th

The 20 International Symposium on High Voltage Engineering, Buenos Aires, Argentina, August 27 – September 01, 2017

STATISTICAL ANALYSIS OF SURFACE DISCHARGES IN ROTATING

MACHINE STATOR INSULATION SYSTEM
1* 1* 2
Ramesh. P. Nair B. Nageshwar Rao and B. V. Sumangala
1
Central Power Research Institute, Bengaluru, India
2
Dr. Ambedkar Institute of Technology, Bengaluru, India
*Email: [email protected], [email protected]

Abstract: Partial Discharge (PD) phenomenon in rotating machine is one of the most
degrading mode that affects the insulation system. Even though the main insulation of
rotating machine is Mica, which is highly PD resistant, the prolonged discharges that take
place at the surface of stator bars are highly detrimental to insulation system and can
eventually lead to machine insulation failure. In this paper, PD measurements are
conducted on Mica-Epoxy-Glass laminates using a cylindrical-plane electrode system. A
distinct Phase Resolved Partial Discharge (PRPD) pattern was observed for surface
discharges in the sample-electrode arrangement. The discharge pattern in the sample is
similar to the surface discharges (end-winding discharge) pattern in actual rotating
machine insulation system. Statistical analysis of the PD parameters like charge
magnitude (q), phase angle of occurrence (ɸ) and charge frequency (n) is used to classify
the surface discharges. In-order to classify the surface discharges, a distribution is made
using the PRPD pattern. The mean discharge magnitude distribution which represents
average charge magnitude as a function of corresponding phase angle is found useful to
correlate the surface discharge with PRPD patterns. This distribution has distinct shape in
the positive and negative half cycle of voltage waveform. The shape of distribution is
described using statistical operators: skewness and kurtosis. Finally to confirm the
patterns, known surface defect was simulated in a 6.6 kV stator coils to simulate surface
discharges (end-winding discharge). PRPD patterns obtained on mica-epoxy-glass
laminate and 6.6 kV stator bars for similar type discharges were compared and found to
have similar characteristics.

1 INTRODUCTION The detection and monitoring of partial discharges

in rotating machine is very important to avoid any
The insulation system of any HV equipment is forced outage of machine during service, which
more prone to deterioration compared to its other results in huge generation/production loss. Since
component under varying stresses. In rotating the partial discharges in machine insulation system
machines, the insulation system consists of MICA- can be the cause as well as the symptom of other
main insulation, GLASS FIBRE-backing material deteriorating mechanism, the trend analysis of
and RESIN-binding material. The insulation system partial discharges will be helpful to monitor the
ages as the machine continuous to operate. The condition of machine insulation. The availability of
aging of insulation is due to normal operating digital Partial discharge detectors with Phase
stresses like thermal, electrical, ambient and Resolved Partial Discharge pattern display has
mechanical stresses [1]. These four stresses act in made it possible to classify the partial discharge
combination which results in aging of the insulation sources.
system. During transient operating conditions,
these stresses will be very high and the insulation The partial discharges that occur in the voids
ages severely, even though it is momentarily. present inside the bulk volume of the winding
insulation are commonly called internal discharges
In modern insulation system of rotating machine, [3]. Surface discharges occur on the end winding
micro voids will be present inevitably. These voids portion of the winding insulation which is outside
under sufficient electrical stress start to ionize, the slot. These discharges occur when the
leading to partial discharges in the insulation electrical contact between stress grading coating
system. Partial discharges are localized electrical and the semi-conducting coating is lost. These
discharge that only partially bridges the insulation surface discharges are popularly known as end-
between conductors and which can or cannot winding discharges. This leads to the ionization of
occur adjacent to a conductor [2]. Partial discharge the air-gap between the two coatings. It is basically
is a deteriorating phenomenon, which will a corona activity [4]. In the present work, similar
eventually lead to breakdown of insulation and surface discharge is simulated on two 6.6 kV, class
thereby the failure of machine. F stator bars and on a mica-epoxy-glass laminate.
The partial discharge patterns are compared and
statistical parameters are used to identify such used. The experimental arrangement is as shown
surface discharge. in Figure 2.

2 EXPERIMENTAL DETAILS

The details regarding the test specimen used, PD

measurement test set-up and method of simulation
of surface discharges are summarised in the
following sections.

2.1 Test specimen

In the present investigation mica-epoxy-glass

tapes (B-stage) were used to prepare the sample
plaques. The laminate is prepared by half lapping
the tape to form a single layer and the desired
sample thickness is attained by building up the
layers. The B-stage insulation sample is hot
pressed at 180°C for 30 min and cooled to room
temperature. The prepared insulation sample
thickness is 2 mm. Figure 1 shows the laminate
and cylindrical-plane electrode arrangement used Figure 2: Test set-up for PD measurement on
for PD measurement. Two 6.6 kV, Class F stator insulation sample.
bars are also used for the experiments.
For PD measurement on the stator bars, a thin
aluminium foil of thickness 0.5 mm was wrapped
along the slot portion of stator bar and grounded.
High voltage was applied at the copper conductor
of stator bar. Care is taken to minimize the voids
between the metal foil electrode and semi-
conductive coating on the bar.

2.3 Simulation of surface discharge

To simulate surface discharges (end-winding

discharge) on the stator bars, a small portion of 5
mm width of semi-conducting coating is removed
near the junction of stress grading coating and
semi-conducting coating as shown in the figure 3,
thereby both coatings are not in direct contact and
surface discharge will occur in the air-gap between
the two coatings when sufficient electric field is
available to ionize the air. In the laminate, the
sharp edge of the electrodes leads to surface
discharge.

Figure 1: Laminate and cylindrical-plane electrode

arrangement.

2.2 PD measurement test set-up

The PD measurements are conducted using

Omicron make MPD 540 digital PD detector. The
test source used is a 15 kV Phenix make PD free
voltage source. PD free 1000 pF capacitor is used
as a coupling device. The whole test arrangement
is accommodated inside the Faradays’ cage. The
measured background noise was less than 5 pC.
To conduct the PD measurement on laminate, a
cylindrical-plane steel electrode arrangement is Figure 3: Stator Bar with surface defect.
The internal discharges of smaller magnitude 4 RESULTS AND DISCUSSIONS
which are very likely to occur in these samples are
considered not to influence the surface discharge 4.1 Surface discharge on laminate
measurement.
PD measurement was carried out on laminate
sample. The PRPD pattern on this sample is as of
surface discharge is shown in Figure 4.
3 STATISTICAL ANALYSIS

Surface discharge was identified by examining the

distribution of PRPD parameters. The voltage
waveform is divided into several phase windows
and mean discharge magnitude distribution which
represents average charge magnitude as a
function of corresponding phase angle H qn (f ) is
determined [5]. These distributions in positive and
negative half cycle of voltage waveform have
distinct shapes. To describe the shape of
distribution, two statistical operators are
considered: skewness and kurtosis. Skewness
represents asymmetry of distribution.
Figure 4: PRPD pattern for surface discharge on
If distribution is symmetric, Sk = 0, if it is
laminate.
asymmetric to left, S k > 0, and if asymmetric to the
It is observed from the Figure 4 that the PRPD
right, S k < 0. pattern is asymmetric, with higher magnitude
discharge occurring on the negative half cycle of
Skewness is given by, the voltage waveform. This PRPD pattern is
different from the well known internal discharge

å (x - µ ) . pi
pattern, which is a symmetric PRPD pattern with
3

Sk =
i almost equal discharge magnitude on both positive
s3 (1) and negative half cycle of voltage waveform [6].

Where: Apart from the PRPD pattern, it is also clear from

the charge distribution plot that the charge
xi Measured value magnitude in the negative cycle is more compared
to the positive cycle. Further statistical analysis can
pi Probability of frequency of appearance of
be done on this charge distribution to classify the
that value of xi in window i discharge sources. The charge distribution plot for
µ Mean Value surface discharge is given in Figure 5.
s Variance

Kurtosis represents sharpness of distribution. If

distribution has same sharpness as normal
distribution, K u = 3. If it is sharper than normal
distribution, K u > 3 and if it is flatter than normal
distribution, K u < 3.

Kurtosis is given by,

å (x - µ ) . pi
4
i
Ku =
s4 (2)
These operators are applied to the distributions for
positive half cycle and for negative half cycle of
voltage waveform, to identify surface discharges. Figure 5: Charge distribution plot for surface
discharge on laminate.
4.2 Surface discharge on stator bars distribution plot is similar to the charge distribution
of surface discharges on insulation sample. The
Surface discharges (end-winding discharges) charge distribution is shown in Figure 8 and 9 for
pattern were also obtained on two stator bars, Bar Bar 1 and 2 respectively.
1 and Bar 2, of voltage rating 6.6 kV and class F
insulation. The PRPD pattern obtained is shown in
Figure 6 and 7. The patterns obtained on stator
bars are similar to the pattern that was obtained on
mica-epoxy-glass laminate sample described
earlier.

Figure 8: Charge distribution plot for surface

discharge on stator bar 1.

Figure 6: PRPD pattern for surface discharge on

stator bar 1.

Figure 9: Charge distribution plot for surface

discharge on stator bar 2.

4.3 Results of statistical analysis

Figure 7: PRPD pattern for surface discharge on
The distinct PRPD patterns will be useful in
stator bar 2.
identifying surface discharges but in order to
quantify the distinct nature of patterns, a mean
From Figure 6 and 7 it is clear that the PRPD
discharge magnitude distribution which represents
patterns were asymmetric for both stator bars, with
average charge magnitude as a function of phase
higher magnitude discharges occurring on the
angle is made and to describe the shape of
negative half cycle of voltage waveform. Similar distribution in positive and negative half cycles,
asymmetric pattern was obtained for surface statistical parameters like skewness and kurtosis
discharges on mica-epoxy-glass laminate. Hence,
are used.
the distinct asymmetric pattern will be useful in
identifying surface discharges (end-winding
The mean discharge magnitude distribution for
discharge) on the rotating machine insulation
surface discharges on laminate, stator bar 1 and
system.
stator bar 2 are given in Figure 10, 11 and 12
respectively.
The charge distribution for surface discharges is
plotted for both the stator bars. The charge
 In all the three distribution plots shown in Figure
10, 11 and 12, the mean discharge magnitude is
higher in the negative half cycle.

In order to describe the shape of distributions in

positive and negative half cycles, two statistical
parameters: skewness and kurtosis are calculated
for both half cycles. The calculated parameters for
surface discharges for both positive and negative
half cycles are given below in Table 1.

Table 1: Statistical parameters for surface

discharge

Surface discharge on
Statistical
Figure 10: Mean discharge magnitude distribution Parameter Insulation Stator Stator
for surface discharge on laminate. sample Bar 1 Bar 2

Positive Skewness 3.31 4.88 6.65

half
cycle Kurtosis 24.35 42.56 80.76

Negative Skewness 28.48 8.09 12.80

half
cycle Kurtosis 1112.8 114.15 246.83

From Table 1 it can be noticed that, for surface

discharges, which is an asymmetrical discharge
skewness and kurtosis are having different values
for positive and negative half cycle. The change in
the kurtosis value is significant since there is a
sharp increase in the magnitude of charge in the
Figure 11: Mean discharge magnitude distribution
negative half cycle. The variation of parameters is
for surface discharge on stator bar 1.
similar for surface discharge occurring on
insulation sample, stator bar 1 and 2.

5 CONCLUSIONS

In this paper, an attempt is made to identify surface

discharges occurring on rotating machine
insulation system, by analysing the Phase
Resolved Partial Discharge (PRPD) patterns and
statistical analysis of PD parameters. By simulating
the surface discharge on the laminate and two
rotating machine stator bars, the PD measurement
has confirmed that PRPD patterns for surface
discharges are asymmetrical, with higher
magnitude of charge occurring at negative half
cycle. Further statistical analysis of PD parameters
Figure 12: Mean discharge magnitude distribution is useful to quantify the distinct PRPD patterns.
for surface discharge on stator bar 2. Statistical parameters like skewness and kurtosis
are good tools to describe the distinct shape of
The mean discharge magnitude distribution PRPD patterns for surface discharge.
obtained for surface discharges on laminate is
identical to the distribution obtained for stator bars.
 ACKNOWLEDGMENTS
The authors thank the Central Power Research
[4] Carlos Azuaje, William Torres, “Experience in
Institute management for providing the facilities for
identification of partial discharge patterns in
testing and allowing to publish the paper.
large hydrogenerators”, CIGRE session A1-
209, 2004.
REFERENCES
[1] G. C. Stone, “Condition monitoring and [5] E. Gulski, ‘‘Discharge pattern recognition in
diagnostics of motor and stator windings- A high voltage equipment’’, IEE Proc.: Sci.,
review.” IEEE TDEI, Vol. 20, Issue.6, pp. 2073- Meas. Technol., Vol. 142, pp. 5161, 1995.
2080, 2013.
[6] Hudon, C, Bélec M, “Partial discharge signal
[2] IEC 60270-1.Ed.3.0, “High-voltage test interpretation for generator diagnostics” IEEE
techniques - Partial discharge measurements, TDEI, vol. 12, no. 2, pp. 297–319, April 2005.
International Standard”, 2010-12.

[3] IEEE Std. 1434, “IEEE Guide for the

Measurement of Partial Discharges in AC
Electric Machinery," 2014.