PCIM Europe digital days 2021, 3 – 7 May 2021

Modular Multilevel Converter for Variable Speed Operation of

Pumped Storage Hydropower Plants
Raghbendra Tiwari, Roy Nilsen, Arne Nysveen

Norwegian University of Science and Technology (NTNU), Norway

Corresponding author: Raghbendra Tiwari, raghbendra.tiwari@ntnu.no

Abstract
This paper presents the Modular Multilevel Converter topology as a potential alternative for the full-size
converter fed synchronous machines in range of 100 MVA in a pumped storage power plants. The suitability
of the converter has been analyzed based on the torque it can provide at standstill to drive the turbine
in pump mode with water in the turbine casing. The basis of analysis is the conduction and switching
losses incurred in the semiconductor devices at standstill compared to the rated operation to determine
the possible output current from the converter, which consequently transforms to torque availability. The
analysis also takes into account the ripple in the submodule capacitor voltages occurring at low or zero
frequency operation.

1 Introduction match the stator power rating and the transformer

at both ends of the converter are installed to match
Pumped storage power plants are normally realized the machine and grid transformer voltages.
with fixed speed synchronous machine as in a
Three-level voltage source converter and Modular
conventional hydropower plant with a reversible
Multilevel Converter (MMC) topologies are
pump turbine (RPT). Such plants are only used for
suggested as possible solutions for this application
pumping with fixed power for a long term operation,
[4]. Due to the modular structure of MMC, any
e.g. weekly or seasonal in some cases. To achieve
output voltage to meet the stator voltage rating can
the dynamic regulation of power in pump mode
be realized by connecting several submodules in
and to run the RPT at optimal efficiency, variable
series. But, it is a well known problem of MMC
speed operation of pumped storage plants has
at very low and zero frequency operation that
been realized using doubly fed induction machines
the arm currents become close to dc current
(DFIM). A list of large scale plants around the world
and monotonically either charges or discharges
with DFIM technology is listed in [1]. The DFIM
the submodule capacitor. In [5] and [6], the
technology takes several minutes (5–10 minutes)
control strategy for low frequency and standstill
for start up of the system as the water inside the
operation of MMC have been proposed such that
turbine casing needs to be compressed below the
the amplitude of submodule capacitor voltage
turbine level [2]. Considering the future growth of
remains within a pre-specified limit of ±10 − 15%.
renewable energy sources, very fast start up of the
In [6], it has been shown that 40% of rated torque
plant in pump mode or a transition from generating
can be achieved from a PWM modulated MMC at
to pumping mode or vice versa is imminent. Such
standstill without exceeding the ripple in submodule
dynamic feature can only be achieved by installation
capacitor voltage more than 20%.
of a full-size converter between the stator of
the machine and the grid. Such converter fed In this paper, a full size modular multilevel converter
synchronous machine (CFSM) technology exists for a 100 MVA, 13–15 kV synchronous machine in
at Grimsel-2 plant in Switzerland where one of the a pumped storage power plant is analyzed based
100 MW units with synchronous machine has been on its starting torque requirement in pump mode
supplied with a full-size converter [3]. Two active and in case of fast transition from generating mode
neutral point clamped converter are employed to to pump mode. Analytical loss expressions have

ISBN 978-3-8007-5515-8 1361 © VDE VERLAG GMBH · Berlin · Offenbach

Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
 PCIM Europe digital days 2021, 3 – 7 May 2021

been used to determine the losses in the devices output current (io ) and the dc-link currrent (idc ) at
at rated speed and at standstill. The starting torque steady state operation are as:
capability has been determined based on the total
idc SM 1
loss at low frequency or standstill operation which T1
must not exceed that at rated operation. In addition, SM 2
D1
the capacitor voltage ripple must also be within a
SM N uc
limit of 20% as suggested in [5] and [6]. T2
L usm D2
iau
Udc R
2 Application requirements and ia

Converter Topology
R ial

2.1 Torque–Speed Characteristics of L

Reversible Pump Turbine SM 1

SM 2
The design of the RPTs are normally optimized for
pump operation and therefore, are more efficient SM N
in that region. From a typical torque speed
characteristics presented Fig. 1, the torque required Uan Ubn Ucn

to start the RPT in pump mode from standstill Fig. 2: Schematic diagram of a Modular Multilevel
is about 12% with minimum guide vane opening Converter with half-bridge sub-module
configuration. In this figure, ”SM” stands
(α = 1◦ ), i.e. the converter needs to supply dc
for submodule.
current around 12% of its rated current to start the
RPT in pump mode when the turbine is submerged 1
in the water. The torque required for fast transition iu = io + Iz,dc + iz,ac (1)
2
from generating to pumping mode varies from 40
– 120% depending upon the guide vanes opening. 1
il = − io + Iz,dc + iz,ac (2)
Therefore, the mode reversal from generation to 2
pump to adapt fast variation of renewable sources where,
can be demanding for many converter topologies. io = iu − il = Iˆo sin(ωt + ϕ) (3)

1
icirc = (iu + il ) = Iz,dc + iz,ac (4)
2
M Iˆo (5)
Iz,dc =
cos ϕ
4
Here, Iˆo is the peak of output current, M is
the peak of modulation index and cos ϕ is the
power factor. The circulating current (icirc )
consists of dc component (Iz,dc ) and ac harmonics
component (iz,ac ) which predominantly consists of
Fig. 1: Torque speed characteristics of a reversible 2nd harmonics of the output current in an open
pump turbine at different guide vane openings
loop phase shifted PWM modulated MMC. There
(α). The speed is positive in generating
mode and negative in pumping mode. The
exists several control methods to eliminate these
shaded region shows the rated operating region. current harmonics, known as circulating current
Courtesy: Water Power Laboratory, NTNU, suppression (CCS) control [8]–[12] such that only
Trondheim. the dc-component of the circulating current flows
through the arms. Then the equations (1) and (2)
2.2 Modular Multilevel Converter becomes:
The schematic of the MMC is presented in Fig. 2
[7]. The relation between the arm currents (iu,l ), the Iˆo (6)
iu,l = ± sin(ωt + ϕ) + Iz,dc
2

ISBN 978-3-8007-5515-8 1362 © VDE VERLAG GMBH · Berlin · Offenbach

Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
 PCIM Europe digital days 2021, 3 – 7 May 2021
The CCS control remains active in normal operation
to eliminate the higher harmonics current through
the arm which causes additional losses in the
semiconductor devices and passive elements in
the arms.
Unlike in the voltage source converters (VSCs),
the load current itself flows through the submodule
capacitors in an MMC. Hence, the amplitude and
frequency of the output current have direct impact
on the voltage ripple in these capacitors. The
expression for ripple in submodule capacitor voltage
(ũc,u ) has been approximated in [13] as:
Iˆo  M 2 cos ϕ
ũc,u = cos ωt − cos (ωt − ϕ)
8πfo C 2 (7)
M 
+ sin (2ωt − ϕ)
4 should be scaled, the arm currents for phase “a”
From (7), it is quite obvious that the capacitor
voltage ripple increases as the output current (Iˆo )
increases or output frequency (fo ) decreases which
is the case in drives application.
2.3 Low Speed Operation
All semiconductor devices have either maximum
conduction loss or maximum switching loss or both
at zero speed. The total loss in the devices decide
the current limit the converter can yield.
Considering V /f control of the machine where the
amplitude of the applied voltage to the motor is
decreased as the speed decreases to maintain
constant (1 pu) flux in the stator of the motor, at very
low frequency or zero frequency, M is very small
(0.03 – 0.05 pu) to compensate only the drop in the
resistance (converter components, stator winding
of the motors and others). According to (5), the dc
component of the arm current (Iz,dc ) approaches
zero for very low value of M. In this case,
M · vdc
V̂ph,peak = (8)
2
Iˆo Iˆo
iu,a = and il,a = − (9)
2 2
The output current in all phases and all arms will be
dc current but the instantaneous sum of the three
phase output currents will still be zero. As shown
in Fig. 2, if the output current is positive dc as in
case of bridge leg “A”, the current flows through the
ISBN 978-3-8007-5515-8
Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
capacitor via D1 and charges it or it bypasses the
capacitor when the switch T2 conducts. Thus, the
capacitor in the upper arm is either only charged
or stays at same value. A longer duration (at low
frequency or zero frequency operation) can charge
the capacitor to an unacceptably high value above
its rated value and the opposite will happen in the
lower arm. To avoid this continuous charging and
discharging of the capacitors, the arm current needs
an ac component superimposed with the dc current
such that the resultant current both charges and
discharges the capacitor and keeps the ripple within
the permissible limit.
Assuming the ac component as sinusoidal current
of angular frequency ωcom , and constants k1 and k2
as the factor by which the dc and ac components
can be formulated as:
Iˆo
iu,a = k1 + k2 Io sin(ωcom t)
2 (10)
Iˆo
il,a = − k1 + k2 Io sin(ωcom t)
2
The simple way to solve this problem could be to
induce common mode sine wave current in the arms
in such a way that the arm currents in (10) become
equal to (6). Assuming the induced common
mode current has the same frequency as the rated
nominal frequency, the ac current will now circulate
within the arms whereas the dc current will flow out
of the bridge legs into the windings of the motor.
By flipping the position of the ac and dc part of the
currents in (6), it can be observed that the maximum
DC output current can be half of the nominal output
current to keep the peak of the arm current same
as in the case of rated frequency operation. For
example, with Iˆo = 100A, M = 1 and cos ϕ = 1,
the instantaneous upper arm current using (6) is
iu = 50 sin ωt + 25 A. Now, at zero speed, when the
output current is only dc current, a similar waveform
to match the peak of the upper arm current at rated
frequency operation could be achieved by common
mode sinusoidal current injection as in (10) and can
be made as: iu = 25 + 50 sin ωcom t A. In the latter
equation, 25 A is the maximum dc current from
each arm and hence a total of 50 A from a bridge
leg that can be delivered from the converter without
exceeding the current rating of the devices. Hence,
the theoretical limit of the output torque from a half
1363 © VDE VERLAG GMBH · Berlin · Offenbach
 PCIM Europe digital days 2021, 3 – 7 May 2021

bridge MMC at zero frequency can be maximum 3.1.1 Conduction Loss

of 50% if the devices are designed for rated speed
The conduction loss in a semiconductor device is
operation.
calculated using (18):
A high frequency sinusoidal current injection in
2
the arm currents was first proposed in [5] and Pcon,loss = UCE0 · Iavg + RCE,on · Irms (18)
later a square wave harmonic injection method has
followed in [6]. The amplitude of the common mode From (18), the conduction loss can be calculated
current through each arm as presented in [6] is as with the knowledge of the average and rms current
follows. through a device. The derived expression for the
average and rms current through each device of a
  submodule are as follows:
∗2
1 vo,ref vdc
iz,ac,ref = − io sin ωcom t
Vcom vdc /2 2
(11) Iˆo
∗
vo,ref io IT 1,avg = 4 − 2M 2 cos2 ϕ cos θz − M cos ϕ sin 2θz
+ − Iz,dc 16π
vdc (19)
The equation can be simplified to The average current through the switch T1 and diode
  D1 over a cycle are equal which results in keeping the
m2n −1 mn
iz,ac,ref = sin ωcom t + io − Iz,dc (12) average value of submodule capacitor voltage constant.
Mcom 2

where,
Iˆo 
Vcom IT 2,avg = 2 + M 2 cos2 ϕ cos θz
mn = M sin ωt and Mcom = (13) 8π
vdc /2 M  (20)
+ cos ϕ sin 2θz + M (π − 2θz ) cos ϕ
2
From (1) and (12):

(1 + m) m2 − 1 Iˆo 
ia,u = io + sin ωcom t (14) ID2,avg = 2 + M 2 cos2 ϕ cos θz
2 Mcom 8π
M  (21)
+ cos ϕ sin 2θz − M (π + 2θz ) cos ϕ
And, the total modulation signal excluding the 2
contribution from the CCS control becomes:
2 Io2  3 4M
m = M sin ωt + Mcom sin ωcom t (15) ID1,rms =− M cos 3ϕ cos θz − cos ϕ cos 3θz
128π 3
+ M 3M 2 − 4 cos ϕ cos θz
At low speed, as ω → 0 and M → 0, upper arm + 4 M 2 cos2 ϕ − 1 sin 2θz
current can be rewritten as: 
  + 2 (π − 2θz ) M 2 cos2 ϕ − 2
Iˆo 2
ia,u ≈ 1− sin ωcom t (16) (22)
2 Mcom
Io2  3 4M
and, IT2 2,rms = M cos 3ϕ cos θz − cos ϕ cos 3θz
128π 3
m ≈ Mcom sin ωcom t (17)
+ M 3M 2 + 28 cos ϕ cos θz
+ 4 M 2 cos2 ϕ + 1 sin 2θz
3 Analytical Loss Equations 
+ 2 (π − 2θz ) 3M 2 cos2 ϕ + 2
3.1 Rated Speed Operation (23)
3.1.2 Switching Loss
The authors have derived the expressions for
The switching loss characteristics as function of current,
calculating conduction and switching losses in each esw (i), from an experimental result can be approximated
device of a submodule. The loss expressions are as in (24).
based on the assumption that the CCS control is
enabled. esw (i) = k1T · i + k2T · i2 (24)

ISBN 978-3-8007-5515-8 1364 © VDE VERLAG GMBH · Berlin · Offenbach

Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
 PCIM Europe digital days 2021, 3 – 7 May 2021

where, k1T and k2T are the curve fitting coefficients

determined using the experimental results or  
manufacturer’s datasheet. The same coefficients Iˆo2 2
IT2 1,rms,dc
= 2
− 1 (π − 2θx )
for diodes are k1D and k2D . 16π Mcom
 
1
−2 − Mcom cos θx (31)
Mcom
Uc Io fsw   
PT 1sw = 4k1T (M (π + 2θz ) cos ϕ − 4 cos θz ) 1 2
32π  +2 2
− 1 sin 2θx − cos 3θx
Mcom 3Mcom
+ k2T Io (π + 2θz ) M 2 cos2 ϕ + 2

− 4 cos θz (sin θz + 2M cos ϕ)  
(25) 2 Iˆo2 2
ID1,rms,dc = 2
− 1 (π + 2θx )
16π Mcom
 
1
+2 − Mcom cos θx
Uc Io fsw 
(32)
Mcom
PT 2,sw = 4k1T (M (π − 2θz ) cos ϕ + 4 cos θz )  
32π  1 2
+ k2T Io (π − 2θz ) M 2 cos2 ϕ + 2 −2 2
− 1 sin 2θx + cos 3θx
 Mcom 3Mcom
+ 4 cos θz (sin θz + 2M cos ϕ)
(26)  
Iˆo2 2
where, IT2 2,rms,dc
= 2
+ 3 (π + 2θx )
16π Mcom
θz = sin−1 (−M cos ϕ/2) (27)  
7
+2 + Mcom cos θx (33)
Mcom
 
The diodes D1 and D2 follow the same expressions 1 2
−2 + 1 sin 2θx − cos 3θx
as that for switches T2 and T1 respectively but 2
Mcom 3Mcom
with loss coefficients k1D and k2D . These
expressions are used in later sections to calculate
the conduction and switching loss in a MMC at rated  
2 Iˆo2 2
speed operation. ID2,rms,dc = + 3 (π − 2θx )
16π 2
Mcom
 
7
3.2 Zero Speed Operation −2 + Mcom cos θx (34)
Mcom
 
In this section, the equations for average and rms 1 2
+2 2
+ 1 sin 2θx + cos 3θx
currents to calculate the conduction loss at standstill Mcom 3Mcom
condition are derived.
where,
  
Iˆo 2  
IT 1,avg,dc = − Mcom + sin θx cos θx Mcom
4π Mcom θx = sin−1 (35)
2
(28)

Iˆo  1   
IT 2,avg,dc = π + 2θx − sin 2θx Uc fsw Iˆo 8
4π  2  PT 1sw,dc = k1T (π − 2θx ) − cos θx
 (29) 8π Mcom
2 
+ + Mcom cos θx  
Mcom 2
+ k2T Io (π − 2θx ) 2
+1
Mcom
Iˆo  1 
ID2,avg,dc = π − 2θx + sin 2θx 8 2
4π  2  (30) − cos θx + 2 sin 2θx
2  Mcom Mcom
− + Mcom cos θx (36)
Mcom

ISBN 978-3-8007-5515-8 1365 © VDE VERLAG GMBH · Berlin · Offenbach

Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
 PCIM Europe digital days 2021, 3 – 7 May 2021

Tab. 1: Key Electrical Parameters of IGCT 5SHY

65L4521 and Diode FRD 5SDF 28L4520 at
   140◦ C [14].
Uc fsw Iˆo 8
PT 2sw,dc = k1T (π + 2θx ) + cos θx
8π Mcom
   Parameters Values
2
+ k2T Io (π + 2θx ) 2
+1 On-state Knee Voltage of IGCT (UCE0 ) 1.11 V
Mcom
 On-state Resistance of IGCT (RCE,on ) 0.297 mΩ
8 2 On-state Knee Voltage of Diode (UF 0 ) 1.10 V
+ cos θx − 2 sin 2θx
Mcom Mcom On-state Resistance of Diode (Rd ) 0.47 mΩ
(37)
Similar to the case at rated speed operation, the Tab. 2: Switching loss curve fitting coefficients of the
selected IGCT and Diode at 140◦ C.
expressions for switching loss in diodes D1 and
D2 are be same as that for switch T2 and T1
respectively. Parameters Values
k1T 4.7 × 10−3 J/A
4 Results k2T 3.17 × 10−3 J/A2
k1D 1.303 × 10−2 J/A
The loss analysis has been carried out using k2D −1.33 × 10−6 J/A2
phase shifted modulation technique such that no
additional capacitor balancing algorithm is required. fitting technique as expressed in (24) and presented
The converter discussed here is for a pumped in the Table 2.
storage power plant application where synchronous
machine is employed. The machine side converter The stack of the semiconductor devices are
in a back-to-back converter system will always be assumed to have double side cooled with 5 l/min
operated at unity power factor (cos ϕ = 1) in order to water (with 50 % Glycol) which gives parameters in
achieve the smallest possible size of the converter . the thermal circuit as shown in Table 3 [15].
Tab. 3: Thermal resistance of different section of the
The total loss in each device in a converter is
cooling circuit with double side cold plates [15].
the sum of average conduction loss and average
switching loss in that device over one fundamental
cycle as in (38). Parameters Values
Tambient 40 ◦ C
Ptot = Pcon + Psw (38) Rth,j−c 6.8 ◦ K/kW
For the analysis of the losses in the semiconductor Rth,c−h 2.2 ◦ K/kW
devices of the MMC, the IGCT 5SHY 65L4521, Rth,h−w 5.5 ◦ K/kW
2800 V, 6500 A from Hitachi ABB and the matching
diode diode FRD 5SDF 28L4520 are selected. 4.1 Rated Speed Operation
The IGBTs have higher switching loss compared
The total loss at rated frequency and unity power
to IGCTs at the same switching frequency, and
factor has been calculated for each device in a
therefore, are not favoured for this analysis. The
submodule using the expressions (19)–(27) and
key electrical parameters of the selected switch and
is presented in Fig. 3. A peak output current
diode for loss analysis are presented in Table 1.
of Iˆo = 5.5 kA leads to the total loss in each
The switching frequency (fsw ) is chosen as 250 Hz
device which leads to an steady state temperature
to have good control of the arm current and to incur
below the maximum operating temperature of the
less switching loss. The analysis is done assuming
device. The IGCT 5SHY 65L4521 positioned as
same type of switches and diodes at each position
T2 incurs the maximum loss and the switching
in the converters.
devices must be rated as per the rating of this
The switching loss coefficients of the selected IGCT device. The conduction and switching losses in the
and Diode have been calculated using the curve device T2 are 2.944 kW and 2.392 kW respectively.

ISBN 978-3-8007-5515-8 1366 © VDE VERLAG GMBH · Berlin · Offenbach

Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
 PCIM Europe digital days 2021, 3 – 7 May 2021
The total loss of 5.336 kW in the device with the
double side cooling arrangement as in Table 3 can
raise the junction temperature of the device up to
117◦ C which is well within the maximum operating
temperature of the device (140◦ C).
Fig. 3: Loss at rated frequency output and Iˆo = 5500 A.
4.2 Zero Speed Operation
Similar loss analysis as in case of rated speed
operation has been carried out for zero speed
operation. As discussed earlier, the modulation
index of the sinusoidal common mode injection
signal has been chosen as Mcom = 0.5 because
the study in [6] shows that similar value gives a
submodule voltage ripple of less than 20% in the
low and zero speed region. The total loss in each
device in a submodule has been calculated using
(28) –(37). The iterative loss calculation with various
dc output current from the bridge leg shows that the
loss with 35% of peak current at rated operation (Iˆo )
yields the same loss as in case of rated speed
operation. The loss at Iˆo = 0.35Io,ratedSpeed =
0.35 · 5500A = 1925 A is presented in Fig. 4.
4.3 Proposed Converter Configuration
The application requirement of 100 M V A at 13 −
15 kV leads to a need of 4.44 − 3.85 kA of rms
current output, i.e. Iˆo = 6.28 − 5.44 kA. As the
analysis at rated speed operation shows, a peak
current of Iˆo = 5.5kA i.e. Irms = 3.89kA can be
achieved using the selected IGCTs. Hence, a power
rating of 87 – 100 MVA can be managed depending
upon the stator voltage rating of the synchronous
machine in a pumped storage plant. The voltage
rating of the selected IGCT device is 2.8 kV and
ISBN 978-3-8007-5515-8
Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
Fig. 4: Loss at zero frequency output and Iˆo =
0.35Io,ratedSpeed = 1925 A.
hence to meet the DC link voltage of 25 kV, 10
submodules are required. The total number of
submodules will be 60 in the machine side converter
alone.
The total loss in one submodule at rated speed
operation is 11.5 kW (using Fig.3). The loss
in the semiconductor devices of the converter
would then be 690 kW which gives an efficiency
higher than 99% excluding the losses in the
passive components (submodule capacitors and
arm inductors), gate driver circuits and cooling
system.
5 Conclusion
The loss analysis based on analytical loss
equations presented in this paper shows that
modular multilevel converters of 100 MVA, 13–15
kV range can be achieved with the submodules
of available IGCTs and diodes with single devices
at each position. The start up torque of 12%
with submerged turbine is also achievable without
overheating the devices as the MMC can yield up
to 35% of full load current at zero speed. The study
also shows that the transition with full flow requires
around 120% of rated torque at reduced speed
and therefore, the fast transition from generating
to pumping mode can be realized only at partial
opening of the guide vanes.
Acknowledgment
The authors would like to thank HydroCen
(https://www.ntnu.edu/hydrocen) for supporting this
work.
1367 © VDE VERLAG GMBH · Berlin · Offenbach
 PCIM Europe digital days 2021, 3 – 7 May 2021
References
[1] M. Basić, P. C. O. Silva, and D. Dujić, “High
Power Electronics Innovation Perspectives for
Pumped Storage Power Plants,” Hydro 2018,
p. 10, Oct. 2018.
[2] J. Pérez-Dı́az, G. Cavazzini, F. Blázquez,
C. Platero, J. Fraile-Ardanuy, et al.,
“Technological developments for pumped-
hydro energy storage, Mechanical Storage
Subprogramme, Joint Programme on Energy
Storage,” European Energy Research
Alliance, Technical Report, May 2014.
[3] H. Schlunegger and A. Thöni, “100MW full-
size converter in the Grimsel 2 pumped-
storage plant,” Innsbruck, Hydro, 2013.
[4] P. K. Steimer, O. Senturk, S. Aubert, and S.
Linder, “Converter-fed synchronous machine
for pumped hydro storage plants,” in 2014
IEEE Energy Conversion Congress and
Exposition (ECCE), 2014, pp. 4561–4567.
[5] A. J. Korn, M. Winkelnkemper, and P.
Steimer, “Low output frequency operation of
the Modular Multi-Level Converter,” in IEEE
Energy Conversion Congress and Exposition
(ECCE), 2010, pp. 3993–3997.
[6] M. Hagiwara, I. Hasegawa, and H. Akagi,
“Start-Up and Low-Speed Operation of an
Electric Motor Driven by a Modular Multilevel
Cascade Inverter,” IEEE Transactions on
Industry Applications, vol. 49, no. 4, pp. 1556–
1565, 2013.
[7] A. Lesnicar and R. Marquardt, “An innovative
modular multilevel converter topology
suitable for a wide power range,” in 2003
IEEE Bologna Power Tech Conference
Proceedings, vol. 3, Jun. 2003. DOI:
10.1109/PTC.2003.1304403.
[8] A. Antonopoulos, L. Angquist, and H.-P. Nee,
“On dynamics and voltage control of the
Modular Multilevel Converter,” in 2009 13th
European Conference on Power Electronics
and Applications, ISSN: null, Barcelona,
Spain: IEEE, Sep. 2009, pp. 1–10.
[9] Q. Tu, Z. Xu, and L. Xu, “Reduced Switching-
Frequency Modulation and Circulating
Current Suppression for Modular Multilevel
Converters,” IEEE Transactions on Power
Delivery, vol. 26, no. 3, pp. 2009–2017,
ISBN 978-3-8007-5515-8
Authorized licensed use limited to: Norges Teknisk-Naturvitenskapelige Universitet. Downloaded on August 16,2021 at 13:23:21 UTC from IEEE Xplore. Restrictions apply.
Jul. 2011, Conference Name: IEEE
Transactions on Power Delivery. DOI:
10.1109/TPWRD.2011.2115258.
[10] K. Ilves, A. Antonopoulos, L. Harnefors, S.
Norrga, and H.-P. Nee, “Circulating current
control in modular multilevel converters
with fundamental switching frequency,” in
Proceedings of The 7th International Power
Electronics and Motion Control Conference,
vol. 1, Jun. 2012, pp. 249–256. DOI: 10.1109/
IPEMC.2012.6258844.
[11] X. She, A. Huang, X. Ni, and R. Burgos, “AC
circulating currents suppression in modular
multilevel converter,” in IECON 2012 -
38th Annual Conference on IEEE Industrial
Electronics Society, ISSN: 1553-572X, Oct.
2012, pp. 191–196. DOI: 10.1109/IECON.
2012.6388809.
[12] G. Konstantinou, J. Pou, S. Ceballos, R.
Picas, J. Zaragoza, and V. G. Agelidis,
“Control of Circulating Currents in Modular
Multilevel Converters Through Redundant
Voltage Levels,” IEEE Transactions on Power
Electronics, vol. 31, no. 11, pp. 7761–7769,
Nov. 2016. DOI: 10 . 1109 / TPEL . 2015 .
2512842.
[13] M. Hagiwara, K. Nishimura, and H. Akagi, “A
Medium-Voltage Motor Drive With a Modular
Multilevel PWM Inverter,” IEEE Transactions
on Power Electronics, vol. 25, no. 7, pp. 1786–
1799, 2010.
[14] Asymmetric and reverse conducting (IGCTs)
| Hitachi ABB, [online] https://www.hitachiabb-
powergrids.com/offering/product-and-
system/semiconductors/integrated-gate-
commutated-thyristors-igct/asymmetric-and-
reverse-conducting.
[15] Webra High Quality Thermal Solutions
- Cold Plate Catalogue, [online]
http://www.webra.se.
1368 © VDE VERLAG GMBH · Berlin · Offenbach