1
An Innovative Modular Multilevel Converter
Topology Suitable for a Wide Power Range
A. Lesnicar, and R. Marquardt
Abstract-- This paper presents a new multilevel converter
topology suitable for very high voltage applications, especially
network interties in power generation and transmission. The
fundamental concept and the applied control scheme is
introduced. Simulation results of a 36MW–network intertie
illustrate the efficient operating characteristics. A suitable
structure of the converter-control is proposed.
Index Terms— HVDC converters, high voltage transmission,
multilevel converter, space-vector PWM
I. INTRODUCTION
T
HE deregulation of international energy markets and the
trend to decentralized power generation are increasing the
demand for advanced power electronic systems. For this
application field multilevel converters with a high number of
voltage levels seem to be the most suitable types, because of
the need for series connection of semiconductors in
combination with low voltage distortion on the line side
[1]-[3]. Besides these points, a lot of other important aspects
have to be taken into account for these applications. Main
technical and economical aspects for the development of
multilevel converters are:
•
•
•
•
•
Modular realization:
- scalable to different power- and voltage levels
- independent of the state of the art of fast
developing power devices
Multilevel waveform:
- expandable to any number of voltage steps
- low total harmonic distortion
- dynamic division of voltage to the power devices
High availability:
- use of approved devices
- redundant operation
Failure management:
- fail safe operation on device failures
- avoidance of mechanical destruction (high
current magnetic forces and arcing)
Investment and life cycle cost:
- standard components
- modular construction
II.
A. Principle of M2LC
In order to fulfil the above mentioned requirements, a
converter system solely composed of an arbitrary number of
identical submodules was a prerequisite. For the sake of
stringent modular and scalable realization, additional “central”
components have to be avoided. The DC-link capacitor of
conventional voltage source inverters presents an example of
such a component – independent of its realization out of a
number of series connected capacitors or not. The subsystems
of the new concept are two terminal devices composed of
switches and a local DC-storage capacitor (C0). No additional
external connection or energy transmission to the submodules
is needed, for full 4-quadrant operation of the converter
system.
Fig. (1) illustrates an inverter leg consisting of n
submodules in each arm. In a first step, the submodules can be
considered as a controlled voltage source. Regardless of the
sign of the current ia,i, the terminal voltage Vx,i of each
submodule can be switched to either 0 V or to the voltage VC.
For easier explanation all voltage values VC are assumed to be
equal to V0. By switching a number of submodules in the
upper and lower arm, the voltage Vd is adjusted. In a similar
manner, the voltage VAC can be adjusted to a desired value.
P
i a,1
V x1,1
SM
Vd
2
Vd
V x1,n
SM
V x2,1
SM
V x2,n
N
The authors are with the Institute for Power Electronics and Control,
Universität der Bundeswehr München, 85577 Neubiberg, Germany (e-mail:
anton.lesnicar@unibw-muenchen.de, rainer.marquardt@unibw-muenchen.de)
CONCEPT OF THE NEW MODULAR MULTILEVEL
CONVERTER M2LC
i AC
V AC
Vd
SM
i a,2
Fig. 1. Inverter leg consisting of 2n submodules
2
2
The following equation shows the limitation of the voltages
Vd(t) and VAC(t) respective to the number of submodules per
arm:
Vd (t ) + 2 ⋅ V AC (t ) ≤ 2 ⋅ n ⋅ V0
(1)
When choosing Vd = const. and:
Vd = n ⋅ V0
(2)
then the amplitude of the output voltage is restricted to:
VˆAC ≤ n ⋅ V0
(3)
A suitable and simple realization of the submodule is given
in Fig. 2. The interface is composed solely of two electrical
terminals and one bi-directional fibre-optic interface. This
reduces the costs for manufacturing and maintenance, too. The
voltage of any submodule can be freely controlled by
software. The individual voltages of the submodules may even
be chosen unequal. This can be used to increase the number of
resulting voltage steps (e.g. together with PWM-operation). In
contrast to the conventional VSI a common central capacitive
storage is for the concept of M2LC dispensable. This
advantage eases the protection of the converter against
mechanical destruction in case of a short circuit, significantly.
In addition, a defective submodule can be replaced by a
redundant submodule in the arm by control action without
mechanical switches. This results in an increased safety and
availability.
ia
SR
Vx
SF
C0
+
VC
-
Fig. 2. Structure of a submodule
Table 1 shows the commonly used control states of a
submodule in failure-free operation. When the IGBT SF is
switched on, the voltage Vx = 0. To apply the voltage VC to
the terminals, the IGBT SR has to be switched on. In case of
switching off both IGBTs, the impressed voltage to the power
devices is limited by the capacitor voltage VC.
TABLE 1
COMMONLY USED CONTROL STATES OF A SUBMODULE TABLE
Mode
1
2
3
4
SF
SR
OFF ON
OFF ON
ON OFF
ON OFF
ia
VX
dVC/dt
>0
<0
>0
<0
VC
VC
0
0
>0
<0
0
0
B. Concept of controlled voltage balancing
In order to keep the capacitors on the same voltage level
and to ensure equal stress for the power devices the following
algorithm is applied for each arm:
The voltages of the capacitors are periodically measured
with a typically sampling-rate in the millisecond-range.
According to their voltage the capacitors are sorted by
software. In case of positive current the required number of
submodules, determined by output state controller, with the
lowest voltages are switched on. Therefore, the selected
capacitors are charged. When the current in the corresponding
arm is negative, the demanded number of submodules with
highest voltages are selected. By this method, continuous
balancing of the capacitor voltages is guaranteed. Inherently,
this concept supports an optimized utilisation of the stored
energy and evenly distributed power loses for the installed
electrical devices. Additionally, the power losses can be kept
low by switching the submodules solely when a change of the
output state is requested.
III. CONTROL SCHEME
With regard to the modular and scalable topology of the
M2LC, the applied control scheme should be easily
expandable to any number of levels. With this in mind, the
space-vector PWM is a suitable control scheme. In general all
concepts, based on the space-vector PWM theory, are
compatible with the following fundamental algorithm:
The first task is the transformation of the three phase
voltages into the two-dimensional space by using the equation
(4).
1
1 

−   vaY 
1 −
 vRe  2 
2
2 ⋅ v

 = ⋅
(4)


3
3   bY 

v
3
 Im 
−
0 +
 v
2
2   cY 

The second step is to find the three active switching vectors
adjacent to the reference vector vref. The three active switching
points next to the reference vector have to be located to
minimize the harmonics. Finally, due to the expression of the
reference vector by the determined surrounding vectors, the
correspondent dwelling times have to be calculated [3]-[6].
3
2,89
Im
2,31
1,73
1,15
vref
X
0,58
0,00
Re
-0,58
-1,15
-1,73
-2,31
-2,89
-3,33
-2,67
-2,00
-1,33
-0,67
0,00
0,67
1,33
2,00
2,67
3,33
Fig. 4 Space-vector trajectory (5-level topology) for a possible
three-phase output with sine wave common-mode voltage
Fig. 3 Space-vector diagram of a 5-level converter
The number of nSS adjustable switching states for a threephase nlevel – multilevel converter is simply given by:
3
n SS = nlevel
(5)
The number of different voltage vectors nV for a threephase
nlevel – multilevel converter can be calculated
from:
nV = 3 ⋅ nlevel ⋅ (nlevel − 1) + 1
(6)
Applied to the topology of M2LC, Fig. 3 shows a 5-level
space-vector diagram, assuming that the capacitor-voltages of
the submodules are equal and scaled to the value 1. From this
it follows that the normalized terminal-voltages VX of the
submodules have either the value 0 or 1. The nV points
illustrate the possible voltage vectors in the two-dimensional
space.
The M2LC offers the degree of freedom to control the
common-mode voltage of the three-phase system. The
possibility to vary the common-mode voltage with unchanged
phase voltages can be utilized to avoid high rate of rise or
high amplitude of the common-mode voltage, which leads to
strong insulation stress or other drawbacks. The proposed
control scheme adopts the fundamental principles of the
abovementioned general concept of space-vector theory. In
addition to the two dimensions, the common-mode voltage is
taken into account. Thus a three-dimensional space arises,
shown in Fig. 4.
For synthesizing the reference vector by using the most
convenient switching states a suitable algorithm shall be
broadly described:
In the same way, like in the plane representation, the phase
voltages have to be transformed (4). The third dimension is
given by the common-voltage axis. The next surrounding
switching-state vectors to the reference vector have to be
located. Compared to the plane graph (surrounding triangle
built by 3 points) now 4 points have to be determined, which
encase the reference vector by a tetrahedron (Fig. 5).
Analogue to the conventional duty cycle computation the four
dwelling times of their correspondent switching-state vector
can be calculated.
common-mode
voltage
Re
Im
Fig. 5 Reference vector encased by a tetrahedron
Simulation results of the three-phase voltages and the
common-mode voltage using the proposed space-vector
PWM are shown in Fig. 6 an 7.
4
5.0
4.0
SM
SM
SM
SM
SM
SM
3.0
P0
SM
SM
SM
SM
n=22
2.0
V 11
1.0
V 12
V 20
0.0
IT 11
-1.0
Vd
IT 12
V 13
IT 20
-2.0
IT 13
-3.0
SM
SM
SM
SM
SM
SM
SM
SM
SM
SM
-4.0
-5.0
0
3.33
6.67
10.00
time[ms]
13.33
16.67
20.00
N0
Fig. 6 Normalized line-to-line voltages (5-level topology)
Fig. 8. Simulation model
1.33
Starting under de-energized condition, the output terminal
of this supply is connected to the DC-bus of the multilevel
converter. Per inverter leg a number of (2n-1) IGBTs SF are
triggered. The IGBT SF of the capacitor, which has to be
charged and all remaining IGBTs SR in the inverter leg are not
triggered. When the capacitor reaches the operation voltage,
the appropriate IGBT SF has to be triggered. Simultaneously
another IGBT SF of the same leg is no longer triggered. In that
way all capacitors in the arms are gradually charged to the
operation voltage (Fig. 9). Finally, the voltage source has to
be disconnected by series diodes or mechanical switching.
1.00
0.67
0.33
0.0
-0.33
-0.67
-1.00
-1.33
40
0
3.33
6.67
10.00
time[ms]
13.33
16.67
35
20.00
30
A network intertie model has been built and tested to verify
the concept. To this purpose the simulator program
SIMPLORER was used. In the model 10 inverter legs are
connected with a DC-link to one three-phase power converter
(power supply) and one one-phase power inverter (load). Each
arm is composed of 22 identical submodules. No additional,
central capacitive energy storage at the DC-side is installed
(Fig. 8).
In practice difficulties occur, when a common network
intertie has to be put out of the de-energized condition into
operating condition (″black start″). The new power converter
topology allows a simple and safe black start. In the following
a possible process of charge per inverter arm is described. In
order to accomplish this procedure, only one auxiliary voltage
source with a relatively low output voltage (VLoad ≈ VC) is
necessary.
Voltage [kV]
Fig. 7 Example of controlled, normalized common-mode voltage (5-level
topology)
IV. SIMULATION RESULTS
Vd
25
20
15
10
5
0
V Lo ad
V C1
0
1
2
V C22
3
4
5
tim e [m s]
6
7
8
Fig. 9 Process of charge per inverter arm
The curves shown in Fig. 10a and 10b are under steady
state conditions. The single phase and the three phase voltages
are synthesized by 21 levels (+1 redundant submodule). The
ripple content of the 2nd harmonic on the DC-link voltage has
no effect on the input and output power characteristic (Fig.
10b). The capacitive energy storage of all submodules is the
same as used in a conventional network intertie of the same
power rating (Pd = 36 MW).
5
50
40
Vd
30
V20
Voltage [kV]
20
10
0
-10
-20
-30
-40
-50
0
20
40
time [ms]
60
80
100
Fig. 10a. Output voltage and DC-link voltage
80
Pout
70
60
50
The converter cubicles are equipped with a number of
identical submodules which are connected solely by duplex
optical-fibre cable to the central control unit. Each submodule
receives the correspondent switching commands, optoelectronically and sends its capacitor-voltage back to the
central control unit, periodically. The power supply voltage
for data transfer and drive circuit of the IGBTs is supplied by
the capacitive energy storage of the submodule [7]. The
measuring system for the branch current measurement is
connected fibre-optically, too. The fault detection
differentiates between faults, which have to lead to fault
indication with or without a disconnection of the system. For
instance, in cases of overcurrent or power loss the failure
management has to start the failsafe operation, automatically.
Whereas, corrigibly failures don’t induce an automatic
disconnection. These failures may appear in data transmission.
The source of error can be located and saved by the fault
diagnosis system. At a later time, in accordance with the
scheduled maintenance, the fault can be cleared. This results
in a distinct improvement of the availability and
maintainability.
Supervisory
Computer
40
Converter
Cubicle
Central Control Unit (CCU)
Submodule
20
T1,T2
10
Vc1
Pin
ribbon
cable
-30
Interface
flat
-20
duplex opticalfibre cable
0
-10
Interface
Power [MW]
30
•
•
•
Submodule
T1,T2
Vc1
-40
-50
0
20
40
time [ms]
60
80
Arm-Circuit
Current
Measurement
100
Fig. 10b. Input and output power variation curve
Sim. results obtained with the parameters:
Transmitted real power Pd = 36 MW; C0 = 2 mF;
one-phase-system: V1,rms =25.0 kV, f1= 25 Hz;
three-phase-system: V3,rms =26.5 kV, f3 = 60 Hz
V. STRUCTURE OF M2LC-CONTROL
Consisting of three different fundamental units, the
following control structure for M2LC is proposed (Fig. 11).
The supervisory unit takes over the feedback and supervisory
control of the entire system. According to the setpoint values
the feedback control supplies the central control unit in realtime mode. A digital signal processor (DSP) or FPGA is well
suited to solve these tasks. The presented concept of voltage
balancing is implemented in the modulator. The output state
controller which has to determine the optimized output states
and the operating sequence for the next PWM-period is
integrated, too. The PWM-generator calculates the dwellingtimes for the appendant switching states.
A
Fig. 11. Structure of M2LC
VI. CONCLUSION
This paper introduced the topology of the new modular
multilevel converter M2LC and its relevant characteristics.
The modular concept allows the application for a wide power
range. The proposed control scheme is well suited for a
different number of voltage levels. It is shown that the control
scheme allows to control phase-voltages and the common
mode voltage at the same time, independently. Simulations
have demonstrated a good performance of the M2LC-concept.
The start under de-energized condition can be safely realized.
The structure of M2LC-control enables good separation of the
low voltage units and the high voltage units of the converter
cubicles. Presently, a prototype of a 2 MW – converter system
is being built. This modular system will enable further
experimental investigations.
6
VII. REFERENCES
[1]
Jih-Sheng Lai, Fang Zheng Peng, “Multilevel
Converters – A new breed of power converters”,
IEEE Trans. Ind. Applicat., vol. 32, pp 509-517,
May./June. 1996
[2]
Rainer Marquardt, Anton Lesnicar, Jürgen Hildinger,
“Modulares
Stromrichterkonzept
für
Netzkupplungsanwendung bei hohen Spannungen”,
ETG-Fachtagung, Bad Nauheim, Germany, 2002
[3]
José Rodriguez, Jih-Sheng Lai, Fang Zheng Peng,
“Multilevel Inverters: A survey of topologies,
controls, and applications”, IEEE Trans. Ind. Electr.,
vol. 49, pp 724-738, Aug. 2002
[4]
A. R. Bakhshai, H. R. Saligheh Rad, G. Joos, “Space
vector modulation based on classification method in
three- phase multi-level voltage source inverters”,
IEEE Trans. Ind. Applicat., vol. 1, pp 597 – 602,
Sep./Oct. 2001
[5]
Nikola Celanovic, Dushan Boroyevich, “A fast
space-vector modulation algorithm for multilevel
three-phase converters”, IEEE Trans. Ind. Applicat.
vol. 37, pp 637 – 641, March/April 2001
[6]
Fei Wang, “Sine-triangle versus space-vectormodulation for three level PWM voltage-source
inverters“, IEEE Trans. Ind. Applicat. vol. 38,
pp 500 – 506, March/April 2002
[7]
Jürgen Hildinger, Rainer Marquardt, “Erzeugung
stabilisierter Hilfsspannungen aus dem Leistungsteil
von U-Umrichtern”, ETG-Fachtagung, Bad Nauheim,
Germany, 2002
Anton Lesnicar was born in Munich, Germany, in
1971. He received the Dipl.-Ing. (M.Sc.) degree in
2000 from the University of Munich.
He is currently working toward the Ph.D. degree
in Power Electronics at the “Universität der
Bundeswehr / München”, Munich, Germany in the
Institute for Power Electronics and Control. He is
engaged in research and development of new
power electronic systems for power generation and
transmission.
Rainer Marquardt was born in Hannover,
Germany in 1953. He received the Dipl.-Ing.
(M.Sc.) and Dr.-Ing. degree (Ph.D.) in electronic
communication and power electronics respectively
from the University of Hannover, before joining
Siemens AG/Erlangen. He performed numerous
industrial research and development projects for
high power applications in power transmission and
advanced AC-Drive systems for traction
applications. He has filed 48 patents in these areas.
Currently, he leads the Institute of “Power
Electronics and Control” as an ordinary Professor
at the “Universität der Bundeswehr / München” in
Germany.