IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO.

4, OCTOBER 2009

2261

Analytical Modeling of a Square-Wave-Controlled

Cascaded Multilevel STATCOM
Ronny Sternberger, Student Member, IEEE, and Dragan Jovcic, Senior Member, IEEE

AbstractThe aim of this paper is to present an analytical, statespace model of an indirect, voltage-controlled cascaded-type multilevel static synchronous compensator (STATCOM) with square
wave control. The multilevel converter model is segmented into
a dynamic and static part in order to accurately represent all internal feedback connections. Each voltage component is analyzed
in detail and described mathematically by an averaged expression
with an equivalent capacitance. The STATCOM model is linearized
and linked with a DQ frame ac system model and the controller
model, and implemented in MATLAB. The controller gains are selected by analyzing the root locus of the analytical model to give optimum responses. The validity and accuracy of the proposed model
are verified against non-linear digital simulation PSCAD/EMTDC
in the time and frequency domain. The model is very accurate in
the subsynchronous range, and it is adequate for most control design applications and practical stability issues below 100 Hz. Furthermore, the developed model can be used for multilevel cascaded
converters which exchange real power.
Index TermsModeling, multilevel converter, static synchronous compensator (STATCOM), state-space methods, static
VAR compensators.

I. INTRODUCTION

N recent years, multilevel converters have become increasingly popular in high-power transmission/distribution systems and in industry applications [1][5]. In contrast to a conventional two-level voltage-source converter (VSC) that works
with pulsewidth modulation (PWM), such multilevel converters
use a number of (low voltage) series-connected capacitors to
generate high ac voltage. This allows higher power-handling capability with reduced switching power losses and harmonic distortion [3].
The main types of multilevel converters are diode clamped,
flying capacitor, and cascaded inverter [3][5]. By comparing
these different topologies, while considering the harmonic level,
losses, and component costs, the cascaded multilevel converter
with square wave control is found to be the optimum solution for static synchronous compensator (STATCOM) applications [3], [5][8]. The modular structure of this converter, with
a number of identical H-bridges, makes this converter very flexible in terms of power-handling capability. The use of square
wave control results in a single switch on and off per cycle for
each switch, which brings benefits of low switching losses. In
addition, since the control angle at each capacitor can be used to
Manuscript received August 25, 2008; revised October 21, 2008. Current version published September 23, 2009. Paper no. TPWRD-00617-2008.
The authors are with the Engineering Department, University of Aberdeen, Aberdeen, AB24 3UE, U.K. (e-mail: r.sternberger@abdn.ac.uk;
d.jovcic@abdn.ac.uk).
Digital Object Identifier 10.1109/TPWRD.2009.2021045

cancel one harmonic, low harmonic distortion can be achieved

[6], [9].
The dynamics of such complex converters are mostly analyzed by using Electromagnetic Transients Program (EMTP)type programs, such as PSCAD/EMTDC [10]. However, these
simulators provide only trial-and-error-type studies in the time
domain. The demanding analysis and design tasks become very
time-consuming since a labored search has to be executed to find
the best solution.
Alternatively, a suitable and accurate analytical system model
would provide faster design routes. This type of model would
provide the ability to use dynamic systems-analysis techniques
(e.g., eigenvalue and frequency-domain methods) and modern
control design theories, resulting in shorter design time and advanced controller configurations [8], [11].
Compared to conventional VSCs, the analytical modeling
of multilevel converters is more challenging. A conventional
PWM-controlled VSC has a fxed structure and the switching
frequency is high which implies that intervals between switchings are short and conventional averaging approaches can
represent converter dynamics [12]. A multilevel converter
with square-wave control, on the other hand, fundamentally
changes its structure as the capacitors are switched in and out
of the current path and, therefore, these changes may have a
significant influence on the system dynamics. Further, each cell
is switched only twice per cycle and, therefore, variables may
undergo notable excursions within a single conduction interval.
One of the primary challenges in the dynamic modeling of
a square-wave converter is therefore the adequate dynamic
representation of cell variables between the switching instants.
The multilevel converter model in [8] is in convenient statespace form but it adopts overly simplified equivalent capacitance which is only valid for a particular number of levels and
does not consider firing angles on individual cells. Similarly,
the state-space model in [9] is of little practical use, since the
equivalent capacitance has to be tuned for every change in converter parameters, using identification methods. The researchers
in [13] developed a frequency-domain model of a square-wavecontrolled converter. Since it only considers a single cell with
uncontrollable, full 180 conduction periods, this model does
not address the aforementioned modeling challenge. The research presented in [14] represents the structure of a multilevel
converter at a wide range of frequencies in order to provide a
model for phasor studies and harmonic analysis. However, this
is essentially a static model, and it completely ignores the control influence. Moreover, since the model neglects the transfer
of active power, it cannot be used for dynamic modeling.
This paper aims to develop a dynamic analytical model of a
multilevel-cascaded STATCOM converter that is convenient for

0885-8977/$26.00 2009 IEEE

2262

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 4, OCTOBER 2009

Fig. 2. Ideal stepped output voltage of the 13-level converter (! = 2 3  3 f ).

Fig. 1. Multilevel STATCOM test system with indirect control.

system stability studies and for the analysis of interactions with

ac systems. We attempt to establish generic modeling principles
that are applicable for a range of multilevel-cascaded converters,
even those which exchange real power. A modular modeling
approach is adopted to represent the complexity of the system
with the benefit that each individual subsystem can be analyzed
independently. This analytical model employs all parameters
and variables with physical meaning and, therefore, it can study
variations in ac and dc system structures.
II. MULTILEVEL TEST STATCOM CONFIGURATION
The selected test STATCOM, which contains a cascaded
(chain-circuit) multilevel converter and its connection to the
ac system, is shown in Fig. 1. The multilevel STATCOM
system includes a three-phase transformer and three strings
of single cells that are connected in series. A single cell here
is a full-bridge, single-phase inverter with a capacitor as the
dc source [15]. Note that only chain circuit A with its insulated-gate bipolar transistors (IGBTs), diodes, and capacitors is
shown, but similar groups of single cells are also connected in
branches B and C.
The STATCOM parameters that determine the main circuit
to
are the number of cells per phase , dc capacitances
, buffer inductance
, transformer leakage inductance
, and transformer ratio
. The relationship between the
number of single cells per string and the number of levels
for a cascaded multilevel converter is given by
.
In further studies, it is assumed that all capacitances are equal
and that the steady-state voltages
across these capacitors are also identical
(with the help of a suitable controller). To ensure low converter losses and dc voltage ripple, the STATCOM
system parameters are optimized by following the algorithm
proposed in [16] and their values can be found for the selected
test systems in Table I.
At the point of common coupling (PCC), the STATCOM is
and
connected to an ac circuit represented by resistance
. The ratio between
and
is
inductance
assumed to be constant
.

Fig. 3. STATCOM phasor diagram where the general reference frame is linked
with V .

The converter switching losses are reduced by the use of

square wave control where each device has only one turn-on
and turn-off per cycle. The harmonic distortion is reduced by
calculating the firing angle at each cell to cancel one harmonic.
In view of the fact that triplen harmonics are cancelled in the
converter delta connection, the firing angles for the individual
cells are calculated for cancellation of the lowest odd, nontriple
harmonics [9], and their values are shown for selected test
systems in Table II. Note that these calculated firing angles
can cancel only steady-state harmonics but as loading changes,
harmonics increase [14]. The ideal stepped output voltage of a
13-level converter is shown in Fig. 2.
The STATCOM uses indirect control [15], as shown in Fig.
1. This control has, as a benefit, a constant modulation ratio
, resulting in constant control angles and, therefore,
in low harmonics even at low converter voltages. The dc voltage
(i.e., the active power
is affected by controlling the angle
transfer) [15].
III. CASCADE MULTILEVEL CONVERTER MODEL
A. General Voltage and Current Representations
The phasor diagram of the STATCOM system is shown in
Fig. 3. It is assumed that the system is balanced and all variables
are represented with a fundamental frequency phasor. Note that
all ac voltages are shown as line-neutral magnitudes.
With reference to Fig. 3, the instantaneous ac voltage
and the ac current
can be presented as
(1)
(2)
Equation (2) can also be represented by using components
frame, where and indices appropriately
in the rotating
label axes
(3)

STERNBERGER AND JOVCIC: ANALYTICAL MODELING OF A CASCADED MULTILEVEL STATCOM

2263

Fig. 4. Single cell of a cascaded converter.

where
(4)
(5)
B. Instantaneous Nonlinear Dynamic Model of a Single Cell
In the first part, we investigate a single cell of a cascaded
multilevel converter, as shown in Fig. 4. The single cell consists
of one capacitor and four switching devices (each one contains
an IGBT and an antiparallel diode).
The instantaneous ac voltage of this type of single cell
is shown in Fig. 5 together with the ac voltage at the PCC
, the capacitor voltage
, and the converter ac
. The variables
and
(all variables in
current
the top graph of Fig. 5) are at the system level for a multilevel converter. These variables are assumed to be the ideal
sine shape due to the averaging action of multiple levels and
filtering action of passive components.
Since a single cell is switched into the electrical circuit within
, (S1,S4 ON
the interval
in time domain can
or S2,S3 ON), the converter ac voltage
be represented as

elsewhere.
(6)
It is also seen that the converter is supplied by a rectified
ac current (positive: S1,S4 ON; negative: S2,S3 ON) and that
the current
, which flows through the capacitor, has the
following form:

Fig. 5. Currents and voltages of a STATCOM system in steady state.

The single-cell variables in Fig. 5 are shown under steadystate conditions for the ideal converter. Under dynamic transients and assuming some internal losses, the instantaneous variables will have different waveforms. Fig. 6 shows the single-cell
, which results
variables assuming a nonzero control angle
in an active power transfer. The transient curves in Fig. 6 have
different traces compared with Fig. 5 since:
;
the current will have some nonzero average value
is not constant, but it is increasing
the dc voltage
along a sine curve in the conducting interval (from the ini).
tial value
can be represented as (8),
The capacitor dc voltage
stands for
shown at the bottom of the page, where
the dynamic voltage component which depends on the current
and the control angle, and has therefore a direct influence on
the converter dc and ac voltage
is the average steady-state value of dc voltage of each
is constant and can be
single cell. It is shown in [16] that
found using static (power flow) equations

(10)
elsewhere.
(7)

with

(11)

(8)

elsewhere

(9)

2264

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 4, OCTOBER 2009

The voltage expression in (9) cannot be directly averaged

since the essential dynamic properties of the converter would
be lost. First, (9) is differentiated, which gives a nonlinear but
algebraic expression on the right side. The resulting expression on the right side of (9) can be averaged as in (12). Following these steps, we obtain a dynamic equation for the average dynamic component of STATCOM voltage in the s-domain (LaPlace transformation):
(14)
Fig. 6. Dynamic dc current and dc voltage of a single cell.

where
defines the converter modulation ratio (
if
STATCOM is using indirect control).
expression does not involve any
We observe that the
dynamics, but it will have a direct static influence on the magnitude of the converter ac voltage. This component depends on
the current magnitude , the firing angles , and capacitance
, but not on the control angle
.
C. Dynamic DC Voltage Component
The dynamic component of the STATCOM ac voltage in (9),
shown at the bottom of the previous page, plays an important
role in the dynamics of the overall system. Since the nonlinear
current waveform in (7), which is driving the dynamic compo, cannot be represented directly in a continuous
nent
model, average dc current
will be derived first. This
average current, also illustrated in Fig. 6, will result in the
average value of the dynamic dc voltage component
that shows the same cumulative dc voltage change over one
half-cycle as the instantaneous dynamic component
(i.e.,
). Since system dynamics are
assumed to be slower than structural changes within one-half
cycle, there will be no loss in accuracy if all variables are represented as constant/linear curves, which give accurate values
of
at the end of each half cycle. The average dc current
over one half cycle can
the nonlinear current waveform
be found to average (7)

where denotes the coefficient of the Laplace transformation.

in (14) uses a linear apThe average dynamic voltage
proximation to represent a sine-shaped dynamic voltage component, as seen in Fig. 6. Replacing (13) in (14), the formula for
dynamic dc voltage component is obtained as

(15)
Assuming a typical STATCOM application, we will have
and it can be seen in (15) that the dynamic ac voltage
and the control
will depend mainly on the active current
angle . Therefore, in a typical STATCOM application, the
in (15) can be neglected with a limited loss
component
in accuracy. However, the converters with active power flow
should include the complete model.
D. Converter Dynamic Model at the Fundamental Frequency
The fundamental ac voltage of the switched dc voltage components can be found if the Fourier Transformation at frequency
is used. We calculate first coefficients of the Fourier series to obtain and axis components of the ac voltage

(16)

(17)

(12)

The Fourier coefficients in (16)(17) are chosen considering

the coordinate frame alignment in Fig. 3. Replacing (8) in
(16)(17), the ac voltage will consist of two components
(18)

Placing (3) in (12), we can derive an expression for the averaged current. At this stage, we assume that
and
in
(3) are constant over the integration period in (12). This can be
justified since the current dynamics are slower than the voltage
dynamics and can be neglected over the short segment of half a
fundamental cycle. The dc current equation becomes
(13)

(19)
: In order to preserve
1) Dynamic Voltage Component
the dynamic nature of the converter voltage in (15), we need to
apply Fourier transformation to the differential of the converter
voltage. For a given function , the following property of Fourier
transformation is employed:
(20)

STERNBERGER AND JOVCIC: ANALYTICAL MODELING OF A CASCADED MULTILEVEL STATCOM

2265

Placing (15) in (16)(17), using (20) and assuming a conacross the integration interval, the
stant current magnitude
dynamic equations for the ac voltage components are

(21)
(22)
(23)
2) Static Voltage Component
: Placing (10) in (16)(17),
the corresponding static ac voltage components become
(24)
(25)

Fig. 7. Structural diagram of the STATCOM system model.

(26)
The expressions for equivalent capacitance in (11) and (23)
represent some imaginary capacitance which enables us to represent multilevel converter dynamics using simple equations for
a single-capacitor system. It is observed that these capacitances
and the control angle ,
depend only on the capacitance
which are constant for a given STATCOM topology. Therefore,
the equivalent capacitance will not change as the operating conditions change.
It is noted that the components in (22) and (25) will be small
in a typical STATCOM.

The final component

of the multilevel converter model
is therefore obtained by using (21) and (24)

(31)
(32)
and using (22) and (25), the q-component

is

(33)
(34)

E. Nonlinear Dynamic Model for a Multilevel Converter

The aforementioned modeling approach can be adopted for
each cell of a multilevel converter. The fundamental component
of ac voltage for a cascaded STATCOM with X cells is obtained
as
(27)
(28)
The capacitor voltage balancing is normally used with multilevel converters and this control function is very fast compared
with main control responses [17]. Consequently, the dc voltages
on individual cells can be assumed identical. This enables us
to use the same equations for each cell (21)(26), and only the
firing angles will differ. Replacing (21)(26) in (27) and (28), it
can be shown that the fundamental ac voltage components of a
multilevel converter are similar to (21)(22) and (24)(25). The
and the parameter for a multilevel
equivalent capacitance
converter are different, however, and result in

IV. ANALYTICAL STATCOM SYSTEM MODEL

A. Model Structure
The STATCOM system consists of the multilevel converter,
a transformer, a control system with phase-locked loop (PLL),
and it is connected with an ac system.
The STATCOM model schematic diagram is shown in Fig. 7
with all linking variables between the subsystems. It can be seen
that the STATCOM model is segmented in two main subsystems: 1) an ac system and 2) a dc system, which is a conventional approach with converter modeling [11]. Each subsystem
is represented as a stand-alone state-space model, and the dc
system includes the converter model, the controller model, and
the PLL model.
The final state-space model can be written in the following
matrix form [9], [11]:
(35)
(36)
where

(29)
(30)
where

are the control angles on individual cells.

(37)
In (35) and (36),
is the state-variable vector,
represents the input variable vector (including reference voltage

2266

Fig. 8. Equivalent ac circuit.

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 4, OCTOBER 2009

The delay filter in Fig. 7 is introduced to represent firing control circuit delay and does not represent actual system dynamics.
This filter accounts for the delay in the signal transfer in a discrete system [18], and the delay time constant is determined as
, where 50 represents the fundamental
frequency, 3 is the number of branches, and 2 considers two
switchings per cycle per branch.
The main role of the PLL is to synchronize the converter firing
signals with the ac system. The latest D-Q-Z type [18], which
has the advantage of accurate phase information even with a
distorted ac system voltage waveform, is used here.
V. MODEL VERIFICATION

Fig. 9. Dominant branch of the root locus for test system 1.

and disturbance voltage

), and
stands for the
output variable vector. The submatrices in the system matrix
are associated with the individual subsystems: dc for the dc
subsystem and ac for the ac subsystem. The indices with input
matrix and output matrices indicate the actual model (first
index) and the linking subsystem (second index).
B. AC Model
Assuming a symmetrical and balanced ac system, the ac grid
can be modeled as a single-phase dynamic system as shown in
Fig. 8. The instantaneous circuit variables are used as the states.
cannot be measured directly in a
Since the ac grid voltage
state-space model, an artificial capacitance
, which has a
very small value (no dynamic effects to the system), is used to
accommodate state-space modeling.
It is shown in [16] that conduction losses are dominant in a
converter with square-wave control and, therefore, the converter
losses are represented in the model as a series-connected resis.
tance
rotating
The final model is subsequently converted to the
frame using Parks transformation as described in [11].
C. Converter Model
The final STATCOM model is presented in the dc submodel
and the ac voltage is the sum of dynamic and static voltage components (31)(34). Since these expressions are strongly nonlinear, they are linearized in a usual manner.
The linearized STATCOM model is linked with the PLL and
the controller model, as shown in the next section, to create the
final dc model.
D. Controller and PLL
The STATCOM system uses indirect control, which consists
of a proportional plus integral (PI) controller, to regulate the ac
voltage as seen in Fig. 7. The PI controller produces the phase
angle
, which is common for all cells. The reference signal
is provided by an external or a system control [15].

The derived analytical model is implemented in Matlab and

tested in the time-domain against the detailed, nonlinear simulation software PSCAD/EMTDC.
Three, largely different test systems are simulated in order to
verify the accuracy and flexibility of the analytical model, and
all parameters are given in Table I. The test system 1 is a 30-kV,
STATCOM system, whereas test
70-Mvar 13-level
system 2 is operating at 15 kV and has a rated reactive output
power of 25 Mvar. To confirm the model accuracy with different ac system parameters, the grid strength in test system 2
is reduced by the factor of 3, compared with test system 1. A
third test system is introduced to study the model robustness
for a STATCOM system with a different number of levels. This
system is similar to test system 1, but it involves a nine-level
. The operating point of all systems is set
converter
. The MATLAB model
close to their rated reactive power
enables an insightful study of changes in parameters or structure
since all parameters are directly reflected in the symbolic form
in the final model.
The perturbation signal used in step responses is the reference
, as seen in Figs. 1 and 6. In disturbance tests, a
voltage
in
step on the remote source voltage magnitude, given by
Figs. 1 and 7, is introduced.
A. Selection of Controller Gains
The controller gains are selected in this study by analyzing
the dominant eigenvalues of the system and they are tuned to
meet the following requirements:
rise time less than three cycles of fundamental frequency;
overshoot below 10%;
settling time less than ten cycles of the fundamental frequency.
Fig. 9 shows the root locus of test system 1 as the open-loop
gain changes with the nominal system configuration. The
controller zero is selected for this particular test system at
2.8, which is close to the slowest system pole,
in order to improve speed of response and robustness.
It can be seen that for small open-loop gains, the dominant
eigenvalues are only real poles resulting in slow rise times and
large damping. It can also be seen that large controller gains will
result in significant overshoots, faster rise times, and reduced
damping. Best controller gains are then selected for each individual test system from its root locus to give optimum responses

STERNBERGER AND JOVCIC: ANALYTICAL MODELING OF A CASCADED MULTILEVEL STATCOM

Fig. 10. Test system 1 response following a 0.02-kV voltage reference step. (a)
System voltage. (b) AC current I . (c) Control angle ' .

2267

Fig. 12. Test system 3 response following a 0.1-kV voltage reference step. (a)
System voltage. (b) AC current I . (c) Control angle ' .

Fig. 13. Test system 1 response following a 0.02-kV disturbance step change.

that the developed model shows good accuracy for all considered variables (i.e., the voltage magnitude, the STATCOM cur).
rent, and the control angle
Fig. 11 shows the model verification for test system 2 (13level converter). It is concluded that the analytical model shows
very good accuracy even for different system topologies.
2) Nine-Level Converter: Fig. 12 shows the system response
when test system 3 is used (i.e., with a nine-level converter). It
is seen that excellent model matching occurs for all compared
variables.
C. Reference Disturbance Input
Fig. 11. Test system 2 response following a 0.1-kV voltage reference step. (a)
System voltage. (b) AC current I . (b) Control angle ' .

according to the aforementioned requirements, and final controller gains are given in Table I.
B. Reference Step Input
1) 13-Level Converter: A reference step response for the test
system 1 (13-level converter) is shown in Fig. 10. It can be seen

1) 13-Level Converter: The perturbation on the remote ac

source tests the models ability to represent perturbations on a
remote ac source and the coupling between the ac and dc submodels. The model verification for disturbance inputs is shown
in Figs. 13 and 14 for test system 1 and test system 2, respectively. It can be seen that good response matching exists for both
systems, and similar matching was confirmed for all of the other
variables that are not shown.
2) Nine-Level Converter: Fig. 15 shows the ac voltage at the
if test
PCC in response to a disturbance step change in
system 3 is used. It is concluded that the linear model readily

2268

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 4, OCTOBER 2009

Fig. 14. Test system 2 response following a 0.02-kV disturbance step change.

Fig. 17. Test system 3 frequency response of the linear model compared with
PSCAD/EMTDC frequency response.
Fig. 15. Test system 3 response following a 0.1-kV disturbance step change.
TABLE I
SYSTEM PARAMETER VALUES

Fig. 16. Test system 1 frequency response of the linear model compared with
the PSCAD/EMTDC frequency response.

accommodates structural changes (number of cells) in the multilevel converter.

D. Frequency Response of the AC Grid With STATCOM
Since a time-domain study cannot reveal the frequency bandwidth for model accuracy, test systems 1 and 3 are also tested
against PSCAD in the frequency domain. Since PSCAD does
not possess the frequency-response ability, the individual frequency components are injected (1-Hz step) with the measureat the
ment of the gain and phase of the system voltage
PCC. The experimental frequency response is performed in the
100 Hz, and using
as the
frequency range 1.6 Hz
injection point. The readings at individual frequencies are then
linked to create gain and phase curves as shown in Figs. 16 and
17 for test system 1 and 3, respectively.
Compared with the PSCAD model, it can be observed that
very good accuracy can be achieved for both test systems in
40 Hz, and reasonably good accuracy
the frequency range
at frequencies below 100 Hz. It is important that the dominant
complex poles, which are located just above 50 Hz, are well
represented by the linear model. PSCAD, however, points to
the existence of additional complex poles and zeros close to the

TABLE II
SWITCHING ANGLES FOR A 9- AND 13-LEVEL CONVERTER

dominant poles. The slope of the bode curves is well matched

by the analytical model indicating that the order of the analytical
model is accurate.
VI. CONCLUSION
A suitable and accurate analytical model of an indirectly controlled cascaded multilevel STATCOM with square-wave control is presented in this paper.
The converter voltage components are analyzed in detail for a
single-cell and the results are then generalized for a multilevelcascaded converter. The converter ac voltage waveform is of
a nonlinear, discrete, and dynamic nature, which is described
mathematically by appropriate averaged expressions.

STERNBERGER AND JOVCIC: ANALYTICAL MODELING OF A CASCADED MULTILEVEL STATCOM

The dynamic, analytical state-space model is built of subsystems to enable model application to a wide range of system configurations and various dynamic studies.
The developed STATCOM model is linearized and implemented in MATLAB. Eigenvalue studies are conducted for each
particular test system in order to select optimum open-loop controller gains.
The validity and accuracy of the proposed model is verified
against nonlinear PSCAD simulations, and good matching is
observed in the time domain for a range of outputs and for three
different test systems. The model is also tested in the frequency
domain and it is concluded that the presented model can be used
for dynamic studies below 100 Hz.
REFERENCES
[1] G. Joos, X. Huang, and B.-T. Ooi, Direct-coupled multilevel cascaded
series var compensators, IEEE Trans. Ind. Appl., vol. 34, no. 5, pp.
11561163, Sep. 1998.
[2] R. Betz, B. Cook, T. Summers, R. Palmer, A. Bastiani, S. Shao, and
K. Willis, Design and development of an 11 kV H-bridge multilevel
STATCOM, presented at the Australasian Univ. Power Eng. Conf.,
Perth, Western Australia, Dec. 2007.
[3] C. Hochgraf, R. Lasseter, D. Divan, and T. A. Lipo, Comparison of
multilevel inverters for static var compensation, in Proc. Conf. Rec.
Ind. Appl. Soc. Annu. Meeting, Denver, CO, 1994, vol. 2, pp. 921928.
[4] J. S. Lai and F. Z. Peng, Multilevel converterA new breed of power
converter, IEEE Trans. Ind. Appl., vol. 32, no. 3, pp. 509517, May
1996.
[5] F. Z. Peng, J. W. McKeever, and D. J. Adams, A power line conditioner using cascade multilevel inverters for distribution systems,
IEEE Trans. Ind. Appl., vol. 34, no. 6, pp. 12931298, Nov. 1998.
[6] S. Sirisukprasert, J. S. Lai, and T. H. Liu, Optimum harmonic reduction with a wide range of modulation indexes for multilevel converters,
IEEE Trans. Ind. Electron., vol. 49, no. 4, pp. 875881, Aug. 2002.
[7] J. D. Ainsworth, M. Davies, P. J. Fitz, K. E. Owen, and D. R. Trainer,
Static var compensator (STATCOM) based on single-phase circuit
converters, Proc. Inst. Elect. Eng., Gen. Transm. Distrib., vol. 145,
no. 4, pp. 381386, Jul. 1998..
[8] D. Soto and R. Pena, Nonlinear control strategies for cascaded
multilevel STATCOMs, IEEE Trans. Power Del., vol. 19, no. 4, pp.
19191927, Oct. 2004.

2269

[9] R. Sternberger and D. Jovcic, Small signal multi-level STATCOM

model, presented at the Power Eng. Soc. Gen. Meeting, Montreal, QC,
Canada, Jun. 2006.
[10] PSCAD/EMTDC 4.1 Users Manual, Manitoba HVDC Research
Center, 2004, Tutorial manual.
[11] D. Jovcic, N. Pahalawaththa, M. Zavahir, and H. Hassan, SVC dynamic analytical model, IEEE Trans. Power Del., vol. 18, no. 4, pp.
14551461, Oct. 2003.
[12] A. Yazdani and R. Iravani, An accurate model for the DC-side voltage
control of the neutral point diode clamped converter, IEEE Trans.
Power Del., vol. 21, no. 1, pp. 185193, Jan. 2006.
[13] M. Mohaddes, A. M. Gole, and S. Elez, Steady-state frequency response of STATCOM, IEEE Trans. Power Del., vol. 16, no. 1, pp.
1823, Jan. 2001.
[14] D. Jovcic and R. Sternberger, Frequency domain analytical model for
a cascaded multi-level STATCOM, IEEE Trans. Power Del., vol. 23,
no. 4, pp. 21392147, Oct. 2008.
[15] N. G. Hingorani and L. Gyugyi, Understanding FACTS: Concepts and
Technology of Flexible AC Transmission Systems. Piscataway, NJ:
IEEE, 2000.
[16] R. Sternberger and D. Jovcic, Theoretical framework for minimizing
converter losses and harmonics in a multi-level STATCOM, IEEE
Trans. Power Del., vol. 23, no. 4, pp. 23762384, Oct. 2008.
[17] Y. Li and B. Fu, A novel DC voltage detection technique in the CHB
inverter-based STATCOM, IEEE Trans. Power Del., vol. 23, no. 3, pp.
21392147, Jul. 2008.
[18] D. Jovcic and G. P. Pillai, Discrete system model of a six-pulse SVC,
in Proc. Conf. IASTED EuroPES, Sep. 2003, pp. 318323.
Ronny Sternberger (S06) received the Dipl.-Ing.(FH) degree in electrical energy engineering from the Fachhochschule fr Technik und Wirtschaft, Berlin,
Berlin, Germany, in 2004 and is currently pursuing the Ph.D. degree at the University of Aberdeen, Aberdeen, U.K.
His current research interests are flexible ac transmission devices, multilevel
converters, and control systems.

Dragan Jovcic (S97M00SM06) received the D.Eng. degree in control engineering from the University of Belgrade, Belgrade, Yugoslavia, in 1993 and
the Ph.D. degree in electrical engineering from the University of Auckland,
Auckland, New Zealand, in 1999.
Currently, he is a Lecturer with the University of Aberdeen, Aberdeen, U.K.,
where he has been since 2004. He was also a Lecturer with the University of
Ulster, Newtownabbey, U.K., from 2000 to 2004 and a Design Engineer in the
New Zealand power industry from 1999 to 2000. His research interests are in the
areas of flexible ac transmission systems devices, HVDC, and control systems.