IEEE Transactions on Power Delivery, Vol. 14, No.

2, April 1999 393

A simplified model for zinc oxide surge arresters.

P. Pinceti M. Giannettoni
Electrical Engineering Dept. Electrical Engineering Dept.
University of Genoa University of Genoa
Genoa ITALY- Genoa - ITALY

Abstract: A model for metal-oxide surge arresters, derived f?om The goal of this paper is to present a model for MOA, and
that one recommended f?om the IEEE W.G. 3.4.11, is presented. to propose a simplified and/or better procedure for its
The main innovation introduced by the paper lays in the simplicity parameter identification. The structure of the proposed model
of the criteria proposed for the model’s parameter identification. is based on the well known frequency-dependent model
Such criteria allow to calculate the model parameters directly fiom
the standard data reported in the arrester data-sheets with a simple recommended by the IEEE W.G. 3.4.11 in [2], with minor
and straightforward procedure. The effectiveness of the model was modifications useful to make the computer solutions more
tested for several arresters of different manufacturer both for robust. This method represents with a good accuracy the
Medium Voltage and for High Voltage application. The discharge dynamic behavior of MOA. The procedure to identifj the
test results obtained by the manufacturers were compared with the model parameters matches the following requirements:
results of simulations performed with the Alternative Transient
Program (ATP). Effectiveness and simplicity of use make the I. all necessary data are reported on manufacturer’s
proposed model a useful tool for insulation coordination studies catalogues or data-sheets;
involving steep f?onttransients. 11. no iterative correction is necessary to identify the
parameters;
Key words: Metal-Oxide Arrester, Electromagnetic transients,
EM”, Modeling Lightning, Frequency-Dependent Model, 1II.model’s performances match the real device
Insulation Coordination, Overvoltages behavior for surge arresters of different model and
fiom various manufacturers.
1 . INTRODUCTION
The basic requirement of the job was to find out a tool to
Although metal-oxide arresters (MOA) were introduced on define a MOA model starting fiom fieely available data. The
the market several years ago, their modeling is still a problem proposed solution is based on extremely simple mathematical
[l, 2, 31. Several accurate models have been proposed to relationships between device’s performances and model’s
describe the arrester behavior for different kinds of stress. parameters.
The hard point is the identification of the model parameters, Effectiveness of the proposed tool was verified comparing
and the need of field tests or of trial-and error procedures to the data of standard discharge tests [4] with the results of the
determine acceptable values. As a matter of fact, due to these proposed model calculated with the Alternative Transient
difficulties only arrester manufacturers or specialized Program. In particular, calculated residual voltages were
laboratories have today the real possibility of performing compared with those measured by manufacturers and
overvoltage ceordination studies. published on catalogues.

2. THE IEEE MODEL

The model recommended by IEEE W.G. 3.4.11 [2] is

shown in Fig. 1. In this model the non-linear V-I
characteristic is obtained by means of two non-linear resistors
(tagged A0 and Al) separated by a R-L filter. For slow surges
the filter impedance is extremely low and & and AI are
practically connected in parallel.
On the contrary, during fast surges, the impedance of the
filter becomes significant, and causes a current distribution
between the two branches. For precision sake, the current
PE-039-PWRD-0-05-1998 A paper recommended and approved by
the IEEE Surge Protective Devices Committee of the IEEE Power
trough the branch & rises when the fiont duration decreases.
Engineering Society for publication in the IEEE Transactions on Since resistance is greater than AI resistance for any
Power Delivery. Manuscript submitted September 10, 1997. given current, the faster the current surge, the higher the

0885-8977/99/$10.00 0 1998 IEEE

 394

4. PARAMETERS DEFINITION

It will be shown that the proposed model can be easily

defined by adopting the two following rules:

- the definition of non-linear resistors characteristics (&

and A,) is based on the curves shown in Fig. 3. These
U) L1 curves derives fiom the curves proposed by IEEE W.G.
3.4.11, and are referred to the peak value of the residual
voltage measured during a discharge test with a 10 kA
lightning current impulse (V~,ZO);

1.3
1.2
Fig. 1 IEEE Frequency-DependentModel
-1.1
The comparison of the calculated peak values with the % 1
measured values shows that the fiequency dependent model
gives accurate results for discharge currents with times to
-20.9
0

crest between about 0.5 ps and 45 ps. p 0.8

The main problem of this model is how to identify its 0.7
parameters. The W.G. 3.4.11 suggests an iterative procedure 0.6
where corrections on different elements are necessary until a 1E-05 0.0001 0.001 0.01 0.1 1 10 100
satisfactory behavior is obtained. The starting values can be Current FA]
obtained through formulas that take into account both the
electrical data (residual voltages), and the physical Fig. 3 Static characteristics of the non-linear elements. The
parameters (overall height, block diameter, columns voltage is in p.u. referred to the Vr8/20.
number). - to define the inductances, the following equations can be
used (values are in pH):

3. THE PROPOSED MODEL

The model here presented derives fi-om the standard model

[2], with some minor differences.
yr8/20
LO L1 where:
V,, is the arrester rated voltage
=
Vrlm residual voltage at 10 kA hst fi-ont current
=
surge (l/Tz ps). The decrease time is not
explicitly written because different
manufacturers may use different values. This
fact does not cause any trouble, since the peak
value of the residual voltage appears on the
rising fi-ont of the impulse
Vs,zo = residual voltage at 10 kA current surge with a
8/20 ps shape.
Fig. 2 The proposed model The proposed criteria does not take into consideration any
By comparing the models in Fig.1 and in Fig.2, it physical characteristic of the arrester. Only electrical data are
can be noted that: needed. This approach differs fiom other ones presented in
0 the capacitance is eliminated, since its effects on model
[ 1,2,5], where other data are needed.
behavior is negligible, The equations (1) and (2) are based on the fact that
parameters Lo and L, are related to the roles that these
0 the two resistances in parallel with the inductances are
elements have in the model. In other words, since the
replaced by one resistance R (about 1 Mn) between the input
function of the inductive elements is to characterize the
terminals, with the only scope to avoid numerical troubles.
model behavior with respect to fast surges, it seemed logical
The operating principle is quite similar to that of the IEEE
to define these elements by means of data related to arrester
fiequency-dependent model.
behavior during fast surges.
 395

5. MODEL PERFORMANCES The model was set with the successive iterations procedure
for the ABB MWA 16 arrester, then a proportional
With the above model relative errors between calculated relationship with device's height was introduced to calculate
and measured residual voltages are lower than 4.5% for the parameters of the arresters of different voltage.
discharge currents with time to crest ranging from 1 to 30 It's easy to observe that these criteria always need a
microseconds. The error decreases when the considered times correction procedure to keep errors within acceptable limits.
to crest are not greater than 8 ps.
These results are obtained considering arresters from Traditional Criteria
different manufacturers, with different nominal discharge
currents and rated voltage levels.
Although such performances are similar to that considered
satisfactory in [2], it's important to observe that they have
been obtained at the first trial. No iteration is required,
differently fiom what is described in [2].
When necessary the accuracy of the model can be enhanced
by reconsidering A0 and A1 profile, because it slightly
changes from one manufacturer to another.

6 . MODEL VALIDATION
Fig. 5 Relative errors on residual voltages with
Fig. 4 shows the circuit implemented on ATP to simulate inductive elements defined in relation to physical
dimensions of the device.
the behavior of the arrester with different current impulses.
Simulations considered discharge tests with fast surge current Differently, Fig. 6 shows the results adopting the proposed
impulses (1/5 ps or 1/2 IS),lightning current impulses (U20 rules. A considerable homogeneity can be observed as a proof
ps) and switching current impulses with amplitudes between of their efficacy.
1kAand20kA.
Proposed Criteria
_I_____-- . -

MWAType

Fig. 6 Relative errors on residual voltages with

Fig. 4 Simulated test circuit. mductive elements defined 111 relation to measured
perfor"C43.
To put into evidence the efficiency of the proposed model,
the relative errors between calculated residual voltages and The results obtained by considering arresters fiom different
those reported on manufacturer's catalogue are shown. manufacturers give a firther confirmation of the model
The important point is the criteria for choosing of the efficiency (see Fig. 7). Technical data of such arresters are
values of inductive parameters while the definition of non- reported in Table 1.
linear resistors characteristics is done according to the IEEE Table 1 Technical data of the considered arresters
W.G. 3.4.1 1 indications.
Rsidual wltage with wavehm..
The comparison is made on a set of arresters manufactured
at a discharge cunmt d..
by ABB; the considered arresters have Maximum Continuos
8/20us
Operating Voltage (MCOV) between 4 kV and 37 kV. The h t e r vdtage lOkA 5kA lOkA 20kA
results obtained with switching current impulses are not kV kV kV kV kV
reported because they are not relevant for the dynamic A-GECAlslhan 20 55 48 51 57
behavior of metal oxide surge arresters. Fig. 5 shows the B-SIEMENS I11 282 250 266 298
c-SIEMENS 111 300 266 283 317
performances of the model when traditional criteria are used
to define the inductive elements.
D-SIEMENS I 120 I 286 I 254 I 270 I 300
396

Appendix
Example of Parameter Selection
for a Metal Oxide Arrester

-4
-6 1.- ” ... . .. .. ” . .~...~.__.,_~.I..”... ., ../
Arrester

Fig. 7 Relative errors on residual voltages obtained from

the generalized model at the first attempt for some Reted Max Cont. Residual voltage with waveform...
arresters manufactured from Siemens and GEC vdtage Operating at a discharge current of
Alsthom. Voltage

7. CONCLUSIONS
115 us

1okA
(fast front)

20kA
I 8/20 us (lightning surge)

5kA 1okA 2okA

[kV] [kV] Up[kV] Up[kVJ UpFV] up FV] up FV]
To allow easy but nevertheless accurate overvoltage studies, 20.0 16 50.7 57.4 44.2 46.2 51.0
a simple model for surge arresters is presented. The model is Fig. AI: Arrester data (ABB MWA 16).
based on the well know IEEE one, with minor modifications,
but a new method of identifying its parameters is described. A1 ARRESTER INFORMATIONREQUIRED
The model can be completely identified starting f?om the
standard manufacturer’s data-sheets that report the residual vn - arrester rated voltage
voltages for different current pulses.
The paper shows the complete modeling procedure as well V,,O - residual voltage measured for a 10 kA lightning
as some tests performed on arresters of different current impulse
manufacturers to check its accuracy.
VI,, - residual voltage at 10 kA fast front current surge
REFERENCES
From Fig. A1 we obtain:
[I] Hileman, J. Roguin, K.H. Weck - Protection performance of metal oxide
surge arresters - Electra No. 133, pp.132 - 143, December 1990. v, = 20kV
[2] IEEE W.G. 3.4.1 1 of Surge Protective Devices Committee - Modeling V,,o 46.2 kV
of metal oxide surge arresters - IEEE Trans. on Power Delivery, Vol. 7, Vrlm= 50.7 kV
-
NO. 1, pp. 301 309, January 1992
[3] Ikmo Kim, Toshihisa Funabashi, ... - Study of ZnO arrester model for
steep h n t wave - IEEE Trans. on Power Delivery, Vol. 1 1 , No. 2, pp. A2 DETERMINING THE MODEL PARAMETERS
834 - 84 1, April 1996
[4] IEC 99-4 , “Metal oxide surge arresters without gaps for A.C. system”,
February 9 1
A2.1 NONLINEAR RESISTORS
-
[5] Durbak - Zinc-oxide arrester model for fast surges EMTP Newsletter,
Vol. 5, No. 1, January 1985 Nonlinear resistors A0 and A1 are modeled with equations
[6] ATP Rule Book - V.E. - Exponential ZnO surge arrester R(i)
[7] ATP Rule Book - XIX.1. - ZNO FITTER to punch Type 92 ZnO branch
of the form I = B V in the ATP, using the Type 92 branch
cards cards. Correct values of parameters B and q are calculated
providing some points of the V-I characteristic to a dedicated
subroutine [6,7]. The points of the V-I characteristics
defining the curves shown in Fig. 3 in the paper, are listed in
Table A2.1. Voltage values are in p.u. referred to Vfi,20.

Table A2.1: V-I characteristic for A0 and A I

I N ] I A0 Ip.u.1 I A1 Ip.u.1
2.106 I 0.810 I 0.623
0.974 0.788
1.052 0.866
1.108 0.922
1.195 1.009
20 1.277 1.09 1
 397
Scaling voltage values by the factor,,
V of the arrester to ones is below 2%. In addition the figures show the dynamic
be modeled, A0 and A1 characteristics are defined by the response of the model for discharge currents of different
points listed in Table A2.2. shapes.
Table A2.2: V-Icharacteristic used for A0 and 1.2 7 , I I
AI in the arrester model
1.1

IpG4] I AO@cv] I A1 kv] -3 1

2 10" I 37.4 I 28.8
*
4 0.9

0.8
51.2 42.6 0.7
55.2 46.6
20 50.4 0.6
0 5 10
WAI
A2.2 INDUCTIVE ELEMENTS Fig. A3.1. Dynamic behavior of the MWA 16 model with 10
kA current pulses.
Using the equations (1) and (2)presented in the paper, the
values of inductive elements are straightforward determined
as follows:

r, =4-
1 Vr1, -'r~,20
Vr,,
.v =-.41 50.7-46.2
46.2
.20 = 0.487fl

Y ,OII

A2.3 CONVERGENCE RESISTOR 0 10 20 30 40 50 60

Time bl
As previously stated in the paper, a 1Mn value for the Fig. A3.2. Residual voltage of the MWA 16 model for a
resistor R is recommended to avoid numerical troubles. 20 kA 8/20 ps discharge current.

A3 MODEL TESTING

In this paragraph, the results of discharge tests simulated

with the ATP using the above defined model are presented.
The results in terms of peak values of residual voltages and
errors related to the arrester performances are shown in Table
A3.
Table A3: Calculated residual voltages for the ABB MWA 16 arrester.

waveform 1/5 ms waveform 8/20 ms

0 2 4 6 8 10 12 14 16
Time C1
Fig. A3.3. Residual voltage of the MWA 16 model for a
20 kA 115 ps discharge current.

Fig. A3.1 - Fig. A3.3 show the results of discharge test

simulations in the two standard planes:
- currentlvoltagevs. time,
- current vs. voltage.
The model fits with a high accuracy the arrester
performances reported in the data sheet. The difference
between the calculated residual voltages and the declared
398

Authors’ biography.

Paolo Pinceti has got a Ph.D. degree at the Genoa University in 1987 in electrical
engineering. He now teaches “Industrial Power System” at the same University.
His research activities are in the area of power system protection, control and
management. He is member of IEEE and ISA.

Marco Giannettoni has got a degree in electrical engineering at the Genoa

University in 1996, where he’s now a Ph.D, student. His research activities are in
the area of communication systems (fieldbus) for industrial automation.