You are on page 1of 9

Digital Twin for a 10 MW Electrical Drive System

for Future Electric Aircraft Applications


Xinjun Liu Christian Bentheimer Florian Hilpert
Drives and Mechatronics Aerospace Electronics Aerospace Electronics
Fraunhofer Institute for Integrated Fraunhofer Institute for Integrated Fraunhofer Institute for Integrated
Systems and Device Technology IISB Systems and Device Technology IISB Systems and Device Technology IISB
Erlangen, Germany Erlangen, Germany Erlangen, Germany
xinjun.liu@iisb.fraunhofer.de christian.bentheimer@iisb.fraunhofer.de florian.hilpert@iisb.fraunhofer.de

Maximilian Hofmann Bernd Eckardt Martin Maerz


Drives and Mechatronics Vehicle Electronics Fraunhofer Institute for Integrated
Fraunhofer Institute for Integrated Fraunhofer Institute for Integrated Systems and Device Technology IISB
Systems and Device Technology IISB Systems and Device Technology IISB Erlangen, Germany
Erlangen, Germany Erlangen, Germany martin.maerz@iisb.fraunhofer.de
maximilian.hofmann@iisb.fraunhofer.de bernd.eckardt@iisb.fraunhofer.de

Abstract—Reliable and high power density electric machine and realized by a co-simulation system, which couples the finite
2021 11th International Electric Drives Production Conference (EDPC) | 978-1-6654-1809-6/21/$31.00 ©2021 IEEE | DOI: 10.1109/EDPC53547.2021.9684198

inverter solutions are essential for a variety of applications. element analysis (FEA) and the model of a SiC-inverter with
Digital twin for future electric aircraft applications can multilevel topology to realize the control strategy, to induce
contribute to analyze the system behavior and the effects and characterize the effects of SiC-inverters with high
between the fast-switching inverter and the electric machine. switching frequency into electrical drive system for electric
Within this paper, the concept of a "Digital Twin for a 10 MW aircraft application. In our system design, the switching
Electrical Drive System for Future Electric Aircraft frequency of the multilevel SiC-inverter was increased to a
Application" is presented. A digital twin platform is based on an high level to realize a fine current waveform and to reduce
outer-rotor permanent magnet synchronous machine and serval
stator and rotor losses. With the developed digital twin the
interleaved multi-level SiC-inverters, which allows the usage of
high switching frequencies and a high DC link voltage of 1500 V.
effects of the topology, the control method, and the switching
The combination of those components have advantages in future frequency of the multilevel SiC-inverter were used in a co-
electric aircraft applications. Several multi-level inverter simulation of the electrical drive system to predict the system
topologies have been investigated and a first system design for a behavior. Furthermore, the developed digital twin can be used
10.5 MW inverter system is presented. to analyze some basic failure cases for the electrical drive
system [3].
Industry 4.0 Applications in Electric Drives Production
II. MOTOR AND INVERTER SYSTEM DESIGN
Keywords—twin-builder, co-simulation, outer-rotor, PMSM, A. Design of a 10 MW Electric Machine for Aircraft
power electronics, multilevel inverter, SiC-inverter
Applications
I. INTRODUCTION For the aviation electrical drive system, an electric
The major approaches and first results towards the machine was designed, which can be used as a traction motor
implementation of a digital twin of a 10 MW electrical as well as a generator. The overall system requirements were
drive/generator system for future electric aircraft applications based on the assembly space, the electrical power and system
are presented. Analysis has shown that for such applications boundary conditions of an Airbus A400 Aircraft. The
machines and inverters with a power rating in the 10 MW specification of the electric machine, which is shown in Table
range are needed [1] [2]. Development and validation of the I, was oriented on high-power density aircraft application
electric aircraft application require a suitable digital twin of machines.
the electrical drive system. TABLE I. SPECIFICATION OF THE ELECTRIC MOTOR

An outer-rotor permanent magnet synchronous machine Parameter Specification Value


(PMSM) enabling high power up to 10 MW, high power Machine type PMSM
density, and improved efficiency of 95% for motor and Connection type 12-phase configuration
generator modes was designed based on electromagnetic and DC link voltage 1,500 V
mechanical simulations. The electric machine uses direct Output power 10 MW
cooled copper wires to reduce the temperature in the winding Motor phase current 2700 A
and to increase the power density during continuous operation.
Target efficiency 95%
Four galvanically isolated 3-phase winding sets were used,
Rated speed 13,000 rpm
resulting in a 12-phase electric machine. Each of the four 3-
phase systems can be driven by three interleaved (parallel) 3- Rated torque 7,345 Nm
phase SiC-inverters (twelve 3-phase inverters in total) with an Overall machine diameter 0.8 m
efficiency of about 99%. For the target DC link voltage of Thermal class H (180°C)
1500 V, three different multilevel inverter topologies (neutral Coolant topology Direct cooled wound wires
point clamped, active neutral point clamped, and flying
capacitor) are investigated and compared. At Fraunhofer IISB,
the digital twin of a 10 MW electrical drive system was

978-1-6654-1809-6/21/$31.00 ©2021 IEEE

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
According to the high speed requirement and high power TABLE III. COMPARISON OF WINDING FACTOR
density for this application, an electric machine with an outer- Torque-producing Winding
Combination Periodicties
rotor was designed and simulated in FEM software for by harmonic order factor
electromagnetic and mechanical optimization. An outer-rotor 72 slots/ 80 poles 2, 4, 8 5 0.945
electric machine has the advantage that the topology has a 81 slots/ 80 poles - 40 0.955
higher power density due to the larger bore diameter compared 84 slots/ 80 poles 2, 4 10 0.953
to the inner-rotor electric machine. The direction of the 84 slots/ 76 poles 2, 4 19 0.945
centrifugal force for the rotor is always outwards, so the air 84 slots/ 68 poles 2, 4 17 0.913
gap of the electric machine can be designed with a smaller
value, and the permanent magnet can be glued onto the rotor Since the electric machine is not connected to the propeller
without a bandage. directly, but to a gearbox with a transmission ratio of 1:9 the
electric machine with fractional slot was considered to reduce
Thanks to the high inverter switching frequency of 90 kHz, the torque ripple and to keep the power density high [8].
the electric machine can achieve a high electrical frequency Several stator slots and poles combinations were discussed
with a smooth current waveform. The distortion of the current and compared in Table III. A comparison of the winding factor
waveforms, which are caused by inductances of the machine and the periodicities is shown. With a higher winding factor,
and inverter, and by the switching frequency of the inverter, the electric machine can provide more torque based on the
should affect the electrical drive system as little as possible. equation (1) and (2). For the design of this 12-phase electric
The maximal pole pairs of the electric machine can be machine, four 3-times interleaved multilevel SiC-inverters
designed up to 40 pole pairs, which leads to an electrical were designed to provide enough current and voltage, so the
frequency of 8.67 kHz by the rated speed. 84 slots/ 80 poles combination was chosen for the
An electric machine with frictional slots was considered electromagnetic design due to the higher winding factor and
due to the advantages such as high power density, high the possible periodicities to couple with four SiC-inverters.
efficiency, and high slot fill factor [4] [5]. The comparison of
the different winding topologies is shown in Table II. Direct
cooled copper wires were used for the electromagnetic design.
They allow increasing the current density, the continuous
operation and they improve the heat dissipation considerably.
To realize the direct cooling in the winding wires,
concentrated windings were used for this application.
Concentrated windings are simpler than distributed windings
due to the shorter winding head and the higher slot fill factor.
TABLE II. COMPARISON OF DIFFERENT WINDINGS
Concentrated
Distributed Windings
Windings
Typical fill factor 45%-50% More than 50% Fig. 1. Simulated PMSM structure with outer-rotor in Ansys/Maxwell [9]
Winding head Long overlapping Short non-overlapping The electromagnetic design, which is shown in Fig. 1, was
Direct cooled
Difficult Possible realized in Ansys/Maxwell. The materials of the electric
copper wires
machine were chosen to achieve high power density and high
From the literature [6], the electromagnetic torque mechanical strength. The types of used materials are shown in
converted in the synchronous machine can be expressed by the Table IV. According to the low ambient temperature of the
following equation: aircraft application and the direct cooled copper wires from
the specification, the electromagnetic performance of this
=− ∙ sin % (1) electric machine was simulated at 13,000 rpm and a
!
where is the electromagnetic torque, is the number of permanent magnet temperature of 60°C. The electromagnetic
phases, is the synchronous speed of the machine, is torque is shown in Fig. 2.
the phase voltage, is the back-electromotive force (back TABLE IV. MATERIAL TYPES FOR THE ELECTRIC MACHINE
EMF), % is the load angle and phase shift between and , Component Type
! is the “synchronous reactance” of '-axis. Stator lamination VacoFlux 48
The back EMF of an electric machine can be expressed by Rotor lamination NO 25-14 HS
the following equation: Permanent magnet 52 UH
2
4
,) = √2 ∙ , ∙ - ∙ ./,) ∙ 1 2 35,) . (2) Fig. 2 shows that the electric machine has a low torque
0
where 7 is the harmonic order, , is the electrical frequency of
ripple of only 55.1 Nm from the FEM simulation compared
the electric machine, - is the number of winding per phase in
with the average torque of 7,534 Nm. If the rotor losses and
series, 1 is the pole pitch of the rotor, 2 is the length of the
the magnet losses are considered later on, the actual output
torque can be lower than the simulated electromagnetic
electric machine, and ./ is the winding factor of the slot and torque. The simulated induced line voltage is shown in Fig. 3.
pole pairs combination. The maximal value of the induced voltage of the PMSM is

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
1365.7 V. With a DC link voltage of 1500 V the modulation : ;
;
< =
;
< >
;
< ⋯< ;
factor can be calculated to be about 0.91. 89 = A 100 (3)
@

where 89 is the THD value of the line voltage, @ is the


amplitude of the fundamental line voltage, and is the
amplitude of the n-th order harmonic line voltage. The
calculated THD value of the induced line voltage is shown in
Table V.
In reality, the current waveforms are not purely sinusoidal,
and the harmonics depend on the topology, the control
strategy, and the switching frequency. The harmonic orders,
which are generated by the inductances and the switching of
the inverter, have a significant influence on the electro-
Fig. 2. Electromagnetic torque of the PMSM with outer-rotor magnetic torque and the losses of the electric machine.
Due to the high machine speed of 13,000 rpm and the Therefore, it is necessary to simulate and analyze the harmonic
overall diameter of the machine, a critical point during the orders for the electrical drive system.
development was the mechanical strength of the outer-rotor. TABLE V. THD OF INDUCED VOLTAGE AND MACHINE LOSSES
At the machine speed of 15,000 rpm – including a safety factor
Component Value
of 1.15 – the simulated maximum mechanical stress of the
rotor was about 121 MPa (see Fig. 4), which is lower than the THD of the induced voltage 2.4 %
yield’s stress according to the datasheet of Waelzholz [10]. Stator iron losses 18.4 kW
Rotor losses 17.3 kW

The described finite element model for the PMSM was


used in combination with a mathematical model of the inverter
system and the control strategy to develop a co-simulation.
The implementation and realization of the co-simulation are
described in chapter III. It allows simulating the influence of
the inverter topology (2-level or multi-level), the control
strategy, and the switching frequency on the electrical drive
system.
B. Design of a 10.5 MW Multi-Level SiC inverter for
Fig. 3. Induced voltage of the PMSM (line voltage) Aircraft Applications
The fast Fourier transformation (FFT) has been To achieve the rated electrical output power of 10.5 MW,
implemented to analyze specific harmonics. The target of the the inverter system is split into several subsystems. Increasing
FFT is to analyze the harmonics of the induced voltage. Fig. 5 the output power by connecting inverters in parallel is a well-
shows the FFT results of the line to line voltage. known solution for large inverter systems. It raises the system
complexity but also increases the fault tolerance capability,
which is an interesting topic for future investigations. The
electrical output power of an inverter (subsystem) is mainly
limited by the maximum DC link voltage and the maximum
output current, considering that sufficient cooling is applied to
the inverter system. The maximum output current is limited by
the current capability of the power modules and therefore by
the maximum number of power semiconductors bare dies that
can be connected in parallel inside a power module. The
challenge is that the current must be shared equally between
the parallel dies. The parasitic inductances will rise with more
parallel dies, which limits the switching speeds and leads to
Fig. 5. FFT results of the induced voltage from Ansys/Maxwell higher ringing and overvoltages. From previous projects
To validate the performance of the PMSM, the stator iron experiences at Fraunhofer IISB it can be concluded that the
losses and the rotor losses are the most important output values number of parallel bare dies should be limited to a maximum
from the simulation. Those values were calculated by the FEM number of 20. This translates to a current limit of roughly
and are shown in Table V. The total harmonic distortion 1000 Apeak (= 707 ARMS) for a single inverter subsystem for the
(THD) can be used for the evaluation of the harmonic waves, latest SiC MOSFET generation considering the high switching
and it is given by the following equation [11] [12]: frequency and the cooling conditions. For the presented motor
design and parameters this can be translated to a power limit
of roughly 1 MW. The output current of a single inverter
subsystem can be calculated to:

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
H IJ
DEFG =
(IGBTs) and to 65% of the devices breakdown voltage for
∙n ∙ ∙U ∙ cosSφU
(4)
3 i √3 LL,RMS
silicon carbide devices (SiC MOSFETs) to preserve low
failure rates [13].
with H IJ the mechanical output power of the motor, the
number of motor phases (multiple of factor three), nV the
number of parallel/interleaved inverters per 3-phase motor
system, UWW,XYZ the nominal line-to-line voltage, and cosSφU
the motor power factor. The number of motor phases was set
to twelve. The number of parallel inverter subsystems per 3-
phase motor system was calculated with (4) to nV = 3 with an
output current of 882.7 Apeak. So there are twelve identical
inverter subsystems in total. The derived specifications of the
inverter system are given in Table VI.
TABLE VI. MAIN INVERTER SPECIFICATIONS

Specifications of the inverter system


Nominal motor output power Pmech 10.0 MW
Nominal inverter output power Pel 10.53 MW
Nominal inverter apparent power Sel 12.48 MVA
Number of motor phases m 12
Number of interleaved systems ni 3
DC link voltage VDC 1500 V
Nominal motor phase current iRMS 624 ARMS
Nominal line-to-line voltage ULL,RMS 955 VRMS
PWM switching frequency fS 90 kHz
Target efficiency ηInverter 99% Fig. 6. Three interleaved ANPC-3L inverters connected to ¼ of the motor
Motor power factor cos(φ) 0.85 To tackle the problem of necessary high breakdown
Modulation factor mF 0.9 voltages multi-level inverters can be used, which reduce the
Heatsink temperature 65°C voltage stress of individual power semiconductors. In a three-
level inverter the necessary breakdown voltage is reduced by
The three subsystem inverters per three-phase motor 50% and in a five-level inverter the necessary breakdown
system will be used in an interleaved configuration, i.e. they voltage is reduced by 25% compared to a standard two-level
share a common DC-link and the corresponding AC outputs inverter. The choice of power semiconductors and the inverter
are connected through inductors (see Fig. 6). This is very topology are the major issues for the losses of the inverter
advantageous for this application because it helps to reduce system. For a two-level inverter only high voltage devices
the motor current ripple for the high speed and low inductance above 3.3 kV can be used, which have high conduction and
motor. Due to the interleaving, the effective switching switching losses. For a five-level inverter the needed amounts
frequency for the motor is increased by factor three. The of power semiconductors and gate drivers increase
additional inductors also help to reduce the current ripple dramatically, although devices with lower blocking voltages
furthermore. Additionally, the DC link capacitors can be and therefore lower conduction and switching losses can be
reduced, because due to the interleaved operation the capacitor used. As a compromise, it was decided to use a three-level
current and voltage ripple values will be smaller. However, the topology, because the number of necessary devices is limited
interleaved operation increases the common mode circulating and widespread power semiconductors can be used. The DC
currents between the parallel connected inverters. Therefore, link voltage was set to 1500 V in the end, because then the
control techniques for minimizing circulating current have to usage of mature 1.2 kV SiC MOSFETs is possible. This is
be implemented later on [21] [22]. very favorable, because there are a large number of such
The DC link voltage for the 10.5 MW inverter was set to devices from different manufacturers available today and it
1500V because it is a voltage level that seems to be a also allows the building of a first prototype in the near future.
promising candidate for future aircraft applications. In [2] it The number and maturity of available 1.7 kV and 3.3 kV SiC
was shown that the optimal DC-link voltage for future aircraft MOSFET is still very limited today.
applications lies somewhere in the range of VDC = 2 kV. It The usage of silicon carbide devices is necessary for the
represents a good trade-off between the cable masses and the proposed motor design, because of the high electric motor
usable power modules for the inverters. It should be noted that frequency of 8.67 kHz at a speed of 13,000 rpm a high PWM
power semiconductors at high altitudes are prone to cosmic switching frequency of at least 90 kHz for a single subsystem
radiation-induced failures because of the increased particle is required. This is only feasible with SiC MOSFETs because
flux. High energy particles can generate “single event IGBTs can’t handle such high switching frequencies. The
burnout” failures. Measurements show that for aircraft usage of nowadays available GaN devices was not considered
applications the applied DC link voltage should be limited to here, because they have a limited breakdown voltage of only
50% of the devices breakdown voltage for silicon devices 650 V. GaN devices with 1200 V rating are only in the first

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
prototype phase and need further development. inverter topologies depend on the operating point of the
inverter (modulation factor, power factor), the voltage and
For the inverter, three different 3-level inverter topologies current levels, and the switching frequency. The evaluation of
Neutral Point Clamped (3L-NPC), Active Neutral Point the losses shows that the conduction losses outweigh the
Clamped (3L-ANPC), and Flying Capacitor (3L-FLC) have switching losses. This is not surprising as for the investigated
been investigated and compared. One phase leg of each 3-level topologies the motor current has always to flow
inverter topology is shown in Fig. 7. The main disadvantage through two power semiconductors in series. Therefore it is
of the 3-level NPC inverter is the unequal loss distribution beneficial to choose power semiconductors that produce
among the power semiconductors because the neutral current minimal conduction losses. Four different 1.2 kV SiC
path is passively determined by the load current direction. MOSFETs with low RDS(on) values of 9 mΩ and 16 mΩ from
Therefore the Active NPC inverter was introduced to solve different manufacturers have been selected for the loss
this problem. The clamping diodes of the NPC inverter are comparison of the 3-level inverters. SiC MOSFETs A and B
replaced by active switches. These switches enable to use are based on a cascode configuration, where a SiC JFET is co-
different zero states in the commutation strategy and therefore packaged with a cascode optimized low voltage Si-MOSFET.
the loss distribution can be changed and balanced [14] [15]. For the loss calculation of the NPC clamping diodes a 50 A
Compared to conventional 2-level inverters, 3-level SiC diode has been chosen. The extracted datasheet
inverters have some distinct advantages and disadvantages. parameters of the power semiconductors are listed in Table
VII and Table VIII.
For each topology, the total losses were analytically
determined for the operating point of 10 MW mechanical
output power. Several justified simplifications were made for
the calculations to get a first rough system design and
comparison, which can then be refined later on with a more
accurate modelling, as discussed in [16]. The conduction and
switching losses have been calculated with the datasheet
parameters of the chosen SiC devices. The driving losses,
blocking losses and dead time losses are small and thus have
been neglected here.
For the switching losses a linear dependence of the
Fig. 7. Inverter legs of different inverter topologies: 2L, NPC-3L, ANPC-3L switching energies from the voltage and current levels have
and FLC-3L been assumed and they are scaled by the given reference
values from the datasheet (see Fig. 8). For the conduction
A few should be given here. It is clear that the 3-level losses of the clamping diodes in the NPC inverter, the voltage
inverters need a higher number of semiconductors and gate drop of the diodes is linearly approximated with a constant
drivers. The control effort for the 3-level inverters is also term Vf0 (knee voltage) and a linear term Rf = ∆v/∆i. For the
higher, as the DC link voltage levels have to be balanced. But calculation of the conduction losses of the SiC MOSFETs, the
3-level inverters enable to use a third voltage level for the parts of the formulas with the knee voltage are dropped
generation of the output voltage. Thus the output voltage is because they are unipolar devices. There are no external anti-
approximated closer to the desired sine waveform. Therefore parallel diodes and synchronous rectification is used for the
the THD value of the output current can be reduced, which SiC MOSFETs, i.e. there would be only a current flow through
helps to reduce the motor losses. the intrinsic body diodes during the dead time, which is
The main goal in the system design phase was to find the neglected here. This also implies that the conduction losses
DC/AC inverter topology, which produces minimal losses. during reverse conduction and the reverse recovery losses of
The necessary total die area and the number of needed gate the intrinsic body diodes are attributed to the corresponding
drivers are further evaluation criteria. The total losses of the SiC MOSFET.

TABLE VII. COMPARISON OF DIFFERENT SIC MOSFTETS

Die Size in VDS in RDS(on),25°C in RDS(on),125°C in Eon in Eoff in Eges in Iref in Uref in
Package
mm2 V mΩ mΩ mJ mJ mJ A V
SiC MOSFET A Bare-Die 37.5 1200 9 13.3 3.50 0.70 4.20 100 800
SiC MOSFET B Bare-Die 20.0 1200 16 24.8 2.82 0.15 2.97 80 800
SiC MOSFET C Bare-Die 26.0 1200 16 21.6 2.12 0.62 2.74 75 800
SiC MOSFET D Bare-Die 24.0 1200 16 22.4 6.25 1.10 7.35 100 800

TABLE VIII. 1.2 KV SIC SCHOTTKY DIODE FOR NPC INVERTER

Die Size in VDS in Rf,125°C in Err in Iref in Uref in


Package Vf0,125°C in V
mm2 V mΩ mJ A V
SiC Diode Bare-Die 24.0 1200 0.809 24.9 0.3 36 600

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
I] !
/[\IJ S U= <
It is also assumed that the on-state resistance for the
/ (5)
MOSFETs is approximately the same for the first and third
quadrant. It is also assumed that the on-state resistance for the with /[\IJ the total losses of an inverter switch with the
MOSFETs is approximately the same for the first and third number of parallel bare dies, / the switching losses and
quadrant.
I] ! the conduction losses under the assumption that one
The formulas for the loss calculation of the 2-level inverter bare die carries the whole current. The power loss of a single
are e.g. given in [17], for the 3-level inverters NPC and ANPC bare die has to be limited to not exceed the allowed maximum
they can be e.g. found in [18] and for the FLC inverter in [19], junction temperature. To increase the lifetime of the power
[20]. For a fair comparison, the PWM frequency of the FLC modules and the number of power cycles a maximum
inverter was halved to 45 kHz to get the same output current allowable mean junction temperature ^,H averaged over one
ripple. For the ANPC inverter, the PWM modulation strategy sinus period is defined. For this study, the value was set to
“PWM 4” from [18] was used, which reduces the conduction ^,H = 125°C and the heatsink temperature was defined as
losses in the inner switches S2, S3, S5, and S6. All four J = 65°C. Therefore the maximum allowable temperature
switches in the clamping path are turned on together and the difference is calculated to ∆ = ^,H − J = 60 K. With the
current can flow in either direction through both clamping die size e![ and the area specific thermal resistance from
paths, which reduces the conduction loss in the corresponding junction to heatsink f\ ,g , the power loss of a single bare die
SiC MOSFETs. The losses for the outer switches S1 and S6
![ can be expressed as:
are not influenced and equal to the NPC-3L inverter. ∆ ∙ e![
![ =
f\J,h
(6)
For the high output currents, a parallel connection of n bare
dies for each inverter switch is necessary. It is assumed that From (5) the power loss of a single bare die ![ can be also
the current is equally shared between the dies, so that each expressed as:
/[\IJ S U
bare die carries 1/n of the total current and generates 1/n of the
/ I] !
![ = = < ;
total switching losses. Accordingly, the switching losses of (7)
one inverter switch is the sum of the switching losses of the
parallel dies and becomes independent of the number of By combining (6) and (7) and solving for n, we get:
parallel dies under the assumption that the switching losses are
;∆ ∙ e![
<i <4∙ ∙ I] !
linearly dependent on the current and the simplification that / / f\J,h
= (8)
∆ ∙ e![
the switching losses are zero for a switched current of zero
2∙
f\J,h
ampere (see Fig. 8). The conduction losses for a unipolar
switch with n parallel bare dies are reduced by the factor 1/n.
Equation (8) allows calculating the minimum number of
parallel bare dies, which have to be connected in parallel to
not exceed the given temperature difference ∆T or the given
mean junction temperature ^,H . Since only discrete numbers
of bare dies are available, n has to be rounded up to a whole
number. The switching and conduction losses are calculated
with the loss equations for the different inverter topologies and
the semiconductor parameters are taken from Table VII and
Table VII regarding the desired mean junction temperature of
^,H = 125°C. For the area specific thermal resistance f\ ,g a
value of 16 K∙mm2/W was chosen, which can be reached with
an advanced power module design with single-sided water
cooling.
Fig. 8. Example for linearization of switching losses

TABLE IX. COMPARISON OF DIFFERENT INVERTER TOPOLOGIES AND SIC MOSFETS

SiC MOSFET A SiC MOSFET B SiC MOSFET C SiC MOSFET D


NPC- ANPC- FLC NPC- ANPC- FLC NPC- ANPC- FLC NPC- ANPC- FLC-
Inverter Topology
3L 3L -3L 3L 3L -3L 3L 3L -3L 3L 3L 3L
Efficiency 98.49% 98.83% 98.83% 98.53% 98.84% 98.86% 98.48% 98.80% 98.85% 98.01% 98.36% 98.41%
Total losses in kW 161.1 124.6 125.0 156.8 123.6 121.3 162.2 128.0 122.9 213.5 175.4 170.3
SiC MOSFET area in cm2 375.3 402.1 375.3 331.4 345.8 345.8 337.2 356.0 374.7 483.3 517.8 483.3
SiC Diode area in cm2 103.7 -- -- 103.7 -- -- 103.7 -- -- 103.7 -- --
No. of dies for S1 and S4 9 9 7 14 14 12 11 11 10 20 20 14
No. of dies for S2 and S3 5 4 7 9 6 12 7 5 10 8 6 14
No. of dies for S5 and S6 6 2 -- 6 4 -- 6 3 - 6 4 -
No. of gate drivers 144 216 144 144 216 144 144 216 144 144 216 144
Total No. of SiC MOSFETs 1008 1080 1008 1656 1728 1728 1296 1368 1440 2016 2160 2016
Total No. of SiC Diodes 432 -- -- 432 -- -- 432 -- -- 432 -- --

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
AM4
The comparison of the losses for the different SiC

A
MOSFETs and inverter topologies is shown in Table IX. The +
V S11_1 S21_1 S31_1
U_ZK_1
most efficient inverter is the flying capacitor inverter FLC-3L S15_1 S12_1 S25_1 S22_1 S35_1 S32_1
for all four different SiC MOSFETs. The ANPC-3L inverter +
S16_1 S13_1 S26_1 S23_1 S36_1 S33_1

is a little bit less efficient than the FLC-3L inverter but distinct 0 V

U_ZK_2
more efficient than the NPC-3L inverter. The SiC MOSFET B S14_1 S24_1 S34_1
U1_1 V1_1 W1_1
with a nominal RDS(on) value of 16 mΩ generates the lowest
losses and has the lowest required die area for all three 3-level S11_2 S21_2 S31_2
inverter topologies. However, for this SiC MOSFET the
S25_2
efficiency difference between the FLC-3L and the ANPC-3L S15_2
S16_2
S12_2
S13_2
S22_2
S23_2
S35_2
S36_2
S32_2
S33_2
is very low and the total bare die area is identical. Therefore S26_2

the two topologies FLC-3L and the ANPC-3L are both viable S14_2 S24_2 S34_2
candidates for the final system design. The main drawback of U1_2 V1_2 W1_2
the FLC-3L inverter is the need for an additional (flying)
capacitor per phase leg, i.e. 36 in total for the complete S11_3 S21_3 S31_3

inverter, compared to the ANPC-3L inverter. The capacitance


S15_3 S12_3 S25_3 S22_3 S35_3 S32_3
was sized via simulation and it was determined to be in the S16_3 S13_3 S26_3 S23_3 S36_3 S33_3
range of 500 µF, which will clearly add weight and volume to
the inverter system and reduces the inverter power density. S14_3 S24_3 S34_3

The main drawback of the ANPC-3L inverter is the need for U1_3 V1_3 W1_3

two additional active switches per phase leg compared to the Fig. 9. Co-simulation model for the 3-times interleaved ANPC-3L SiC-
FLC-3L inverter, which increases the number of gate drivers inverter in Simplorer
and associated signals. Targeting a high power density design, i_U1_1 i_U1_2
the ANPC-3L inverter was chosen for the final system design FEM
AM5
and it is used for the buildup of the co-simulation model. U1_1 p

A
AM6 PhaseU PhaseB
X0

U1_2 Y0
p

A
AM7 Z0

III. DIGITAL TWIN FOR FUTURE ELECTRIC AIRCRAFT U1_3


p

A
ROT2

i_U1_3
i_U1_ist
APPLICATION i_V1_1 i_V1_2

AM8
A. Overview of the Co-Simulation for the Electrical Drive V1_1 PhaseC
p

AM9 PhaseV
System V1_2
p

A
AM10
V1_3
p

A digital twin was developed using a co-simulation that i_V1_3 i_V1_ist


i_W1_1 i_W1_2
combines finite element models for the electromagnetic
behavior of the 10 MW electric machine and mathematical AM11 PhaseA

W1_1
p

AM12 PhaseW
models for the power electronic (ANPC-3L) and the control W1_2
p

AM13
method [23] [24]. The 10 MW electric machine was realized W1_3 i_W1_ist
p

i_W1_3 ROT1

by the finite element method (FEM) and the multilevel Mechanical_Angle


+

+
Π

inverter was realized with 54 switches for each quarter Speed_Sensor


machine and 216 switches in total for the whole machine,
which allows virtualizing the behavior of the multilevel SiC- Fig. 10. Co-simulation model for the 10 MW electric machine (FEM) in
inverters. The electromagnetic and electric circuits were Simplorer
coupled using Ansys/Simplorer as an interface (see Fig. 9 and
Fig. 10). The FEM model allows coupling between non-linear
magnetic and electric circuits and provides a realistic
performance characteristic for the co-simulation [25]. This is
especially interesting for the characterization of the multilevel
SiC-inverter and the influence of the switching frequency on
the motor behavior within high power density and high power
drive systems. The torque ripple and iron losses of the motor
are highly dependent on the current ripple and therefore on the
PWM switching frequency of the inverter system.
The co-simulation model requires a current controller
model to implement the control strategy, which allows the
electric machine to provide the maximum power and controls
Fig. 11. Co-simulation model for current controller and modulation
all the switches for the multilevel inverter for the generation systems for the multilevel SiC-inverter
of the desired motor currents. The current controller that
generates the control signals for the switches is shown in Fig. Three subsystems with 120 degree phase shift for the
11. The current controller calculates the duty cycles for the interleaving of the inverters take the calculated duty cycles
semiconductors depending on the desired and actual motor and generate the control signals for the semiconductors. The
currents. Space Vector Modulation (SVM) is used as control signals for the different phases are connected to the
modulation scheme, which adds a zero sequence to the duty power electronic model, which produces the required motor
cycles and allows the usage of the complete DC link voltage phase currents for the FEM-Model of the 10 MW electric
range. machine. The co-simulation model can calculate the realistic
inverter output currents and voltages including the impact of

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
the additional harmonics, which are generated by the
multilevel semiconductor switching. The magnitude and
frequency of these harmonics depend on the inductances of
the multilevel inverter, the inductances of the electric
machine, and the PWM switching frequency of the multilevel
inverter. These parameters have a strong influence on the
overall machine electromagnetic torque, torque ripple, and on
the additional losses of the electric machine.
B. Comparison of the Simulation Results
Fig. 12 shows the simulated phase current waveforms and
the voltage waveforms for the 10 MW electrical drive system
with a PWM switching frequency of 90 kHz. The FFT results
for the phase currents and line voltages with a switching
frequency of 90 kHz are shown in Fig. 13. The results show
that the main harmonics of the electrical drive system are
90 kHz, 180 kHz, and 270 kHz. As described in chapter II A,
the THD value can be used for the evaluation of the harmonic
waves. The THD results are shown in Table X.

Fig. 13. FFT results for the current waveforms (top) and the voltage
waveforms (bottom) for a inverter switching frequency of 90 kHz
Due to the harmonics from the interaction of the inverter
with the electric machine, the electric machine generates more
torque ripple which leads to a higher load on the gearbox.
Additionally, the harmonic-induced additional stator and rotor
losses must be considered for the thermal design and the
efficiency calculation of the electric machine. Fig. 14 shows
the electromagnetic torque from the co-simulation model. The
comparison for the torque ripples and the stator and rotor
losses between the normal FEM model and the co-simulation
are shown in Table XI. The comparison shows that the
electromagnetic torque from the co-simulation is lower and
the torque ripple is higher compared to the normal FEM
simulation. The inverter has a strong influence on the
electromagnetic torque and the torque ripple. Only the co-
simulation can take the phase currents with the inverter caused
harmonics into account.

Fig. 12. The simulated current waveforms (top) and the voltage waveforms
(bottom) for an inverter switching frequency of 90 kHz

TABLE X. THD VALUES OF MOTOR CURRENTS AND VOLTAGES


THD values Current Voltage
Normal FEM - 2.4 %
Co-simulation 0.6 % 21.6 %

A co-simulation was also developed for an automotive


application with a 6-phase 150 kW IPMSM motor and a two-
level SiC-inverter with a PWM switching frequency of Fig. 14. Electromagnetic torque of the PMSM with outer-rotor
40 kHz, which has about 3.4 % higher current THD values. TABLE XI. COMPARISION OF THE MACHINE LOSSES
The electrical drive system for the aircraft application with the Component Normal FEM Co-simulation
three-times interleaved ANPC-3L inverter topology and the
Torque 7530 Nm 7275 Nm
friction slot electric machine has clearly smaller harmonic
values than the automotive application. Torque ripple 55.1 Nm 307.4 Nm
Stator iron losses 18.4 kW 19.1 kW
Rotor losses 17.3 kW 17.5 kW

The torque values are treated as important parameters to


validate the design of the electric machine. The stator and
rotor losses for the co-simulation show only slightly higher
values compared to the FEM results. The interleaved

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.
multilevel inverter topology generates quite low THD current [6] A. Binder, “Betriebsverhalten der Vollpolmachine am starren Netz,” in
values, which results in low additional losses in the PMSM Elektrische Maschinen und Antriebe, Springer, Band 2., 2017, pp.539-
540.
due to the harmonics. The harmonics losses lead to more
[7] G. Mueller, K. Vogt and B. Ponik, “Stromverdrängung,” in
heating and lower efficiency of the drive system and can’t be Berechnung elektrischer Maschien, WILEY-VCH, Band 6., 2008, pp.
neglected. By combing the FEM motor model with the 385-423, 438.
inverter model, the co-simulation shows to be a better method [8] A. Binder, “Permanetmagnet-Synchronmaschinen mit Zahnspulen und
to calculate and validate the electrical drive design. It was also High-Torque-Synchronmaschinen,” in Elektrische Maschinen und
shown that even for a high power 10 MW electrical machine Antriebe, Springer, Band 2., 2017, pp.708-737.
and a quite complicated inverter system the creation of a [9] ANSYS, https://www.ansys.com/products/electronics/electric-motors
digital twin with a co-simulation is possible and allows to [10] Waelzhholz, Elektrobandwerkstoffe und Lacksysteme (2019).
verify and optimize the system design. Accessed: July 19, 2021. Available online:
https://waelzholz.s3.amazonaws.com/website/Website-
IV. SUMMARY AND CONCLUSION DE/Documents/Broschueren/waelzholz-elektroband-2021-06-de.pdf
[11] Narayanan, G.; Ranganathan, “Analytical Evaluation of Harmonic
The concept and the realization of a digital twin for a Distortion in PWM AC Drives Usingthe Notion of Stator Flux Ripple,”
10 MW electrical drive system and the comparison with the IEEE Trans. Power Elctron, 2005, vol. 20, pp. 466–474.
traditional FEM simulation results are presented within this [12] Y. M. You, “Optimal Design of PMSM Based on Automated Finite
paper. The digital twin development was implemented for an Element Analysis and Metamodeling,” MDPI Energies 2019, 12, 4673,
outer-rotor electric machine with high power density and a doi.org/10.3390/en12244673.
SiC-based interleaved multilevel inverter system enabling the [13] C. Felgemacher, S. V. Araújo, P. Zacharias, K. Nesemann and A.
Gruber, “Cosmic radiation ruggedness of Si and SiC power
use of high PWM switching frequencies. Different three-level semiconductors,” 2016 28th International Symposium on Power
inverter topologies have been compared and it was shown that Semiconductor Devices and ICs (ISPSD), 2016, pp. 51-54, doi:
the ANPC-3L or the FLC-3L can be used for the buildup of a 10.1109/ISPSD.2016.7520775.
10.5 MW inverter system. The FEM and circuit simulation [14] H. Liu, J. Jiang and W. Luo, “A Comparative Analysis of the Three-
based approach for the model was described including the Level NPC and ANPC Converter Loss Distribution”, J. Electrical
interfaces and coupling between the different simulation Systems 11-3 (2015), pp. 271-280.
environments. [15] Y. Jiao and F. C. Lee, “New Modulation Scheme for Three-Level
Active Neutral-Point-Clamped Converter With Loss and Stress
A comparison between the co-simulation results and the Reduction,” in IEEE Tran. on Ind. Elec. , vol. 62, no. 9, pp. 5468-5479,
normal FEM results was given to validate the digital twin. The Sept. 2015, doi: 10.1109/TIE.2015.2405505.
improved co-simulation model showed a higher ratability to [16] H. Hafezi and R. Faranda, “A New Approach for Power Losses
standard FEM simulation models and it can analyze the Evaluation of IGBT/Diode Module,” Electronics, vol. 10, no. 3, p. 280,
Jan. 2021.
harmonics of such a complicated drive system. It could be
[17] R. Yapa, A. J. Forsyth and R. Todd, “Analysis of SiC technology in
seen that the switching frequency of the inverter has a two-level and three-level converters for aerospace applications,” 7th
significant influence on the electromagnetic torque ripple and IET International Conference on Power Electronics, Machines and
the machine losses. However, the high switching frequency Drives (PEMD 2014), 2014, pp. 1-6, doi: 10.1049/cp.2014.0498.
will also add losses in the multilevel inverter. Therefore a [18] J. Dodge, “3L-ANPC vs.3L-NPCInverters”, Application Note AN0023
combined investigation including the electric machine and the United SiC, 2020, Available online: https://unitedsic.com/appnotes/
inverter system is addressed in future work in order to achieve UnitedSiC_AN0023_inv_comp.pdf (accessed on 25 July 2021)
a much higher power density for a 10 MW electrical drive [19] F. Gebhardt, H. Vach and F. W. Fuchs, “Analytical derivation of power
semiconductor losses in MOSFET multilevel inverters,” 2012 15th
system. Further work will also address a more detailed study International Power Electronics and Motion Control Conference
on the rotor magnets, which have high magnet losses due to (EPE/PEMC), 2012, pp. DS1b.18-1-DS1b.18-6, doi:
the higher switching frequency. 10.1109/EPEPEMC.2012.6397218.
[20] V. Dargahi, A. Khoshkbar-Sadigh and K. Corzine, “Analytic
V. REFERENCES determination of conduction power losses in flying capacitor multicell
power converter,” 2015 IEEE Applied Power Electronics Conference
[1] P. Kshirsagar et al., “Anatomy of a 20 MW Electrified Aircraft: Metrics and Exposition (APEC), 2015, pp. 2358-2364, doi:
and Technology Drivers,” 2020 AIAA/IEEE Electric Aircraft 10.1109/APEC.2015.7104678.
Technologies Symposium (EATS), 2020, pp. 1-9..
[21] G. J. Capella, J. Pou, S. Ceballos, J. Zaragoza and V. G. Agelidis,
[2] Filipenko, Mykhaylo & Biser, Stefan & Boll, Martin & Corduan, “Current-Balancing Technique for Interleaved Voltage Source
Matthias & Noe, Mathias & Rostek, Peter. (2020). Comparative Inverters With Magnetically Coupled Legs Connected in Parallel,” in
Analysis and Optimization of Technical and Weight Parameters of IEEE Transactions on Industrial Electronics, vol. 62, no. 3, pp. 1335-
Turbo-Electric Propulsion Systems. Aerospace. 7. 107. 1344, March 2015, doi: 10.1109/TIE.2014.2345345.
10.3390/aerospace7080107.
[22] J. Burkard, M. Pfister and J. Biela, “Control Concept for Parallel
[3] G.Roeder, X. Liu, M. Hofmann, and M. Schellenberger, “Cognitive Interleaved Three-Phase Converters with Decoupled Balancing
Power Electronics for Intelligent Drive Technology,” in 10th Control,” 2018 20th European Conference on Power Electronics and
International Electric Drives Production Conference (EDPC), Applications (EPE'18 ECCE Europe), 2018, pp. P.1-P.9.
Ludwigsburg, Germany, 2020, pp.1-1, doi:
10.1109/EDPC51184.2020.9388182. [23] T. Inoue, Y. Inoue, S. Morimoto, and M. Sanada, “Maximum torque
perampere control of a direct torque controlled PMSM in a stator flux
[4] A. M. EL-Refaie, Z. Q. Zhu, T. M. Jahns, and D. Howe, “Winding linkage synchronous frame,” IEEE Trans. Ind. Appl., vol. 52, no. 3, pp.
inductances of fractional slot surface-mounted permanet magnet brush- 2360–2367, May/Jun. 2016.
less machines,” in Cof. Rec. IEEE IAS Annu. Meeting, Edmonton, AB,
Canada, Oct. 2-5, 2008, pp. 1-8. [24] T. Inoue, Y. Inoue, “Mathematical Model for MTPA Control of
Permanent-Magnet Synchronous Motor inStator Flux Linkage
[5] A. M. EL-Refaie, “Fractional-Slot Concentrated-Windings Synchronous Frame,” IEEE Trans. Ind. Appl., vol. 51, no. 5, pp. 3620-
Synchronous Permanent Magnet Machines: Opportunities and 3628, Sep. 2015.
Challenges,” IEEE Trans. Ind. Elec., 2010, Vol.57, p.107-12
[25] A. A. Arkadan and B. W. Kielgas, “The coupled problem in switched
reluctance motor drive systems during fault conditions,” Magnetics,
IEEE Transactions, vol. 30, issue 5, pp. 3256-3259, 1994.

Authorized licensed use limited to: Indian Institute of Space Science And Technology. Downloaded on February 07,2024 at 12:16:57 UTC from IEEE Xplore. Restrictions apply.

You might also like