Design and Analysis of Adaptive Neural Controller for Voltage Source Converter for STATCOM

ACEEE Int. J. on Control System and Instrumentation, Vol. 02, No. 01, Feb 2011

Design and Analysis of Adaptive Neural Controller
for Voltage Source Converter for STATCOM
Aman Ganesh1, G.K.Singh2, and Ratna Dahiya3
1
Department of Electrical and Electronics Engineering, GRIMT Radaur, India
2
Department of Electrical Engineering, IIT Roorkee, India
3
Department of Electrical Engineering, NIT Kurukshetra, India
Email: gksngfee@iitr.ernet.in, ratnadahiya@yahoo.co.in

Abstract—Usually a STATCOM is installed to support power function based PWM and correcting optimized
system networks that have a poor power factor and often poor PWMswitching patterns were implemented to obtain
voltage regulation. It is based on a power electronics voltage- waveform with minimum harmonic content from VSC. The
source converter. Various PWM techniques make selective effect and analysis of switching pattern, modulation index
harmonic elimination possible, which effectively control the
variation and load variations on voltage and current
harmonic content of voltage source converters. The distribution
systems have to supply unbalanced nonlinear loads transferring
harmonics were carried out [13-16] to cancel low order
oscillations to the DC-side of the converter in a realistic harmonics. But the harmonic output at PCC was not
operating condition. Thus, additional harmonics are modulated discussed. Therefore, a strategy was proposed [17] for
through the STATCOM at the point of common coupling voltage balancing of distinct dc buses in cascaded H- bridge
(PCC). This requires more attention when switching angles are rectifier. The method was used to ensure that the dc bus
calculated offline using the optimal PWM technique. This capacitor voltage converges to the reference value, even
paper, therefore, presents the artificial neural network model when the loads attached to them extracts different power.
for defining the switching criterion of the VSC for the Both stepped and PWM methods were used to reduce the
STATCOM in order to reduce the total harmonic distortion
current and voltage harmonics. In all such methods discussed
(THD) of the injected line current at the PCC. The model takes
into the account the dc capacitor effect, effects of other possible
above, it is observed that the conventional methods can give
varying parameters such as voltage unbalance as well as good control capability over a wide range of operating
network harmonics. A reference is developed for offline conditions. But they need a mathematical model of the system
prediction and then implemented with the help of back to be controlled, which in most cases cannot be obtained
propagation technique. easily.
In the above context, this paper presents a VSC
Index Terms— STATCOM, voltage source converter, selective based STATCOM using ANN in which voltage regulation is
harmonic elimination, adaptive neural controller, total achieved, and the input voltage harmonics are minimized
harmonic distortion. by selective harmonic elimination technique (SHEM) to
comply with IEEE Std. 519-1992 [18]. A brief introduction
I. INTRODUCTION has been provided for the role of VSC for STATCOM
The Static Synchronous Compensator (STATCOM) has applications and various PWM techniques to minimize
the capability of generating and /or absorbing reactive power voltage and current harmonics of VSC. Section II contains
at a faster rate [1]. It can control the line voltage at the point the inside view of the STATCOM followed by the
of common connection to the electric power network [2, 3]. introduction of SHEM. In section III the design of neuro
The 3-phase output voltage produced by the voltage source controller is explained. Finally, the validation and comparison
converter (VSC), which forms the power stage of of results is described in section IV, followed by conclusion
STATCOM, is nearly in phase with, and coupled to the in section V.
corresponding AC grid voltage through a relatively small
reactance [4-6]. In case of VSC, a capacitor bank is placed II.CONVENTIONAL STATCOM MODEL
into the DC link, which has the capabilities of harmonic The basic decoupled conventional control scheme of a
filtering and load balancing in addition to reactive power STATCOM to be considered is shown in Fig. 1. The two PI
compensation. This leads to the generation of harmonic controllers, PIv and PIDC, are for regulating the line voltage
components superimposed on fundamental voltage at the point of common coupling (PCC) and the dc link
component at the AC side of VSC due to the turning on and voltage inside the device. The deviations in the line voltage
turning off, of the IGBT switches at frequencies much higher ÄV and the dc link voltage ÄVdc are passed through these
than the supply frequency [7]. Many methods [8-12] such two decoupled PI controllers in order to determine the
as selective harmonic elimination, variable modulating inverter modulation index m a and the phase shift á,
respectively.

© 2011 ACEEE 1
DOI: 01.IJCSI.02.01.31


Active and reactive power consumed by the STATCOM from
the supply can be expressed as in (3) and (4) according to
power sink convention.

A. Reactive Power Control
Fig. 2 represents a simplified Single phase Y-
equivalent model of STATCOM.

Modulation index (m) can be defined as in (9) for SPWM
Where or SHEM.
VS : RMS line to neutral AC grid voltage
VC : RMS line to neutral STATCOM fundamental
voltage
Is : RMS source fundamental current
IL : RMS load fundamental current
IC : RMS STATCOM fundamental Current
QS : Source reactive power
QL :
Load reactive power
QC : STATCOM reactive power
R :Y- equivalent of total loss resistance including
coupling transformer losses, input filter losses and
converter losses
X :Y-equivalent of total reactance including source
reactance, leakage reactance of coupling transformer, and
input filter reactance.
Fig.3 represents the phasor diagram for lossy systems. If STATCOM were a lossless system, Pc, and hence
ä would be zero, while supplying constant Qc to constant
Vs’ in the steady state. However, for a practical VSC
STATCOM, system losses can be supplied by keeping ä at a
very small value, while delivering or absorbing constant Qc
in the steady state. Therefore, cosä can be approximated to
unity for STATCOM operating at constant reactive power.
In view of the above findings and (11), reactive power can
be controlled by one of the following techniques:
(i) varying modulation index m, while keeping DC
link voltage Vd constant
Equations (1) & (2) are obtained by resolving VR and VX (ii) varying Vd, while keeping m constant
along the real and imaginary axis. (iii) varying both m and Vd
The highly mathematical PWM method has two
base terms modulation index and modulation ratio.
Modulation index determines the output voltage fundamental
component, while the modulation ratio determines the
incident (location) of harmonic in the spectra.

2
© 2011 ACEEE
DOI: 01.IJCSI.02.01.31


B. Selective Harmonc Elimination A. Block Representation of Neuro Controller
Harmonic content of STATCOM line current ic in ANC maintains a population of possible neuro-
Fig.1 should comply with IEEE Std. 519-1992. This can be controller solutions that serve as reinforcement learning
achieved by applying a proper PWM technique such as evaluations, similar to EVOP experiments [21]. The neuro-
SHEM. In this research work, since fixed dc link voltage controller is evaluated individually over a number of sensor
and variable modulation index has been chosen, harmonic sample periods while interacting with a dynamic process
content of line current and load voltage will be minimized as in Fig. 5.
by SHEM technique. At the expense of coupled active and
reactive power control, and relatively slower transient
response, it is advantageous to apply SHEM to VSC topology.
A general 2 level 3 leg VSI is shown in Fig. 4.

Initially, the process may be at an arbitrary operat-
ing point (state, st). The neuro-controller observes the cur-
rent process operating point at sample, t, and selects a con-
The switching pattern of IGBT switches are defined trol action, at. The control action changes the operating point
on the basis of the modulation index (m) and the phase shift to st+1. A reward, rt, is assigned based on the economic
(á). The two patterns are then given to the VSC circuit via value of this new operating point. The objective is to maxi-
PWM module which produces switching pattern of IGBTs mize the total reward over a series of control actions,
to produce a corresponding or desired 3 phase voltage while maintaining a specified control response. An opti-
waveform for STATCOM application. mization algorithm adapts the neural network weights based
The SHEM allows optimized PWM pattern, its the reward feedback from each evaluations. The above con-
implementation is much easy. The switching frequency of troller can also be represented as shown in Fig. 6 while main-
power semiconductor, fs is directly related to the number of taining the above explained generality.
harmonics to be eliminated [19], as given below.
f s = (2 N + 1) f1 (12)
where f 1 is the grid frequency and hence the
fundamental frequency.
The ic waveforms are found by Vc waveform arising
from SHEM, by simulating this circuit in MATLAB. This is
because in application where STATCOM, is going to be
utilized as VAr compensator, or terminal voltage regulator,
point of common coupling (PCC) should be taken as common
connection of STATCOM. Therefore, total harmonic
distortion (THD) values of STATCOM line current and
individual magnitude of harmonic components at PCC are
of the designer’s direct interest, and should comply with the
IEEE std. 519-1992.

III.NEURO CONTROLLER
Neuro-controllers may originate from various sources.
Neural networks may be trained to mimic the control action The shown network can be trained for a given set
of existing VSC controllers, thereby distributing the inverter of plant parameter data-base. The training will result in
functionality over several neurons. Neuro-controllers are also optimum set of weights and bias, for a predefined number
developed utilizing evolutionary reinforcement learning of neurons, selected for achieving the selected/ or the defined
techniques [20]. Neural networks are beneficial to an performance and the control criterion.
adaptive scheme, such as generalization and graceful
degradation. Neural network controllers are collections of IV.VALIDATION AND RESULT DISCUSSION
neurons, with each neuron specifying the weights from the The switching test circuitry used is shown in Fig. 4.
input layer (process states) to output layer (control actions). The analysis is comparative in nature as the other circuitry
Neuro-controller parameters are the neural network weights. is
© 2011 ACEEE 3
DOI: 01.IJCSI.02.01.31


nearly the same except that the firing pulses for the VSC are
now generated by the adaptive neuro controller, which forms
the part of the feedback circuit that has its reference from
the conventional VSC circuitry (ref. fig. 5) and the 3 phase
voltage output is fed back to the ANN controller and then
the neuro controller generates the firing pulses for the VSC
(ref. fig 6).
The comparison is made between the current and voltage
outputs at the load and at the PCC. The simulations are carried
in MatLab and the waveforms obtained are shown below.

Figure 8.RMS fundamental load current

From the simulation results, it is found that the neural
network based VSC performs in a better way. In order to
make it more clear, the results for the odd harmonics current
components are given in Table 1.

If individual harmonic magnitude and also their
THD values are compared with the limit given in IEEE Std.
519-1992, it can be concluded that SHEM technique is very
successful in approximating the line current waveform of
STATCOM to a pure sine wave. The same can be extended
to the voltage waveforms also.

V.CONCLUSION
The paper describes the design, implementation, and
performance of an adaptive neural controller of a VSC based
STATCOM. Voltage regulation is achieved by optimizing
the modulation index and VSC harmonics are eliminated by

4
© 2011 ACEEE
DOI: 01.IJCSI.02.01.31


SHEM technique. The proposed system complies with the [10] J.Y. Lee and Y.Y.Sun, “Adaptive harmonic control in PWM
IEEE std. 519-1992 The selective harmonic elimination inverters with fluctuating input voltage,” IEEE Transaction
technique applied to voltage source converter based static on Industrial Electronics, vol. 33, pp. 92-98, Feb. 1986.
synchronous compensator leads to reduction in switching [11] P. N.Enjeti, and W.Shireen, “A new technique to reject dc-
frequency, the same method can also be used for reactive link voltage ripple for inverters operating on programmed
PWM waveform,” IEEE Transaction on Power Electronics,
power control with slight modification in design and
vol. 7, pp. 171-180, Jan. 1992.
implementation. [12] Sirisukprasert Siriroj, Lai Jih- Sheng and Liu Tina-Hua,
“Optimum harmonic reduction with a wide range of
REFERENCES modulation indexes for multilevel converters,” IEEE
Transaction on Industrial Electronics, vol. 49, no.4, pp. 2094-
[1] J. Dixon, L. Moran, J. Pontt, “Reactive power compensation
2099, 2000.
techniques”, Proc. IEEE, vol. 93, no 12, pp. 2144-2164,
[13] Soto Diego and Pena Ruben, “Nonlinear control strategies
December 2005.
for cascaded multilevel STATCOMs,” IEEE Transaction on
[2] G. Ree, J. Paserba., “SDG&E Talega STATCOM project
Power Delivery, vol. 19, no. 4, pp. 1919-1927, Oct. 2004.
system analysis, design, and configuration”, IEEE/PES,
[14] Filizadeh Shaahin and A. M.Gol, “Harmonic performance
Transmission and Distribution Conf. and Exhibition, vol.2,
analysis of an OPWM-Controlled STATCOM in network
pp 1393-1398, October 2002.
applications,” IEEE Transaction on Power Delivery, vol. 20,
[3] C.Schauder, M. Gernhardt, et. al.., “Development of a ± 100
no. 2, pp. 1001-1008, April 2005.
MVA static condenser for voltage control of transmission
[15] Chen Ben-Sheng and Hsu Yuan-Yih, “An analytical approach
systems”, IEEE Transaction on Power Delivery, vol.10, no.
to harmonic analysis and controller design of a STATCOM,”
3, pp. 1486-1496, October 2002.
IEEE Transaction on Power Delivery, vol. 22, no. 1 pp.423-
[4] H. F. Biligin, M. Ermis, “Reactive power compensation of
432, Jan.2007.
coal mining excavators by using a new generation
[16] Song Qiang, Liu Wenhua and Yuan Zhichang, “Multilevel
STATCOM”, IEEE Transaction on Industrial Application, vol.
optimal modulation & dynamic control strategies for
43, no 1, pp. 97-110, January 2007.
STATCOMs using cascade multilevel inverters,” IEEE
[5] Cetin, “Design and implementation of a voltage source
Transaction on Power Delivery, vol. 22, no.3 pp.1937-1946,
converter based STATCOM for reactive power compensation
July 2007.
and harmonic filtering”, Ph.D dissertation, Graduate School
[17] Eini Hossein Iman, Schanen Jean-Luc, Farhangi Shahrokh
of Natural and Applied Sciences, METU, Ankara, Tukey,
and Roudet James, “A modular strategy for control and voltage
2007.
balancing of cascaded H-Bridge rectifiers,” IEEE Transaction
[6] L.T. Moran, P.D. Ziogas, G. Joos, “ Analysis and design of a
on Power Electronics, vol. 23, no. 5, pp. 2428-2442, Sept.
three phase synchronous solid state VAr compensator”, IEEE
2008.
Transaction on Industrial Application, vol. 25, no 4, pp. 598-
[18] IEEE Recommended Practices & Requirement for Harmonic
608,1989.
control in Electrical Power Systems, IEEE 519-1992, 1993.
[7] A.L. Julian, G. Oriti, T.A. Lipo, “Elimination of common mode
[19] P. Enjeti, P.D. Ziogas, J.F. Lindsay, “Programmed PWM
voltage in three phase sinusoidal power converters”, IEEE
technique to eliminate harmonics: A Critical Evaluation”,
Transaction on Power Electronics, vol. 14, no 5, pp. 982-
IEEE Transaction on Industrial Application, vol.26, no. 2,
989,1999.
pp. 302-316, 1990.
[8] Ran L., L Holdsworth., and G.A.Purtus, “Dynamic selective
[20] C.H. S’equim and R.D. Clay, “Fault tolerance in artificial
harmonic elimination of a three-level inverter used for static
neural networks”. International Joint Conference on Neural
VAr compensation,” Proc. Inst. Elect. Eng., Gen., Trans., Dist.,
Networks (San. Diego), vol 1, pp. 703-708, 1990.
vol. 149, pp. 83-89, Jan. 2002.
[9] D. W. Sandells and T. C.Green, “Firing angle optimization [21] L. P. Kaelbling, M.L. Littman and A.W. Moore,
for chain cell converters,” in Proc. 32nd IEEE Annual Power “Reinforcement Learning: A survey”, Journal of
Electronics Specialists Conference, pp. 1549-1554. June 2001 Artificial Intelligence Research, vol. 4, pp.237-285,
1996.

© 2011 ACEEE 5
DOI: 01.IJCSI.02.01.31

Design and Analysis of Adaptive Neural Controller for Voltage Source Converter for STATCOM

More Related Content

What's hot (19)

Viewers also liked (19)

Similar to Design and Analysis of Adaptive Neural Controller for Voltage Source Converter for STATCOM (20)

More from IDES Editor (20)

Recently uploaded (20)

Design and Analysis of Adaptive Neural Controller for Voltage Source Converter for STATCOM