A NEW TECHNIQUE FOR THE REMOTE PARTIAL DISCHARGE

MONITORING OF THE STATOR INSULATION OF HIGH-VOLTAGE

MOTORS LOCATED IN Ex (HAZARDOUS) LOCATIONS
Copyright Material IEEE
Paper No. PCIC
Lee Renforth

Russell Armstrong

David Clark

Steven Goodfellow

Paul. S. Hamer

Member, IEEE
Managing Director
HVPD Ltd
Manchester, UK
[email protected]

E&I Reliability Supervisor

Tengizchevroil
Tengiz, Kazakhstan
erelengr@tengiz
chevroil.com

Student Member, IEEE

PhD Student
Manchester University
Manchester, UK
[email protected]

PD Test Engineer
HVPD Ltd
Manchester, UK
[email protected]

Fellow IEEE
Senior Consulting Engineer
Chevron Energy Techn. Co
Richmond, CA, USA
[email protected]

Abstract The authors present a paper on the application

of a new technique for the remote on-line partial discharge
monitoring of the stator insulation condition of in-service,
high-voltage rotating machines. The technique applies
wideband, ferrite-based high frequency current transformer
sensors and high resolution measurement technology. This
remote partial discharge monitoring technique has significant
advantages when monitoring motors which are located in Ex
hazardous gas zones in oil and gas and petrochemical
facilities. The paper includes a technical review of published
papers covering some of the history of the development of
modern, on-line partial discharge sensors and measurement
systems for rotating machines. This is followed by a
comparison of the various partial discharge sensor options
available and an introduction to the remote wideband partial
discharge monitoring measurement technique employed by
the authors. A case study from a recent pilot project (August
2011) where the new remote partial discharge monitoring
measurement techniques were successfully trialed to
measure partial discharge activity of in-service 10 kV motors
in an oil processing facility is also presented.

over 50 years [1]. This paper includes a brief technical

review of some selected published papers on this topic along
with the sensor options available for PD monitoring of
rotating HV machines.

Index Terms Partial Discharge, On-line Partial Discharge,

Phase Resolved Partial Discharge, PD wave shapes, PD
Diagnostics, Ex motors, rotating machines, hazardous gas
zones.

The main benefits in applying the OLPD technology are:

To improve the reliability of the networks by identifying

sites of insulation degradation before they fail.

To provide insulation condition data to provide an early

warning against incipient insulation faults.

To provide plant condition data to support reliable life

extension programs of ageing networks.

To avoid unplanned outages and minimise downtime.

To provide qualitative information for a condition-based

motor replacement and repair programmes.

I.

INTRODUCTION

The on-line partial discharge (OLPD) assessment of inservice high voltage (HV) plant is now becoming more
widespread in the oil and gas and petrochemical process
industries. Particular focus is made on the condition
assessment of the stator insulation of rotating high-voltage
motors which drive the critical processes. The on-line partial
discharge (PD) sensors, test and monitoring techniques
described in this paper have been applied to a wide range of
utility and industrial HV plant owners by the authors in test
projects from around the world over the past 15 years. Many
of these projects have been initiated following the
experience of an in-service HV insulation failure which has
resulted in loss of business revenue to the client. Partial
discharge testing of HV rotating machine stator windings has
been used to assess the condition of the stator insulation for

The paper includes a review of wideband high frequency

current transformer (HFCT) sensors that can be located in
the machines cable box or placed remotely from the
machine under test (at the central switchgear lineup). This is
made possible by the low-frequency response (down to
around 100 kHz) of the HFCT sensor, providing a
measurement range of up to around 1.0 km in polyvinyl
chloride (PVC) insulated cables, 1.5 km in paper insulated
lead covered (PILC) cables and around 2.5 km in cross
linked polyethelene (XLPE) cables. These varying
measurement ranges for different cable insulation types are
due to the different dielectric properties, permittivity and
geometry of these three main types of power cable [2].

II.

DRIVERS FOR APPLYING ON-LINE PARTIAL

DISCHARGE TESTING

Oil and gas supply chain operators worldwide are

increasingly deploying OLPD technology to test, monitor and
manage a wide range of HV plant including power cables,
switchgear, transformers, and rotating HV machines [3]. The
business drivers for carrying out OLPD testing and
monitoring of the HV networks include:

Health and Safety most diagnostic OLPD test projects

are carried out further to insulation faults and/or
equipment component failures occurring within an HV
network. In this case there is an immediate requirement

to carry out condition spot tests on sister plant to ensure

there is no immediate risk of failure and thus danger to
staff or the public.

To support Reliable Life Extension projects as many

in-service HV machines are reaching the end of their
ascribed design-life (typically of 20-25+ years) operators
have important asset management decisions to make
as to whether to extend the life of these existing assets
(using condition-based maintenance (CBM) techniques)
or to replace the machines with new units (time-based
replacement).
To Avoid Unplanned Outages and Downtime
generation, petrochemical and industrial process
operators are focussed on the effect of any plant failure
induced downtime on their generation capacity, process
or service. This is the strongest financial driver for these
production and process industries.

Studies on power generation, petrochemical and process

customer networks show that the main cost driver which
dominates is to avoid any unplanned outages. This is simply
due the cost of an interruption to the process being normally
much higher than the capital replacement/repair cost of any
cable/plant/machine in the network.
Oil and gas and chemical processing facilities have HV
networks which support critical processes with potentially
large losses to business for unplanned outages. Such HV
plant owners have little difficulty in justifying the cost of
regular PD diagnostic measurements or complete-system,
permanent PD monitoring installations.

III.
THE HISTORY OF PD TESTING OF
ROTATING HIGH VOLTAGE MACHINES
As stated in the introduction section of this paper, the
condition assessment of stator insulation using PD testing
has been used for over 50 years. Pioneering developments
in Canada/USA (using HV coupling capacitor (HVCC)
sensors, Resistive Temperature Detector (RTD) sensors
and Stator Slot Couplers (SSC) sensors) and in the
UK/Europe (using both HVCC and air-cored, Rogowski Coil
(RC) sensors) in the 1960s showed that it was possible to
make effective measurements of PD activity in the HV stator
windings of the machines [1]. Over the past 50 years, the PD
testing and monitoring of rotating machines has been
discussed by many authors as a key tool to understanding
the causes of stator insulation failure. Today, the continuous
monitoring of PD activity in rotating machines has now
become widely accepted as an effective method to identify
sites of localised damage or degradation ahead of scheduled
preventative maintenance outages.
Historically, periodic (typically annual) PD spot testing of
rotating machine stator insulation has been used to provide
long-term trending of PD activity. Whilst useful for the
detection of long-term insulation deterioration due to ageing,
these annual spot tests can sometimes fail to diagnose
certain load-related, environmental (temperature/humidity),
machine duty-cycle or seasonal-related insulation problems.

Permanent PD monitoring of rotating machines overcomes

many of the limitations of the annual, PD spot-tests, by
providing continuous monitoring of the PD activity under
normal (and abnormal) operating conditions. As a result of
this, on-line continuous diagnostic PD monitoring over time
has now become accepted as the most effective means of
providing the necessary condition data suitable for planning
and scheduling of preventative maintenance activity to
support CBM strategies.
The relevance of PD monitoring within rotating machines is
discussed by Stone et al in [4] and [5] as an effective
technique to determine the condition of the stator winding
insulation. In particular cases the authors use pre-installed
80 pF, HVCC sensors, one or more per phase, to monitor
both phase-to-phase and phase-to-earth PD activity.
Interpretation of the severity of any PD activity has
traditionally been carried out by considering the peak PD
level (typically given in mV) and also the number of PD
pulses across the 50/60 Hz power cycle.
Analysis of PD measurement severity by many authors [6]
considers the energy density of the PD activity across the
power cycle as the most effective way to measure the PD
severity. A number of authors [1, 4, 5, 7] also refer to the
effectiveness of comparing measured PD activity from
similar types and voltage-classes of HV cables, plant and
rotating machines through the use of databases and
comparing measurements against other similar plant. The
relevance of the rotating machines PD database and
comparing PD measurements between similar classes of
machine has been reported in [4, 5] where the authors
reviewed peak PD level, PD energy density, winding age,
and the voltage class of the rotating machine to analyse how
these parameters affect PD levels and the reliability of the inservice rotating HV machines.
With regards to the application of the OLPD monitoring of
HV motors located in Ex hazardous gas zones, a recent
study by S. Haq et al in [3] discussed minimum discharge
levels required to produce a possible spark risk. This study
was made on rotating machines operating within different
gas groups and reports that static charge within the stator
winding should be kept below 10 nC for a machine operating
in gas group B (Hydrogen) to minmise the risk of gas
ignition (10 nC equates to a minimum ignition energy of
approximately 0.019 mJ). This provides a benchmark,
maximum level for static charge in such machines whilst also
emphasizes the importance of monitoring partial
dischargeactivity in this type of motor. It should be noted that
whilst PD is not a purely static charge in this sense, it is
known that PD activity can lead to active charge surface
tracking of 100 nC+ (100,000pC) in extreme cases.

IV. MODERN ON-LINE PD TESTING AND

MONITORING TECHNOLOGY
There are now a wide range of technologies available in the
market for OLPD testing and monitoring of in-service HV
cables, plant and rotating machines. These range from
simple-to-use, handheld PD screening test units (for

screening large numbers of HV assets quickly and easily) to

continuous PD monitoring technology. It iss now becoming
increasingly popular for rotating HV macchine owners to
install permanent PD sensors and monito
oring systems at
suitable points in the HV networks which drive their industrial
and gas refining.
processes, chemical processing and oil a
The OLPD technology is used to provide a
an early warning
against insulation failure through the detection of incipient
insulation faults i.e. faults yet to occur. PD detection is now
commonly regarded as the best indicattor of insulation
degradation of in-service HV cables and pla
ant. By combining
PD screening tests (Phase 1), PD diagnostic spot-tests
monitoring (Phase
(Phase 2) and extended, continuous PD m
3) technologies, it is possible to carry out:

of HV cables and
Routine, regular walkby assessments o
rotating plant using simple, handheld PD
D test units.
Diagnostic PD testing and location o
of the PD site(s)
using portable diagnostic spot-test techn
nology.
Permanent monitoring of key substatio
ons, circuits and
rotating machines on the network to tre
end PD activities
with time/load cycle changes

V. PARTIAL DISCHARGE SENSO

OR OPTIONS
There are a number of sensor options ava
ailable for the online detection of PD activity in cables, switchgear,
transformers plant and rotating machines.. These sensors
include SSC, RTD, HVCC, HFCT, RC, and
d Transient Earth
Voltage (TEV) sensors. The four main types of on-line PD
sensors used are shown in Table I. The wid
deband frequency
response of the three sensor types used to
o test rotating HV
machines is shown in Fig. 1.
It can be noted from Table I that at 10 MHz, the HVCC
he HFCT sensor,
sensor is the most sensitive, followed by th
the TEV sensor and then the RC sensor. Fig. 1 shows the
frequency response along with the approxximate spectra of
PD activity occurrence from 10 kHz to 100
0 MHz. The most
suitable sensor solution for any application
n will depend on
the cable/plant/machine to be tested and tthe most suitable
point of attachment (POA) for the sensor on the network.

Fig. 1 Normalised Frequency Response of four types of PD

um
Sensor with PD Pulse Spectru

TABLE I
PARTIAL DISCHARGE SEN
NSOR OPTIONS
PD Sensor Options
Sensor

High Voltage
Coupling
Capacitor
Ferrite-cored
High
Frequency
Current
Transformer
Transient
Earth
Voltage
Rogowski
Coil

Picture

Coupling
C
Method

Relative
Sensitivity
at 10 MHz

Capacitive
C

100

Inductive
I

30

Capacitive
C

Inductive
I

nter frequency and wide

With reference to Fig. 1, the cen
bandwidth of the new HFCT sen
nsor ensures that lower
frequency PD signals, which hav
ve been attenuated by
dispersion within the power cable and propagation through
n still be detected with
the machine stator windings, can
sufficient sensitivity to make reliable measurements. The
HFCT sensor can also detect highe
er (> 10 MHz) frequency
components of the PD signal which appear closer to the
uch as in the switchgear
point of origin of the PD activity su
cable box or machine terminal box. This wideband frequency
he sensor can be located
response of the HFCT means that th
remotely from the rotating machin
ne under test whilst still
being able to detect sufficient PD pulse energy to ensure
that the PD pulse is represented ac
ccurately in the time and
frequency domains. This allows for the discrimination
n and pulse wave-shape.
between PD types based on location
sed in the IEEE and IEC
It should be noted that, as discuss
standards [8] and [9] (sections 11 and 5 respectively), the
s always exists due to
risk of misinterpreting PD signals
interference exhibiting similar cha
aracteristics as the PD
signals. These IEEE and IEC stan
ndards discuss time and
frequency domain methods of noise separation, types of
interference that can be expected in OLPD testing, and the
importance of distinguishing between the origins of the PD
activity. The remote monitoring technique developed by the
sification knowledge rules
authors has been built on PD class
described in [6] and [7] that use tim
me and frequency domain
parameters to distinguish betw
ween PD types and
interference. This new, remote PD monitoring
m
technique with
HFCT sensors located at the central switchboard can also
reduce the risk of Variable Speed Drive (VSD) and inverter
g classified as noise, as
drive pulses at the machine being
such pulses are attenuated by the low-pass filtering effect of
the power cable from the machines under test.
A brief description of the four main types of sensor used for
V cables, switchgear, bus
the OLPD detection of in-service HV
duct, transformers and rotating machines is given below.

A.

High Voltage Coupling Capacitor Sensors

A typical installation of three HVCC sensors, one per phase

in a generator terminal box is shown in Fig. 2. These
sensors require a galvanic connection to the HV terminals of
the machine and provide the greatest measurement
sensitivity of all the PD sensor options when installed here.
HVCC sensors of different voltage and capacitances are
available in the marketplace ranging from 6.6 kV to 36 kV in
voltage with typical capacitance ratings available of 80 pF,
500 pF, 1 nF and 2 nF. High voltage coupling capacitor
sensors are recommended for PD monitoring of larger
rotating machines (e.g. 20 MW+ at 11 kV) and are now
widely employed to monitor PD activity in HV generators and
larger HV motors. A limitation in the use of these sensors is
sometimes seen with smaller motors (<5 MW) supplied with
concentric neutral, 3-core triplex cables as it is sometimes
difficult to fit the HVCC sensor in the smaller cable boxes on
these machines. In such cases HFCT sensors are preferred
as they require less space to fit.

the lower frequency PD pulses that have travelled down the

HV cable from the machine.

Fig. 3 Permanent HFCT sensor inside cable boxes

Left: HFCT sensor around Cable + Earth/Drain.
Right: HFCT sensor on Earth Drain Wire
It is known that attenuation is a function of frequency [10],
which for a PD pulse is directly related to the distance
propagated from the machine stator winding. With their low
frequency response and wide measurement range, HFCT
sensors can be used to monitor PD activity in the stator
windings of rotating machines by connecting the sensor at
the central switchboard at up to 2.5 km/1.5 miles from the
rotating machine under test in XLPE cables, as illustrated
below in Fig. 4.
Effective Measurement Range:

Fig. 2 Permanent On-line High Voltage Coupling Capacitor

installation in an large 11 kV Generator Terninal Box
B.

Up to 1.0 km for PVC

Up to 1.5 km for PILC
Up to 2.5 km for XLPE

High Frequency Current Transformer Sensors

Split-core, ferrite HFCT sensors have now become the de

facto sensor of choice for the on-line PD testing of in-service
HV cables. The HFCT works inductively to detect PD
currents in the either the HV cable earth/drain wire or in the
HV cables conductor. Permanent HFCT sensor installations
such as those shown in Fig. 3 are also now becoming more
popular with HV plant owners for the application of PD
monitoring of smaller rotating machines (sub 10 MW) due to
their ease of installation and lower cost compared to the
more conventional HVCC sensors. The HFCT sensors are
installed inside the terminal box and are installed to intercept
the PD current on the conductor of each phase (i+) or the PD
current on the earth drain/electrostatic shield (i-). Another
advantage of the HFCT sensor, as discussed in the previous
section of this paper, is that they have a suitable low
frequency response (down to approximately 100 kHz, as
shown in Fig. 1) whilst also being capable of detecting high
frequency PD signals (up to approximately 30 MHz+).
Due to this wideband frequency response (from 100 kHz to
30 MHz), the HFCT sensors are suitable for permanent
installation within either the machine terminal box or
switchgear cable box. Installation of these sensors at the
switchgear cable end enables remote PD monitoring of
rotating HV machines to be made as they are able to detect

Up to 2.5 km

Fig. 4 Measurement range for an HFCT sensor connected

at the switchboard cable end for the three main cable
insulation types (PVC, PILC, XLPE)
C.

Rogowski Coil Sensors

Air-cored, RC inductive sensors have been used in the UK

power generation, oil and gas and petro-chemical industries
to monitor PD in rotating machines for many years. They
were originally adapted for the application of monitoring PD
activity in large generators by scientists at the UKs Central
Electricity Generating Board (CEGB) in the 1960s. These
RC sensors have also been used in the UK North Sea
offshore oil and gas industry and a number of designs have
become Atmosphere Explosive (ATEX) approved for
installation within hazardous gas zones. They are normally
permanently installed in the cable boxes of the motor or
generator and are located on each phase, around the HV
cable cores. Whilst proven to have a very low spark-risk, the
main drawback to the RC sensor is that it has a low

sensitivity to PD signals, typically around 30 to 100 times

less than the HFCT and HVCC sensors respectively. This
means that only very significant levels of PD activity are
detectable by the RC sensor with the early stage detection of
PD activity more difficult due to the poor signal-to-noise ratio
of this sensor.

TABLE II
PD LEVEL GUIDELINES FOR HV ROTATING MACHINES
Assessment

D.

Transient Earth Voltage Sensors

Transient Earth Voltage sensors are used on the outside of

the metal-clad plant under test and are used predominantly
to test HV switchgear for internal PD activity. The occurrence
of PD within the HV insulation system induces a high
frequency voltage pulse (the so-called transient earth
voltage) on the inner surface of the earthed housing. These
electromagnetic signals emerge onto the outer skin of the
switchgear through breaks in housing such as vents, joints
or seams. The TEV sensors are thus placed on the outside
of the switchgear panel to capacitively couple to these
induced PD signals originating from inside.
There are a number of handheld PD screening devices in
the market which incorporate TEV sensors which have been
shown to be very useful in identifying and locating PD
activity in HV switchgear panels and bus ducting (example
shown in Fig. 5). Handheld PD test technologies for this type
of simple look-see PD screening test are now becoming
more popular and can now also be used for the testing
rotating HV machines which are equipped with built-in
permanent PD sensors (HVCC, HFCT or RC sensors).

V. PD LEVEL GUIDELINES FOR HV ROTATING

MACHINES
Table II gives guidelines for condition assessment vs PD
levels for measurements on rotating machines using both
HFCT sensors (PD levels in pico-coulombs (pC)) and HVCC
sensors (PD levels in mV) for rotating machines in the 10-12
kV voltage class. The PD level guidelines in Table II are
based on the authors own experience and also information
published in the IEEE Transactions on Industry Applications,
see [3, 4].

PD Level (mV)
HVCC sensors

Excellent

< 2000

< 20

Good

2000 4000

20 40

Average

4000 10000

40 100

Still Acceptable

10000 15000

100 250

15000 25000

250 600

> 25000

> 600

Inspection
Recommended
Unreliable

Fig. 5 A handheld PD test unit with TEV sensor

shown testing a switchgear panel from the outside

PD Level (pC)
HFCT sensors

Whilst the guideline PD levels given in Table II provide a

good basis for the initial assessment of the condition of the
machines stator insulation, it should be noted that the PD
condition assessment of rotating machine stator insulation
should not be based on PD magnitude levels alone. It is the
trend in the PD activity which is generally considered as
more important.
Alongside PD levels, it is thus also important to measure the
PD magnitude trend over time including trends in:

Average PD Pulse Charge Content, Q (in pC or mV).

Total PD Activity across the power cycle i.e. the Energy
Density measurement (nC/cycle or mV levels multiplied
by N - number of pulses, the NQN factor).

VI. WIDEBAND ON-LINE PD TEST TECHNIQUE

In order to carrry out fully diagnostic PD testing in the field it
is necessary to apply wideband PD diagnostic test and
continuous PD monitoring technology. This requires PD
measurements based around wideband (0 - 400 MHz) test
units, 4-channel synchronous data acquisition which can
also provide high high-resolution Analogue-to-Digital
conversion of a minimum of 100 Mega-Samples per second
(MS/s) for monitoring and up to 500 MS/s for diagnostic
testing and PD site location. Such high resolution A to D
hardware allows for entire 50/60 Hz power cycles to be
sampled synchronously on all three phases of the HV
machine at once to an accuracy of 1 sample per 2 ns with all
channels synchronous to within 2 ns.
High resolution capture of the PD signals in this way can
then be followed by data analysis using software algorithms
to categorise and analyse the PD pulse data collected.
Software data analysis techniques employed can use pattern
recognition techniques including the well-established PhaseResolved Partial Discharge (PRPD) plots across the 50/60
Hz power cycle.
More advanced PD signal extraction techniques are now
also being employed which utilise pulse waveshape analysis
to separate and analyse the PD pulses from noise and other
interferences whilst also differentiating phase-to-earth and
phase-to-phase discharge activity.

CASE STUDY EXAMPLE

The HVCC type of sensor, mounted within the terminal box,

has been used at the facility for a number of years to monitor
the stator condition of the gas-turbine generators on the site.
This conventional approach has been applied as the gas
turbines are located in a non-hazardous location. However,
as most of the HV motors at the facility are located in Ex
(hazardous gas) locations, an alternative PD monitoring
solution was sought. A solution using wideband HFCT
sensors was thus proposed, since this allowed the PD
sensors to be located at the switchgear end of the cable, in a
non-hazardous area. A large part of this work involved the
measurement of the PD pulses at each end of the PVC
motor cable feeders (which were around 350 m in length) to
take into account the attenuation of the high-frequency
partial discharge signals as they passed along the cables. It
is known from previous on-line PD tests that these PVCinsulated cables significantly attenuate PD pulses (far more
than XLPE cables for instance) and it was therefore
important to evaluate whether the PD pulses could be
measured reliably at the switchgear end of the cable.
A pilot test was carried out on two motors: Motor 1: a 10 kV,
800 kW four-pole induction motor driving a pump, and Motor
2: a 10 kV, 2400 kW two-pole induction motor driving an air
blower. The first tests carried out involved injecting
calibration pulses at the motor terminal boxes to calibrate the
pulse propagation time along the cable and the pulse
attenuation coefficients of the cables. Secondly, on-line PD
data was collected from both the switchgear end and motor
terminal end of the cables with the motors running using
clamp-on HFCT sensors. The PRPD plots from the
measurements made at the switchgear end of the cable
tests are shown below in Fig. 6. The plots are referenced to
the blue phase 60 Hz power cycle voltage (a sine wave from
0 to 360 degrees).

10,000
PD Magnitude (pC)

A crude oil producing field and treatment facility in

Kazakhstan began operation in 1993, and has grown
significantly since. There are approximately 100 HV motors
at the plant rated 10 kV presently in operation, with about
half of these now nearing an in-service age of 20 years. The
site conditions are severe, with the ambient air temperature
ranging from -30C to +45C. Additionally, almost all of the
motors are Ex-rated for continuous operation in hazardous
gas zones (refer to [3] for definitions of Class, Group,
Division and Zones related to hazardous environments). The
motors are critical to production operations with limited
spares maintained at the site. As the motors are reaching
an in-service age of 20 years, a decision was made by the
operator to employ on-line partial discharge analysis PD
diagnostic testing to measure the condition of the stator
windings to ensure that the motors will continue to work
reliably.

Machine PD

5,000
0
-5,000
-10,000
0

90
180
270
Phase of Pow er Cycle (deg)
Machine PD

360

90
180
270
Phase of Pow er Cycle (deg)

360

40,000
PD Magnitude (pC)

VII.

20,000
0
-20,000
-40,000

Fig. 6 PRPD Plots from PD Measurements of Motors

measured at the Switchboard
Top Motor 1 Phase to Earth PD Activity
Bottom Motor 2 Phase-to-Phase PD Activity
The results presented in Fig. 6 showed that it was possible
to reliably measure the PD pulses at the switchgear end of
the motor feeder cables using the wideband HFCT sensors.
Comparison of the measurements made at each end of the
cable gave an average PD Pulse Retention Factor (PPRF) of
the pC content of the pulses of around 0.50 (this being the
ratio of the switchgear end PD to the motor end PD). The
PRPD plot for Motor 1 on the top of Fig. 6 shows
predominant phase-to-earth PD activity which indicates slot
section discharge activity on the blue phase of motor. This is
characterized by there being predominant signals on this
blue phase, which are not seen on the other two phases at
the same point in time. Conversely, the PRPD plot for Motor
2 on the bottom of Fig. 6 shows predominant phase-tophase PD activity which indicates end winding PD activity in
this motor. This is characterized by there being equal and
opposite signals observed on two phases at once.
The motors were tested with a high-specification, digital
discharge detector based around a fast digital storage
oscilloscope (DSO) with a wide bandwidth (0-400 MHz) and
high-resolution data capture (100 MS/s). This made it
possible to analyse the PD activity based on individual pulse
wave-shape analyses of the high resolution measurements
of all of the individual PD pulses detected.

Fig. 7 PD Pulse Shark-fin Wave Measurement Parameters

The basis for this pulse wave-shape analysis was first
developed in 2005 [11] and is illustrated in Fig. 7 which
shows the time-domain pulse parameters measured for an
individual PD pulse. The pulse wave-shape analysis uses
measurements of the pulse rise-time (measured between 10
90% of the rising edge of the pulse), the pulse width (time
between 50 50% of the rising and falling edges) and the
pulse fall-time (90 10% on the falling edge). These
measurements, combined with the frequency measurement
of the pulse are fed through the knowledge-rule based
analysis algorithm which is based around the PD pulse
analysis of the shark-fin shape as it passes through the HV
network from its source.
Using the parameters from the pulse as shown in Fig. 7 it is
possible to incorporate the Transfer Impedance of the HFCT
sensor ( ) and some basic calculus to derive the charge
content ( ) of the pulse using (1).
1

Where the limits of integration;

and
define the time of
the PD event waveform. The charge content is effectively the
area underneath the PD pulse. The PD pulses shown in
Fig. 8 and Fig. 9 below are from measurements made with
HFCT sensors on the two 10 kV motors where the PD
pulses have been measured at both the motor and
switchgear ends of the 350 m, PVC insulated motor feeder
cables.
The three different lines in Fig. 8 and Fig. 9 represent the
three phases of the motor/cable circuit with the PD signals
detected using a split-core HFCT sensor connected around
phase. The PD pulse parameters for a phase-to-earth PD
signal measured at each end of the cable (from Motor 1) are
shown in Fig. 8 and for a phase-to-phase PD signal (from
Motor 2) in Fig. 9. By comparing the pulse wave-shape
measurements made at each the cable, the effects of
attenuation and dispersion of the pulses as they travel along
the cable feeder can be assessed. In this case it can be
noted that this leads to a general broadening of the pulses.

Fig. 8 PD Measurements at each end of Motor 1 Cable

Top Phase-to-Earth PD Measured at the Motor Terminal
Bottom Phase-to-Earth PD at the Switchgear End

Fig. 9 PD Measurements at each end of Motor 2 Cable

Top Phase-to-Phase PD at the Motor Terminal
Bottom Phase-to-Phase PD at the Switchgear End
The PD pulse parameters for the individual PD pulses shown
in Fig. 8 and Fig. 9 above are provided in Table III and Table
IV respectively. The measurements of the shark-fin PD
pulses have been made using the measurement parameters
shown in Fig. 7. The characteristics of the individual PD
pulses illustrate how the PD pulse emanating from the motor
stator attenuates in magnitude and disperses (pulse
broadening) as it travels down the cable to the switchgear
end. The measurements made at each end of the cable
agreed with previous PD measurements on cable-fed HV
plant carried out by the authors and also with the data
reported in [12].
TABLE III

PHASE-TO-EARTH PD PULSE PARAMETERS MEASURED AT

THE HV MOTOR AND SWITCHGEAR END OF THE FEEDER
Motor-End PD
Switchgear-End PD
Measurements
Measurements
Parameter of
PD Pulse
Phase
Phase
L1
L2
L3
L1
L2
L3
Rise Time
29.4
49.2
58.3
150.5
122.4
166.9
(ns)
Fall Time
59.6
54.8
47.7
192.2
322.3
306.7
(ns)
Pulse Width
90
100
100
340
470
480
(ns)
Charge
3.84
5.18
10.93
2.56
3.79
5.03
(nC)
Amplitude
55.6
57.6
409.6
26.9
24.6
54.4
(mV)

TABLE IV
PHASE-TO-PHASE PD PULSE PARAMETERS MEASURED AT
THE HV MOTOR AND SWITCHGEAR END OF THE FEEDER
Machine-End PD
Switchgear-End PD
Parameter
Measurements
Measurements
of PD
Phase
Phase
Pulse
L1
L2
L3
L1
L2
L3
Rise Time
88.2
24.9
50.4
198.3 214.8 192.5
(ns)
Fall Time
95.6
74.6
88.2
291.0 285.4 278.3
(ns)
Pulse
190
100
150
580
550
510
Width (ns)
Charge
20.48
3.97
19.65
10.05
1.62
9.66
(nC)
Amplitude
415.98 160.0 543.98
73.0
11.52 73.60
(mV)

A PPRF for the cable/motor was used to take into account

signal attenuation along the cable feeder and thus convert
the remote PD measurements to the at machine-end
equivalent. As stated previously, the PPRF for any cable
feeder depends on length of the cable, its geometry and the
type of cable insulation used (PVC cables produce the
greatest pulse attenuation followed by paper-insulated
cables whilst XLPE cables produce the least attenuation).
In this project it was possible to carry out calibration pulse
injection tests to compare against the PD pulse analysis
summarized in Tables III and IV. This was possible to do
with the motor de-energized with low-amplitude (60 mV)
phase-to-earth and phase-to-phase calibration pulses
injected onto the stator windings of the motors at their
terminal box. The charge content of the injected calibration
pulses were measured at each end of the cable and the
results from the calibration testing of Motor 1 are shown in
Fig. 10. This shows the pulse injected onto the motor
terminals in the top trace and the same pulse measured at
the switchgear end of the cable in the bottom trace. The
feature observed at approximately 6 s on the timebase is
the reflection of the pulse from the end of cable. It can be
noted that inversion of the reflected pulse is due the
impedance of the stator winding being less than that of the
feeder cable, and hence the reflection coefficient being
negative.

The parameters of the calibration pulse injections on Motor 1

for both phase-to-earth and phase-to-phase pulses are
summarised below in Table V. The same calibration tests
were also carried out on Motor 2
TABLE V
PULSE PARAMETERS FROM MOTOR 1 END CALIBRATION
PULSE INJECTIONS MEASURED AT BOTH CABLE ENDS
Phase-to-Earth Cal.
Phase-to-Phase Cal.
Pulse Injection
Pulse Injection
Parameter
At
Measured
At
Measured
Machine
at Swgr
Machine
at Swgr
Rise Time
124.4
219.4
93.2
119.8
(ns)
Fall Time
77.6
223.8
89.4
220.6
(ns)
Charge
0.82
0.37
2.46
1.43
(nC)
Pulse
278.6
451.4
224.6
442
Width (ns)
Amplitude
19.52
3.56
53.12
22.54
(mV)

The calibration tests on Motor 1 shown above in Table V

yielded a PPRF for phase-to-earth pulses of 0.45 and a
PPRF of 0.58 for phase-to-earth pulses along 350 m of PVC
motor cable feeder.
Calibration test results on Motor 1 and Motor 2 were
compared with the results from the OLPD measurements
made at each end of the motor feeder cables these are
summarized in Table VI. It can be noted from Table VI that
there is a good correlation between the average PPRF
values obtained by direct calibration pulse injection and the
PPRF values calculated from the on-line PD measurements.
The calibration PPRF of 0.45 for phase-to-earth PD on Motor
1 was in close agreement to the average PPRF (0.62)
calculated from the on-line PD testing. Likewise the
calibration PPRF of 0.58 for phase-to-phase PD on Motor 2
was in close agreement to the average PPRF (0.46) from the
PD pulse measurements. On average, the 350 m-long PVC
cables had a PPRF of around 0.50 i.e. 50% of the PD pulse
energy is retained at the switchgear end of the cable.
TABLE VI
PULSE RETENTION FACTORS FROM PHASE-TO-EARTH AND
PHASE-TO-PHASE PULSE INJECTIONS
PPRF Ph-to-Earth, Motor 1
PPRF Ph-to-Ph, Motor 2

Fig. 10 Pulse Injection Waveforms Motor 1

Top Pulse Injected at Motor Terminal
Bottom Measured at Switchgear

Cal
Pulse
Injection

L1

L2

L3

L1

L2

L3

PD

Cal
Pulse
Injection

PD

PD

PD

PD

PD

0.45

0.66

0.73

0.46

0.58

0.49

0.40

0.49

Calibration tests confirmed that it was possible to monitor PD

activity reliably using wideband HFCT sensors located at the
switchgear end of the motor feeder cable. The results also
confirmed the importance of using the PPRF factor as
without it the remote measurements would be out by a factor
of two. It is thus recommended that calibration pulse
injections are made on the cable to measure the PPRF for

each motor feeder cable as part of any remote HFCT sensor

and PD monitor commissioning tests.
Further to the success of the remote PD monitoring trial
project described above, a new project has been initiated to
install HFCT sensors within the switchgear motor feeders for
all of the 10 kV motors at the processing facility to enable
continuous PD monitoring of the motors. Since most of the
motors are of similar ratings, this is expected to lead to a
program where replacement candidates can be prioritized
and spares or replacements procured. This new project will
provide continuous PD monitoring to identify variations
related to operational, environmental and seasonal
variations to provide trends in data. The results of this
continuous monitoring program of a large population of
motors, using HFCT sensors mounted at the switchgear end
of the cable, may be a future topic of another technical paper
on this application into the future.

IV. CONCLUSION
The project has shown that it is possible to reliably detect the
PD signals emanating from the stator windings of the motors
located in the Ex zone using remotely located HFCT sensors
at the central switchboard. It is also proposed that the
individual pulse wave-shape analysis method described in
this paper allows for a much more detailed analysis of the
PD pulses emanating from the machines windings to be
made. Most crucially, this type of waveshape-based,
diagnostic measurement provides the ability to discriminate
between phase-to-phase PD (from the end windings of the
machine) and phase-to-earth PD (from the slot sections of
the machine). This has a significant bearing on the level of
maintenance required as the repair of slot insulation requires
removal of the rotor (a big job) whilst end-winding discharges
can be repaired in many cases with only having to remove
the motor end caps to clean and touch-up the end winding
insulation with stress-grading paint.
Partial Discharge diagnosis algorithms developed by the
authors over 15 years ago for PD testing cables,
transformers and switchgear, allowed a similar analysis to be
made for HV rotating machines. To better understand pulse
propagation, retention of energy and multiple reflections of
PD signals in rotating machines, the authors will continue to
carry out pulse injection tests and remote-end PD testing of
in-service HV motors and generators. Such tests provide an
understanding of how the machine construction and cable
geometry and affects pulse attenuation and distortion from
source to sensor. The accurate location of PD sites within a
rotating machine will be dependent upon the combined
knowledge of the propagation paths throughout the windings
and the application of suitable differential equations to
describe the pulse propagation process and ultimately locate
the source(s) of PD.
Developing relationships between the different sites of PD
activity within rotating machines, and the PD pulse signals at
different locations, is dependent upon having knowledge of
the frequency response of the PD sensor, and an
understanding of certain types of differential equations which
permit so-called travelling wave solutions. Such equations

help in describing the PD pulse propagation process as they

allow decaying modes of solution that model how a wave
travels in a dispersive medium. Further work is planned to
examine the validity of using differential equations to predict
PD pulse attenuation, dispersion and reflections as a
function of distance travelled through the windings.

REFERENCES
[1] G. C. Stone and J. Kapler, Stator Winding Monitoring,
in IEEE Cement Industry Conference, 1997.
[2] E. Wen Shu and S. Boggs, Dispersion and PD
Detection in Shielded Power Cable, Dielectrics and
Electrical Insulation Society, Vol. 24, No. 1, January
2008.
[3] S. Haq, B. Mistry and R. Omranipour, How Safe is the
Insulation of Rotating Machines Operating in Gas
Groups B, C & D?, IEEE-PCIC, Toronto 18-20
September 2011.
[4] G. C. Stone and V. Warren, Objective Methods to
Interpret Partial Discharge Data on Rotating-Machine
Stator Windings, in IEEE Transactions on Industry
Applications, Vol. 42, No. 1, January 2006.
[5] V. Warren, G. C. Stone and M. Fenger, Advancements
in Partial Discharge Analysis to Diagnose Stator
Winding Problems, in IEEE International Symposium
on Electrical Insulation, Anaheim, CA USA, April 2000.
[6] R. Mackinlay and C. Walton, 2001, Diagnostics for MV
cables and switchgear as a tool for effective asset
management", Proceedings CIRED June 2001.
[7] M. Seltzer-Grant, D. Denissov, R. Mackinlay, F. Petzold,
L. Renforth and H. Schlapp, On-Line Continuous
Monitoring for In Service Distribution Class
Cables and
st
Switchgears, Proceedings of the 21 International
Conference on Electricity Distribution (CIRED),
Frankfurt, 6-9 June 2011.
[8] IEEE Std. 1434: Trial Use Guide to the Measurement of
Partial Discharges in Rotating Machinery, 2000.
[9] British Standards Institution. BS EN: 60034-27, Off-Line
Partial Discharge Measurements on the Stator Winding
Insulation of Rotating Electrical Machines, London,
2011.
[10] Y. Asiri, A. Vouk, L. Renforth, D. Clark and J. Copper,
Neural Network Based Classification of Partial
Discharge in HV Motors, IEEE Conference and
Exhibition on Electrical Insulation, Annapolis, MD, USA,
5-8 June 2011, pp.333-339.
[11] R. R. Mackinlay, New Methods in On-line PD Detection
for HV Plant in UHVnet Conference, Cardiff, UK,
January 2005.
[12] G. C. Stone, S. R. Campbell and H. G. Sedding,
Characteristics of Noise and Interphasal PD, in IEEE
Electrical Insulation Conference, Annapolis, MD USA,
June 2011.

VITA
Lee A. Renforth studied Electrical and Electronic
Engineering at the University of Manchester in the UK
between 1986 and 1990. He was sponsored throughout this
4-year industrially-linked course by BICC cables where he
received his first training in the testing of HV cables. He
graduated with a BSc, MEng degree in 1990 and went on to
study for a PhD in the field of high voltage insulation
breakdown under sponsorship by the National Grid
Company, the transmission network operator for the UK. He
was awarded his PhD in 1993 when his thesis on the topic of
the breakdown of high voltage insulation was published, also
at Manchester University. In 1994 he set-up a new
technology start-up company, IPEC Ltd, with other members
of Manchester Universitys Power Systems Group. He was
Managing Director at IPEC Ltd for 10 years from the
companys start-up before leaving to set-up HVPD Ltd in
2004. He is currently Managing Director of High Voltage
Partial Discharge (HVPD) Ltd of Manchester, England, UK
which has now established itself as one of the market
leaders in the growing field of on-line partial discharge test
and monitoring technology for high voltage plant, cables and
rotating machines. He oversees the relationships with
HVPD's customers and partners in more than 80 countries
worldwide whilst remaining involved in the companys R&D
activities in conjunction with Manchester University, UK.
Steven Goodfellow received an HND in Electrical &
Electronic Engineering in 2003 from Tameside College in
Manchester, achieving a distinction level in his final year
project, the study of HV full-bridge resonant power supply.
With 8 years electronics manufacture and field experience
working as a HV test engineer, previous projects have
ranged from LASER power supplies to HV Marx Generators.
Prior to joining HVPD in Jan 2009, he worked as an
electrical test engineer at AMTAC laboratories and since
joining HVPD, he has been focused on field trials of the
HVPDs range of PD monitoring technology. With over 1000
hours of at-site field test experience with HVPD, Steven
carried out the on-line test trials reported in this paper. He is
an offshore certified engineer and is a member of the HVPD
online field test team providing PD test services to industrial
customers worldwide.
th

David Clark was born in Canterbury, Kent, UK, on 4

October 1980. He received the B.Sc. degree in Mathematics
from the University of Kent in 2009, and the M.Sc. degree in
Electrical Power Engineering from the University of
Greenwich in 2010. At present he is pursuing a Ph.D. degree
at the University of Manchester with sponsorship from HVPD
Ltd. His research interests include partial discharge pulse
propagation in machines, applications of solution theory and
travelling wave solutions to partial discharge interpretation
and the development of knowledge-based rules for partial
discharge (PD) classification.

Paul S. Hamer (S70M74SM89F97) received the

B.S.E.E degree from Virginia Polytechnic Institute and State
University, Blacksburg, and the M.S.E.E. degree from
Oregon State University, Corvallis, in 1972 and 1979,
respectively. In 1979, he joined Chevron Corporation, where

he is currently a Senior Consulting Engineer, Electrical

Machinery and Power Systems, with Chevron Energy
Technology Company, Richmond, CA. His primary
responsibilities include power system, motor, and generator
application and consultation. He has worked on many
refining, chemical, and oil production projects during his
career with Chevron. In 2008, he was appointed a Chevron
Fellow, Chevrons highest recognition for individual
accomplishment. As a member of the American Petroleum
Institute (API) Subcommittee on Electrical Equipment, he
has contributed to the API standards for induction and
synchronous machines and the API recommended practice
on electrical area classification. He has represented the API
on the National Electrical Code, Code-Making Panel 11, and
on the Technical Committees for NFPA 70E and NFPA 780.
He is now an IEEE-PCIC liaison to the Maintenance Team
for IEC 60079-7 on increased safety equipment.
Russell Armstrong is the E&I Reliability Supervisor at
Tengizchevroil (TCO), Tengiz, Kazakhstan. He worked on
the project described in this paper in conjunction with HVPD
and Chevron, applying his knowledge of the 10 kV network
under test to identify the most suitable points and attachment
(POAs) for connection of the HFCT sensors and supervising
access and switching of the network to allow for the
installation of temporary sensors and off-line tests.