DEGRADATION OF METAL OXIDE SURGE ARRESTERS UNDER SIMULATED NATURAL

CONDITIONS
P.M. Vipin, N.K. Klshofe, G.R. Nagabhushana and B.N. Jayaram,
Departmentof High Voltage Engineering,
Indian Institute of Science,
Bangalore 560 012
ABSTRACT

2-EXPERIMENT

Th,& degradation of metal oxide surge

arresters
(MOSA)
under
continous
operating voltage ( C . 0 . V ) ,with time, is
a well known fact. But the contribution
of bransient overvoltages in the form of
lightning
surges,switching
surges
etc..against which the arrester is used
as a prot.ection device, to the arrester
degradation has not been studied much.
The simulation of a situation, wherein
and
the arrester is kept uder C . 0 . V
different surges are applied on it, is
somewhat
complex.
A
method
for
simulation of field operating conditions
of an arrester in the laboratory is
presented in this paper. Results of two
seperate
degradation
studies
on
different samples under A.C, and D.C
excitations with superimposed transient
stresses are given. A com$arison of
observed data with calculated ones is
also made.

2.1

Scheme of experiment

The representation o f thi. 5 1 . 1 v i i . 1

conditions is achieved by c s t iiiit>t i i i t j I l i t *
stress cycle per year. T h i s c y c l i - i x ;
then applied to the surye
al-icstr*i
element over one day, to simii1.3Le tIv>
electrical stresses coming I I L ~ I I t I I P
arrester
under field condit ions.
A
typical transient stress cycle c)vi-r i t i i r year has been broadly esLiamIi~t1 .Ifollows:
(i)Lightning transients
(a) direct stroke, s i n i u l a t c t l Iig . I
surge current of 4/1O)is,
50kA
postulated to occur OIICC- c.vi*r.g
5 years,

(b) indirect strokes, simu1;ltid 11y

surge
currents
01
n/70)1s
duration and magnitudes 5 , I O
ana
20kA
occtii-i II'J
respectively; at fl-cc/[[r:rlc: i v s
of roughly 18,5 and 2 t i n i c s i i i i
year.

1.INTRODUCTION
Developments in material technology
have spurred quantum advances in power
system equipment protection within the
past two decades. Eversince its advent,
around 1970's, metal oxide based surge
arresters have emerged as the
most
reliable device against power system
traiisient overvoltages. It is
being
increasingly used in surge protection
due
to
its
superior
nonlinear
characteristics
and
greater
energy
handling capabilities. But, due to the
absence of any series isolating gaps,
the
surge
arrester
elements
are
continously
stressed by the
system
voltage, which causes the flow of a
small leakage curent (about SO*
or sol.
This current as well as the energy
absorption
at the time of
impulse
discharges, cause gradual degradation of
theelements. In order to predict the
period of reliable operation of a metal
oxide surge arrester, or its life, the
degradation phenomena is studied the
worldover. Still, a definite procedure
is yet to be postulated to
assess
degradation under both A.C and
D.C
excitations. Here we present the results
of work done-in our laboratory to study
the
degradation of
surge
arrester
element-s
under
simulated
natural
conditions.

(iilswitching transients siriwIa1:c.tl l l y

J
rnct:angular current. L n i 1 * : P
r,t
1kA
magnitude
and
1- .4 111:;
pulsewidth,.occurring oricc! i r i t i 1 1
year.
While
the above
transients
arc
applied the arrester is kept
undi~i~
continous operating voltage ( C . 0 . V ) at a
0

temperature of 60 C .
Experiments were
carried out with D.C and A . C cscitations
applied to the arrester elements. T l i c
0

elevated temperature of G O C siiiiitla,l.es

the maximum probable element t cmp r a t CII'(in field service, taking into a c c - i - ~ i 1 I i l
the influence of ambient teniper;11:urp,
solar
radiation
and
inE11~c1ic.i~
[ 1 1.
pollution on the arrester Iioiisirig
Further, one year field l j r r . iir 1 1 1 1 .
arrester is simulated within ani. 11.1y
fo11owing
t he
acce1eratcd
a q i 11.3
procedure described in clause 6 1 . 7 cif
IEC T C 37 WG-4 AUGUST 1980 docr~rnriil- f t l !
MOSA. Accordingly, after the abq.>li i - . > I - i t ) i > ,
of surges the sample was Iica!cil 1 1 1 . I

165

cycles of aging, a comparison lict\;cc?ii

the
A.C
and D.C
leakage
cul-rc~~l'
variations for a 10 year p r r i o c l O C
simulationran be.done. This is depjr:t.ed
in Fig.2 in the form o f normal i s c d
curves (normalised with respect .to tllc
initial leakage current values U I I ~ C I . A . C
and
D.C
respectively).
IVhilc
tllc
increase- in leakage current is a b r ~ \ i t 5
fold in case of A . C voltage (after a
,ljeriod of 2 0 year simultion), tllat j n
the case of D . C - excitation is r o u g h l y 4
times
(after 10 days = 10
ypt3t.s,
.accelerated aging). It is also ol>st?rvt?tl
that the magnitude of leakage curl.ei1~s
in case of D . C is more than double t h ; ~ t
under A . C [Figs.3(a) & 4 t a ) l .

temperature of 115 + / - 4 C and kept at

across
thak temperature with the C . 0 . V
it,, ,till the start of the next aging
cycle.,A 2 0 day simulation on the above
lines was carried out with
minimum
A.C
operating
interruption
of the
voltage across the arrester clement.
Another 10 day experiment with
D.C
opefkking voltage on a different sample
was.a l s o conducted.
2 . 2 .Instrumentation

.The' lightning impulse current's were

gen,ek-ated using'a capacitor discharge.
One'generator was capable of giving upto
4OkA, .'.8/20psand another generator of
200kA,.'4/10ps. The switching impulse
current' was obtained from . a 8 stage
1adder:network transmission line model.
It had, a surge impedance of 3.6 ohms and
the output pulse width was 2.4ms. The
A.C/D.C' excitation for the
arrester
elements was obtained from a 230V/3.6-03.6kV, O.SkVA, oil filled transformer
and * half-wave
rectifier. &
filter
capacitor
arrangement. The
arrester
element was kept inside a temperature
controkled oven to maintain thF elevated
temperatures. The oven had a maximum

3.2 Barrier Height:

T h e barrier height is calculated f r c > # t i
the low voltage D . C characteristics of
the arrester element after each cyc11: uF
aging. The calculation is based on tlic
following equation:

where

resistivity in

vb

k,,

barrier height i n cv
Boltzman constant,
-5
0 ,
8.625*10
eV/ K

temperature i n

temperature

capability of 30.0 C with

+ / - 2 ,C regulation.

The
measuring
instrumentation,
coniprised of a 17.5mnco-axial current
shunt with Gould 4074 looMHz Digital
Storage Oscilloscope , for measuring
impulse
currents and
a
resistance
potentjal
divider
to
monitor
the
residual voltage across the arrester as
well , a , s the C.O.V. Typical
8/20ps
lightning impulse .current and
2.4ms
iswitching impulse current
waveforms,
obtained using the above arrangement are
shown in Fig.1. The leakage current
through
the
arrester
element
was
monitored using a series resistancedi.gita1 . multimeter
scheme.
The
capacitance and tan delta 'variation with
aging..was measured using GENRAD RLC
Digibridge Type 1657

'111

1;

The variation in barrier height. i s

found to be quite small initially under
both A . C and D . C voltages. But, arter
about
6 to 8 cycles of
aging
a
significantly
abrupt
decrease
is
observed in h o t h cases ( F i g s . 5 b 6 1 .
I

3 . 3 Capacitance and Tan delta:

The
capacitance
and
tan . c l c l ta
measurements were made at lkHz
without.
any
external
bias
voltage.,
The
capacitance
values
show
gdlleral
decreasing trends under both D . C and A . C
operating voltages. On the otlicl-li.iird
measured tan delta values at thc! etid,of
each cycle 'of aging show a randainly
varying
spread, which is
hard
explain. While it exhibits a s h i
.spectrum
of values , w i t h
ovrr,al I
increase, in case of D . C , ,the vai-.Lat i o n
is highly erratic, with some p c , a I c s , i i i
case of A . C .
4.DISCUSSION

3 .RESULTS

The
variation in
the
different
degradation
indices
monitored
are
graphically depicted in Figs. 2 to 10 as
a
iunction
of the
aging
cycles.
Eventhough
the 1eakag e current
and
barrier height were the main indices
monitored, capacitance and tan delta
variations are also shown.

As
a real .life simulation o f > I O S A
under field conditions, it is fcll. 1.11;11.
much more need to be done to havr: a c_loud
representation
of
the
operat3 iir~
atmosphere. Even then, the present sc:lic:!iiir:
offers a guidelin6,on how to p l a n a
laboratory simulatbn of PlOSh
ric1c.I
life, to study the complexities o r r 1 . ; 1 1
life stress gegrada-tion.
The leakage current seems to bc t h c
best index of degradation, showirig a
predictable variation with SUCCCSS~VC.
cyclds
. of c x p c ? c t . e d
.
.of aging. The . values

I(

3.1 'Leakage current:

1.ns.piteof the scatter observed, the
leakqge
current
shows . significant
overall increase after successive aging
cycles under both A . C and D . C operating
voitages. since the a .c leakage current
i s fo11nd b.0 s t a h i l i s ~ somewhat. aCte1.
10

Po. - constant

.1

166

leakage;, currents were calculated based

on an, expression given by Y.Fujiwara
et a1[21.

found to fo'llbw a predictable \ ? d i i . j , i i . i t i n

opening
up
possibilities
of
1 i
prediction on the basis of t l i c i i r < l c * i , * i f
its variation. Eventhough thc gi>tit.i-.i1
trends
of variation in
i3<,qi.qiil.li i r i i i
indices is similar under botli A . C . i i i t l
D.C
voltages. evidence of
~ ~ 1 ~ ~ 1 1 1 ~ ~ 1 ~ 1 ~
different phenomena of aging i s c . t l i s c s i . \ . c - t l
from the tan delta and
c.,lLaai<.i t : i i i ( . ~ ~
variations.
fi-

It i s observed that the calculated

values of leakage currents show matching
trend
with
the
observed
values,
eventhough
slight
deviations
are
present. The following features could be
observed
from
the
calculated
and
observed graphs [Figs. 3(b) & 4(b)l:

AKNOWLEDGEMENTS

(i) While the calculated values are

more
conservative
than
the
observed values in case of D.C
voltage aging, it is
reverse
in case of A.C voltage aging.

The assistance of Blr. B.

setting up the experimenCa1
is gratefully acknowledged.

11

(ii) The leakage current variation

with respect to number of aging
'cycles is determined by the surge
current
density,
. element
,.temperature and
number
of
absorbed surges - which were used
to compute the theoretical values
under both A.-C and D.C vol'tages.

1I I

III

I*\

IEC TC 37 WG-4 AUGUST 1.r)ee "surwr;

ARRESTERS - Part 3: Metal O s i d r S L I I - ~ J I '
arresters without gaps for A . C s y s t c i r i s "

1 2 1 Y Fu j iwa ra , Y .S 11ibuy a , P1 . T mi 1. ;I IC i S
T.Nitta,
"Evaluation
or
~ I I l ~ ~
Degradation
of
Metal
Osidr:
Siii~~.l~Arresters" ,
IEEE
Trans.
on
l'~~h'c?t~
Apparatus and Systems, Vol. PAS-101,
No.4, pp. 9 7 8 - 9 8 5 , April 1 9 0 2 .

,
Considering
the
barrier
, height
variation, the almost steady value and
then the sudden drop evident from the
Figs. 5 . & 6 , indicate that irreversible
changes in the microstructure of the
MOSA material happen only after some
cycles .of aging. In the present case. it
is observed that such a change happens
after about 8 cycles of aging in case of
D.C & 6"cycles of aging in case of A.C.
But i t , .is.notable that the order of
change in case'D.C is far more than that
in case of A.C. This cl.early indicates
the p o 1 a r i z a ~ ; ~e~f nf e c t s '..me into play
in case of
. . an unidirectional excitation
131.
'

C;3(.i

REFERENCES

(iii,) While the effect of surges on

degradation
is found
to
be
pronounced
at
the
elevated
temperature
and
operating
voltage, the contr.ibutionof'the
thermal stress to aging is less
than that of the former.

RarJiiutLiin

values
The::: decreasing capacitance
with number of aging cycles might be the
result's:".. of
microstructural
changes
happerling at the depletion region near
the ZnO,'grain-graininterfaces, in the
arreste,r',material.
This as well as the
erratic,'..nature
of tan delta variation
will have to be explained based more on
the
material characterisation
after
aging and associated changes in qiiantum
mechanical phenomena, which were outside
the s,cope of the present work. Hence the
decision.to consider the leakage current
and barrier height as the main indices
of aging.'
5.CONCLUSIONS
A simulation of the field conditions
under which surge arresters-are supposed
to operate has been achieved to a fairly
good extent in the laboratory.
The
leakage current; through the arrester is
167

J ~ '

I
I
I

AQC

odc

NO: OF AGING CYCLES

: Comparison between a c &

dc observed leakage currents

continous operating voltage (normalised).

Fig.2

+
K

200

*I

I
I

I
I

" I

II

1
1------I

50$-;;;--

at

1 1
I

I I

10

NO: OF AGING CYCLES

I
I
I

20
NO: OF AGING CYCLES

(a 1

NO: OF AGING CYCLES

NO: OF AGING CYCLES

(b 1
FicJ.3 : ( a ) A . C leakage current at

(b 1
Fig.4

C.0.V

:(a) D.C

leakage current at C . 0 . v
0

(measured at 60 C )
( b ) Normalised dc leakage c u r r e n t s
(observed & calculated)

(measured at 60 C )
( b ) Normalised ac leakage currents
(observed 6, calculated)
168

0.5I
I

I
. I
I

-_

- - 1*

!E

----

5 0.3

tEi

I
I
I

K
U

0.1
F1g.5

NO: OF AGING CYCLES

Barrier height variation after
a g i n g under a c , e x c i t b t i o n .

I
I

I
I

-_-___
1I - - - - *- - I
I
I

0.80:
0

20

NO: OF AGING CYCLES

:.

a g i n g under d c e x c i t a t i o n .

Y 0.85

Fig.7

: Barrier height variation aftcl-

10

NO: OF AGING CYCLES

Fig.6

I
1

0:53

II
I

NO:

Capacitance v a r i a t i o n a f t e r
aging under a c e x c i t a t i o n .

Fig.8

5
OF AGING CYCLES

10

C a p a c i t a n c e v a r i a t i o n after
aqing under dc e x c i t a t i o n .

0.041

2 0.021

10

0.001

20

NO: OF AGING CYCLES

Fig.9

Tan d e l t a v a r i a t i o n a f t e r a g i n g
undel' ac e x c i t a t i o n .

Fig.10

169

NO: OF AGING CYCLES

Tan d e l t a v a r i a t l o n d f t r r a c J l n g
under d c e x c i t a t i o n .