Evaluation of Mathematical Models of Metal Oxide Surge

EVALUATION OF MATHEMATICAL MODELS OF METAL OXIDE SURGE ARRESTER FOR ENERGY ABSORPTION STUDY
A. Goudarzi ( I ) and H. Mohseni ( I )
(1) University of Tehran ,Iran
ABSTRACT
Metal oxide surge arresters (MOSA) are widely used to protect insulation of a power system or part of i t against excessive stress due to over voltages. Metal oxide elements are capable to dissipating energy of transient over voltage in term of heat. MOSAs have very nonlinear V-I characteristic that include a dynamic behavior, so their modeling is still a problem. There are several models for surge arresters. Appropriate modeling of their dynamic characteristics is very important for insulation coordination, arrester location and energy absorption studies. The objective of this paper is comparison study between several models of surge arrester applicable for energy absorption study. Keywords: Metal-oxide surge Arrester, IEEE Model, Insulation coordination, energy absorption.
INTRODUCTION
Metal-oxide surge arresters (MOSA) are extensively used in power systems due to their good performance in overvoltages protection. The metal-oxide arrester implements a highly nonlinear resistor used to protect the insulation of equipments in electrical systems against internal and external over voltages. Surge arresters have an extremely high resistance during normal system operation and a relatively IOW resistance during transient over voltages. That is, it has non-linear voltage versus current (V-I) characteristic. Experimental data have shown the voltage across MOSA increases as time to crest of the arresters current decreases and that the MOSA voltage reaches a peak before the arresters current reaches its peak. Therefore, the MOSA cannot only be modeled by a nonlinear resistance.[ 1],[2] Typically, the predicted residual voltage for an impulse current with time-to-crest of 1 ps is 8-12% higher than that due to a time-to-crest of 8-ps lightning current wave. For longer time-to-crests of between 45 and 60ps, the voltage is 2 4 % lower than that for the 8 ps current impulse [ 5 ] . Indeed, the voltage across the arrester is not only a function of the discharge current, but also o f the rate of its rise. Surge arrester modeling is still a problem. [ 1,2,3,4]Several accurate models have been proposed to describe the MOSA behavior for different kind of electrical stress. Appropriate modeling of their dynamic characteristics is very important for insulation coordination, arrester location and energy absorption studies. In addition Metal oxide elements dissipate energy of transient over voltage in term o f heat. Electrical field stress and heat loss are two major causes of arrester aging and outage. For calculating of heat energy absorption we need a proper model of MOSA. The purpose of this paper is to review and comparative study between several models of arrester, and using them for energy absorption study. For this reason we have selected a distribution system with a distribution class polymeric surge arrester with Uc=9.0kV which protects insulation of a transformer. Data of this surge arrester are shown in table 1. We apply various models to adopted arrester and compare the results of heat loss calculation and levels of protection voltage of arresters.
NON-LINEAR RESISTOR MODEL IN PSCADKMTDC
In EMTDC-a transient state analysis program-, a large resistive branch is used to model a metal-oxide arrester. The resistance of the branch is changed in a piecewise !inear fashion to represent the slope of the arrester characteristic. Also, the series voltage source is modified when needed. All switching is done with interpolation. In Figure 1, the series voltage in the branch EDC is changed to E l , E2, E3, etc. as the operation portion of the characteristic changes from conductance values G1, G2, G3, etc. This is accomplished in EMTDC using subroutine VZNINT. It is only necessary to input arrester voltage rating and number of parallel stacks for modeling. That is possible using default data or entering I-V characteristic .[6j
21 1
LO
L1
Figure 4 : Pinceti proposed model

I,-i
1,
Figure1:Piece-wise linear approximation of metal oxide arrester characteristics in PSCADKMTDC
MODEL PROPOSED BY PINCETI ET AL

The model presented by the authors denves from the IEEE recommended model of the previous section, with some minor differences [3J. This model is shown in Figure. 4 1.The capacitance is eliminated due to its little effect on the model behavior 2. The two resistances in parallel with the inductances are replaced by one resistance R (about Im)between the input terminals; this resistance has the only scope to avoid numerical troubles. The operating principle is quite similar to that of the IEEE recommended model. Adjustment procedure of parameters is described in [3].
IEEE FREQUENCY DEPENDENT MODEL

The proposed model by the working group 3.4.11 of IEEE is referred to as the frequency-dependent model. For this model the non-linear V-I characteristic of an arrester is represented with two section of nonlinear resistance designated A, and A , as shown in Fig2 The two section are separated by a R-L filter. For slow front surges, this R-L filter has very little impedance and the two non-linear section of the model are essentially in parallel. For fast-front surges the impedance of the R-L filter becomes more significant. This results in more current in the non-linear section designated A, than in the section designated Al. Since characteristic A, has a higher voltage for a given current than A, (see Fig.31, the result is that the arrester model generates a higher voltage. Since metal-oxide arresters have a higher discharge voltage for fast- front surges, the model matches the overall behavior of a metal-oxide arrester. Adjustment procedure of parameters is described in [2].
L O
0
MODEL PROPOSED BY FEFWANDEZ ET AL
L1
The proposed model is shown in Figure 5 and derives 4 ] .It is intended for the simulation of the from that in [ dynamic characteristics for discharge currents with front times starting from 8ps. Between the non-linear resistances A0 and A I only the inductance L1 is taken into account. RO and LO are neglected. CO represents the terminal-to-terminal capacitance of the arrester. The resistance R in parallel to A0 is intended to avoid numerical oscillations. The model in figure 5 works essentially in the same way as that proposed in [2] but this model and previous are simpler rather than IEEE model because IEEE model need some more data of arrester such as height of arrester and implement a iterative correction to identify the parameters.
Figure 2: LEEE Frequency-dependent model
4 ] . Adjustment procedure of parameters is described in [
Figure 5 : Fernsodez proposed model

Current (Amper)
Figure 2: Non-linear characteristic for &and Al. The voltage is in p.u. referred to the Urm0
212
I
U,
Table I: Arrester characteristics.

Raychem HDA-09NA
1
1
modeled simply as a network of resistances representing the surge impedance. Surge Impedances of Transmission Line is 269. I ohm: Transformer faced with current impulse can be modeled only by stray capacitances of the transformer; so two capacitors are model of transfonner. Model of arrester is putting in steep front surge arrester subpage. The LV connection and the neutral are combined since the LV service wire and the neutral wire are twisted together and are strongly coupled. Transmission lines represent the service cable and ground cable models with 100-meter length. Results of simulation are shown in table 2. Dissipated heat and protection voltage are two interested parameters that listed in the table. Dissipated heat is time base integration of product of voltage and current of arrester, and protection voltage is maximum voltage occurring on both side of arrester during surge current. In this paper we suppose IEEE model as base and determining the percentage of errors based it, since IEEE model apply more data of arrester and use more points for V-I characteristic. As seen from table 2, all of frequency dependent model have acceptable result but conventional model or nonlinear resistance has more result out of range. There is good agreement for protection voltage with maximum error about 3.1% for three models and about 10% for heat loss calculation.
uMcOvVWZ
V~VIOO D
Number of
(1 O k A ) (500A) (Height) column
I l l k V 1 9.0kV I30.0kV 123.lkV I0.212m I

MODELING AND SIMULATION
Our case study includes a network with distribution

transformer that is protected with a surge arrester against lightning current impulse. Figure 6 shows the desired distribution network, which simulated in PSCAD. The characteristic of arrester is shown in table 1. A lightning current impulse with 8/20ys waveform and 20kA amplitude is applied to the network. The current impulse (8/20ps) generator is modeled according with this equation :
I ( ( ) = A . I, . 1 3 .e-('") where A=0.01244 T=3.911 ps Ip=desired peak current magnitude.

Since the time period over which we are interested is less than the travel time of the line, the line can be
K
fi
afthe transfmnw
T
It
Etank
A OD1243 T 3911 us
.........
...........................................................
. . . . I -
Figure 6: Simulated circuit for lightning current impulse
21 3
Table 2
Arrester model
lEEE Pinceti
Energy
Absorption(kj) 9.54 10.453 10.5
Error of energy
Absorption(%)
Protection
Voltage(kV)
Error o f protection Voltage(%)

0
0
0 9.6
Fernandez Non-linear
10
-20.86
7.55
32.3 32.3 33.3 22.62
3,1 -30
CONCLUSION:
In this paper three frequency dependent model and nonlinear resistance model applied to a distribution class arrester. Accuracy validity of such models for energy absorption calculation were interested. Frequency dependent models have good agreement for protection voltage because these models have developed mainly for proper residual voltage calculation; therefore for energy absorption there is a little increscent error. Of course Pinceti and Femandez models for simplifying use less data and V-I point for low current through the arrester. Non-linear model has shown that doesnt have good accuracy for steep front surges like lightning current impulse, because it doesnt include the dynamic effects of voltage-cutrent characteristic of arrester.
Address of the Author The first author can be contacted at Department of Electrical Engineering of University of Tehran North Karegar Tehran Iran a.goudaai@ece.ut.ac.ir
REFERENCES
1. W.Schmidt, J.Meppelink, B.Richter, K.Feser, L.Kehl, D.Qui, Behavior of MO-Surge-Arrester Blocks to Fast Transient ,IEEE Transaction on Power Delivery,Vol.4, No. 1, pp.292-3OO,January, 1989 2. IEEE WG 3.4.1 I, Modeling of metal oxide surge arresters,IEEE Trans. On Power Delivery, vol. 7, pp.
302-309, Jan. 1992

3. P. Pinceti, M. Giannettoni, A simplified model for zinc oxide surge arresters , IEEE Trans. on Power Delivery, Vol. 14, No. 2, pp.393-398, 1999. 4. Femandez F., D i u R., Metal oxide surge arrester model or fast transient simulations paper 144,Intemational conference on power system transients, IPSTOl, 20 -24 June 2001. 5. IEEE Guide for the application of Metal-oxide surge Arrester for Altemating Current Systems,IEEE Standard C62.22,1997 6. Manitoba HVDC Research Center Inc. PSCADEMTDC Simulation M a n u a l .
214

Evaluation of Mathematical Models of Metal Oxide Surge

Uploaded by

Copyright:

Available Formats

Evaluation of Mathematical Models of Metal Oxide Surge

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Evaluation of Mathematical Models of Metal Oxide Surge

Uploaded by

Copyright:

Available Formats

EVALUATION OF MATHEMATICAL MODELS OF METAL OXIDE SURGE ARRESTER FOR ENERGY ABSORPTION STUDY

A. Goudarzi ( I ) and H. Mohseni ( I )

(1) University of Tehran ,Iran

Figure 4 : Pinceti proposed model

Figure1:Piece-wise linear approximation of metal oxide arrester characteristics in PSCADKMTDC

MODEL PROPOSED BY PINCETI ET AL

IEEE FREQUENCY DEPENDENT MODEL

MODEL PROPOSED BY FEFWANDEZ ET AL

Figure 2: LEEE Frequency-dependent model

4 ] . Adjustment procedure of parameters is described in [

Figure 5 : Fernsodez proposed model

Table I: Arrester characteristics.

(1 O k A ) (500A) (Height) column

I l l k V 1 9.0kV I30.0kV 123.lkV I0.212m I

Our case study includes a network with distribution

I ( ( ) = A . I, . 1 3 .e-('") where A=0.01244 T=3.911 ps Ip=desired peak current magnitude.

Figure 6: Simulated circuit for lightning current impulse

Error o f protection Voltage(%)

32.3 32.3 33.3 22.62

302-309, Jan. 1992

You might also like