You are on page 1of 6

,&3'(1

SIMULATIONS OF LIGHTNING IMPULSE


RESIDUAL VOLTAGE TEST OF SURGE
ARRESTERS IN MATLAB-SIMULINK
Kaveri Bhuyan Saibal Chatterjee
Department of Electric Power Engineering, Department of Electrical Engineering,
Norwegian University of Science and Technology North Eastern Regional Institute of Science and Technology,
Trondheim, Norway Nirjuli, Arunachal Pradesh, India
kaveri.bhuyan@ntnu.no saibalda@ieee.org

Abstract— Assessment of the dynamic characteristics of surge


arrester is an important aspect of insulation coordination. II.MODELS DESCRIPTION
Computer simulation of residual impulse voltage test of surge The IEEE [6], Pinceti–Gianettoni [7] and the Fernandez–
arrester models have been done using Alternative Transients
Diaz [8] surge arrester models are considered for
Program (ATP) and Power systems computer aided design
(PSCAD) design softwares. In this paper, the models have been investigation in this work. The referred models [6-8] are
developed in MATLAB Simulink. The accuracy of the models are described below:
ascertained by comparing the results with the manufacturer’s
design data. I. The IEEE model [6]

Index Terms— Surge arresters, Residual voltage, Impulse,


Simulations, MATLAB Simulink

I.INTRODUCTION
Metal oxide surge arresters are used to protect medium
and high voltage systems and insulation of equipment
against over voltages. Surge arresters are often subjected
to high voltage stresses due to high earth resistance of the
grid grounding system, sealing defects and Fig. 1 IEEE model [6]
environmental contamination, resonance and switching
and lightning over voltages, repeated transients of long The IEEE model [6] (Fig. 1) has two non-linear resistances
duration and large magnitude [1, 2]. Metal-oxide surge A0 and A1, and five linear elements L0, R0, L1, R1 and C0. The
arresters are designed to be highly resistive under normal function of the inductive elements is to characterize the
operation and conductive under transient over voltages. model behavior with respect to fast surges [6, 8]. The non-
The duration of arrester current flow depends on linear elements are represented by non-linear resistances
overvoltage duration [3]. Thus, surge arresters have a (A0 and A1) whose variations are given by equation (1) and
nonlinear voltage versus current (V-I) characteristic [4- the nonlinear V-I characteristics are shown in Fig. 2 [6].
5]. The per unit voltage values are referred to the peak value of
the residual voltage measured during a discharge test with a
The most widely used frequency-dependent surge 10 kA, 8/20 μs lightning current impulse.
arresters models for accurate simulation of surge arrester V–I characteristic of the surge arrester depend upon the
behavior under different kinds of stress are the IEEE [6], wave shape of the arrester current and is expressed as [10,
Pinceti–Gianettoni [7] and the Fernandez–Diaz model 11]:
[8]. The protective performance of these surge arrester i = p (v/ Vref ) q (1)
models are determined by the residual voltage test [9]. In where i and v are the arrester discharge current and residual
this paper, surge arresters from different manufacturers, voltage respectively, p and q are constants of device and Vref
nominal discharge currents and rated voltage levels are is an arbitrary reference voltage near to the rated voltage,
simulated using MATLAB SIMULINK. The main Vr.
objective is to perform lightning impulse residual voltage Vref is calculated using the expression given as [12]:
test and compare the simulated results with manufacturer Vr = kVref (2)
data. The proposed approach will help to select optimum Where value of k varies between 0.9 to 0.98
arrester model for transient studies.

‹,(((
The constants p and q for the segment between (I0,V0) to into account any physical dimension of the arrester for
(I1,V1) in V-I characteristics can be obtained from the parameter calculation. This model can give satisfactory
following expressions [12]: results for discharge currents within a range of times to
p = I0 / (V0/Vref) q (3) crest for 1 μs to 30 μs [7].
q = ln (I1/I0)/ ln (V1/V0) (4)
III. Fernandez–Diaz model [8]

Fig. 4 Fernandez–Diaz model [8]

Fernandez-Diaz model [8] (Fig. 4) consists of three linear


elements L0, R0, and C and two non-linear resistors (A0 and
Fig. 2 V-I characteristics for non linear resistors A0 and A1). It is also a simplified IEEE model [6]. R1-L1 filter
A1 circuit between the non-linear resistors A0 and A1 (as shown
in Fig. 1) is neglected in Fernandez-Diaz model [8] and
only the inductance, L0 is taken into account. Capacitance,
The inductance, L0 represents the inductance associated
C is used to represent the terminal-to-terminal capacitance
with the magnetic fields in the surrounding and the resistor,
of the arrester. The resistance, R0 is connected in parallel to
R0 is used to obtain numerical stability. The capacitance, C0
A0 to avoid numerical instability and oscillations. This
represents the external capacitance due to height of the
model can give satisfactory results for discharge currents
arrester [5]. This model can give satisfactory results for
within a range of times to crest from 8 μs.
discharge currents within a range of times to crest for 0.5 μs
to 45 μs [6-8]. The values L0, R0, L1, R1 and C0 are
determined using the physical parameters related to overall III.MODELLING AND SIMULATIONS
height, diameter of block, number of columns etc. of the
arrester. The simulations are performed in MATLAB SIMULINK
using 8/20 μs standard lightning current impulse waveform
II. Pinceti–Gianettoni model [7] having amplitude of 5, and 10 kA on surge arresters models
of 15 kV, 120 kV and 150 kV rated voltages. The framework
of investigation has been shown in Fig. 5.

Fig. 3 Pinceti–Gianettoni model [7]

The Pinceti–Gianettoni model [7] (Fig. 3) consists of three


linear elements L0, R0, and L1 and two non-linear resistors
(A0 and A1). It is a simplified IEEE model [6]. Comparing
the IEEE model and the Pinceti–Gianettoni model, it is seen
that in the later model the capacitor, C0 is eliminated as it
has negligible effect on the model behavior. Also, the two
resistances R0 and R1 connected in parallel with the
inductances L0 and L1 are replaced by one resistance R0
(about 1 MŸ) between the input terminals to obtain
numerical stability. The electrical data of the arrester are
used to determine the model parameters. It does not take Fig. 5 Framework of Investigation
Table III. Model parameters for 120 kV surge arresters [6-
The reference voltage of the surge arrester elements are 8, 10]
adjusted with respect to the rated voltage and manufacturer’s
design data. To select the optimum surge arrester model,
relative errors between observed residual voltages and
manufacturer data are calculated and compared. Formulas to
calculate parameters of the circuits shown in Fig. 1, 3, and 4
were initially suggested in [6-8]. Recently, many different
procedures based on numerical optimization techniques have
been presented in literature for estimating the parameters in
less time and more accurately. In this work, optimized
values are obtained using Genetic Algorithm [13],[11], Table IV. Model parameters for 150 kV surge arresters
Powell’s Optimization method [14] and Modified Particle [6-8, 14]
Swarm Optimization (MPSO) algorithm [10]. The electrical
data of the tested surge arresters are given in Table I [10, 13,
14]. The model parameters are given in Table II, III and IV
for the three surge arresters under consideration.

Table I. Surge arrester design data [10, 13, 14]

Reference Rated voltage 15 kV


MCOV 12 kV IV.OBSERVATIONS
[13] Max.Residual 5 kA 36.72 kV Discharge currents of 5 kA, and 10 kA of 8/20 μs is applied
voltages with 10 kA 38.88 kV
8/20 μs 20 kA 43.37 kV
to the models. The residual voltages obtained are shown in
Height 183 mm Fig. 6. The maximum value of the voltage at the arrester
Reference Rated voltage 120 kV terminals (residual voltages) is recorded and compared with
MCOV 94 kV the manufacturer data to validate the model. The
[10] Max.Residual 5 kA 275 kV observations are tabulated in Table V and VI.
voltages with 10 kA 294 kV
8/20 μs 20 kA 319 kV
Height 1140 mm
Reference Rated voltage 150 kV
MCOV 120 kV
[14] Max.Residual 5 kA 367 kV
voltages with 10 kA 396 kV
8/20 μs 20 kA 449 kV
Height 1330 mm

Table II. Model parameters for 15 kV surge arresters [6-8,


13]

(a) Residual voltage obtained with initial model parameters


for 15 kV surge arrester model against 5 kA, 8/20 μs current
impulse

(b)Residual voltage obtained with optimized model


parameters for 15 kV surge arrester model against 5 kA,
8/20 μs current impulse
(c) Residual voltage obtained with initial model
parameters for 120 kV surge arrester model against (g) Residual voltage obtained with initial model
5 kA, 8/20 μs current impulse parameters for 15 kV surge arrester model against
10 kA, 8/20 μs current impulse

(d) Residual voltage obtained with optimized model


parameters for 120 kV surge arrester model against (h) Residual voltage obtained with optimized model
5 kA, 8/20 μs current impulse parameters for 15 kV surge arrester model against
10 kA, 8/20 μs current impulse

(e) Residual voltage obtained with initial model


parameters for 150 kV surge arrester model against (i) Residual voltage obtained with initial model
5 kA, 8/20 μs current impulse parameters for 120 kV surge arrester model against
10 kA, 8/20 μs current impulse

(f) Residual voltage obtained with optimized model


parameters for 150 kV surge arrester model against (j) Residual voltage obtained with optimized model
5 kA, 8/20 μs current impulse parameters for 120 kV surge arrester model against
10 kA, 8/20 μs current impulse
V.CONCLUSIONS
The IEEE W.G. 3.4.11, Pinceti–Gianettoni and Fernandez–
Diaz models are developed in MATLAB Simulink. These
frequency dependent surge arrester models are subjected to
5 kA, and 10 kA of 8/20 μs lightning impulse current
waveform. The impulse current waveforms have been
generated in MATLAB. The residual voltage appearing
across the arrester terminals are recorded. The effectiveness
of the developed models was tested by comparing with the
manufacturer’s residual voltage test results. The accuracy of
(k)Residual voltage obtained with initial model the models is ascertained from the relative error values of
parameters for 150 kV surge arrester model against 10 the residual voltages which do not exceed 1.02 %.
kA, 8/20 μs current impulse
ACKNOWLEDGEMENT

Kaveri Bhuyan is funded by ERCIM Alian Bensoussan


Fellowship for carrying out her research works at NTNU,
Norway. The second author acknowledge the support of
DST-FIST Program project number ETI-211/2007, dated
28-04-2008 for this work.

REFERENCES
[1] Mohammad Akbar, and Monir Ahmad, "Failure Study of
Metal-Oxide Surge Arresters," Electric Power Systems
(l) Residual voltage obtained with optimized model Research, vol. 50, pp. 79-82, May 1999.
parameters for 150 kV surge arrester model against 10 [2] E. C.Sakshaug, J.J.Burke, and J.S.Kresge, "Metal Oxide
kA, 8/20 μs current impulse Arresters on Distribution Systems: Fundamental
Considerations," IEEE Transactions on Power Delivery, vol.
Fig. 6 Residual voltages obtained against discharge currents 4, pp. 2076-2089, October, 1989.
of 5 kA, and 10 kA of 8/20 μs impulse applied to surge [3] G. L. Goedde, L. J. Kojovic, M. B. Marz, and J.J.
arrester models of 15 kV, 120 kV and 150 kV rated voltage. Woodworth, "Series Graded Gapped Arrester Provides
Reliable Overvoltage Protection in Distribution Systems,"
Table V. Error percentage obtained using the initial IEEE Power Engineering Society Winter Meeting, vol. 3, pp.
parameter values 1122-1127, 2001.
[4] J. A. Martinez, and D.W Durbak, "Parameter Determination
for Modeling Systems Transients-Part V: Surge Arresters,"
IEEE Transactions on Power Delivery vol. 20, pp. 2073-
2078, July 2005.
[5] A. Bayadi , N. Harid, K. Zehar, and S. Belkhiat "Simulation
of Metal Oxide Surge Arrester Dynamic Behavior under Fast
Transients ", The International Conference on Power Systems
Transients - IPST, New Orleans, USA, 2003.
[6] IEEE Working Group 3.4.11 "Modeling of Metal Oxide
Surge Arresters," IEEE Transactions on Power Delivery, pp.
302-309, 1992.
Table VI. Error percentage obtained using the optimized [7] P. Pinceti, and M. Giannettoni, "A Simpli¿ed Model for Zinc
parameter values Oxide Surge Arresters," IEEE Transactions on Power
Delivery, pp. 393-398, 1999.
[8] F. Fernandez, and R. Diaz, "Metal Oxide Surge Arrester
Model for Fast Transient Simulations," International
Conference on Power System Transients, Rio De Janeiro,
Brazil, 2001.
[9] A. Haddad, P. Naylor, I. A. Metwally, D. M. German, and
R.T. Waters "An Improved Non-Inductive Impulse Voltage
Measurement Technique for Zno Surge Arresters," IEEE
Transactions on Power Delivery, vol. 10, pp. 778-785, April,
1995.
[10]M. Nafar, G.B. Gharehpetian, T. Niknam, "Improvement of [13]C.A. Christodoulou, I.F. Gonos, and I.A. Stathopulos,
Estimation of Surge Arrester Parameters by Using Modified "Estimation of the Parameters of Metal Oxide Gapless Surge
Particle Swarm Optimization," Energy, Volume 36, pp. 4848- Arrester Equivalent Circuit Models Using Genetic
4854, August, 2011. Algorithm," Electric Power Systems Research, vol. 81, pp.
[11]A. Bayadi "Parameter Identification of Zno Surge Arrester 1881-1886, October, 2011.
Models Based on Genetic Algorithms," Electric Power [14]C.A.Christodoulou, V.Vita, L. Ekonomoub, G.E. Chatzarakis,
Systems Research, vol. 78, pp. 1204-1209, July, 2008. and I.A. Stathopulos "Application of Powell’s Optimization
[12]M. Garcia-Gracia, S. Baldovinos, M. Sanz, and L. Montanes, Method to Surge Arrester Circuit Model Parameters", Energy
"Evaluation of the Failure Probability for Gapless Metal 35,pp.3375-3380,2010.
Oxide Arresters," IEEE Transmission and Distribution
Conference, vol. 2, pp. 700-705, April, 1999.

You might also like