172

CONTROL OF DYNAMIC VOLTAGE RESTORER

FOR VOLTAGE SAG MITIGATION
Glanny Martial Christiaan Mangindaan1, 2, Mochamad Ashari1, Mauridhi Hery Purnomo1
1)
Electrical Engineering, Sepuluh Nopember Institute of Technology
Kampus ITS Sukolilo Surabaya 60111
2)
Electrical Engineering, Sam Ratulangi University
Kampus Unsrat Bahu Manado 95115
email: [email protected]

ABSTRACT voltage source PWM converter. Current control

(CC) techniques for voltage source [1], a high
A dynamic voltage restorer (DVR) is power performance current regulation scheme [2], a
electronics equipment that can protect a sensitive hysteresis control method of current controlled VSI
load from the effect of voltage sag. In this paper PWM inverters to minimizes mutual interferences
present the comparative performance of dynamic among the control of the various phases [3],
voltage restorer using PI, fuzzy logic and neural Voltage distortion caused by slowly varying
network control. Design of controller to control the harmonic current obtained by using synchronous
dynamic voltage restorer (DVR) that can result a frame harmonic regulators [4], combination of
quick and precise respond should be present in this supply voltage feed-forward and PI d-q load voltage
paper. feedback [5] voltage space vector PWM [6], three
phase resonant PI regulator [7], fuzzy logic control
Keywords: Dynamic voltage restorer (DVR), [8] Narma-L2 controller neural network [9].
voltage sag, PI, fuzzy logic, neural network.

1 INTRODUCTION
Supply Critical
In power system, disturbance, such as voltage Voltage Load
sag, occur on some occasions that inevitable. These
disturbances are usually as result of faults in the
power system. Dynamic voltage restorer can
provide the most cost effective solution to mitigate
voltage sags by establishing the proper voltage
quality level that is required by customer. When a
Energy Power
fault happens in a distribution network, sudden Storage Converter
voltage sag will appear on adjacent loads. DVR
Dynamic Voltage Restorer (DVR)
installed on a critical load, restores the line voltage
to its nominal value within the response time of a
few millisecond thus avoiding any power disruption Figure 1. Dynamic Voltage Restorer Operation
to the load, as shown in Figure 1. Voltage sags are
characterized by a reduction in voltage, but the load
is still connected to the supply. Sags are inmost This paper compares the counterpropagation
cases considered less critical compared to neural network with fuzzy logic and PI controller.
interruptions, but they typically occur more Counterpropagation was originally proposed as a
frequently. Voltage sags have in several cases been pattern-lookup system that takes advantage of the
reported as a threat to sensitive equipment and have parallel architecture of neural networks [10]. Two
resulted in shutdowns, loss of production and a types of layers are used: The hidden layer is a
hence a major cost burden. Kohonen layer with competitive units that do
unsupervised learning; the output layer is a
The DVR employ IGBT solid state power Grossberg layer, which is fully connected with the
electronics switching devices in a pulse width hidden layer and is not competitive.
modulation (PWM) inverter structure. Various
control strategy have been proposed for three phase

ISSN 1858-1633 @2008 ICTS

Control of Dynamic Voltage Restorer for Voltage Sag Mitigation - Glanny Martial Christiaan Mangindaan 173

2 DYNAMIC VOLTAGE The voltage regulator is important part for

RESTORER control a DVR. Reference input is compared with
the voltage on the source side of the DVR to
The power circuit of DVR in Figure 1 can be produce an error. The controller in the voltage
divided into six parts shown in Figure 2: A— regulator can contain of the neural network, fuzzy
voltage regulator; B—PWM generator; C—energy logic or PI controller. Detail of voltage regulator
storage; D—IGBT inverter; E—filter; and F— shown in Figure 3.
series transformer.

A. Voltage Regulator

Figure 2. Dynamic Voltage Restorer Power Circuit

B. PWM Generator The DC Voltage Source block implements an

The output pulses are a vector (with values=0 or ideal DC voltage source. The positive terminal is
1). Depending on the selected "Generator Mode", represented by a plus sign on one port. The voltage
the output vector contains: For a 3-arm bridge: Six applied is 100 V, and structure of energy storage is
pulses. Pulses 1, 3 and 5 are respectively for the for half bridge DC/AC converter.
upper switches of the first, second and third arm.
Pulses 2, 4 and 6 are for the lower switches. The D. IGBT Inverter
carrier frequency and sampling time are 10 [kHz] The gate input for the controlled switch devices.
and 5.10-6 [s] respectively. The pulse ordering in the vector of the gate signals
corresponds to the switch. For the diode bridges,
C. Energy Storage the pulse ordering corresponds to the natural order

ISSN 1858-1633 @2008 ICTS

174 4th International Conference Information & Communication Technology and System

of commutation. The IGBT inverter is controlled injection transformers. In this paper, the filter
with a voltage regulator. placed on the inverter side, with the advantaged low
voltage side and close to harmonic source. With
E. Filter this scheme high order harmonic current are
Filtering in DVR can be placed either on the prevented from penetrating to series transformer.
high-voltage side or the inverter side of the series

Figure 3. Voltage Regulator Block

F. Series Transformer impedance at faulted line. u sag , differs from the

Series transformer is three single phase prefault voltage u pre-sag by u missing .
transformers that can be connected to distribution
line star/open winding or delta/open winding. The
=
usag u pre − sag − umissing (2)
star/open winding allows the injection of positive, Voltage sag can be symmetrical or
negative and zero sequence voltages, where the nonsymmetrical. However, most voltage sags are
delta/open winding only allows positive and nonsymmetrical. Figure 4 shows the variation
negative sequence voltages. In this paper, use voltage magnitudes and durations in IEEE 1159.
series transformer connected star/open winding.
2.2 Control Methods
2.1 Voltage Sags The performance of the DVR depends on the
Voltage sags are usually caused by a short quality of applied control strategy. Therefore,
circuit current flowing into a fault on a transmission control methods of DVR are important subject in
or distribution line, where the magnitude and phase this paper.
of the faulted voltage usag during the sag at the
A. PI Controller
point of common coupling (PCC) are determined by The PI parameters are determinable from
the fault and supply impedances using the equation simulink response optimization simulation. The
[11]: constraints on the behavior of the response signals
Zf and the tuned parameters are set within the signal
usag = E (1) constraint blocks.
Z f + Zs
Where u sag is voltage during the sag at PCC, E is
supply voltage, Z s is supply impedance, Z f is

ISSN 1858-1633 @2008 ICTS

Control of Dynamic Voltage Restorer for Voltage Sag Mitigation - Glanny Martial Christiaan Mangindaan 175

network may not generalize sufficiently. With too

Transient
many, the processing time takes too long.
Swell Overvoltage
110%
For the network to work properly, the input
Normal operating voltage
Event magnitude

90 %
vector must be normalized. This means that for
every combination of input value, the total length of
the input vector must add up to one. This can be
Notch/transient

Voltage Sag Undervoltage

done with a preprocessor, before the data is entered
into the counterpropagation network, or a
10 %
Momentary Temporary
normalization layer added between the input and
Sustianed interuption
0.5 cycle 3 sec
Kohonen layer. During training, the weights are
1 min
Event duration adapted as follows [10]:
Figure 4. Variation Voltage Magnitudes and Durations 1. A winner of the competition is chosen in
response to an input pattern.
2. The weights between the input layer and
The first step in the response optimization the winner are adjusted according
process is to choose which signals in simulink to:
model would like to constrain and to attach signal
constraint blocks to these signals. Although specify
= wic (t ) + α (oi − wic (t )) .
wic (t + 1)
the constraints for each signal individually, set the All the other weights remain unchanged.
remaining settings such as tuned parameters and 3. The output of the network is computed
optimization settings within one signal constraint and compared to the target pattern.
window as they apply to the whole project. The last 4. The weights between the winner and the
step is run the response optimization, and it stop output layer are updated according
until the parameters pass the signal constraints. to:
1) wck (t ) + β (ok − wck (t )) .
wck (t +=
B. Fuzzy Logic Controller All the other weights remain unchanged.
The fuzzy logic control derived from fuzzy set
theory introduced by Zadeh in 1965. Fuzzy logic
controllers are an attractive when precise 3 SIMULATION RESULT
mathematical formulation are not possible. Other A detailed simulation has been carried out using
advantages of fuzzy logic are: 1) it can work with MATLAB software to verify the three control
less precise inputs; 2) it doesn’t need fast strategy. The total simulation period is 0.09 s.
processors; 3) it needs less data storage in the form Voltage sag generated with three phase faults, two
of membership functions and rules than phase faults with ground connected and single
conventional look up table for non linear phase fault, simultaneously.
controllers; and 4) it is more robust than other non The three phase fault simulated with 0.6-p.u.
linear controllers. remain voltage and from 0.011 to 0.061s. Two
phase fault with 0.4-p.u. remain voltage from 0.025
A reference input is compared with the voltage to 0.045s, and single phase fault with 0.2-p.u.
on the source side of the DVR to produce an error. remain voltage from 0.031 to 0.051s.
The error is fed to the fuzzy logic controller, which
performs calculations to generate output [8].

C. Neural Network Controller

Robert Hecht – Nielsen developed the counter-
propagation network as a means to combine an
unsupervised Kohonen layer with a teachable
output layer [12]. The operation
counterpropagation network is similar to that of the
Learning Vector Quantization network in the
middle that Kohonen layer acts as an adaptive look-
up table, finding the closest fit to an input stimulus
and outputting its equivalent mapping. The size of
the input layer depends upon how many separable
parameters define the problem. With too few, the

ISSN 1858-1633 @2008 ICTS

176 4th International Conference Information & Communication Technology and System

In Figure 5 – 7, showed the simulation result of

this three control strategy. The upper side of each
figure is the voltage source with sag, and the below
is the voltage at the critical load. All type of this
control strategy can mitigate the voltage sag at the
critical load side.

When compensate the voltage sag at the critical

load, DVR produce a harmonics distortion fed from
series transformer as an injection voltage to the
critical load. Using the FFT analysis to analyze the
THD for the voltage signal, shown in Figure 8 – 10,
result that the fuzzy logic (0.82%) are better than PI
(1.44%), and the counterpropagation neural
network(0.79%) slightly better than fuzzy logic for
the THD at the one phase with the same fault.
Figure 5. Simulation result with PI controller

Figure 6. Simulation result with fuzzy logic controller Figure 8. FFT analysis for PI controller

Figure 7. Simulation result with counterpropagation neural

Figure 9. FFT analysis for fuzzy logic controller
network controller

ISSN 1858-1633 @2008 ICTS

Control of Dynamic Voltage Restorer for Voltage Sag Mitigation - Glanny Martial Christiaan Mangindaan 177

with Constant Modulation Frequency, IEEE

Transaction on Industry Application,
Vol.26, No.1, 1990, pp.88-92.
[4] Mattavelli, P., (2001), Synchronous-Frame
Harmonic Control for High-Performance
AC Power Supplies, IEEE Transaction on
Industry Application, Vol.37, No.3, pp.864-
872.
[5] Nielsen, J. G., Newman, M., Nielsen, H.,
and Blaabjerg, F., Control and Testing of a
Dynamic Voltage Restorer (DVR) at
Medium Voltage Level, IEEE Transaction
on Power Electronics, Vol.19, No.3, 2004,
pp.806-513.
[6] Zhan, C., Ramachandaramurthy, V. K.,
Arulampalam, A., Fitzer, C., Kromlidis, S.,
Barnes, M. and Jenkins, N., Dynamic
Figure 10. FFT analysis for counterpropagation neural network
controller
Voltage Restorer Based on Voltage-Space-
Vector PWM Control, IEEE Transaction on
Industry Application, Vol.37, No.6, 2001,
4 CONCLUSION pp.1855-1863.
[7] Zmood, D. N., Holmes, D. G. and Bode, G.
This paper has presented result from DVR H., Frequency-Domain Analysis of Three-
simulation at the low voltage (0.38 kV) distribution Phase Linear Current Regulators, IEEE
line. The DVR is designed to protect the critical Transaction on Industry Application,
load against voltage sag from 0.9-p.u. up to 0.1-p.u. Vol.37, No.2, 2001, pp.601-610.
remain voltage. [8] Jurado F., Fuzzy Logic Control of a
Counterpropagation neural network control and Dynamic Voltage Restorer, IEEE
fuzzy logic control provides good voltage Symposium on Industrial Electronics,
regulation capabilities than a PI controller, that the Vol.2, 2004, pp.1047 – 1052.
control parameters are fixing. [9] Jurado, F., Neural Network Control For
Both artificial intelligent results a good Dynamic Voltage Restorer, IEEE
performance to mitigate the voltage sags. The Transaction on Industrial Electronics,
hybrid method of artificial intelligent to be applied Vol.51, No.3, 2004 pp. 727-729.
is the interesting field for future research. [10] Anonymous, Stuttgart Neural Network
Simulator (SNNS) User Manual ver.4.1.,
Germany, University of Stuttgart 1995.
[11] Nielsen, J. G. and Blaabjerg, F., A Detailed
Comparison of System Topologies for
REFERENCE Dynamic Voltage Restorers, IEEE
[1] Kazmierkowski, M. P. and Malesani, L., Transaction on Industry Applications,
Current Control Techniques for Three- Vol.41, No.5, 2005, pp.1272-1280.
Phase Voltage-Source PWM Converters: A [12] Purnomo, M. H. and Kurniawan, A.,
Survey, IEEE Transaction on Industrial Supervised Neural Networks dan
Electronics, Vol.45, No.5, 1998, pp.691- aplikasinya, Yogyakarta, Graha Ilmu, First
703. edition, 2006.
[2] Lee, D. C., Sul, S. K. and Park, M. H., High
Performance Current Regulator for a Field-
Oriented Controlled Induction Motor Drive,
IEEE Transaction on Industry Application,
Vol.30, No.5, 1994, pp.1247-1257.
[3] Malesani, L. and Tenti, P., A Novel
Hysteresis Control Method for Current-
Controlled Voltage-Source PWM Inverters

ISSN 1858-1633 @2008 ICTS