Professional Documents
Culture Documents
3, July 1997
1168
B. Richter
ASEA Brown Boveri
Switzerland
D.A. Roby
ASEA Brown Boveri
Australia
ABSTRACT
Multiple stroke lightning ground flashes can impose surges of exceptional severity on exposed distribution surge arresters. This
paper describes a series of tests on metal-oxide (MO) arresters
and varistors to study surface flashover mechanisms. The results show that the more likely causes of surface flashover were
- plasma generation, manufacturing defects of the varistor surface coating, dielectric properties of the surface coating and the
electrode contact system. For multipulses, plasma enhancement
and varistor surface coating were found to play a dominant role
in surface flashover.
Plasma could be created near the edges of a varistors metallisation, accumulating with each successive puls
multipulse. The generated plasma would not deionise in the
inter-pulse time intervals of about 35 ms, whereas deionisation
could take place in the
dard lightning impulse test with
single pulses separated b
cause of plasma production is
metallisation and the resulting increase in current density at the
edge.
9. Introd~ction
In our earlier study [l],several makes of gapless metal-oxide
(MO) arresters were tested using single and multiple impulse
currents under the representative service conditions and also
under the conditions specified in the Standards [2,3]. Comparisons of the effects of multipulse and standard current impulses
show important differences. The result of major significance
is that multipulse currents can give rise to changes in varistor
characteristics (including failure) that are not evident during
tests with single impulses [l].While multipulse currents caused
changes in the electrical characteristics of the varistors including
some that lead to thermal instability, the most common limiting effect was to cause surface flashover of the varistors. This
indicates that the effects of multipulses may not be purely a
thermal energy problem, rather discrete failure mechanisms (by
surface flashover) appear to be involved.
This paper describes tests on MO varistors from one manufacturer; different surface finishes were used to examine surface
flashover phenomena. These flashover studies are discussed in
section 2 . Based on this study, a multipulse test proposal for
distribution arresters is described in section 3.
2. Flashover Studies
The probable causes of surface flashover of varistors are - a)
plasma created near the edge of metallisation, b) the varistor
outer surface (glass) possibly affected by humidity, c) the likelihood of material defects at the outer edges of the varistor and
d) partial discharges produced at the edges and surface possibly
affecting the dielectric strength. Studies were conducted to examine these four possible causes of surface flashover and these
are reported in this section.
Experimental tests
Five types of tests were c
ed out to determine the comparative performance of five varistor surface finishes. The tests included three applied essentially in accordance with IEC 99-4 [3],
namely, (i) single 8 / 2 0 p s impulse currents applied to the varistors (initially at room temperature) at one-minute intervals; (ii)
single impulse operating duty tests, in which two groups of five
8/20ps impulses were applied at one-minute intervals with the
test object pre-heated to 6OoC and energised at rated voltage,
thirty minutes separated the two groups; after each current application and for thirty minutes after the final application, 1.05
times the continuous operating voltage was maintained on the
test object to check for thermal stability; (iii) two single 4/10ps
high lightning impulse currents, the first applied with the test
objects at room temperature and the second with the sample
pre-heated to 60C. After the second impulse, thermal stability
was checked in the usual way. The other two types of tests used
multipulse currents, in particular quintuple (five-pulse) 8 / 2 0 p s
impulse currents, applied in such a way to provide a comparison
with the single-impulse tests and the single-impulse operating
duty tests; namely, (iv) multipulse currents (inter-pulse intervals of 35 ms) applied to test samples at ambient temperature
and pre-heated to 60C. When more than one quintuple sequence was used, they were applied at intervals of 30 minutes;
1169
(v) multipulse operating duty tests carried out in a similar way
to the single-impulse operating duty tests, except that a quintuple sequence (time intervals 35 ms), without point-on-wave
timing control, was applied in place of each group of five single
impulses.
Each of the twenty five test samples contained a single varistor diameter 38.2 mm; rated current 5kA, rated voltage 6.25kV and
continuous operating voltage 5kV. The samples had five types
of surface finishes - three were on varistors alone and two were
assembled as arresters. The five types are described and the
samples are identified as follows A) varistor with glass coating (samples L1 to L6)
B) varistor with glass coating plus silicone varnish (samples 1
to 6)
C) as in B), but assembled as an arrester but without a polymer
housing (samples AS1 to 3, AS6)
D) varistor with glass coating bonded to a silicone moulding
(samples S1 to 3, S6)
E) complete silicone moulded arrester containing a glass coated
varistor bonded to the silicone housing (samples AF1 to 3).
Type C and E samples were tested using the fittings as supplied.
Type A, B and D samples required the use of a special fitting,
which included disc electrodes (similar to those in the type C
and E samples) and a supporting jig comprising two aluminium
end plates (80x8Qx11.3mm)and four insulated clamping screws.
The varistor samples were placed between the end plates and
clamped with suitable pressure.
90kA applied to eight test samples of all five types. The two
glass coated type A varistors failed by surface flashover at 65kA.
One of the type B varistors (glass coating plus silicone varnish) exhibited partial surface flashover damage at 72kA and
the other failed by surface flashover at 90kA. The one type
C sample showed a burn mark near an electrode after tests
at 90kA; its residual voltage was unaffected, but the DC and
AC reference voltages were reduced significantly (by about 20
percent). In contrast, the samples with glass coated varistors
bonded to silicone mouldings (types D and E) showed no GOSS
effects from the 90kA tests, remained thermally stable, experienced little change in the residual voltage, but had significantly
reduced reference voltages.
1170
Surface defects
During the fabrication of varistors, an outer insulating surface
is provided by spraying glass or insulating material on to the
varistor material and fusing it at aboul 500C. After surface
coating, varistor ends are usually metallised with aluminium
and bonded at about 50OoC. Because of manufacturing procedures and material handling, some varistors may have surface
contamination or irregularities on or near surface.
It is likely that surface irregularities or contamination could contribute to the initiation of processes leading to surface flashover.
An irregularity might take the form of a localised defect in the
outer insulating material or at its boundary with the varistor
material. Such a defect would produce a localised concentration
in the electrical field, which could cause a localised (partial)
discharge and begin the process which ultimately resulted in
surface flashover. It is of course well known that contamination
can degrade the dielectric properties of an insulating surface
and this can also lead to surface flashover.
o The change in the reference voltage was significant and depended on the stress the
was seen after the high c
impulse test. This phenomenon
is well knowii and is a CO
m e b both the large magdude
and the high steepness of the high current impulse.
The second (surface dielectric) dimension showed the failure
mode of the tested samples; all failures (independent of the
impulse stress) were surface flashovers. Here again two main
groups were noticed.
1171
This type of phenomenon was also observed on the tested sample surfaces, which failed during multipulse tests.
Single and multipulse impulse tests on varistors reveal that relatively low magnitude multipulse currents, when compared to
single impulse currents, could cause surface flashover. This indicates that a comparable plasma enhancement should have occurred with multipulses. In the case of multipulses, plasma enhancement very likely plays a dominant role in the breakdown
initiation.
It may be, especially at higher temperatures, that at the edges
of the blocks, where naturally the highest field strength is, the
possibility of ionisation is higher when the impulses are very
frequent as they are with multipulses. This is possibly due to an
increase of free electrons and ions forming a plasma, which does
not have time to neutralise between the individual multipulses.
These effects need further and more in depth investigation.
The varistor results show how important it is to have a good
electrode contact system and a good surface coating around the
varistors with high dielectric strength when exposed to extreme
conditions such as high humidity. For these reasons, the complete arresters with one varistor seem to perform very well in
the conducted tests.
1172
(b) There is no need to time the application of the quintuple
impulses with respect to the power frequency voltage waveform.
(c) Rated voltage should be applied at least 10 s prior to the
application of first lightning impulse current and the arrester
should be pre-heated to 60f3C. The applied voltage should
be reduced to 1.05 continuous operating voltage 10 s after each
quintuple set of current impulses, and should be maintained
for the 25 to 30 minutes between groups of quintuple currents.
Rated voltage should be re-applied before the application of the
next set of quintuple currents. After the fourth quintuple set,
1.05 times the continuous operating voltage should be maintained for 30 minutes to check thermal stability.
The criteria for assessing the effects of the multipulse tests are
(evident within thirty minutes), b) subsequent changes in residual voltages (limited to 5 % ) , c) d.c.
reference voltages (limited to 10%) and d) power-losses (limited
to 5%). This operating duty test protocol is a reflection of field
conditions likely to occur on distribution systems in service. In
addition, no significant abnormalities should be found in the
current and voltage oscillograms recorded during the tests.
- a) thermal stability
ents
Support for this work (which was carried out at the University
of Queensland) has been provided by the Australian Research
Council, the Australian Electricity Industry Research Board
and by ASEA Brown Boveri Power Transmission (Australia).
5. References
1. M. Darveniza, D. Roby and L.R. Tumma, Laboratory
and Analytical Studies of the Effects of Multipulse Lightning Current on Metal-Oxide Arresters, IEEE Transactions on Power Delivery, Vol. 9, April 1994, pp. 764-71
BIOGRAPHY
Mat Darveniza, born in Australia in 1932, is a g
University of Queensland (BE, DEng) and of London (PhD).
He is Professor (Personal Chair) in Electrical Engineering at
. His research interests are in lightthe University of Quee
ning, high voltage and
cal insulation. He is a Fellow of the
IEEE, the Institution of Engineers Australia and the Academy
of Technological Sciences and Engineering. He is past-chairman
of the IEEE Australian Council and of the PE Chapter, Queensland.
L.Reddy Tumma, born in India in 1960, obtained his BE from
the Osmania University, India in 1982, ME from Indian Institute of Science, Bangalore, India in 1984 and PhD from the
University Queensland, Australia in 1995. He worked in industry for over four years in the field of switchgear protectio
is a member of the IEEE.
Bernhard Richter, born in Germany in 1950, obtained the Dip1.Ing (FH) in 1973 from the Technical High School (Beuth) and
the Dip1.-Ing. in 1979 from the Technical University of Berlin.
After five years as a scientific assistant at the Institute of Highvoltage of the University of Berlin he joined BBC, now ABB, in
1985, where he works now in the development of surge arresters.
Besides other international activities he is member of the JEC
working groups 06 of T C 81 arid 05 of SC 37 A.
David A. Roby, born in India in 1951, is an electrical engineering graduate from the Indian Institute of Technology, Madras
(B.Tech.). He is the Business Unit Manager - Surge Arresters
with ABB Power Transmission Pty Limited and a member of
subcommittee EL/7/3 - Surge Arresters. His research interests
are in gapless metal-oxide technology, system protection and
low energy self blast circuit breakers.
1173
Lt
Type of
surface
coating
A
3
AS1
B
C
s2
Sample
No.
I D.C. reference I
Type of
surface
coating
Stress applied
No.
AF3
Sample
AS3
S6
C
D
AF2
voltages 1.4mA
+ive 1 -ive
0.959 I 1.008
A.C. reference
voltages 1.4mAp
+ive 1 -ive
0.992 1.029
0.955
1.045
,995
1.032
1.015
1.031
0.971
0.954
1.022
1.043
0.992
0.991
1.035
1.017
1.026
1.005
1.022
1.001
0.942
1.006
0.995
1.031
1.03
1.016
Discharge
residual voltage
+ive I -ive
1,019 1.026
D.C. reference
voltages 1.4mA
+ive I -ive
1.01 1 1.03
A.C. reference
voltages 1.4mAP
+ive I -ive
1.03 I 1.02
Discharge
residual voltage
+ive I -ive
1.03 1 1.02
1.024
1.029
1.03
1.048
1.021
1.05
1.006
1.023
1.012
1.026
1.016
1.022
1.W
0.989
1.05
1.016
1.017
1.032
1.03
1.05
1.014
1.03
1.018
1.021
1.006
1.075
1.018
1.032
1.014
1.00
Remarks
t results
AS3
IB
IC
s3
AF1
ID
IE
I90kA
I 90kA
I-
I-
I 0.795 I 0.884
I
I
I
I 90kA I 0.791 I 0.909
I 90kA 1 0.734 I 0.859
I
II
II
I 0.991 I 1.00
I failed
I 0.992 I 1.009 I
I 1.00 I 1.021 1
HV electrode
1174
ISCUSSlON
JOHN B. F'OSEY (Westfield Centei, OH) The authors have
reported failures of unhoused blocks caused by multipulse
discharges Also reported is "Blocks which were completely
assembled and moulded in silicone (types D and E) did not fail any
of the tests" and "showed no evidence of gross damage or
degradation
in
electrical characteristics"
Would the authors please clarify how data from the testing of
assembled arresters supports the suggested change in standards?
The published test data does not seem to support the proposal, but
instead shows unassembled blocks fail but assembled arresters are
immune to the proposed multipulse test procedure.
Manuscript received August 14, 1996.
Dr.
3.
The reported study was performed on metal-oxide
elements of only one manufacturer.
The elements were
directly coated with a glass collar. Other manufacturers
use other collar materials, both inorganic and organic.
Have the authors conducted simlar multipulse tests on
continuing work?
4.
The authors previous work (and the work of others)
indicates that
a major
cause
of porcelain-housed
distribution arrester failures is moisture ingress. While
most polymer-housed arrester designs contain little free
air space and moisture ingress should be less of a
concern, it is still necessary to have some means of
detentuning that arrester performance is not degraded by
moisture ingress.
Many polymer-housed designs do not
permit simple disassembly to inspect for moisture and
therefore some electrical test evaluation means are
usually employed. Do the authors have an opinion on the
merits of a multipulse test as such an evaluation means?
1175
seems to be no compelling reason for them from field
experience with metal-oxide arrester failures. We are not
convinced that this is correct, eg. we agree that problems due
to moisture ingress should be less for polymer housed
arresters than for those with porcelain housings. But only
time in service will prove if this is really so. In fact, our
continuing research project is to correlate failure modes
produced in the laboratory with failure modes encountered by
arresters in the field. This work is being conducted now with
the co-operation of about fifteen utilities, who will send us
their failed and sumect arresters for examination and test.