You are on page 1of 6

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO.

1, JANUARY 2002 123

An Improvement for the Selection of Surge Arresters


Based on the Evaluation of the Failure Probability
Luis Montañés, Miguel García-Gracia, Member, IEEE, Mariano Sanz, and Miguel Angel García

Abstract—An improvement on the typical selection procedure


of arresters is presented. The analysis shows that the risk of failure
of an arrester depends on several parameters (the striking point,
the lightning current waveform, the arrester itself and the system
configuration). Therefore, the selection of the optimal arrester de-
pends on how well its stresses can be estimated. After the appli-
cation of the typical selection procedure, the power system simu-
lations are carried out using the Electromagnetic Transients Pro-
gram (EMTP) for each suitable arrester. From these results the
failure probability of each arrester is calculated, which permits the Fig. 1. Lightning stroke with a triangular wave shape.
optimal selection from several valid arresters by the comparison of
their risk of failure. • the probability of the stroke hitting on the ground wire
(OGW) and producing the failure of the arrester.
Index Terms—Energy absorption capability, failure probability,
lightning, overhead ground wire, surge arresters. The failure probability has been obtained using EMTP, and
by applying the procedure developed in this paper. The resulting
failure probability gives rise to a further selection criteria for the
I. INTRODUCTION arrester. After applying the typical selection procedure, it per-
mits the discrimination between several valid arresters from the
S URGE arresters are vital elements in the protection of
equipment in substations against different types of over-
voltages [1].
comparison of their risk of failure. The procedure has been ap-
plied to a power substation as an example, however its applica-
As the protective ratio increases, the energy demands on the tion could be extended to power line protection.
arrester also increase. The selection procedure of an arrester for
a specific application is basically a compromise between protec- II. LIGHTNING CURRENT PARAMETERS
tive ratio and energy capability. The selection of an arrester with There are a wide variety of stroke current waves, and each
an adequate energy capability is not obvious due to the highly lightning flash may contain more than one stroke. A simplifica-
nonlinear voltage stress and current characteristics ( – ) of the tion in the calculation is achieved if we consider the first one as
arresters. the most severe [3].
In this paper, a method to evaluate the risk of failure of a surge The waveform of a lightning stroke has been approximated by
arrester is proposed. Frequently, it is accepted that an arrester a triangular wave (Fig. 1), which is usually represented by three
with a higher energy capability reduces the risk of failure [2], parameters: peak value of stroke current ( ), rise time ( from
however, the evaluation of the failure probability is more com- 0 to ), and time-to-half value ( up to 50% of ).
plex. It depends on several parameters such as: The magnitude of the current waveform , and the rise and
• statistical data of the lightning current waveform; fall times ( and ) will vary with every lightning stroke. The
• the proximity of the stroke from the arrester (further parameters that describe the waveform of a lightning stroke are
distances result in less severe overvoltages due to studied from their experimental statistical distribution.
attenuation); The front steepness of a lightning stroke is much faster
• the type of arrester selected (its energy capability, its – than the tail time, so it has a negligible influence on the
characteristic, etc.); energy absorption in comparison with the peak current and
• the system configuration. the time-to-half value. Therefore, in order to obtain a more
The failure probability of surge arresters is evaluated as the simplified method, the rise time value is fixed at 2 s [4].
The cumulative frequency distribution of the peak current can
sum of:
be approximated by a fractional type distribution, a logarithmic
• the probability of the stroke hitting on the phase conductor normal distribution or an exponential type [5]. In this study, the
and producing the arrester failure; first two distributions have been considered without finding rel-
evant differences between both of them. For simplicity, only the
Manuscript received August 10, 1999. fractional distribution is presented, where the expression for the
L. Montañés is with ERZ, Grupo ENDESA, Zaragoza, Spain (e-mail: peak current is given by
LMontanes@erz.es).
M. García-Gracia, M. Sanz, and M. A. García are with the Departa-
mento de Ingeniería Eléctrica, Universidad de Zaragoza, Spain (e-mail: (1)
mggracia@posta.unizar.es).
Publisher Item Identifier S 0885-8977(02)00540-X.

0885–8977/02$17.00 © 2002 IEEE

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on September 26,2022 at 07:31:35 UTC from IEEE Xplore. Restrictions apply.
124 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 1, JANUARY 2002

and adopting the same distribution to express the cumulative


frequency distribution of the tail duration we have

(2)

where the constants and must be deduced from experi-


mental observations, as well as and .

III. FAILURE PROBABILITY OF AN ARRESTER


A. Probability of the Stroke Hitting on the Phase Conductor or
on the OGW
Fig. 2. Vulnerability of phase conductor.
The OGW intercepts lightning strokes that would otherwise
have terminated on the phase conductors. However, complete
where and are represented in Fig. 2, and can be evaluated
shielding is not always achieved, and so it is necessary to cal-
for each peak current ( ) solving (4)–(6).
culate the probability of strokes hitting the phase conductor and
those hitting the OGW.
B. The Probability of the Stroke Hitting on the OGW and
A phase conductor or OGW have an area in which it will
Producing the Failure of the Arrester
draw strikes on itself. This area is limited by a circle of radius
, which is defined as the critical distance, and centered on the Considering a fractional type distribution for the peak current
object. There are different expressions to calculate [6], and in given by (1) and the time-to-half given by (2), the proba-
this paper we have adopted the IEEE Working Group [3] bility distribution is obtained by the derivative of the cumulative
distribution.
(3) • Probability distribution of the peak current

The probability of the stroke hitting on the ground, the phase (10)
conductor or the OGW, depends on the value of the peak current,
and the location of the ground wires with respect to the phase
conductor.
According to [7] we construct the diagram of Fig. 2, and the • Probability distribution of the time-to-half :
following equations are obtained:
(11)
(4)
(5)
Then, the probability of the stroke hitting on the OGW and
(6)
producing damage to the arrester is given by
where is a multiplication factor to account for the different at-
tractiveness of the phase wire and the ground, and can be eval- (12)
uated by the expression [6]:
where is the minimum peak current value which implies the
(7) arrester failure, and is dependent on each time-to-half value .
The peak current that causes the energy dissipated by the
The intersection of these circles permits the shielding failure arrester to be greater than its energy capability is evaluated by
distance ( s) to be calculated, and the probability of the stroke simulation (EMTP) for each . Thus, the expression
hitting on the phase conductor or the OGW.
For a given peak current we can calculate the probability
of the strike on phase conductor or OGW in the form: (13)
• stroke to OGW:
represents the probability of a stroke hitting on the OGW and
(8) . Multiplying by the probability for
each time-to-half value, and adding all these terms, (12) is
obtained.
• stroke to line:
The flow chart corresponding to the calculation of the prob-
ability of the stroke hitting on the OGW and producing the ar-
(9) rester failure is shown in the Appendix.

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on September 26,2022 at 07:31:35 UTC from IEEE Xplore. Restrictions apply.
MONTAÑÉS et al.: AN IMPROVEMENT FOR THE SELECTION OF SURGE ARRESTERS 125

Fig. 3. The distribution of the energy absorption capability is unknown, and it


is approximated by a step funcion.

C. The Probability of the Stroke Hitting on the Phase


Conductor and Producing the Arrester Failure
In the same way that the probability of the arrester failure is
evaluated when the stroke hits on the OGW, the probability of
the stroke hitting on the phase conductor and producing damage
to arrester can be calculated:
Fig. 4. Single-circuit line of the substation of Alcañiz (Spain).
(14)

where is the minimum peak current value which implies


the arrester failure, considering that the stroke now hits on the
phase conductor.

D. The Whole Failure Probability


To summarize, the failure probability of an arrester is
evaluated as

(15)
Fig. 5. Configuration of the 132-kV line of “La Oportuna.”

E. Distribution of the Energy Absorption Capability


As this probabilistic function is not usually available, a step
As it is well known, the energy capability of the arrester is a function has been considered in the analysis (see Fig. 3). It
probabilistic function [8], [9]. However, in this study we have should be noted that the energy absorption capability is a value
considered that the probability of failure is 100% if the energy that will be higher than the energy discharge capability pub-
exceeds the published capability, and 0% otherwise, which is a lished by manufacturers and considered here. So, again, the ap-
limitation on the precision of the results presented. Therefore, proximation taken into account has a conservative character.
as it is indicated in other works [10], the energy capability test
should perhaps involve the whole surge arrester subjected to IV. CASE STUDY
different currents, and data published by manufacturers should
include the distribution of the energy absorption capability of A. System Power Configuration
the surge arresters. The proposed method has been applied to a real case
Considering that this distribution function were known, as (“Alcañiz” substation, Spain), which the single line diagram
is shown in Fig. 3, we would have for each value of energy is shown in Fig. 4. The voltage of the system is 132 kV with
higher than a value of probability between 0 and 1. There- 550-kV BIL insulation. The configuration of the single-circuit
fore, in the expressions (12), (14), and (15) and would line with one shield wire corresponding to the 132-kV line of
be the values of the peak current corresponding to . In “La Oportuna” is shown in Fig. 5.
order to complete these expressions, we need simply to intro- In the computer simulations, the lightning is assumed to strike
duce the failure probability associated to each value of the en- the first span outside the substation on 132-kV line. If the light-
ergy capability. ning stroke is captured by the shield wire, the voltage across

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on September 26,2022 at 07:31:35 UTC from IEEE Xplore. Restrictions apply.
126 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 1, JANUARY 2002

TABLE I TABLE II
CONSTANTS OF CUMULATIVE FREQUENCY DISTRIBUTION OF LIGHTNING SURGE ARRESTERS DATA
STROKE CURRENT WAVEFORMS

the insulator strings rises. When a backflashover takes place, a


surge propagates toward the substation. The severity of the over-
voltages at the substation depends on how far the strick is away
from the substation.
The amplitude and front of wave are reduced by the corona
effect. This effect limits dangerous overvoltages in the substa-
tion to those originating in the closest part of the line, which
we define as the influence zone. Furthermore, the lightning
impulse is normally reduced before entering the substation
by the lightning current paths to ground through shield wires
or by flashovers across the line insulators. Therefore, the risk
of failure of the surge arresters is greater when the striking
point is situated closer to the substation. In other words, larger
distances result in less severe overvoltages due to stronger
distortion and attenuation effects.
Lightning overvoltages in the phase conductor can be due to
direct strokes. On the other hand, direct strokes to the substation
have been ignored considering that the substation is perfectly
shielded.

B. EMTP Simulation Considerations


Fig. 6. Probability of failure for the study case, and when a well-shielding or
In the EMTP simulations, the total simulation time was no shielding is considered in the example.
chosen to be sufficiently large for a complete analysis, being set
to 0.7 ms, and considering a 0.1-ns time-step. Decreasing the energy demands on the arrester also increase. Correct applica-
time step, identical results have been observed, which validate tion requires a judicious compromise.
this election. According to the actual application practice, we found that
The tower is represented as a distributed parameter line several arresters were suitable. This problem is illustrated by
( ) terminated by a resistance representing the means of the following example.
tower footing impedance ( ). At the position in Fig. 4, from a database containing more
At the substation the power transformer is represented by an than six hundred arresters, we found that the arresters shown in
open circuit where the footing impedance of the substation is the Table II could be selected according to the actual application
1 . practice.
Ten spans have been considered. Each span has been modeled There are different acceptable arresters for this position, thus
by a multi-phase model, considering the distributed nature of the selection between them can be based on an engineering judg-
the line parameters. The surge impedance values considered are ment, or by a systematized method as is suggested in this paper.
700 for the OGW, and 420 for the phase conductors. Finally, The selection of the optimal arrester from the valid arresters
the insulators are modeled with their flashover characteristics. shown in the Table II is not obvious. Because of highly nonlinear
voltage-current ( – ) characteristics of the arrester, the energy
C. Statistical Values discharge capability provided by the manufacturer can never be
considered as an estimation of the risk of arrester failure.
Statistical parameters must be deduced from experimental
The failure probability of each surge arrester shown in the
observations [5], [11]. In this paper, the constants and
Fig. 6 is evaluated by the application of the procedure devel-
considered are and according to [12].
oped above. For example, from these results we found that the
Constants of cumulative frequency distribution of the lightning
arrester Q132 has an energy discharge capability 108 kJ (22%)
stroke current waveform and time-to-half value are shown in
higher than the arrester Q108, however its failure probability is
Table I [12].
only 0.4% higher. On the other hand, the protective ratio of the
arrester Q132 is 34% worse than the protective ratio of the ar-
D. Results rester Q108.
In a typical application of a surge arrester the highest pro- The optimal arrester, considering those from Table II, must
tective ratio is desirable. However, as this margin increases the exhibit a high protective ratio and a low probability of failure.

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on September 26,2022 at 07:31:35 UTC from IEEE Xplore. Restrictions apply.
MONTAÑÉS et al.: AN IMPROVEMENT FOR THE SELECTION OF SURGE ARRESTERS 127

Fig. 7. Representation of the protective margin (protective ratio minus 1 and


expressed as a percentage) and of the probability of nonfailure (1 minus the
failure probability).

Fig. 9. Flow chart.

Fig. 8. Influence of the statistical parameters (T and I ) in the probability


of failure. rester failure is obtained by application of the method proposed.
As is shown in the results, the probability of failure cannot be
From the results obtained (shown in the Fig. 7) the arrester Q108 associated to the energy discharge capability due to the highly
is selected as the best solution. Furthermore, a quantification of nonlinear characteristics of the arrester. According to the actual
the performance level for each arrester is obtained. application practice, several arresters are found to be suitable.
This method could be applied to investigate the energy stress On the basis of this study, a quantification of the performance
on line arresters. In this case, the major difference is to give a level for each of these arresters is obtained, which permits an
higher weighting to the probability of failure than to the protec- optimal selection.
tion ratio.
The influence of the statistical parameters ( and ) in
the evaluation of the failure risk is analyzed in Fig. 8 for several APPENDIX
arresters. As it can be observed in this figure, the highest influ-
ence is due to the value of the peak current . In Fig. 9, the flowchart corresponding to the calculation of the
On the other hand, the failure probability can be related to the probability of the stroke hitting on the OGW and producing the
average life of the arresters as follows: failure of the arrester is shown. This program has been devel-
oped in C++ language in which the calls to EMTP simulations
are controlled.
(16)
To determinate (the minimum peak current of the stroke
that causes the arrester failure when the stroke hits on the
where is the length of the influence zone, is
OGW for each ) a Newton–Raphson algorithm is used. The
the whole probability failure and is the number of
process starts with an initial value . Considering
flashes/100 km/year.
s and for each , the circuit is simulated with EMTP.
is varied using the Newton–Raphson algorithm, where the
V. CONCLUSION energy dissipated ( ) by the arrester is compared with its
Based on experimental statistical distribution of the parame- energy capability ( ). This process is used for each from
ters that describe the lightning current waveform, the risk of ar- (4 s) to (200 s), using a step of 1 s.

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on September 26,2022 at 07:31:35 UTC from IEEE Xplore. Restrictions apply.
128 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 1, JANUARY 2002

The method used in the case of the stroke hitting on the OGW Luis Montañés was born in Barcelona, Spain, on January 11, 1973. He re-
is similar to the case of a hit on the phase conductor, and so only ceived the B.S. degree in electrical engineering, in 1997, from the University
of Zaragoza, the M.S. degree in electrical engineering, in 1998, from the Uni-
the first one is presented in this Appendix. versity of Zaragoza.
He has been working as a research engineer in of the electrical section of
CIRCE (Research center in electrical energy), and he is presently an engineer
in ERZ-Grupo ENDESA. His main research interests are high voltage, power
REFERENCES system protections, lightning protections and electrical transients.

[1] A. Greenwood, Electrical Transients in Power Systems, 2nd ed. New


York: Wiley, 1991.
[2] ABB, “Selection guide for ABB HV surge arresters,” HV Switchgear Miguel García-Gracia (M’99) was born in Saint-Brieuc, France, on April 23,
SESWG/A—2300E, Asea Brown Boveri, 1990. 1963. He received the B.S. (1989), M.S. (1990), and Ph.D. (1996) degrees in
[3] IEEE Working Group on Lightning Performance of Transmission Lines, applied physics from the University of Zaragoza.
“A simplified method for estimating lightning performance of transmis- Since 1990, he has been teaching and researching in the Department of Elec-
sion lines,” IEEE Trans. Power App. Syst., vol. PAS-104, pp. 919–932, trical Engineering, University of Zaragoza, where he is an assistant professor.
Apr. 1985. His main research interests are in the field of electrical transients, power system
[4] K. Nakada et al., “Energy absortion of surge arresters on power distribu- protection, high voltage, lightning protections and dielectrics.
tion lines due to direct lightning strokes,” IEEE Trans. Power Delivery,
vol. 2, pp. 1779–1785, Oct. 1997.
[5] K. Miyake, T. Suzuki, and K. Shinjou, “Characteristics of winter light- Mariano Sanz was born in Santander, Spain on December 22, 1940. He re-
ning current on Japan Sea Coast,” IEEE Trans. Power Delivery, vol. 7, ceived the Dipl. Eng. in electrical engineering, in 1960, from the University of
pp. 1450–1457, July 1992. Zaragoza, the B.S. degree in electrical engineering, in 1970, from the University
[6] IEEE Working Group on Lightning Performance of Transmission Lines, of Barcelona, and the Ph.D. degree in electrical engineering, in 1990, from the
“IEEE working group report estimating lightning performance of trans- University of Zaragoza.
mission lines II—Updates to analytical models,” IEEE Trans. Power De- In 1970, he was director of AITEL S.A. In 1993, he became director of De-
livery, vol. 8, pp. 1254–1267, July 1993. partment of Automatic and Electronic of ITA. Since 1994, he is the director of
[7] J. G. Anderson, Lightning Performance of EHV-UHV Lines, 2nd the electrical section of CIRCE (Research center in electrical energy), and since
ed. Palo Alto, CA: Electrical Power Research Institute, 1982. 1996 he is the director of the Electrical Engineering Department, University of
[8] T. Imai, T. Udagawa, H. Ando, Y. Tanno, Y. Kayano, and M. Kan, “De- Zaragoza. He is also a member of the energy commission of the “Consejo de
velopment of high zinc oxide nonlinear resistors and their application la Energía del Gobierno de la Comunidad Autonoma de Aragón.” His research
to surge arresters,” IEEE Trans. Power Delivery, vol. 13, no. 4, pp. interests include high voltage engineering, and power systems.
1182–1187, Oct. 1998.
[9] K. G. Ringler, P. Kirkby, and C. C. Erven, “The energy absortion ca-
pability and time-to-failure of varistors used in sation-class metal-oxide
surge arresters,” IEEE Trans. Power Delivery, vol. 12, pp. 203–212, Jan. Miguel Angel García was born in Zaragoza, Spain on May 10, 1962. He re-
1997. ceived the B.S. (1989) and the M.S. (1992) degrees in electrical engineering
[10] M. Bartkowiak, M. G. Comber, and G. D. Mahan, Failures, “Failure from the University of Zaragoza. He is currently working toward the Ph.D. de-
modes and energy absorption capability of ZnO varistors,” IEEE Trans. gree from the University of Zarazoga.
Power Delivery, vol. 14, pp. 152–162, Jan. 1999. Since 1992, he has been teaching and researching in the Department of Elec-
[11] K. Berger, R. B. Anderson, and H. Kröninger, “Parameters of lightning trical Engineering, University of Zaragoza. Since 1994 up to 1997, he has been
flashes,” Electra, no. 41, July 1975. director of Quality Assessment of the Electrical Metrology Laboratory (Univer-
[12] J. L. Tora, Transporte de la Energía Eléctrica. Madrid, Spain: UPCO sity of Zaragoza). His main research interests are in the field of power systems,
(ICAI-ICADE), 1997. electrical transients lightning protections and power transformers modeling.

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on September 26,2022 at 07:31:35 UTC from IEEE Xplore. Restrictions apply.

You might also like