Some aspects of system modelling for the

estimation of lightning performance of

high voltage substations
J.P. Bickford, B.Sc, M.Sc, Ph.D.. C.Eng., M.I.E.E., and M.S. Savic. B.S..
M.S., Ph.D.
Indexing terms: Lightning and lightning protection, Modelling

Abstract: The computation of lightning overvoltages in a gas-insulated substation is carried out for different
overhead line equivalent circuits. Models of the back-flashover characteristics of the line insulation in front of
the substation are analysed, and an investigation is made into the influence of the number of overhead line
spans represented in the equivalent circuit. The results show that the inclusion of three overhead line spans in
the equivalent circuit is sufficient for the estimation of the substation lightning performance. The influence of
the instantaneous value of the power frequency voltage at the moment of lightning strike is also analysed.
Direct strikes to a phase conductor and to the tower top with back flashover are simulated. For relatively small
values of lightning current, the phase voltage at the moment of strike has a significant influence.

1 Introduction is determined. A curve of limiting current magnitude as a

function of slope of the current wavefront may then be
Interest in the lightning performance of power substations constructed. In the zone above this curve all lightning dis-
has increased in the last decade due to the application of charges will cause insulation failure.
recently developed gapless metal-oxide surge arresters on
the one hand, and to the use of gas-insulated high-voltage This method does not require a large number of transient
substations on the other hand. The literature contains a calculations. On the other hand, the only stochastic par-
number of papers describing investigations in this field, ameters incorporated into the model are the lightning
and three approaches have been adopted: current amplitude and its rate of rise. All other parameters
(a) In the classical approach the deterministic lightning are taken as deterministic.
parameters (current magnitude and rate of rise) are taken In this paper an effort is made to investigate the
as input data, and an effort is made to obtain the influence minimum equivalent circuit required to compute lightning
of different substation and lightning parameters on the overvoltages with sufficient accuracy. The influence of the
maximum lightning overvoltages. The influence of the gas- instantaneous value of power frequency voltage on the
insulated substation length on overvoltage magnitude has maximum lightning overvoltages is investigated to deter-
been investigated [1-4], and the effect of surge arrester mine its importance. Frequently this random parameter is
rating has also been considered [1, 2]. The transient neglected entirely, or represented by assuming extreme
response of a large substation to a direct lightning stroke values of equal and opposite polarity to the surge voltage.
to a phase conductor in front of the substation and to a The computation is performed by an electromagnetic
back flashover using various current waveforms has been transient program based on travelling-wave techniques
described [5]. [10], which is outlined in Appendix 7.1. The method by
(b) In the stochastic approach by Monte Carlo simula- which the effect of corona is represented is given in Appen-
tion [6-8], random parameters such as lightning discharge dix 7.2.
characteristics, location of strike, flashover characteristics
of the line insulation, substation insulation breakdown 2 Choice of equivalent circuit
characteristics and the instantaneous power frequency
voltage at the moment of discharge can be incorporated 2.1 Description of equivalent circuit
into the computation. Using random numbers, stochastic In the estimation of lightning performances of overhead
lightning and system parameters are used in a large transmission lines and substations a number of different
number of simulations of the transient process to compute equivalent circuits can be used. In the study of back-
the risk of insulation failure. The disadvantage of this flashover overvoltages in a substation, two different tower
approach is the large number of calculations that are models are investigated. A simple tower model with back
required to obtain sufficient computational accuracy. flashover simulated by direct connection of the earth wire
(c) The method of constructing a curve of lightning- and phase conductor at the point of the stroke has been
discharge limiting parameters which produce insulation proposed [2], and is shown in Fig. 1.
failure is another probabilistic approach [8, 9] to the esti- A more accurate model may be obtained by simulation
mation of the basic lightning impulse insulation level. In
this method transient calculations are performed using a
lightning current wave with a linear front whose slope is
varied. For each value of the slope of the wavefront, the
current magnitude necessary to produce insulation failure

Paper 3285C (P8, P9), first received 21st October 1983 and in revised form 10th
April 1984
Dr. Bickford is, and Dr. Savic was formerly with the Department of Electrical
Engineering and Electronics, University of Manchester Institute of Science and
Technology, PO Box 88, Manchester M60 1QD, England. Dr. Savic was a Lever-
hulme Postdoctoral Visiting Fellow and has now returned to the University of Bel-
grade, Yugoslavia Fig. 1 Simple tower model
204 IEE PROCEEDINGS, Vol. 131, Pt. C, No. 5, SEPTEMBER 1984
of the line insulation fiashover characteristics. When the Height of towers:
lightning overvoltage is increased above the insulation substation entrance tower hl = 25 m
withstand voltage, fiashover occurs. Such a model is pre- other towers h2 = 50 m
sented in a schematic way in Fig. 2. In both models the The lightning surge is modelled by a triangular current
wave, shown in Fig. 4. The steepness of the front of the
wave is taken to be 20 kA//is in all computational exam-
ples. The entrance tower footing resistance is usually much
smaller than the footing resistances of the other towers.
For this reason it is varied in the range from 1 fi to 6 Q.
The substation is protected by a metal-oxide surge arrester
with a rated voltage of 276 kV. The residual voltage char-
acteristic is shown in Fig. 5.

Fig. 2 Tower model with exact fiashover characteristic 600

.400
towers are represented by short lines with distributed par-
ameters. 200
The influence of number of towers in the equivalent 0
circuit of the transmission line for the case of a lightning 10 15
stroke to the top of the second tower in front of the sub- kA
station is investigated further. In Fig. 3, different equiva-
lent circuits which may be used for the calculation of Fig. 5 ZnO arrestor residual voltage characteristic

2.2 Results of computation

Using the simplified model shown in Fig. 1, the maximum
substation overvoltages as a function of number of towers
represented in the equivalent circuit are presented in
Tables 1, 2 and 3 for first tower footing resistances of 2 ft,
3 ft and 6ft, respectively, n is the number of overhead line
spans in the equivalent circuits of Fig. 3.
Table 1 : Maximum overvoltage expressed as a function of
the number of spans in the equivalent circuit
/, kA
n 30 60 90 120 150 180 210
1 578.8 636.7 658.5 744.1 844.0 947.3 1066
Fig. 3 Various equivalent circuits of the overhead line in front of the 2 578.8 631.9 641.4 683.4 805.9 902.6 1007
substation 3 578.8 631.9 638.3 690.6 804.7 898.0 1000
4 578.8 631.9 638.3 690.6 804.7 898.0 1000

lightning overvoltages in gas-insulated substations are pre- Entrance tower footing resistance = 1 Q.
Overvoltages in kV.
sented. The 345 kV system data taken from Tominaga et
al. [2] are given below: Table 2: Maximum overvoltage as a function of number of
spans in the equivalent circuit
BIL of the GIS is 1050 kV
/, kA
Earth wire surge impedance Zg = 350 ft n 30 60 90 120 150 180 210
Phase conductor surge impedance Zp = 350 ft
Tower surge impedance Zf = 100 ft 1 579.2 637.0 660.7 751.0 849.7 954.3 1081
2 579.2 632.1 643.6 690.3 810.1 908.9 1014
Tower footing resistance of all towers, except the first, 3 579.2 632.1 640.6 692.0 808.9 904.3 1008
R2 = 10 Q 4 579.2 632.1 640.6 692.0 808.9 904.3 1008
Tower footing resistance for the first tower is varied
Entrance tower footing resistance = 3 Q .
between 1 ft and 6 ft Overvoltages in kV.
Busbar surge impedance Zs = 80 ft
Length of overhead line spans: Table 3: Maximum overvoltage as a function of number of
dx = 110 m spans in the equivalent circuit
d3 = 250 m /, kA
Length of busbars d2 = 96 m n 30 60 90 120 150 180 210
1 579.8 637.3 663.9 759.8 857.4 963.8 1103
2 579.8 633.2 646.4 694.4 815.9 917.3 1025
3 579.8 633.2 643.6 696.2 814.6 912.7 1018
4 579.8 633.2 643.6 696.2 814.6 912.7 1018
Entrance tower footing resistance = 6 Q.
Overvoltages in kV.

It can be seen that the equivalent circuit in Fig. 3c, with

timers
three spans, is quite sufficient for the estimation of light-
Fig. 4 Lightning current waveform ning overvoltage. The voltage maximum in the substation
1EE PROCEEDINGS, Vol. 131, Pi. C, No. 5, SEPTEMBER 1984 205
occurs before the travelling waves reflected from the fifth first tower footing resistances of 1 Q, 3 fi and 6 Q, respec-
tower approach the substation. tively.
The relative errors for the equivalent circuit in Fig. 3a,
compared with the exact solution, are presented as a func- Table 6: Maximum overvoltage as a function of number of
tion of current amplitude in Fig. 6 for different first tower spans in the equivalent circuit
footing resistances. /, kA
n 120 150 180 210
10
1 950.8 1016 1076 1306
2 789.6 948.5 1047 1017
3 788.6 947.8 1028 1012
4 787.6 947.4 1027 1011
Entrance tower footing resistance = 6 D .
Overvoltages in kV.

In Fig. 8, curves showing the maximum overvoltages

against lightning current amplitude for the equivalent
0 circuit in Fig. 3c are drawn for two tower models. Curve a
30 60 90 120 150 180 210 shows the results for the simple tower model and b the
I , kA
tower model with the exact back-flashover characteristics.
Fig. 6 Percentage error in lightning overvoltages
xR=lfi A R = 6ft
oR = 3n
The percentage error in lightning overvoltage computa- en
tion using the equivalent circuit in Fig. 3b is always less a 600
than 1%. In the equivalent circuits in Figs. 3c and d there
are no errors caused by neglecting the other towers in the
A 00
system.
0 30 60 90 120 150 160 210 240
The study was repeated using the more accurate model
of the tower back-flashover characteristics shown in Fig. 2 current kA
with the same equivalent circuits of Fig. 3. The flashover Fig. 8 Maximum lightning overvoltage against current magnitude for
voltage/time characteristic is presented in Fig. 7, taken different tower models
from Ecklin et al. [11]. The maximum overvoltages at the a Simple tower model
b Tower model with exact flashover characteristics
substation as a function of the current amplitude for differ-
ent equivalent circuits are shown in Tables 4, 5 and 6 for
From Fig. 8 it can be seen that the overvoltages are
greater in the case of the tower model with accurate back-
flashover characteristics. The transient voltage at the sub-
station entrance for the simple tower model and for the
2000
accurate tower model are presented in Figs. 9 and 10,

1000

5 10 15 20 25 5 00

time.ps
Fig. 7 Flashover characteristic of line insulation

Table 4: Maximum overvoltage as a function of number of

spans in the equivalent circuit

/, kA
n 120 150 180 210

1 909.5 968.3 1018 1204

2 725.8 942.9 1015 1134 0 1 2 3 4 5 6 7 8 9 10 11
3 724.2 942.8 1015 1101
time, \i s
4 723.6 942.3 1015 1099
Fig. 9 Transient voltage for case of simple tower model
Entrance tower footing resistance = 1 Q.
Overvoltages in kV.
respectively. The wave shapes of the transient voltages
Table 5: Maximum overvoltage as a function of number of differ for the two models. In the case of the model with
spans in the equivalent circuit accurate back-flashover characteristics the wave has a very
/, kA steep front.
n 120 150 180 210 The influence of attenuation and distortion of travelling
1 949.2 972.9 1067 1234
waves due to wave propagation and corona effect is inves-
2 778.5 945.1 1032 1124 tigated. The data for the digital simulation of these effects
3 778.3 945.1 1028 1012 are taken from Cooper and Hileman [9]. In Table 7 the
4 776.3 944.6 1027 1010 percentage differences between the computed maximum
Entrance tower footing resistance = 3 Q. overvoltages with and without corona effect are shown for
Overvoltages in kV. the exact back-flashover tower model.
206 IEE PROCEEDINGS, Vol. 131, Pt. C, No. 5, SEPTEMBER 1984
The transient voltage at the substation entrance is present- second tower in front of the substation, causing a back
ed in Fig. 11, when attenuation and distortion effects due flashover. Two different types of surge arrester are rep-
to corona are considered. resented. The results show the difference in behaviour of
the classical silicon carbide (SiC) arrestor and the gapless
metal-oxide arrestor.
In Fig. 12, the maximum percentage difference between
the lowest and highest values of overvoltage computed for

500

= 4

i i I i
2 3 4 5 6 7 8 9 10 11 10 20 30
t i m e , us current k A
Fig. 10 Transient voltage in circuit with exact tower flashover model Fig. 12 Percentage difference between the maximum and minimum
voltage as a function of current magnitude
Table 7: Percentage difference between computed overvol- Case of a direct stroke to phase conductor; metal-oxide arrestor connected
tages with and without corona effect
/, kA 120 150 180 210 different power frequency voltages is presented as a func-
DV/V, %. 0.5 9.2
tion of lightning current amplitude in the case of a direct
1.3 12.8
strike to a phase conductor. The system is protected with a
metal-oxide arrestor having a rated voltage of 276 kV. Fig.
13 shows the variation of maximum overvoltage in the
substation with phase angle of the power frequency
voltage, at the instant of lightning discharge for a lightning
current amplitude of 30 kA. The instantaneous value of
power frequency voltage is shown on the same diagram.
500 There is a very small increase in the maximum lightning
overvoltage when the power frequency voltage has the
opposite polarity at the instant of discharge.
The calculation was repeated for a lightning strike to
the top of the tower. Fig. 14 shows the maximum differ-
ence between minimum and maximum overvoltage for dif-
0 _J I I ferent currents, and in Fig. 15 the maximum voltage is
0 1 2 3 4 5 6 7 8 9 10 11 shown plotted against phase angle of power frequency
MS voltage for a current magnitude of 60 kA. In this case there
Fig. 11 Computed transient voltage when corona effect is considered is a decrease in the maximum overvoltage when the power

The computation was then repeated, varying the rate of -1200

rise of the lightning current from 2.5 kA/^s to 30 kA//zs. In
all cases it was sufficient to use the equivalent circuit with r 800
three line spans, as the error for this equivalent circuit due
to neglect of the influence of the other towers is less than 400
0.5%.
The overvoltages caused by a lightning discharge to the
third or fourth tower would be lower because of the 50 1C
attenuation and distortion of travelling waves. A strike to phase angle.deg
-400
the first tower was not analysed, because in this case a
back flashover cannot be expected because of the small Fig. 1 3 Maximum overvoltage as a function of power frequency voltage
footing resistance. Case of a direct stroke to phase conductor; metal-oxide arrestor connected

3 Influence of instantaneous value of power

frequency voltage at moment of lightning 12
discharge
^ 9
In studies of lightning performance of overhead lines and 8
substations, the influence of power frequency voltage is c 6
usually assumed to have extreme values (i.e. positive and
negative peak values).
In this study, the influence of the power frequency 30 60 90 120
voltage of one phase on the maximum value of the light- current k A
ning overvoltages is investigated under two sets of condi- Fig. 14 Percentage difference between the maximum and minimum
tions: firstly, for the case of a direct stroke to the phase voltage as a function of current magnitude
conductor, and secondly, for a stroke to the top of the Case of back flashover; metal-oxide arrestor connected

1EE PROCEEDINGS, Vol. 131, Pt. C, No. 5, SEPTEMBER 1984 207

frequency voltage has the opposite polarity. The same did not exceed 22% in the cases analysed. There is a ten-
investigation is made for the classical SiC surge arrester, dency for the influence of the power frequency voltage to

_ . 20

400- 10

E 1 i
D
E 50 100* 150
X 50 100
a phase angle deg " curre n t . k A
E
-400-
Fig. 18 Percentage difference between maximum and minimum voltages
Fig. 1 5 Maximum overvoltage as a function of power frequency voltage as a function of current magnitude
Case of back flashover; metal-oxide arrestor connected Case of a back-flashover; SiC surge arrestor connected

with a rated voltage of 276 kV. The computed differences

between minimum and maximum overvoltages, for differ-
ent instantaneous values of power frequency voltage in the :8oo
case of a direct strike to a phase conductor, are presented
in Fig. 16. In Fig. 17 the maximum overvoltage is plotted 400

10

phase angle , deg

-400

Fig. 19 Maximum overvoltage as a function of power frequency voltage

Case of back-flashover; SiC arrestor connected

0 30
current kA decrease with an increase in magnitude of the lightning
currents.
Fig. 16 Percentage difference between maximum and minimum voltage
as a function of current magnitude
Case of a direct stroke to phase conductor; SiC surge arrestor connected 5 Conclusions
Various line models for the estimation of lightning over-
2000r voltages in a high voltage substation caused by back-
1600
flashover are analysed. A model in which the
> back-flashover characteristic of the line insulation is rep-
^.1200 resented, and which has only three line spans in front of
the substation, is proposed. It is shown that the more
jf 800 remote towers do not influence the transient overvoltages
> 400 in the substation.
The influence of the instantaneous value of power fre-
0 quency voltage at the moment of lightning strike was
50 10C
-A00
phase a n g l e , d e g
investigated. The maximum difference between the highest
and lowest overvoltages for different phase angles at the
Fig. 1 7 Maximum overvoltage as a function of power frequency voltage moment of strike never exceeded 22% in the cases
Case of a direct stroke to phase conductor; SiC surge arrestor connected analysed. It is not possible to determine the general influ-
ence of the instantaneous value of power frequency voltage
as a function of phase angle of power frequency voltage. on the lightning overvoltage maximum.
The current magnitude is 30 kA. When the power fre- The model, with three line spans in front of the sub-
quency voltage has opposite polarity at the moment of dis- station, can be used for the calculation of lightning per-
charge there is an increase in the maximum overvoltage. formance of substations at voltage levels other than 345
In the case of a back flashover, Fig. 18 shows the kV, and may be employed for both single and double
maximum difference between minimum and maximum circuit lines.
overvoltage as a function of current magnitude, and in Fig. In this paper, the effect of footing resistance of the
19 the maximum overvoltage is shown as a function of entrance tower has been investigated under conditions of a
phase angle of power frequency voltage at the instant of lightning strike to the second tower out from the sub-
discharge. In this case, minimum overvoltage occurs when station. It may be concluded that this parameter does not
the power frequency voltage is small, i.e. near the zero have a significant influence on the maximum overvoltage
crossing. The results of the analyses show that the power in the substation, except when the lightning strike is to the
frequency voltage has a significant influence on the entrance tower itself.
maximum value of the overvoltage. The magnitude of this It is the authors' intention to extend the analysis to
effect cannot be determined in advance, but must be deter- examine the influence of tower footing resistance, lightning
mined for each case. The maximum difference between the parameters and line flashover voltage on estimation of the
highest and lowest values of overvoltage due to this effect risk of substation insulation failure.
208 IEE PROCEEDINGS, Vol. 131, Pt. C, No. 5, SEPTEMBER 1984
5 Acknowledgments where
The authors wish to thank the University of Manchester V[t)y = voltage at node y
Institute of Science & Technology for the provision of all n = number of lines connected to node y
research facilities. In addition Dr. Savic wishes to grate- aiy = refraction coefficient for waves propagating
fully acknowledge the award of a Postdoctoral Visiting along line i towards node y, given by
Fellowship by the Leverhulme Trust.
2Zan

6 References
where
1 BURGER, U.: 'Zum problem des uberspannungsschutzes von gekap-
selted SF6 isolierten hochspannungsanlagen und rohrgaskabeln', " 1
Elektrizitaetsverwertung, 1976, 51, pp. 285-291 7 -I V
2 TOMINAGA, S., AZUMI, K., SHIBUYA, Y., IMATAKI, M., FUJI-
WARA, Y., NISHIDA, S.: 'Protective performance of metal oxide
surge arrester based on the dynamic V-I characteristics', IEEE Trans., Z, = surge impedance of line i
1979, PAS-98, pp. 1860-1871 Zeq = equivalent surge impedance of all lines except
3 PETTERSSON, K., HOLMHORN, H., OLSSON, P., and LIND-
BERG, L.: 'Overvoltage protection of the metal enclosed SF6 sub- line i connected to node y
stations considering steep surges'. CIGRE, Paris, Paper 33-05, 1978 The backward voltage component is calculated in the fol-
4 BRAUNER, G.: 'Das verhalten von metalegecapselten und freiluft
schaltanlagen bein naheinschlagen', Elektrotech. Z. ETZ, 1976, 97, pp. lowing way:
748-752
5 ROBINSON, P., EDLINGER, A., SCHILCHT, D., SOUZA vB(t)iy = my - vAt\,
D'AVILA MELO, E., and SALATKO, V.F.: 'Lightning overvoltages
protection of the 550 kV ITAIPU SF6 gas insulated substation'.
CIGRE, Paris, Paper 33-11, 1982 7.2 Model of corona distortion
6 BAUER, H.: 'Simulationprogram zur berechnung der blitzspan- The distortion of the travelling waves due to corona is
nungsbenaspruchung und der dadurch bedingten fehlerhaugigheit modelled by a voltage-dependent additional propagation
von elektroenergieanlagen', Elektrie, 1975, 29, pp. 575-576 time, given by
7 ANDERSON, J.: 'Monte Carlo computer calculation of transmission
line lightning performance', Trans. Amer. Inst. Elect. Engrs., 1961, 80,
pp. 414-420
8 OLSEN, K, LANGE ANDERSEN, D., ULRIKSEN, Th., HUSE, J,
and SCHEI, A.: 'Lightning overvoltages in SF6-installations: risk of
failure analysis'. CIGRE, Paris, Paper 33-01, 1978 where
9 COOPER, J., and HILEMAN, A.: 'A probabilistic approach in esti- c = velocity of light, km/^s
mating the BIL for 1200 kV gas insulated stations'. CIGRE, Paris, K = constant, dependent on the line geometry
Paper 33-02, 1982
10 AMETANI, A.: 'Modified travellingwave techniques to solve elec-
Vcr = critical corona voltage
trical transients on lumped and distributed constant circuits', Proc. V = instantaneous value of the voltage, kV
IEE, 1973, 120, (4), pp. 497-504 d = distance, km
11 ECKLIN, G., ROBINSON, P., and SCHLICHT, D.: 'Lightning
overvoltage protection of the Drakensberg 420 kV SF6 gas insulated If the voltage at node y is greater than the critical corona
substation', Brown Boveri Rev., 1982, 69, pp. 188-189 voltage, the backward travelling wave propagating from
node y is modified as in Fig. 20.
7 Appendix
7.1 Travelling-wave technique used in the
computation of the transient process
The propagation of travelling waves along an ideal mono-
phase line can be expressed by
VF{t)iy = VB(t - A T ) ,
where t j * At
VF(t)iy = forward voltage component propagating time
towards node y along line i Fig. 20 Computation of the voltage wave after an additional time delay
AT = propagation time between nodes x and y of the travelling wave due to corona effect
along line i h is the computation time-step
VB(t)ix = backward voltage component propagating
from node x along line i.
The voltage at any node x can be computed in the follow- The voltage at the time instant tx is computed by a
ing way: linear interpolation of the voltages occurring at the pre-
vious time step {tx h) and the time instant (tt + At) at
which the voltage, after an additional time delay, should
appear.

IEE PROCEEDINGS, Vol. 131, Pt. C, No. 5, SEPTEMBER 1984 209