2003 Annual Report Conference on Electrical Insulation and Dielectric Phenomena

On the UHF partial discharge measurement in transformers

Raja Kuppuswamy
Research Center
ALSTOM Transmission & Distribution
Saint Ouen, France

Abstract: Partial discharge detection sensitivity of different coupling modes as suggested in the literame
three types of UHF partial discharge (PD) couplers, to understand their influence on PD detection
reportedly used in the literature, has been studied in sensitiviw and PD occurrence frequencies in oil-paper
comparison using experimentally simulated discharges insulation. The coupler immersed directly in oil, that
and those discharges ohserved during PD acceptance used in conjunction with a dielectric window and third
testing of transformers. UHF PD measurements were in open air near the transformer vicinity were
performed by a digital instrument developed for the respectively designated couplers A, B and C.
purpose. An interesting observation has heen that
UHF PD inception and detection sensitivity is linked The influence on the sensitivity was studied by
to the appearance of PD pulse current impulses of simulating discharges using PD models and by
negative polarity. The occurrence frequencies for PD varying the distance between the discharge source and
and common noise sources do not appear to change the coupler. A 200 kV transformer test tank with
with the type of the PD coupler used, although detachable access covers were used for installing
individual sensitivities vary. In the last part of the couplers A and B. Coupler B was used in conjunction
paper typical UHF PD pattems recorded using with a pre-installed Nylon 6-6 dielectric window on
different couplers during PD acceptance testing of one of the access cover. In the last part of the paper
transformer is presented. typical UHF PD patterns recorded using coupler A
and B during PD acceptance testing of large power
Introduction transformers is presented.

When electrical charges move with anything other UHF PD measurements were performed by a digital
than a constant velocity, electromagnetic radiation instrument developed for the purpose and had the
occurs. The electrical discharges that occur due to ahility for simultaneously sampling and analysing
overstressing of an insulating material are pulses of signals from multiple PD couplers. An optional
current in which electrons are accelerated or software program allowed separation of true PD
decelerated. Radiated signals travel away from their signals from noise corrupted UHF coupler outputs.
point of origin in all directions. In a confined space,
such as a transformer tank, multiple reflections occur Noise Spectra at Factory and Site
at conducting and dielectric surfaces and the signals
spread through the tank by a multitude of paths. This The spectrum shown in Figure la was recorded hy an
complex resonance within the tank dies away as signal UHF FD coupler B installed on an inspection door
energy is dissipated. Signal in the UHF range (200- located on phase B hushing turret of an un-energised
Zoo0 MHz) will he excited when the discharges occur 400 kV transformer. An UHF Pi3 coupler C in open
on a time scale of around IO-’ s or less. UHF PD air inside a transformer substation recorded the
coupler is essentially an UHF antenna that gets excited spectrum shown in Figure Ib. The frequency range
when signals of such frequencies hit the coupler active and resolution was 0.2 to 1.5 GHz and 3.25 MHz for
surface. The frequency range and sensitivity is a the first case and 0.2 to 1 GHz and 8 MHz for the
function of the size of the active part of the antenna. second case.

In the literature, three types of UHF coupler have heen If signal amplitudes of less than IO dB are ignored, the
used. One is a coupler prohe inserted into an opening output of PD couplers B and C are similar, except for
on the transformer tank wall like an oil drain valve; an additional interference around 0.7 GHz for the
the second is using a coupler mounted inside a pre- second case. For subsequent experiments amplitude
installed dielecaic window on the transformer wall; threshold of -60 dB was chosen for data-acquisition
and the third an external UHF antenna in the vicinity and interference above this level was automatically
of the transformer [I]. In this paper, three identical identified and suppressed by a software program.
wide hand couplers were designed and used in three

0-78057910-1/03/$17.00 02003 IEEE 349

__
0,
. .. .. ... .. .

Figure 2 Classical and UHF PD partems of ai corona at M a e m

test voltage levels.

(a)
Test valrage slightly above inception. Positive polarity pulse
rate(K)= 585 pulsedsee; N = 0: No LMF response
(b) Int-diate volrage, K =2649, N=2% UHF response - 10 dB
F i p e 1: Noise panw recorded at factory and site. (X-axis: around 221 MHZ and less than 2 dB spread upto 530 MHz.
hequency in MHz and Y-axis. amplitude in dB)
(c) Relatively high voltlge. K=16691,K=299: UHF response -
~ dB m u n d 1070 MHZ aodlers than 2 dB
l O d B at 229 h C i 6
Influenee of external discharges on UHF mupler spreadupto530hBk.

In any practical PD measurements, the influence of During one investigation, air-corona having
discharges in air outside the transformer tank is overwhelming number of only positive polarity pulses
important. The common noise sources k i n g air (upto 11000 pulsedsecond) was generated and during
corona from unshielded or inadequately shielded this period no UHF PD signals could be detected The
voltage terminations and localised arcing resulting peak PD magnitude measured was around 400 pC. At
from bad contacts and Likes. higher test voltage level, negative polarity PD pulses
appeared (Fig. 2b & 2c) and with it a steady UHF
To investigate the influence of such discharges on response was Seen between 227 and 239 MHz. At still
different PD couplers, air-corona and extemal arcing higher test voltage and consequent increase in PD
discharges were simulated and studied. The air corona pulse rate second steady UHF PD peak between 1070
was simulated by strapping a thin wire onto a high and 1100 MHz was observed. Similar observations
voltage bushar connecting the test transformer and the were found using UHF couplers mounted with and
transformer test tank. Coupler A and B were installed without dielectric windows hut with varying degrees
on the tank as previously described. The distance of attenuation in pulse amplitude.
between the extemal discharge wurce and coupler 4
C and B were respectively less than 1 and 1.5 m. Noise interference due to electrical arcing in air was
External arcing discharge were simulated by next studied. UHF PD inception was instantaneous
maintaining two rounded-edge electrodes at different with the classical PD inception. The polarity of
potentials. Figure 2, shows the classical and UHF PD classical PD pulses was bipolar. The UHF PD
patterns of air corona recorded at three different test Occurrence frequency was spread (mostly below 600
voltages. The UHF PD pattern shown in figure 2 was MHz) and magnitude was highly unstable, perhaps
recorded by coupler C and represents the worst case of reflecting the nature of the defect. The classical peak
interference as the coupler was placed in open air and PD magnitude ranged from 15 pC to several hundred
less than Im from the COIOM source. pC. The notable observation was pick-up of such
signals by an UHF coupler A in oil insulation, placed
The striking observation is that in figure 2a, no UHF approximately at a distance between I and 1.5 m from
response is seen. Further investigations and repeat the discharge source, when discharge levels were over
measurements c o n f m that UHF response is almost 5000 pc.
always absent when negative polarity pulses are
absent in classical PD signatures. Peak PD magnitude To study the influence of UHF sensitivity for large
recorded by classical detector was around 70 pC. separation between the discharge source and the

350
coupler, a high frequency transmitter that emits signal Table2a Influence of Sensar type
of known amplitude and frequency was used. For first
Freq. Calib. Coupler Coupler
series of experiments, the transmitter was placed in
open air and position between the coupler and the Sig.. B A
discharge source varied. Tbe results of this study as a MHz dB 0.5m 0.5m
function of different PD couplers are presented in dB dB
Table 1. 300 34 6.1 3.5
600 31 0 0
Table 1 Influence of distance 900 39 1.7 2.5
Freq. Calib. Coupler B 1200 33 0 0
Siq.,
- .
MHz dB 1.5m 4.5m 10m Table 2b Influence of distance
dB dB dB Freq. Calib. Coupler C
300 34 13 9.9 0 Sig
600 31 8 35 0 MHz dB 3m 5m 10m
300 39 14 35 0 dB dB dB
1200 33 4 35 0 300 34 61 35 0
6 00 31 0 0 0
Coupler A 900 39 17 25 0
1200 33 0 0 0
300 34 55 22 0
600 31 5.7 2.3 0 UHF PD Measurements during Classical PD
900 39 2.5 0.6 0 Testing uf a PD Susjwt Transformer
1200 33 3.1 0.3 0
In this section, typical UHF PD patterns obtained
using coupler A - a n d B during-transformer PD
The study of this table reveals that UHF coupler
acceptance testing is reported. Simultaneous classical
directly in oil and inside the tank is less affected by PD measurements were done to study the results in
external signals than the coupler arrsngernent with comparison. Figure 3% was recorded by coupler B at
dielectric windows. The effect of external signals 1.1 Un test voltage. Classical PD level was within 300
decreases as the distance between the coupler and
pC. Our experience shows that average UHF PD
signal source increases and beyond 5m the influence is patterns are more repeatable and representative. The
very negligible for the UHF coupler directly in oil.
average UHF trace in figure 3a is practically absent
above 600 MHz and signal levels hardly exceed 5 dB.
These results suggest that for effective UHF PD In Figure 3b is shown another UHF PD pattern
measurement necessitate positioning of UHF PD
recorded by the same coupler on the same transformer
couplers every 3 or 4 rn. However, in an enclosed at test voltage of 1.5 Un. The difference between
space like inside a transformer tank multiple figures 3a and 3b is that while the former pattern was
reflections and signal attenuation may change this obtained after filtering out noise signals and the later
limit. Experimental results also suggest that UHF without. The signal peaks seen in the average trace of
pick-up for external discharges in air decrease with Fig. 3b was due to noise sources and this can be easily
distance and consequently can aid in noise free signal deduced if one compares against the noise pattern
acquisition. One publication [ I ] have reported using
recorded before transformer energization (see Fig. la).
an external antenna network to record UHF signals
Classical PD level was within 300 pC. Figure 3c
and localise the fault. To properly understand this presents a case when the measured classical PD level
technique, the transmitter was placed inside the
was around 3200 pC. This PD pattern was acquired
transformer tank and an external coupler in open air using a modified form of PD coupler A to suit
was located at different distances from the discharge
insertion into an oil drain valve. Note that peak UHF
source was used for measuring the signals. The results PD amplitude recorded was around 40 dB.
of this study are reported in Table 2a. Analyses of the
results show a very strong attenuation of the
DiseussionandCondusiou
transmitted simal and almost neelieible sienal nick-
the cOhler positioned
L I

from the;ank Experimental results suggest UHF PD inception and

Table 2b presents comparative UHF PD detection
detection sensitivity is linked to the appearance of PD
sensitivities for a coupler with and without dielectric pulse impulses of negative polarity
window. As expected UHF PD coupler directly in oil
irrespective of the simulated pD or
and inside transformer test tank is relatively more repetition rate, one may be that PD pulse
immune to the simulated external interferences.

351
width of positive polarity may be outside the UHF PD acceptance testing of transformer follows the
frequency range. Exact reason can be known only if amplitude trend of classical PD signatures and usually
time-history behaviour of partial discharge current UHF PD signals are ahsent or weak at frequencies
pulse waveforms and signal propagation exceeding 0.9 GHz.
characteristics are studied. A related observation has
been reported in the literature during UHF PD References
measurements with air cavity discharges [2]. It is
reported that Townseud type of discharges, 1. I. Pomgues, P J. Moore, I A. Glover, "North fleet
characterised by current pulses having rise times in the transformer test report", May 02.
range of 10 to 20 ns and pulse width in the range 0.4 2. A.G. Sellars, 0. Farish, et. al. "Using the UHF
to I ps do not generate UHF PD signals. Our earlier technique to investigate PD produced by defects
study 131 revealed that zero-span or point-on-wave in solid insulation", IEFE Transactions on
UHF PD detection is possible only when the ratio of Dielectrics and Electrical Insulation, Vol. 2 No.
the mean charge density (MCDR = QmEad / (&.-+) is 3, June 1995, pp. 448459.
close to unity The occurrence frequencies for PD and 3. K. Raja, F. Devaux, S. Lelaidier, "PD source
common noise sources do not appear to change with recognition using UHF PD signatures", E E E
type of the PD coupler used, although individual Insulation Magazine, vol. 18 No. 5,
sensitivities vary. UHF PD patterns obtained during SeptembedOctober 2002, pp. 8-14,2002

40

30

20

10

0
2m

Figure 3: UHF and classical PD patterns recorded during PD acceptance testing of different transformers

352