How to Identify Stator Insulation Problems Using On-line Partial Discharge Analysis

H. Zhu, V. Green D. Huynh L. Ravenscroft

ADWEL International PacifiCorp Nebraska Public Power District
Canada USA USA

ABSTRACT

On-line partial discharge (PD) testing of the stator the information to make operation and maintenance
winding of hydro generators, turbine generators, and decisions, take action based upon those decisions and then
large motors has been performed for half a century. repeat the process to determine the effectiveness of their
However, questions remain on how to best use PD data to efforts.
identify stator insulation problems. This paper analyses
why one cannot rely on an absolute PD magnitude to Test Equipment Data Information Decisions Actions
assess the insulation condition of rotating machines. Some
of the factors which affect the diagnosis of stator
insulation problems include the following:
Fig. 1 Closing the loop on diagnosing insulation condition.
• PD calibration problems in rotating machines
• Bandwidth of the PD detector On-line partial discharge (PD) testing has been used to
• PD types and locations provide useful information to diagnose and monitor the
• Differences among machines and PD measurements integrity of stator winding insulation of large generators and
motors for half a century [1]. A number of on-line PD
The paper describes how these factors influence the instruments and measurement systems for rotating machines
assessment of stator insulation condition. PD trend are now available. Some individuals and companies have
analysis is recommended as the best way to identify reported good success in interpreting PD data and taking
machine insulation problems and is well demonstrated by action on rotating machines [2]. However, questions remain
some case studies. Precautions in using a PD database to on how to assess stator insulation problems using PD test
assess the likelihood of insulation failures are given. The data.
paper concludes that a significant increase in PD activity
within a certain period of time is a sign of severe While on-line PD testing is widely used today for condition-
insulation deterioration. based maintenance of rotating machines, many users still ask
the question, “How do I know when my machine has a
problem?” Does a rotating machine with higher PD readings
1. INTRODUCTION fail more quickly than one with lower PD readings? Can one
use an absolute PD magnitude to turn on a red light for stator
As competition increases in the electric power generation insulation deterioration? These questions are critical to
market, utilities and other power producers are facing successful use of on-line PD testing technology.
increasing pressure to reduce the amount and cost of
maintenance and testing efforts yet increase the availability of Naturally, one expects that a criterion of an absolute PD
the power generation equipment. Consequently many power magnitude exists for evaluating stator winding insulation
generators are moving away from OEM maintenance condition. Unfortunately, unlike many other types of high
schedules due to their conservative nature and instead are voltage equipment (such as cables and switchgear), a criterion
now striving to do the right maintenance on the right for acceptable PD does not exist for rotating machines due to
equipment at the right time. PD calibration problems.

A vital step in the maintenance strategy of power generators This paper describes why an absolute PD criterion cannot be
is to “close the loop” on their stator testing and maintenance established for rotating machines. How to identify stator
efforts, as shown in Fig. 1. They need to ensure that their insulation deterioration effectively and reliably will be
efforts do not stop at purchasing suitable test equipment. The discussed and demonstrated with some case studies.
power producers must then use the instruments to acquire
data, analyze the data to turn it into useful information, use
2. WHAT INFLUENCES ASSESSEMENT OF may have a different magnitude and waveform than the
STATOR INSULATION CONDITION BASED UPON PD original PD pulse occurring within the stator winding,
MAGNITUDES? depending on the nature of the stator winding [13]. Therefore,
it is difficult to accurately assess stator insulation condition
The first step in the diagnostic procedure is to obtain PD data. using the PD magnitude measured at the machine terminals.
There are various types of on-line PD testing systems
• PD detectors.
available. Several PD instruments have been developed that
are good at detecting the PD and rejecting noise [3] [4] [5] [6] PD detectors having various frequency bandwidths and centre
[7] [8]. Common types of PD data presentation are pulse frequencies are commercially used. These PD detectors may
height analysis graphs, phase resolved graphs, normalised produce different PD output values for the same PD event.
quantity numbers (NQNs), polar graphs, etc. However, there Further more, when the PD detector is coupled to the stator
are difficulties in setting an alarming threshold based upon winding, a frequency spectrum combining the winding
the data presentation. The difficulties are caused by the frequency spectrum and the detector frequency spectrum is
following factors: produced. The combined frequency spectrum can either
increase the PD magnitude (when the detector is within the
• PD calibration problems in rotating machines
peak values of the winding frequency spectrum) or decrease
A PD test specification for rotating machines has not been the PD magnitude (when the detector is within the valley
established, although such specifications exist in most of high values of the winding frequency spectrum) [14]. After the PD
voltage equipment [9]. The reason for this is the lack of an pulse propagates through the winding and goes into the
effective calibration technique for PD measurement in detector, the original PD pulse has been distorted. The PD
rotating machines. Since PD pulses occurring at PD sources pulse appearing at the output terminal of the detector depends
cannot be directly measured at PD sites, a calibration of PD upon both the winding frequency spectrum and the detector
measurement at machine terminals is required. The frequency spectrum.
“calibration” is to establish a scale relationship between
• PD types and PD locations
readings of a PD measurement system at the machine
terminals and the actual magnitude of PD occurring at PD PD occurs at various locations in the stator winding and in
sites within the stator winding. various types. For example, PD occurs
Ø at the interface between the winding and the slot (slot
The conventional PD calibration approach, as stated in IEC
discharges);
270 and ASTM D1868, is to inject a pulse with a known
Ø within groundwall insulation;
amount of charge into the machine terminal and to determine
Ø at the interface between copper conductors and
the magnitude of the response produced by the PD
groundwall insulation;
measurement system in response to the injected pulse. The
Ø in the endwinding region.
calibration pulse with a voltage V is injected into the winding
via a calibration capacitor C. The injected charge is Q = CV.
The degree of threat to the insulation system of rotating
Then, a scale relationship between the injected charge Q and
machines depends upon the PD type and location. The same
readings of the PD measurement system can be established.
PD magnitude with different PD types and locations in the
This conventional PD calibration approach is not what is
stator winding may result in different degrees of threat to the
defined for PD “calibration” of rotating machines in the
lifetime of the insulation system. For example, the same
above paragraph.
magnitude of PD occurring within the internal insulation
(groundwall discharges) and on the insulation surface (phase-
The conventional PD calibration method is not applicable to
to-phase discharges) results in different degrees of damage to
rotating machines, since a complete stator winding cannot be
the insulation system. Hence, evaluation of stator winding
treated as a pure lumped capacitor. The stator winding
insulation by considering only the PD magnitude without
actually is a complex circuit network with distributed
considering the nature and locations of PD is not reliable.
inductance, capacitance and resistance. The simple
relationship Q = CV cannot be applied to rotating machines.
• Differences among machines and PD measurements.
When a PD pulse propagates through the stator winding, the There are considerable differences among rotating machines
stator winding can attenuate the PD pulse particularly in high in specification, manufacturing, QA and acceptance testing,
frequencies and cause resonance of the PD pulse [10] [11] installation, operation, maintenance, environment, etc. Even
[12]. The PD pulse actually occurring within the stator for the same manufacturer and the same type of rotating
winding is measured and calibrated at the machine terminals. machines, differences in installation, operation, and
Therefore, the PD pulse measured at the machine terminals maintenance may cause different degrees of insulation
degradation and result in varying PD levels. In addition, Once machine problems have been identified, the next
there are differences among measurements caused by varying question is what action should be taken. If the machine under
operating conditions (e.g. temperature, load, and humidity) test is in good condition, no action would be taken. Although
and varying measurement settings (e.g. instrument types, gain nothing is done, consciously doing nothing is different from
settings). Therefore, a direct comparison of PD readings ignorantly doing nothing. Consciously doing nothing means
among different rotating machines is difficult. that you know the machine condition and confidently chose
not to take action rather than just doing nothing out of
In summary, all of the above factors can influence the ignorance and hoping for the best. If the machine under test
absolute PD readings. Therefore, it is not reliable to give a is in bad condition, one should take further testing and
warning of insulation failures based upon the absolute PD inspection to corroborate whatever type and degree of
magnitude. insulation deterioration indicated by PD test data. For
example,
3. HOW TO IDENTIFY STATOR INSULATION
• power factor tip-up tests for internal delamination;
DETERIORATION
• wedge tightness/visual inspection for slot discharge or
Several approaches have been used to assess insulation bar looseness;
condition of rotating machines. For example, • visual inspection for endwinding discharge.
• neural networks/pattern recognition technique has been
applied to recognise PD problems[15]; 4. CASE STUDIES
The following case studies demonstrate trending of PD
• compare PD test data to a PD database; activity in assessing the insulation condition on operating
machines. The first case compares PD readings of two
This approach sets an alarm level by comparing the
“sister” motors in a power plant.
detected PD readings to PD data in a database. For
example, if the PD level is within a range of 10% of the
Motor A: Motor A is a 6.9 kV, 7000 HP induction motor
highest PD readings in the database, the insulation may
used for induced draft fan in a power plant. Motor A had
be considered bad. If the PD level is within a range of
Duraguard, class F insulation in its stator winding and was
75% of the low PD readings in the database, the
rewound in 1994. An on-line PD monitoring system from
insulation is usually regarded as good. One should be
ADWEL International was installed on the motor after
cautious when using this PD database approach to set
rewinding. The trending graphs of the NQN values and of
alarm levels since it uses the absolute PD magnitude to
the maximum PD magnitude are shown in Figs. 2 and 3. The
assess the condition of stator winding insulation.
NQN (Normalised Quantity Number) value is a measure of
total PD activity in the stator winding.
• compare PD test data with similar machines in the same
plant;
This comparison gives a relative condition among
rotating machines (i.e. which one is better and which one
is worse). However, as previously discussed, due to
differences among machines, even those with identical
design made by the same OEM, and in the same
operating and measurement parameters, this method may
not always produce reliable results.

• compare PD test data with previous readings on same

machine.
Since the major strength of on-line PD testing is the
trending ability [16], this approach is usually regarded as
the best way to diagnose insulation problems using on-
line PD data. If a significant increase in PD activity
within a certain period of time is detected, further actions
should be taken. As a general rule of thumb, PD activity
doubling in six months or less is often a sign of severe
Fig. 2 Trending of + NQN and –NQN values of motor A.
insulation deterioration.
In Fig.2 above, the NQN value in phase C increased more Motor B: Motor B is an identical “sister” motor operating at
than three times from October 1998 to March 1999. In Fig. 3 the same site. Motor B was equipped with the same on-line
below, the maximum PD magnitude in phase C increased PD monitoring system in 1994. The trending graphs of the
more than two times over the same period of time. This motor NQN values and of the maximum PD magnitude are shown
failed in service during start-up due to insulation breakdown. in Fig. 5 and in Fig. 6. Comparing Figs. 2 and 3 and Figs. 5
The motor was taken out service and a visual inspection was and 6, motor B had higher NQN values and higher PD
conducted. The visual inspection confirmed that there was a magnitudes than motor A. Yet motor B with the higher PD
puncture in the groundwall insulation leading to the activity has been operating well while motor A with the lower
insulation failure. The location of the insulation failure is at initial PD activity subsequently failed in service.
the interface between the slot portion and the endwinding
portion, as shown in Fig. 4. The motor has been rewound
since the insulation failure.

Fig. 5 Trending of +NQN and –NQN values of motor B.

Fig. 3 Trending graph of the positive maximum PD

magnitude of motor A.

Fig. 6 Trending graph of the positive maximum PD

magnitude of motor B.
Fig. 4 Visual inspection of the insulation failure.
A comparison of the maximum PD magnitude in motor A and
motor B is listed in Table 1 and Table 2. One might expect
that the motor with the higher PD readings would fail first. It
did not actually. This is because the PD activity in this motor running the generator was re-established. The utility has
was stable, even though it was relatively higher than the other performed periodic PDA testing on the generator to monitor
motor. Stable PD activity over time indicates that insulation any progress of insulation deterioration since 1993. The PDA
deterioration has not progressed much. A significant increase trend was steady from 1993 to 1999 and consistent with the
in PD activity within a certain period of time indicates that power-factor test data. A recent PDA reading taken in
the rate of insulation deterioration is accelerated (i.e. a sign of August 1999 is shown in Fig. 7 <1>. There is negative pulse
the rapid progress of insulation deterioration that could lead predominance in the curves. This is a sign of internal PD
to insulation failures). occurring closer to the interface between the copper conductor
and the groundwall insulation.
Table 1 Comparison of positive maximum PD magnitude of
motor A and motor B
Test Date Oct. 1998 Dec. 1998 March 1999
Motor A 20 36 48
Motor B 62 56 65

Table 2 Comparison of negative maximum PD magnitude of

motor A and motor B
Test Date Oct. 1998 Dec. 1998 March 1999
Motor A 0 26 30
Motor B 45 38 48

The second case involves a 730 MW, 20 kV steam turbine

generator. During a start-up in 1993 the generator was
extremely overheated due to a failure of the automatic
hydrogen cooling system. The stator winding temperature
reached 165 C° before the cooling failure was discovered and Fig. 7 Comparison of PDA readings taken in 1999 <1> and
the generator was immediately shut down. The wedges, filler in 1993 <2>.
blocks and bracing ties of both the stator and rotor were
melted or burned black beyond repair. The class B The low PD activity initially detected by the PDA system
thermoplastic stator insulation, however, did not fail. The supported the utility’s decision to continue operating the
opinion of many experts at the time was that, if restarted, the generator with the overheated stator winding. Consistent
stator winding would most likely fail and should be rewound PDA readings from 1993 to date have given the utility
immediately. The confidence to continue running the confidence to keep running the unit and to maximize its
generator was lost. lifetime. The avoided cost of the emergency rewind was
millions of dollars. This case demonstrates that insulation
After examining the data from the off-line stator insulation deterioration has not progress much if the PD trend is stable,
tests, the utility decided to install an on-line condition- that is, the threat of machine insulation failure is not great.
monitoring system for the stator insulation and continue
operating the generator with the same stator winding. A 5. CONCLUSIONS
PDA system from ADWEL International was installed during An absolute PD testing specification for rotating machines
the re-wedge outage in 1993 to enable future monitoring of cannot be established due to the lack of a valid PD calibration
the partial discharge activity in the stator winding. technique. It is not reliable to give a warning of severe stator
insulation deterioration based upon an absolute PD
The utility successfully restarted the generator and began magnitude. Therefore, one should be cautious when using a
closely monitoring the on-line PDA readings to detect any PD database to set alarm levels, since this approach is based
signs of severe insulation degradation. The first on-line PDA upon an absolute PD value. It may be useful to compare PD
readings were taken in July 1993, as shown in Fig. 7 <2>. readings of a rotating machine with those identical to it in a
The PD levels were not alarmingly high. This suggests that PD database to determine relative stator insulation condition
the insulation system was still in good shape, even though the among similar types of rotating machines.
generator had been overheated. The confidence to continue
The trending of PD activity is the best approach to identify
severe stator insulation deterioration. A significant increase [8]. Lloyd, B.A., Campbell, S.R., Stone, G.C., “Continuous
in PD activity within a certain period of time is a sign of On-Line Partial Discharge Monitoring of Generator Stator
severe stator insulation deterioration. The rate of increase in Windings”, IEEE/PES Order Number PE-388-EC-0-2-1988
PD activity is a key factor to determine if a warning should be
given. [9]. J. W. Wood, H. G. Sedding, W. K. Hogg, I. J. Kemp and
H. Zhu, "Studies of partial discharges in HV machines:
initial considerations for a PD specification" IEE Proceedings
6. REFERENCES A: Science, Measurement and Technology, Vol. 140, No. 5,
1993, p. 409- 416.
[1]. J.S. Johnson, “Slot discharge detection between coil
surfaces and core of high-voltage stator windings”, AIEE [10]. I. J. Kemp, B. K. Gupta, G. C. Stone, "Calibration
Transactions Vol. 70, 1951, pp 1993 – 1997. difficulties associated with partial discharge detectors in
rotating machine applications", Proc. 18th Electrical and
[2]. John Lyles, G. C. Stone, M. Kurtz, “Experience with Electronic Insulation Conference, Chicago, USA, October,
PDA diagnostic testing on hydraulic generators”, IEEE 1987, p. 92-97.
Transactions on Energy Conversion, Vol. 3, No. 4, December
1988, p824 – p832. [11]. H. Zhu and I. J. Kemp, “Pulse propagation in rotating
machines and its relationship to partial discharge
[3]. Kurtz, M., Stone, G.C., “Diagnostic Testing of measurements”, 1992 IEEE International Symposium On
Generator Insulation Part III – The Partial Discharge Electrical Insulation, June 7-10, 1992, Baltimore, Maryland,
Analyzer and Coupling Systems”, CEA Research Report USA. p. 411-414.
Contract RP76-17, December 1980**.
[12]. J. T. Holboll and M. Henriksen, “Frequency-dependent
[4]. Fruth, B.A., Gross, D.W., “Partial Discharge Signal PD pulse distortion in rotating machines”. IEEE International
Conditioning Techniques for On-Line Noise Rejection and Symposium on Electrical Insulation, Montreal, Canada, June
Improved Calibration”, 1996 IEEE International Symposium 16-19, 1996. p. 192-196.
on Electrical Insulation, Montréal, June 16 – 19, pp 397 –
400, IEEE 96CH3597-2. [13]. I. J. Kemp, H. Zhu, H. G. Sedding, J. W. Wood, W. K.
Hogg, "Towards a new partial discharge calibration strategy
[5]. Timperley, J.E., “Electrical Spectrum Analysis of based on the transfer function of machine stator windings."
Operating Hydroelectric Machinery”, EPRI Workshop on IEE Proceedings A: Science, Measurement and Technology,
Rotating Machinery Insulation, EPRI EL-2211, December Vol. 143, No. 1, January 1996.
1981.
[14]. H. Zhu, "Analysis of partial discharge calibration
[6]. Itoh, K., Kaneda, Y., Kitamura, S., Kimura, K., difficulties in HV rotating machines", 10th International
Nishmura, A., Tanaka, T., Tokura, H., Okada, I. “New Noise Symposium on High Voltage Engineering, Montreal, Canada,
Rejection Techniques on Pulses by Pulse Basis for On-Line August 1997.
Partial Discharge Measurement of Turbine Generators”,
IEEE Transactions on Energy Conversion Vol. 11 No. 3, [15]. Hoof, M., Lanz, S., “PD Diagnostics on Rotating
September 1996, pp 585 - 594. Machines – Possibilities and Limitations”, Electrical
Insulation Conference, Cincinnati, OH, October 26-28, 1999
[7]. Frogner, M., Hutter, W., Ruhe, J., “Experiences with
Different Discharge Monitoring Systems on Rotating [16]. IEEE P1434: Draft IEEE Guide to the Measurement of
Electrical Machines”, Nordic Insulation Symposium NORD- Partial Discharges in Rotating Machinery.
IS 92, Västeras, June 15 – 17, 1992, 3.1.