The model matches the measurements well, with a mean deviation of
less than 0.2°C for the coolant outlet temperature, and less than 2°C for the stator winding
temperature (Table 11). At this time, only one test has been done to verify the model,
which recorded only parts of the necessary channels. Consequently, the full model be-
havior could not be tested. While the general characteristics of the thermal model look
good, there were no measurements available to verify the full range of modelled temper-
atures or any dynamic operation.
The proposed parameterization methods make it easy to set up the model, pending the
acquisition of specific measurements on a test rig. In case that not all nodes of the thermal
model are needed, the model can be further simplified by reducing the number of lumped
elements.
The model provides the physical interfaces that were required for integration into VSM-
PTM and combination with the battery model. The interfaces of the thermal models were
designed in a way that the order of the components in the cooling circuit can be easily
changed.
4 Results
Since the model is simple, there are several limitations. Due to the required measurements
of motor and inverter, the model is not suitable for electric drive development, which
would require evaluation of the effects of changes in specific machine parameters or con-
trol algorithms. The only way to do this is to make new measurements or do appropriate
simulations in order to get the required input data for the model. Another major limitation
stems from the fact that the model uses measurements of steady state operating points and
does not account for any transient behavior in the machine other than torque delay.
4.2 SIMULATION IN VSM-PTM

To show how the designed model works within VSM-PTM, a demonstration simulation
has been set up. It uses a typical setup for open wheeled single seater cars with a turbo-
charged V6 engine, including two electric motors (see Section 3.4). Because no measure-
ments of electric drives used in racing applications were available, parameters were for-
mulated based on the measurements from the UQM PowerPhase 75 (see Section 3.3.1)
and the 2014 Formula One regulations [4]. The maximum power for the MGU-K is
120kW, with a maximum speed of 50krpm. The maximum power of the MGU-H is
100kW. The simulation was done for one lap on the Italian race track Autodromo Na-
zionale Monza (Figure 53).
Figure 53: Autodromo Nazionale Monza track layout [23]
Some of the simulation results are presented on the following pages. Figure 54 gives a
general overview of the lap with the vehicle speed plotted in the top graph and the corre-
sponding pedal positions below. It shows the four long wide open throttle sections of the
4 Results
track, as well as the characteristic corners, most notably the Curva di Lesmo from 2000m
to 2500m. The two bottom plots in Figure 54 show the torque and rotational speed of the
ICE. Again, these plots reflect the track layout well. The shift operations are visible, caus-
ing only a minor torque reduction during the shift due to the use of a sequential gearbox.
400
Speed / km/h
300
200
100
Vehicle Speed
0
0 1000 2000 3000 4000 5000 6000
100
Position / %
75
50
25
Throttle Pedal Position
0
0 1000 2000 3000 4000 5000 6000
60
Brake Pedal Position
Position / %
40
20
0
0 1000 2000 3000 4000 5000 6000
15
Speed / krpm
10
5
IC-Engine Speed
0
0 1000 2000 3000 4000 5000 6000
400
Torque / Nm
300
200
100
0 IC-Engine Torque
0 1000 2000 3000 4000 5000 6000
Distance / m
Figure 54: VSM Monza Lap Vehicle Data
Figure 55 shows information about the two electric drives in the powertrain (i.e. rotational
speed and torque for MGU-K and MGU-H). The MGU-K is attached to the ICE crank-
shaft through a single speed transmission, and therefore shows the same speed profile as
4 Results
the ICE. In contrast, the MGU-K torque curve differs greatly from the ICE curve because
a negative torque is applied when braking to harness braking energy and recharge the
battery. In the acceleration phases, the MGU-K does not start operating until the ICE
approaches full throttle. Below this threshold, the wheels are mostly operated at the slip
limit and additional torque from the MGU-K would only result in wheel spin.
50
Speed / krpm
40
30
20
MGU-K Speed
10
0 1000 2000 3000 4000 5000 6000
50
Torque / Nm
25
0
-25
MGU-K Torque
-50
0 1000 2000 3000 4000 5000 6000
150
Speed / krpm
100
50
MGU-H Speed
0
0 1000 2000 3000 4000 5000 6000
20 MGU-H Torque
Torque / Nm
10
0
-10
0 1000 2000 3000 4000 5000 6000
Distance / m
Figure 55: VSM Monza lap electric dive data
The MGU-H profiles in Figure 55 reflect the operating point of the combustion engine.
When the ICE is accelerating from low speeds, the MGU-H produces positive torque to
quickly speed up the turbocharger and reduce the turbo lag. In high load operating points,
the power harnessed by the exhaust gas turbine is more than necessary for the compressor
4 Results
turbine, which allows the MGU-H to control the charging pressure by converting the ex-
cess power to electricity.
900
Voltage / V
800
700
600
Battery Voltage
500
0 1000 2000 3000 4000 5000 6000
200
Battery Power
100
Power / kW
0
-100
-200
0 1000 2000 3000 4000 5000 6000
200
Current / A
100
0
-100
MGU-K DC Current
-200
0 1000 2000 3000 4000 5000 6000
200
MGU-H DC Current
Current / A
100
-100
0 1000 2000 3000 4000 5000 6000
Distance / m
Figure 56: VSM Monza lap ERS data
In Figure 56, data about the energy storage is shown. The top two plots show the battery
voltage and power, where the battery voltage is directly connected to the battery power.
At high discharge (positive) power, the voltage drops; at high charge (negative) power,
the voltage rises. Here, the biggest stress on the battery is at the beginning of braking
when MGU-K is fully recuperating and MGU-H is still generating energy from the ex-
haust gas. The discharge power is usually less, because at full throttle the MGU-K gets
its energy from the battery and directly from the MGU-H. This is reflected in the two
4 Results
lower graphs displaying the DC currents of MGU-K and MGU-H, where the MGU-H
current at high loads accounts for about half of the current to the MGU-K.
The losses of both electric drives are plotted in Figure 57. The MGU-K losses are higher
than the MGU-H losses, because it has a higher nominal power than the MGU-H and is
mostly operated at full load. In contrast, the MGU-H is used to control boost pressure and
rarely runs at full load. The losses in the inverters are significantly less than the motor
losses due to their higher efficiency.
10
8
Power / kW
6
4
2 MGU-K Motor Loss
0
0 1000 2000 3000 4000 5000 6000
4
MGU-K Inverter Loss
3
Power / kW
2
1
0
0 1000 2000 3000 4000 5000 6000
10
8
Power / kW
6
4
2 MGU-H Motor Loss
0
0 1000 2000 3000 4000 5000 6000
4
MGU-H Inverter Loss
3
Power / kW
2
1
0
0 1000 2000 3000 4000 5000 6000
Distance / m
Figure 57: VSM Monza lap electric drive losses
5 DISCUSSION
The goal of the project was to develop a Matlab® Simulink® model for the electric drives
in hybrid powertrains for motorsports applications, and their integration in VSM-PTM.
The designed model should be easy to use and parameterize, and blend in with the layout
and design of the existing vehicle simulation model.
The result of the project is a flexible model for electric drives, which can be used with
most types of radial flux machines (e.g. induction machines or permanent magnet syn-
chronous machines). It models the electromechanical and thermal behavior of both motor
and inverter, and accounts for the effects of temperature on the motor efficiency. The
parameterization of the model is largely dependent on prior measurements, in order to
reach high accuracy without requiring extensive knowledge about the design and control
of the modelled drive. Therefore, the designed model is suitable for representing an al-
ready existing electric drive in a vehicle simulation environment, but is not suitable for
most cases of electric drive development.
The designed model was seamlessly integrated in VSM-PTM, functioning well with the
other components of the vehicle simulation system. In addition, an existing battery model
was adapted and integrated into VSM-PTM, completing the components of a hybrid elec-
tric powertrain.
Parts of the model could not be parameterized from real measurements, since the electric
drive that was used as a reference broke during the course of this thesis. Additionally,
some measurements are suspected to contain errors, which cannot be investigated since
the drive is not available anymore. However, the general behavior of the model seems
plausible, and general the measurements that were made match well with the simulations.
Future work on this project should include full verification of the model behavior, both
for the electromechanical model and the thermal model. Use of several different electric
drives for racing applications would also provide information about the accuracy of the
model for various types of electric drives. Since the thermal condition of batteries is a
very important factor in vehicles, a thermal model for the battery would also be valuable
for future applications.
6 REFERENCES
[1] M. Ehsani, Y. Gao and J. Miller: "Hybrid Electric Vehicles: Architecture and
Motor Drives", Proceedings of the IEEE, Vol. 95, No. 4, pp. 719-728, April 2007.
[2] C. Chan, A. Bouscayrol and K. Chen: "Electric, Hybrid, and Fuel-Cell Vehicles:
Architectures and Modeling", IEEE Transactions On Vehicular Technology, Vol.
59, No. 2, pp. 589-598, February 2010.
[3] Fédération Internationale de l’Automobile: "2009 F1 Technical Regulations", 11

July 2008. [Online]. Available: http://argent.fia.com/web/fia-public.nsf/90D88
9BE20961303C1257483004B8AC0/$FILE/1-2009%20F1%20TECHNICAL%20
REGU LATIONS%2011-07-2008.pdf. [Accessed 19 March 2015].
[4] Fédération Internationale de l’Automobile: "2015 F1 Technical Regulations", 03

December 2014. [Online]. Available: http://www.fia.com/sites/default/files/
regulation/file/2015%20TECHNICAL%20REGULATIONS%202014-12-03.pdf.
[Accessed 20 March 2015].
[5] J. Jwan: "www.biser3a.com", [Online]. Available: http://biser3a.com/formula-

1/explained-formula-1s-energy-recovery-system-ers/. [Accessed 19 March 2015].
[6] D. Schröder: “Elektrische Antriebe – Grundlagen”, Berlin Heidelberg: Springer

Vieweg, 2013, ISBN: 978-3-6423-0470-5.
[7] N. Mohan: “Electric Machines and Drives”, Hoboken: John Wiley & Sons, 2012,
ISBN: 978-1-118-07481-7.
[8] T. Finken, M. Felden and K. Hameyer: "Comparison and design of different

electrical machine types regarding their applicability in hybrid electrical vehicles",
Electrical Machines, 2008. ICEM 2008. 18th International Conference on, pp. 1-
5, 6.-9. September 2008.
[9] A. Mathoy: "Die Entwicklung bei Batterien und Antriebstechnik für

Elektroautomobile", Bulletin SEV/AES, pp. 8-13, January 2008.
[10] Z. Zhu and C. Chan: "Electrical machine topologies and technologies for electric,
hybrid, and fuel cell vehicles", Vehicle Power and Propulsion Conference, 2008.
VPPC '08. IEEE, pp. 1-6, 3-5 September 2008.
6 References
[11] R. Fischer: “Elektrische Maschinen”, München, Wien: Carl Hanser Verlag, 2004,
ISBN: 978-3-4464-0613-1.
[12] R. Park: "Two-reaction theory of synchronous machines generalized method of

analysis-part I", American Institute of Electrical Engineers, Transactions of the,
vol.48, no.3,, pp. pp.716,727, July 1929.
[13] D. Schröder: “Regelung von Antriebssystemen”, Berlin Heidelberg: Springer-

Verlag, 2009, ISBN: 978-3-540-89613-5.
[14] J. Pyrhönen, T. Jokinen and V. Hrabovcova: “Design of Rotating Electrical

Machines”, Second Edition, West Sussex, United Kingdom: John Wiley & Sons
Ltd, 2014, ISBN: 978-1-118-58157-5.
[15] A. Krings: “Iron Losses in Electrical Machines - Influence of Material Properties,

Manufacturing Processes, and Inverter Operation”, Ph.D. dissertation, Stockholm:
KTH School of Electrical Engineering, 2014.
[16] J. Schützhold and W. Hofmann: "Analysis of the Temperature Dependance of

Losses in Electrical Machines", IEEE Energy Conversion Congress and Exposi-
tion (ECCE), pp. 3159-3165, 15-19 September 2013.
[17] J.-H. Seo, D.-K. Woo, T.-K. Chung and H.-K. Jung: "A Study on Loss
Characteristics of IPMSM for FCEV Considering the Rotating Field", IEEE
TRANS. ON MAGNETICS, VOL. 46, NO. 8, pp. 3213-3216, August 2010.
[18] J. Mikleš and M. Fikar: “Process Modelling, Identification, and Control”, Berlin
Heidelberg New York: Springer, 2007, ISBN 978-3-540-71969-4.
[19] A. Boglietti, A. Cavagino, M. Lazzari and M. Pastorelli: "A Simplified Thermal

Model for Variable-Speed Self-Cooled Industrial Induction Motor", IEEE
TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 39, NO. 4, pp. 945-952,
July/August 2003.
[20] M. Fénot, Y. Bertin, E. Dorignac and G. Lalizel: "A review of heat transfer
between concentric rotating cylinders with or without axial flow", International
Journal of Thermal Sciences, pp. 1138-1155, March 2011.
[21] K. Yamazaki: "Iron Loss Analysis of Interior Permanent-Magnet Synchronous

Motors—Variation of Main Loss Factors Due to Driving Condition", IEEE
TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 42, NO. 4, pp. 1045-
1052, July/August 2006.
6 References
[22] UQM Technologies Inc.: "UQM PowerPhase 75 Datasheet", [Online]. Available:

http://www.neweagle.net/support/wiki/docs/Datasheets/UQM/PP75.pdf.
[Accessed 9 October 2014].
[23] W. Pittenger: "Monza Track Map - Italien Grand Prix - Wikipedia, the free
encyclopedia", 14 September 2008. [Online]. Available: http://en.wikipedia.org/
wiki/Italian_Grand_Prix#/media/File:Monza_track_map.svg. [Accessed 28 April
2015].
7 APPENDIX
APPENDIX A MATLAB CODE: PARAMETER GENERATION

% **********************************************************************
% Read measurement-data from Excel-File and generate parameter-maps
% Alexander Jedinger
% 10.12.2014
%
% All thermal parameters are fitted to meet the recorder-file measurements
%% Read Excel File
[filename, pathname] = uigetfile( ...

{'*.xls', 'XLS - Excel'; ...
'*.*', '*.* - All Files'}, ...
'Select an EXCEL file');
if filename==0 return; end;
targetfile = [pathname filename];

[A,B] = xlsread([pathname filename], 'Measurement_Data');
xls_n = A(:, 1);

xls_T = A(:, 2);
xls_eta_motor = A(:, 3);
xls_eta_inv = A(:, 4);
xls_Is = A(:, 5);
xls_T_winding = A(:, 6);
%% Motor Parameters
% Base Power [W]

Base_Power = 45160; %To match with 155Nm
% Base Torque [Nm]

Base_Torque = 155; %To cover all the points from the dyno
% Base Speed [rpm]

Base_Speed = 2875;
% Maximum Speed [rpm]

Max_Speed = 8000;
% Minimum Voltage for full power [V]

Vdc_min = 400;
% Number of Phases [-]

Phases = 3;
% Motor inertia [kg.m^2]

inertia = 0.003;
% Motor viscous friction [kg.m^2]

friction = 0.004;
% Average stator winding resistance (per phase) [Ohm]

Rs = 13.48e-3;
% stator winding temperature at resistance measurement [°C]
T_Rs = 25;
% Temperature coefficient of Stator Winding [1/K]
Rs_alpha = 3.9e-3;
% PT1 torque delay characteristics

tau_torque_delay = 5e-3; %Time constant
% [num, den] = tfdata(c2d(tf(1,[tau 1]),param.Solver.StepSize),'v');
% param.Motor.TorqueDelay.Num = num;
APPENDIX A Matlab Code: Parameter Generation
% param.Motor.TorqueDelay.Den = den;
% Maximum Speed Torque Limitation Curve (for plotting it in the figures)

Max_Speed_Torque.n = 0:Max_Speed/20:Max_Speed;
Max_Speed_Torque.T = ...
[Base_Torque*ones(1,sum(Max_Speed_Torque.n<=Base_Speed)) ...
Base_Power./(Max_Speed_Torque.n(Max_Speed_Torque.n>Base_Speed)*pi/30)];
% Maximum Speed-Torque Lookuptable (including dependence on Vdc)

Vdc = 0.2:0.2:1;
Base_Speed_vdc = (Base_Power.*Vdc)/Base_Torque *30/pi;
n_tmp = 0:Max_Speed/100:Max_Speed;
T_tmp = zeros(length(Vdc), length(n_tmp));
for i=1:length(Vdc)
T_tmp(i,:) = [Base_Torque*ones(1,sum(n_tmp<=Base_Speed_vdc(i))) ...
Base_Power*Vdc(i)./(n_tmp(n_tmp>Base_Speed_vdc(i))*pi/30)];
end
figure();
plot(n_tmp, T_tmp, 'LineWidth', 2);
grid on;
axis([0,8000,0,160]);
set(gca,'ytick',0:20:160);
legend('0.2 V_D_C','0.4 V_D_C', '0.6 V_D_C', '0.8 V_D_C', '1 V_D_C');
xlabel('Rotor-Speed / rpm');
ylabel('|Torque| / Nm');
title('Power-limitation due to DC-Voltage');
set(findall(gcf,'type','text'),'FontName','Times New Roman', 'FontSize', 12);
set(gca,'FontName','Times New Roman', 'FontSize', 12);
%print('TorqueLimitOverVDC','-dmeta');
Max_Speed_Torque_LUT.n = n_tmp;
Max_Speed_Torque_LUT.Vdc = Vdc*Vdc_min;
Max_Speed_Torque_LUT.T = T_tmp;
%% Stator Current Lookuptable

xnodes = 0:Max_Speed/20:Max_Speed;
ynodes = -Base_Torque:Base_Torque/20:Base_Torque;
StatorCurrentMap.n = xnodes;
StatorCurrentMap.T = ynodes;
StatorCurrentMap.Is = gridfit(xls_n,xls_T,xls_Is,xnodes,ynodes,'smoothness',0.3);
figure();
v = 0:20:180;
[C,h] = contour(xnodes,ynodes,StatorCurrentMap.Is, v);
clabel(C,h);
hold on; grid on;
plot(Max_Speed_Torque.n,Max_Speed_Torque.T,'black');
plot(Max_Speed_Torque.n,-Max_Speed_Torque.T,'black');
axis([0,8000,-160,160])
set(gca,'ytick',-160:20:160);
xlabel('Rotor-Speed in rpm');
ylabel('Torque in Nm');
title('Stator Current Map');
%% Iron and other Loss Lookuptable
% Calculate Stator Resistance at Temperature of Measurement

Rs_hot = Rs*(1 + Rs_alpha*(xls_T_winding - T_Rs));
% Generation of Lookuptable
% Calculation of Copper Losses (Is is stator current from measurement)
P_loss_copper = 3*Rs_hot.*xls_Is.^2;
% Calculation of Iron Losses

% because of varying definition of eta, the Losses have to be
% calculated differntly, according to direction of energyflow
P_mot = xls_n*pi/30 .* xls_T .*(1-xls_eta_motor)./xls_eta_motor;
P_gen = xls_n*pi/30 .* xls_T .*(1-xls_eta_motor);

P_loss_motor = P_mot.*(xls_T>0) - P_gen.*(xls_T<=0);
P_loss_iron = P_loss_motor - P_loss_copper;
IronLossMap.n = xnodes;
IronLossMap.T = ynodes;
IronLossMap.P_loss =
gridfit(xls_n,xls_T,P_loss_iron,xnodes,ynodes,'smoothness',0.2);
figure();
v = 500:250:3000;
[C,h] = contour(xnodes,ynodes,IronLossMap.P_loss,v);
clabel(C,h);
hold on; grid on;
axis([0,8000,-160,160]);
set(gca,'ytick',-160:20:160);
title('Iron Loss Map');
% Inverter Loss Lookuptable

% Calculation of Inverter-Losses
% because of varying definition of eta, the Losses have to be
% calculated differntly, according to direction of energyflow
P_l_inv_mot = (xls_n*pi/30 .* xls_T + P_loss_motor) .*(1-xls_eta_inv)./xls_eta_inv;
P_l_inv_gen = (xls_n*pi/30 .*(-xls_T) - P_loss_motor) .*(1-xls_eta_inv);
P_loss_inverter = P_l_inv_mot.*(xls_T>0) + P_l_inv_gen.*(xls_T<=0);
InverterLossMap.n = xnodes;
InverterLossMap.T = ynodes;
InverterLossMap.P_loss =
gridfit(xls_n,xls_T,P_loss_inverter,xnodes,ynodes,'smoothness',0.8);
% Make Sure loss is never negative

InverterLossMap.P_loss(InverterLossMap.P_loss<0) = 0;
figure();
v = 0:200:2000;
[C,h] = contour(xnodes,ynodes,InverterLossMap.P_loss,v);
clabel(C,h);
hold on; grid on;
axis([0,8000,-160,160]);
set(gca,'ytick',-160:20:160);
title('Inverter Loss Map');
%% Motor Thermal Parameters
% Distribution of iron losses to teeth and yoke

iron_loss_distribution.teethLUT.x = [0 4000 8000];
iron_loss_distribution.rotorLUT.x = iron_loss_distribution.teethLUT.x;
iron_loss_distribution.yokeLUT.x = iron_loss_distribution.teethLUT.x;
iron_loss_distribution.teethLUT.v = [0.6 0.6 0.55];
iron_loss_distribution.rotorLUT.v = [0.1 0.13 0.25];
iron_loss_distribution.yokeLUT.v = [0.3 0.27 0.2];
% Masses [kg]
mass.rotor = 9.5;
mass.teeth = 5;
mass.winding = 11;
mass.yoke = 12;
mass.coolant = 2;
mass.coolant_radiator = 3;
mass.housing = 2;
% Heat capacities [J/(kg K)]

heat_capacity.rotor = 500;
heat_capacity.teeth = 500;
heat_capacity.winding = 385;
heat_capacity.yoke = 500;
heat_capacity.coolant = 4181; %Water at 50°C
heat_capacity.housing = 897; % assuming Aluminum
% thermal Resistances [K/W]

% Geometry Data [m] (based on the UQM Motor)
airgap_length = 1e-3;
slot_depth = 40e-3;
slot_width_back = 24e-3;
tooth_width = 20e-3;
YokeThickness = 40e-3;
l_Yoke = YokeThickness/2;
number_of_slots = 12;
rotor_diameter = 90e-3;
stator_diameter = 250e-3;
stack_height = 140e-3;
thickness_cooling_jacket = 10e-3;
% heat transition coefficient [W/(m^2 K)]

alpha_Cu2Stator = 180; %Assuming a reasonable value (without any calculations)
alpha_Yoke2Coolant = 580; %From UQM Measurement in Nominal Point
% heat transition coefficient from rotor to stator

nu_air = 14.95e-6; %kinematic viscosity of air at 25°C
a_air = 20e-6; %thermal diffusity of air at 20°C and 1 bar
k_air = 0.0262; %thermal conductivity of air
omega = 2875*pi/30; %(0:Max_Speed/10:Max_Speed)*pi/30; %speed vector
R1 = rotor_diameter/2;
R2 = R1 + airgap_length;
N_Taylor = sqrt(omega^2*R1*((R2-R1)/2)^3 /nu_air^2);

N_Prantl = nu_air/a_air;
N_Nusselt = 0.23*N_Taylor^0.63 *N_Prantl^0.27;
alpha_Rotor2Stator = N_Nusselt*k_air/airgap_length;
% thermal conductivity [W/(m K)]
%using same alpha for Rth coolant to housing

Rth_coolant_housing = 1/(alpha_Yoke2Coolant*stack_height*(stator_diameter+2*thick-
ness_cooling_jacket)*pi);
Rth_rotor_teeth = 1/(alpha_Rotor2Stator*(rotor_diameter/2+air-
gap_length/2)*2*pi*stack_height);
% Adjust parameters to match measurement:

Rth_winding_teeth = 0.023;
Rth_winding_yoke = 0.053;
Rth_teeth_yoke = 0.03;
Rth_yoke_coolant = 0.005;
thermal_resistance.rotor_shaft = 0.2;
thermal_resistance.rotor_teeth = Rth_rotor_teeth;
thermal_resistance.winding_teeth = Rth_winding_teeth; %0.039;
thermal_resistance.winding_yoke = Rth_winding_yoke; %0.1;
thermal_resistance.teeth_yoke = Rth_teeth_yoke; %0.01;
thermal_resistance.yoke_coolant = Rth_yoke_coolant; %0.001;
thermal_resistance.coolant_housing = Rth_coolant_housing; %0.001;
thermal_resistance.housing_environment = 0.3;
% initial temperatures [°C]

initial_temp.rotor = 96;
initial_temp.teeth = 85;
initial_temp.winding = 95;
initial_temp.yoke = 60;
initial_temp.housing = 52;
initial_temp.coolant = 56.5;
temp_environment_offset = 0;
temp_shaft = 65;
%% Inverter Thermal Parameters
% % thermal resistances [K/W]

Inv.thermal_resistance.junction_heatsink = 0.013;
Inv.thermal_resistance.heatsink_coolant = 0.013;
Inv.thermal_resistance.coolant_housing = 0.08;
Inv.thermal_resistance.housing_environment = 0.3;
% Masses [kg]
Inv.mass.junction = 0.05;
Inv.mass.heatsink = 1.5;
Inv.mass.coolant = 1.8;
Inv.mass.housing = 0.5;
% Heat capacities [J/(kg K)]

Inv.heat_capacity.junction = 80;
Inv.heat_capacity.heatsink = 897; %assuming aluminum
Inv.heat_capacity.coolant = 4181; %Water at 50°C
Inv.heat_capacity.housing = 897; %assuming aluminum
% Temperatures
Inv.initial_temp.junction = 92;
Inv.initial_temp.heatsink = 55;
Inv.initial_temp.housing = 52;
Inv.initial_temp.coolant = 54.5;
%% Thermal Properties for Water Circuit
% Masses of coolant
Fluidmass.Radiator = 2;
Fluidmass.toRadiator = 0.5;
Fluidmass.fromRadiator = 0.5;
%% Save Parameter File
%Write to Parameter Struct

MDL.ARES.EPT.MGUH.Base_Power.v = Base_Power;
MDL.ARES.EPT.MGUH.Base_Torque.v = Base_Torque;
MDL.ARES.EPT.MGUH.Base_Speed.v = Base_Speed;
MDL.ARES.EPT.MGUH.Max_Speed.v = Max_Speed;
MDL.ARES.EPT.MGUH.Vdc_min.v = Vdc_min;
MDL.ARES.EPT.MGUH.NumberOfPhases.v = Phases;
MDL.ARES.EPT.MGUH.Inertia.v = inertia;
MDL.ARES.EPT.MGUH.Rs.v = Rs;
MDL.ARES.EPT.MGUH.T_Rs.v = T_Rs;
MDL.ARES.EPT.MGUH.Rs_alpha.v = Rs_alpha;
MDL.ARES.EPT.MGUH.tau_torque_delay.v = tau_torque_delay;
MDL.ARES.EPT.MGUH.MaxSpeedTorqueMap.x = Max_Speed_Torque_LUT.n;
MDL.ARES.EPT.MGUH.MaxSpeedTorqueMap.y = Max_Speed_Torque_LUT.Vdc;
MDL.ARES.EPT.MGUH.MaxSpeedTorqueMap.v = Max_Speed_Torque_LUT.T;
MDL.ARES.EPT.MGUH.StatorCurrentMap.x = StatorCurrentMap.n;
MDL.ARES.EPT.MGUH.StatorCurrentMap.y = StatorCurrentMap.T;
MDL.ARES.EPT.MGUH.StatorCurrentMap.v = StatorCurrentMap.Is;
MDL.ARES.EPT.MGUH.IronLossMap.x = IronLossMap.n;
MDL.ARES.EPT.MGUH.IronLossMap.y = IronLossMap.T;
MDL.ARES.EPT.MGUH.IronLossMap.v = IronLossMap.P_loss;
MDL.ARES.EPT.MGUH.InverterLossMap.x = InverterLossMap.n;
MDL.ARES.EPT.MGUH.InverterLossMap.y = InverterLossMap.T;
MDL.ARES.EPT.MGUH.InverterLossMap.v = InverterLossMap.P_loss;
MDL.ARES.EPT.MGUH.Thermal.Mot.iron_loss_distribution.teeth.v =
iron_loss_distribution.teeth;
MDL.ARES.EPT.MGUH.Thermal.Mot.iron_loss_distribution.yoke.v =
iron_loss_distribution.yoke;
MDL.ARES.EPT.MGUH.Thermal.Mot.iron_loss_distribution.rotor.v =
iron_loss_distribution.rotor;
MDL.ARES.EPT.MGUH.Thermal.Mot.mass.rotor.v = mass.rotor;
MDL.ARES.EPT.MGUH.Thermal.Mot.mass.teeth.v = mass.teeth;
MDL.ARES.EPT.MGUH.Thermal.Mot.mass.winding.v = mass.winding;
MDL.ARES.EPT.MGUH.Thermal.Mot.mass.yoke.v = mass.yoke;
MDL.ARES.EPT.MGUH.Thermal.Mot.mass.coolant.v = mass.coolant;
MDL.ARES.EPT.MGUH.Thermal.Mot.mass.housing.v = mass.housing;
MDL.ARES.EPT.MGUH.Thermal.Mot.heat_capacity.rotor.v = heat_capacity.rotor;
MDL.ARES.EPT.MGUH.Thermal.Mot.heat_capacity.teeth.v = heat_capacity.teeth;
MDL.ARES.EPT.MGUH.Thermal.Mot.heat_capacity.winding.v = heat_capacity.winding;
MDL.ARES.EPT.MGUH.Thermal.Mot.heat_capacity.yoke.v = heat_capacity.yoke;
MDL.ARES.EPT.MGUH.Thermal.Mot.heat_capacity.coolant.v = heat_capacity.coolant;
MDL.ARES.EPT.MGUH.Thermal.Mot.heat_capacity.housing.v = heat_capacity.housing;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.rotor_shaft.v =
thermal_resistance.rotor_shaft;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.rotor_teeth.v =
thermal_resistance.rotor_teeth;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.winding_teeth.v =
thermal_resistance.winding_teeth;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.winding_yoke.v =
thermal_resistance.winding_yoke;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.teeth_yoke.v =
thermal_resistance.teeth_yoke;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.yoke_coolant.v =
thermal_resistance.yoke_coolant;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.coolant_housing.v =
thermal_resistance.coolant_housing;
MDL.ARES.EPT.MGUH.Thermal.Mot.thermal_resistance.housing_environment.v =
thermal_resistance.housing_environment;
MDL.ARES.EPT.MGUH.Thermal.Mot.initial_temp.rotor.v = initial_temp.rotor;
MDL.ARES.EPT.MGUH.Thermal.Mot.initial_temp.teeth.v = initial_temp.teeth;
MDL.ARES.EPT.MGUH.Thermal.Mot.initial_temp.winding.v = initial_temp.winding;
MDL.ARES.EPT.MGUH.Thermal.Mot.initial_temp.yoke.v = initial_temp.yoke;
MDL.ARES.EPT.MGUH.Thermal.Mot.initial_temp.housing.v = initial_temp.housing;
MDL.ARES.EPT.MGUH.Thermal.Mot.initial_temp.coolant.v = initial_temp.coolant;
MDL.ARES.EPT.MGUH.Thermal.Mot.temp_environment_offset.v = temp_environment_offset;
MDL.ARES.EPT.MGUH.Thermal.Mot.temp_shaft.v = temp_shaft;
MDL.ARES.EPT.MGUH.Thermal.Inv.thermal_resistance.junction_heatsink.v =
Inv.thermal_resistance.junction_heatsink;
MDL.ARES.EPT.MGUH.Thermal.Inv.thermal_resistance.heatsink_coolant.v =
Inv.thermal_resistance.heatsink_coolant;
MDL.ARES.EPT.MGUH.Thermal.Inv.thermal_resistance.coolant_housing.v =
Inv.thermal_resistance.coolant_housing;
MDL.ARES.EPT.MGUH.Thermal.Inv.thermal_resistance.housing_environment.v =
Inv.thermal_resistance.housing_environment;
MDL.ARES.EPT.MGUH.Thermal.Inv.mass.junction.v = Inv.mass.junction;
MDL.ARES.EPT.MGUH.Thermal.Inv.mass.heatsink.v = Inv.mass.heatsink;
MDL.ARES.EPT.MGUH.Thermal.Inv.mass.coolant.v = Inv.mass.coolant;
MDL.ARES.EPT.MGUH.Thermal.Inv.mass.housing.v = Inv.mass.housing;
MDL.ARES.EPT.MGUH.Thermal.Inv.heat_capacity.junction.v = Inv.heat_capacity.junction;
MDL.ARES.EPT.MGUH.Thermal.Inv.heat_capacity.heatsink.v = Inv.heat_capacity.heatsink;
MDL.ARES.EPT.MGUH.Thermal.Inv.heat_capacity.coolant.v = Inv.heat_capacity.coolant;
MDL.ARES.EPT.MGUH.Thermal.Inv.heat_capacity.housing.v = Inv.heat_capacity.housing;
MDL.ARES.EPT.MGUH.Thermal.Inv.initial_temp.junction.v = Inv.initial_temp.junction;
MDL.ARES.EPT.MGUH.Thermal.Inv.initial_temp.heatsink.v = Inv.initial_temp.heatsink;
MDL.ARES.EPT.MGUH.Thermal.Inv.initial_temp.housing.v = Inv.initial_temp.housing;
MDL.ARES.EPT.MGUH.Thermal.Inv.initial_temp.coolant.v = Inv.initial_temp.coolant;
MDL.ARES.EPT.MGUH.Thermal.Inv.temp_environment_offset.v = temp_environment_offset;
MDL.ARES.EPT.COOLING.WATER_CIRCUIT.Water_Circulation_Electro.cp_coolant.v =
heat_capacity.coolant;
MDL.ARES.EPT.COOLING.WATER_CIRCUIT.Water_Circulation_Electro.initial_temp_coolant_toRadi
ator.v = initial_temp.coolant;
MDL.ARES.EPT.COOLING.WATER_CIRCUIT.Water_Circulation_Electro.initial_temp_coolant_inRadi
ator.v = initial_temp.coolant;
MDL.ARES.EPT.COOLING.WATER_CIRCUIT.Water_Circulation_Electro.initial_temp_coolant_fromRa
diator.v = initial_temp.coolant;
MDL.ARES.EPT.COOLING.WATER_CIRCUIT.Water_Circulation_Electro.m_coolant_toRadiator.v =
Fluidmass.toRadiator;
MDL.ARES.EPT.COOLING.WATER_CIRCUIT.Water_Circulation_Electro.m_coolant_inRadiator.v =
Fluidmass.Radiator;
MDL.ARES.EPT.COOLING.WATER_CIRCUIT.Water_Circulation_Electro.m_coolant_fromRadiator.v =
Fluidmass.fromRadiator;
%Save and clear all

save('param_v4_UQM','MDL');
clearvars -except MDL;
APPENDIX B MOTOR THERMAL MODEL IN SIMULINK
APPENDIX B Motor Thermal Model in Simulink
APPENDIX B Motor Thermal Model in Simulink
APPENDIX C INVERTER THERMAL MODEL IN SIMULINK
APPENDIX C Inverter Thermal Model in Simulink

Jedinger Alexander 2015

Uploaded by

Copyright:

Available Formats

Jedinger Alexander 2015

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jedinger Alexander 2015

Uploaded by

Copyright:

Available Formats

Alexander Jedinger, BSc

Individual Master’s Degree Program

Graz University of Technology

Dipl.-Ing. Martin Ackerl

Dipl.-Ing. Christoph Leichtfried

Graz, June 2015

Schließlich wurde das Modell in das bestehende Gesamtfahrzeugsimulationsmodell von

Symbol Unit Description

Symbol Unit Description

Symbol Unit Description

3.2 Parameterization ................................................................................... 34

Demand Torque Active Torque

Figure 1: Required model interfaces

Figure 2: Flow of energy for the electric drive model

2.1 HYBRID POWERTRAINS IN VEHICLES

Figure 3: Principal topology of hybrid powertrains [1]

2.1.1 TOPOLOGIES OF HYBRID ELECTRIC POWERTRAINS

2.1.1.1 PARALLEL HYBRID ELECTRIC POWERTRAIN

• Only the ICE delivers power to the transmission

Figure 4: Parallel hybrid electric powertrain [1]

2.1.1.2 SERIES HYBRID ELECTRIC POWERTRAIN

Figure 5: Series hybrid electric powertrain [1]

2.1.1.3 OTHER HYBRID ELECTRIC POWERTRAIN TOPOLOGIES

2.1.2 HYBRID POWERTRAINS IN MOTORSPORTS

2.1.2.1 FORMULA ONE

In 2014, as a part of major changes in regulations, the Fédération Internationale de l’Au-

Figure 6: Formula One power unit in 2014 [5]

2.1.2.2 24 HOURS OF LE MANS

2.1.3 REQUIREMENTS FOR HYBRID POWERTRAINS IN MOTORSPORTS

Table 1: Vehicle parameters for simulations

cd crr m PPU ηPU

2.1.3.2 VEHICLE WEIGHT

2.1.3.3 AERODYNAMIC PERFORMANCE

2.2 MODELING OF ELECTRIC DRIVES

2.2.1 BASICS OF ELECTRIC DRIVES

Figure 14: Block Diagram of a typical Electric Drive System [7]

(+) speed (+) speed

Backward Acceleration Forward Acceleration

Figure 15: Four-quadrant operation of electric drives in vehicles

2.2.2 ELECTROMECHANICAL BEHAVIOR OF ELECTRICAL MACHINES

𝑃𝑃𝑒𝑒𝑒𝑒 = 3 ⋅ 𝐼𝐼𝑆𝑆 ⋅ 𝑈𝑈𝑆𝑆 ⋅ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐, (2.1)

𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚ℎ = 𝑇𝑇 ⋅ 𝜔𝜔, (2.2)

2.2.2.1 EQUIVALENT CIRCUIT

Figure 16: Single-phase equivalent circuit of a SPM

asymmetries or frequencies of higher order. Furthermore, there is no simple mathematical

2.2.2.2 SPACE VECTOR MODEL

𝑈𝑈𝑎𝑎 + 𝑈𝑈𝑏𝑏 + 𝑈𝑈𝑐𝑐 = 0. (2.4)

𝑖𝑖𝑎𝑎 = 𝐼𝐼̂ ⋅ cos(𝜔𝜔𝜔𝜔 + 𝜑𝜑𝑎𝑎 ) = 𝐼𝐼̂ ⋅ 𝑒𝑒 𝑗𝑗𝜑𝜑𝑎𝑎 ,

(a) three-phase system (b) two-phase Cartesian system

Figure 17: Transition from three-phase to two-phase system [6]

2.2.2.3 FINITE ELEMENT ANALYSIS (FEA)

Figure 18: Division into finite elements for a 2D SPM problem

2.2.3 DETERMINATION OF LOSSES IN ELECTRICAL MACHINES

𝑃𝑃𝐿𝐿 = 𝑃𝑃𝑐𝑐𝑐𝑐 + 𝑃𝑃𝑓𝑓𝑓𝑓 + 𝑃𝑃𝑓𝑓𝑓𝑓 + 𝑃𝑃𝑣𝑣 . (2.10)

2.2.3.1 COPPER LOSSES

𝑃𝑃𝑐𝑐𝑐𝑐 = 3 ⋅ 𝐼𝐼𝑆𝑆2 ⋅ 𝑅𝑅𝑆𝑆 , (2.11)

𝑅𝑅(𝑇𝑇) = 𝑅𝑅(𝑇𝑇0 ) ⋅ �1 + 𝛼𝛼𝑐𝑐𝑐𝑐 ⋅ (𝑇𝑇 − 𝑇𝑇0 )�, (2.12)

2.2.3.2 IRON LOSSES

𝑝𝑝𝑓𝑓𝑓𝑓 = 𝑓𝑓 � 𝐻𝐻 𝑑𝑑𝑑𝑑. (2.13)