Simulating and Comparing the Transient Behavior of Different Models

of MV Metal Oxide Surge Arresters

Mahmoud Eshagh Ahmadi, Mohsen Niasati, Mohammad Reza Barzegar-Bafruee

Faculty of Electrical and Computer Engineering
Semnan University
Semnan, Iran
[email protected]

Abstract_The results of measuring residual voltage of metal oxide surge arresters (MOSAs) show that MOSA has dynamic
behavior against switching and lightning overvoltages. For this reason, to obtain the real behavior of MOSA, several models have
been proposed. In this paper, regarding the importance of MOSA modeling and selecting an appropriate model in different studies,
four models-IEEE, Pinceti, Fernandez, and P-K are analyzed under different scenarios. The simulation results are also compared with
the manufacturer’s data and the accuracy of mentioned models are determined in details. This study, which is performed by EMTP-
RV® software, indicates that concentrating on the particular model for all studies does not provide the appropriate accuracy.

Keywords: Metal oxide surge arrester, Arrester models, Lightning, Switching, Transient overvoltages.

investigate a multitude of different structural properties in

1- Introduction the design and construction phase, as well as the expected
general behavior and performance in the application
Power system equipment are vulnerable against phase. Therefore, computer simulations are really a cost-
transient overvoltages which are characterized by the effective method. However, the quality of computer
extreme high amplitude and frequency. Therefore, it is simulations can only be as good as the quality of the
essential the protection of the power system equipment built-in models and the applied data. In particular, MOSA
against these waves. As one of the promising approaches, has a nonlinear nature, and even temperature can also
metal oxide surge arrester (MOSA) has been introduced affect its performance [3]. The modeling of linear
to protect electrical equipment against transient systems is generally so much easier than the non-linear
overvoltages and current discharging [1–2]. Surge systems and the results have lower error, Therefore, the
arresters (SAs) that operate as same as the voltage limiter, modeling of actual transient behavior of MOSA is not
transfer stored energy in transient waves into the ground. easy. Attempts to identify the non-linear characteristic of
Such operation prevents from dielectric breakdown and MOSA and numerous application of MOSA in power
increases reliability. system have been caused the presentation of various
There are different types of SAs; however, due to the frequency-dependent models for simulating MOSA [4–
excellent non-linear V-I characteristic, MOSA has been 10]. In among of the proposed models, IEEE, Pinceti,
preferred with other ones (like silicon carbide). MOSA Fernandez, and P-K models are more common and have
that is formed from the series or parallel non-linear shown acceptable accuracy in different studies. In this
resistance, provides the astounding nonlinear V-I paper, due to necessity, the performance of mentioned
characteristic. In operating voltage, it has very high models is comprehensively evaluated under different
resistance and little loss. However, at lightening or transient waves.
switching over voltages, its resistance reduces quickly. In this paper, regarding the importance of these four
In medium voltage type, to prevent leakage current, models, the parameters of each model and required data
manufacturers add air gap in the structure of MOSA. The are extensively described. Then, referring to the a few
modern high voltage types are also constructed from previous studies in term of comparison of MOSA models
metal oxide disks in porcelain or polymer housing [1-2]. [5], [11–12], the transient behavior of mentioned models
Neglecting the advantages of applying MOSA in are analyzed under standard switching and lightning
power system, computer simulation is a very valuable transient waves with the help of EMTP-RV® software.
tool in many different contexts. It makes it possible to Finally, considering manufacturer’s data, the accuracy of

1
mentioned models is evaluated in details. In related to this model, the above-mentioned
parameters are calculated as follows:

2- The Proposed Models for MOSA L0  0.2d / n ( H)

R 0  100d / n ()
MOSA is made of nonlinear resistors stacks in a one
or more column. The common material used in this surge L1  15d / n ( H) (1)
arresters, are mainly made of Zinc. For this reason, R1  65d / n ()
MOSAs are also known as Zinc Oxide (ZnO) surge C  100n / d (pF)
arresters. The main characteristic of MOSA is an extreme
non-linear V-I curve with high energy absorption; Where d is the estimated height of the arrester in
however, the frequency and temperature can affect the meter (obtained from manufacturer’s data) and n is the
metal oxide substances. Temperature dependence appears number of parallel columns of metal oxide in the arrester.
in low current (less than 10A) and it can be neglected in The parameter L1 has the most influence on the results;
transient overvoltages studies. But, the frequency- however, equation related to L1 computes the initial value
dependent of the metal oxide substances can impact on and does not acceptable accuracy. Hence, to increase the
wave front-time and residual voltage of MOSA. accuracy of the model, L1 should be adjusted by trial and
Experimental results indicate that at the same condition error procedure to match the residual voltages for
for discharge current amplitude, as the crest time of the lightning discharge currents published in the
current wave decreases from 8µs to 1.3µs, the residual manufacturer's catalog [13]. Trial and error process and
voltage approximately increases about 6% [7]. According the need for the physical dimensions of arrester are
to the above description, MOSAs cannot be modelled as a significant restrictions of IEEE model. Therefore, to
nonlinear resistor. Therefore, different frequency- overcome the restrictions, various based on this structure
dependent models to simulate actual transient behavior of models have been suggested in recent years. In this paper,
MOSA has been proposed. In this paper, four prominent the three popular structures that are derived from the
MOSA models (IEEE, Pinceti, Fernandez, and P-K) are IEEE model are comprehensively analyzed.
analyzed in details. Table 1. V-I characteristic for A0 and A1.

A0 A1
2-1- IEEE model I (A)
V (pu) V (pu)
This model presented in [7] is shown in Fig. 1. In 10 0.857 0
IEEE model, A0 and A1 are two non-linear resistors that 100 0.936 0.769
are separated by an RL filter. The values of A0 and A1 are 1000 1.05 0.85
given in Table 1. In discharge currents with high rise 2000 1.088 0.894
time, the effect of RL filter can be neglected. In this case, 4000 1.125 0.925
the two parallel non-linear resistors impact on the static 6000 1.138 0.938
behavior of MOSA. In the case of short front-time, filter 8000 1.169 0.956
impedance is large and, thereby, L1 leads high currents to 10000 1.188 0.969
the A0 resistor branch. 12000 1.206 0.975
In the model, L0 represents the inductance associated 14000 1.231 0.988
with magnetic fields in the immediate vicinity of the 16000 1.25 0.994
arrester. The resistor R0 is also used to stabilize the 18000 1.281 1
numerical integration when the model is implemented in 2000 1.313 1.006
a digital computer program. The capacitance C0
represents the external capacitance that is proportional to 2-2- Pinceti model
the height of MOSA.
Fig. 2 shows the schematic of this model [8]. The
model has been derived from the IEEE model with minor
changes. The definition of non-linear resistors
characteristics (A0 and A1) is completely coincided with
previous subsection. The proposed criteria does not take
into consideration any physical characteristic of the
arrester and just electrical data are needed. The flowchart
of Fig. 3 shows the process of computing L0 and L1.
Fig. 1. IEEE model of MOSA [7]. Where Vn, Vr1/T2 and Vr8/20 are the rated voltage of

2
arrester, residual voltage at 10 kA fast front current surge can be obtained the parameters of mentioned models.
(l/T2 µs), and residual voltage at 10 kA current surge with Table 3 describes the value of parameters for each
a 8/20µs shape, respectively. The decrease time (T2) is arrester.
not explicitly written because different manufacturers
may use different values. R1 is considered to avoid
numerical troubles and its value is 1MΩ.

2-3- Fernandez model

The proposed model is the another simplified model
of IEEE [9]. As shown in Fig. 4, two constant resistors
(R0 and R1) and one inductance (L0) are eliminated from Fig. 2. Pinceti model of MOSA [8].
IEEE model. To determine the inductance (L1) and
capacitance (C1) of the model, equations (2) and (3) are manufacturer's
applicable. Where Vss (kV) is the residual voltage at data
500A and switching wave 60µs/200µs or 30µs /70µs.
Finally, the value of R is considered 1MΩ for numerical
oscillations in a digital computer program. In the same Yes No
1 / T2
way, the computation of capacitance in Fernandez model
is exactly identical to IEEE model.
2V V ss
L1  . r 8/ 20 .Vn ( H) (2) K = V1/T2 / V8/20
5 V r 8/ 20

100 No
C= (pF) (3) K < 1.18
d
Yes
2-4- P-K model
Fig. 5 shows the structure of the P-K model [10]. 1 Vr,1/T2 - Vr,8/20 L1 = 0.03× Vn
L1 = . .Vn
Similar Fernandez model, it is just defined L1 where is 4 Vr,8/20 L0 = 0.01× Vn
determined form equation (4). As well as, the value of
1 Vr,1/T2 - Vr,8/20
parallel resistance is assumed 1MΩ for the analytical and L0 = . .Vn
simulation studies. In order to simulate the dynamic 12 Vr,8/20
characteristic of mentioned model, It is performed by
Fig. 3. The process of computing L0 and L1 [8].
discharge currents with front times starting from 0.5μs to
8μs.
9 Vr1/ T2  Vr 8/ 20
L  . .Vn ( H) (4)
1 10 V r 8/ 20

3- Simulation results
Fig. 4. Fernandez model of MOSA [9].
In this section, transient behavior of mentioned
models is investigated using lightning (8/20 µs) and
switching (30/60 µs) surges. The double exponential
waveform is used in the simulation studies. In order to
evaluate the accuracy of the models, the obtained results
from the simulations are compared with the data provided
in the manufacturer's catalog. As well as, to validate the
obtained results and universalize the topic, three MOSAs
are evaluated at medium voltage level. Table 2 gives the Fig. 5. P-K model of MOSA [10].
MOSAs’ information extracted from the manufacturer's
catalog [14]. Referring to previous section and Table 2, it

3
 Fig. 8. Residual voltages for 0.25kA(30/60)µs wave (current curve
Fig. 6. Residual voltages for 5kA(8/20)µs wave (the scale of
scaled 410 times).
current curve is 26).

Fig. 9. Residual voltages for 0.5kA(30/60)µs wave (the scale of

Fig. 7. Residual voltages for 10kA(8/20)µs wave (the scale of
current curve is 205).
current curve is 13).

Table 2. Manufacturer's data [14]. V V

res( Sim) res( Man)
Err.  100 (5)
Vres(kV) V
res( Man)
Ur(kV) 30/60(µs) 8/20(µs) Where Vres(Sim) and Vres(Man) are the obtained residual
0.25kA 0.5kA 5kA 10kA voltages from simulation and manufacturer's data,
respectively.
10 19 19.7 23.3 24.6
Table 4 gives the maximum residual voltage and
30 56.7 59 69.6 73.7
error percentage of each model when the lightning wave
50 94.6 98.3 116 122.8
is applied to presented MOSAs in Table 2. As a result,
Fernandez model and the P-K model introduce the
By applying lightning wave with 5kA and 10kA maximum and minimum errors, respectively. In the case
amplitude and 8/20 µs characteristic to the different of applying 10kA lightning current to models, IEEE
models of MOSA, the residual voltage of 50kV MOSA model gives minimum error compared with other models.
brings the curves like Figs. 7 and 8 for different models. The reason for the reduction of IEEE error is related to
It can be observed that Pinceti model indicates less L1. Because this parameter is obtained from trial and
residual voltage in both cases; so that the maximum error process.
residual voltage for 5kA and 10kA lightning surges are As switching waves 0.25kA(60/30)μs and
equal to 114.571kA and 119.131kV, respectively. By 0.5kA(60/30)μs are applied to different models of
comparing these values with real data derived from the MOSA, residual voltage of 50kV MOSA experiences the
manufacturer's catalog, the significant error in pinceti curves shown in Figs. 8 and 9. In this case, all the models
model is observed. It can be used the following equation have the same peak value except IEEE model. The
to calculate the amount of the error:

4
behavior of all the models depends on the wave shape P-K and IEEE models also showed the less error in
and front-time of applied wave. Moreover, in lightning comparison of Pinceti and Fernandez models.
wave a recursive mode occurs in the Fernandez model,
whereas it do not happen in switching wave.
The error value of each model for switching wave is References
expressed in table 5. As a result, by comparing error [1] V. Hinrichsen, “Metal-oxide surge arrester,” Siemens AG, 2012.
values of lightning wave with switching ones, it can be [2] IEEE Std C62.22a, IEEE Guide for the Application of Metal-Oxide
found out that all of the models have the less errors for Surge Arresters for Alternating-Current Systems—Amendment 1:
Supplement to Consider Energy Handling Capabilities, vol. 2013, no.
lightning wave, which reflects the better performance of July. 2013.
the models for the lightning wave. The results also [3] J. A. Martinez and D. W. Durbak, “Parameter determination for
indicate that the models of MOSA depend on frequency. modeling systems transients-Part V: Surge arresters,” IEEE Trans.
Indeed, the proposed models have been optimized for Power Deliv., vol. 20, no. 3, pp. 2073–2078, 2005.
[4] K. Mardira and T. Saha, “A simplified lightning model for metal
lightning waves. oxide surge arrester,” Australas. Univ. Power …, pp. 4–9, 2002.
In related to the impact of rated voltage of MOSA on [5] A. Bayadi, N. Harid, K. Zehar, and S. Belkhiat, “Simulation of metal
the error value of the models, it can be observed the oxide surge arrester dynamic behavior under fast transients,” in The
significant effect. Except Fernandez model, in other international Conference on Power Systems Transients (IPST’03) in
New Orleans, USA, 2003.
models the error value of lightning wave approximately [6] I. Kim, T. Funabashi, H. Sasaki, T. Hagiwara, and M. Kobayashi,
decreases as the rated of MOSA increases (given in “Study of ZnO arrester model for steep front wave,” IEEE Trans. power
Tables 4 and 5). In switching wave case, the error value Deliv., vol. 11, no. 2, pp. 834–841, 1996.
increases when the arrester rated voltage increases. [7] R. A. Jones, P. R. Clifton, G. Grotz, M. Lat, F. Lembo, D. J.
Melvold, D. Nigol, J. P. Skivtas, A. Sweetana, and D. F. Goodwin,
In addition, the residual voltage of arrester, “MODELING OF METAL-OXIDE SURGE ARRESTERS,” IEEE
absorption energy related to each model is also dedicated Trans. Power Deliv., vol. 7, no. 1, pp. 302–309, 1992.
in Tables 4 and 5. In all of the models, absorbed energy is [8] P. Pinceti and M. Giannettoni, “A simplified model for zinc oxide
equal. Because the A0 resistance in all of the models is surge arresters,” IEEE Trans. Power Deliv., vol. 14, no. 2, pp. 393–398,
Apr. 1999.
equal. [9] F. Fernandez and R. Diaz, “Metal oxide surge arrester model for fast
transient simulations,” in The Int. Conf. on Power System Transients
IPAT, 2001, vol. 1.
4- Conclusion [10] P. Unahalekhaka, “Simplified Modeling of Metal Oxide Surge
Arresters,” Energy Procedia, vol. 56, pp. 92–101, 2014.
In this paper, four common metal oxide surge arrester [11] G. V. N. Bezerra, L. A. M. M. Nobrega, J. F. S. Junior, G. R. S.
models, IEEE, Pinceti, Fernandez, and P-K were Lira, V. S. Brito, E. G. Costa, and M. J. A. Maia, “Evaluation of surge
arrester models for overvoltage studies,” in 2014 ICHVE International
comprehensively analyzed and compared in different
Conference on High Voltage Engineering and Application, 2014, pp. 1–
scenarios. The components of mentioned models and 4.
their value were fully expressed. Then, the error of each [12] A. Meister, R. A. Shayani, and M. G. A. de Oliveira, “Comparison
model is calculated under lightning and switching waves of metal oxide surge arrester models in overvoltage studies,” Int. J. Eng.
Sci. Technol., vol. 3, no. 11, pp. 35–45, 2011.
for three-surge arrester selected from manufacturer’s
[13] M. Nafar, G. B. Gharehpetian, and T. Niknam, “A new parameter
catalog. As a result, the behavior of all the models estimation algorithm for metal oxide surge arrester,” Electr. Power
depends on the wave shape and front-time of applied Components Syst., vol. 39, no. 7, pp. 696–712, 2011.
wave. In general, all of the models have the less errors for [14] A. B. B. M. W. K. S. A. Datasheet, “Medium Voltage Products &
Systems. Catalog 2013.” 2013.
lightning wave, which reflects the better performance of
the models for this wave. In terms of the amount of error,

Table 3. Calculated parameters of different MOSA models

Un=10kV, d=0.187m, n=1 Un=30kV, d=0.347m, n=1 Un=50kV, d=0.507m, n=1

Parameter

IEEE

Fernandez
Pinceti

P-K
Fernandez

Fernandez
Pinceti

Pinceti
IEEE

IEEE
P-K

P-K

L0(µH) 0.0374 0.0788 - - 0.0694 0.227 - - 0.1014 0.376 - -

L1(µH) 2.805 0.233 0.796 0.841 5.205 0.6818 2.393 2.454 7.605 1.129 3.99 4.067
L*1(µH) 0.71 - - - 2.15 - - - 3.6 - - -
R0(Ω) 18.7 - - - 34.7 - - - 50.7 - - -
R1(Ω) 12.155 - - - 22.55 - - - 32.955 - - -
C(pF) 534.75 - 534.75 - 288.18 - 288.18 - 197.23 - 197.23 -

5
 Table 4. Simulation results for different MOSA models as lightning wave is applied to MOSA (residual voltage (kV) and absorbed energy (kJ))

Ur=10kV Ur=30kV Ur=50kV

Arrester
5kA(8/20)µs 10kA(8/20)µs 5kA(8/20)µs 10kA(8/20)µs 5kA(8/20)µs 10kA(8/20)µs
model
Vres E Vres E Vres E Vres E Vres E Vres E

IEEE 23.18 2.34 24.6 4.791 69.46 7.03 73.7 14.35 115.74 11.72 122.84 23.92
%Err. -0.5 - 0.02 - -0.2 - 0.01 - -0.21 - 0.037 -
Pinceti 22.95 2.34 23.87 4.78 68.76 7.03 71.50 14.34 114.58 11.72 119.13 23.9
%Err. -1.47 - -2.96 - -1.19 - -2.9 - -1.22 - -2.9 -
Fernandez 23.74 2.34 24.68 4.79 71.78 7.03 73.97 14.35 119.89 11.72 123.36 23.91
%Err. 1.89 - 0.35 - 3.13 - 0.37 - 3.35 - 0.46 -
P-K 23.23 2.34 24.78 4.79 69.55 7.03 74.10 14.35 115.85 11.72 123.43 23.91
%Err. -0.29 - 0.74 - -0.06 - 0.55 - -0.1 - 0.51 -

Table 5. Simulation results for different arrester models b as switching wave is applied to MOSA (residual voltage (kV) and absorbed energy (kJ))

Ur=10kV Ur=30kV Ur=50kV

Arrester
0.25kA(30/60)µs 0.5kA(30/60)µs 0.25kA(30/60)µs 0.5kA(30/60)µs 0.25kA(30/60)µs 0.5kA(30/60)µs
model
Vres E Vres E Vres E Vres E Vres E Vres E

IEEE 19.25 0.307 19.84 0.639 57.79 0.922 59.59 1.92 96.37 1.53 99.37 3.2
%Err. 1.32 - 0.75 - 1.75 - 1 - 1.88 - 1.09 -
Pinceti 19.69 314 20.3 0.654 58.99 0.942 60.82 1.96 98.29 1.57 101.34 3.26
%Err. 3.64 - 3.05 - 2.11 - 3.09 - 3.88 - 3.1 -
Fernandez 19.69 0.314 20.3 0.654 58.99 0.942 60.82 1.96 98.29 1.57 101.34 3.26
%Err. 3.64 - 3.03 - 2.11 - 3.09 - 3.88 - 3.1 -
P-K 19.69 0.314 20.3 0.654 58.99 0.942 60.82 1.96 98.29 1.57 101.34 2.26
%Err. 3.64 - 3.05 - 2.11 - 3.09 - 3.88 - 3.1 -