674 Benefits of PD Diagnosis On GIS Condition Assessment

674
Benefits of PD diagnosis
on GIS condition assessment
Working Group
B3.24
January 2017
BENEFITS OF PD DIAGNOSIS ON
GIS CONDITION ASSESSMENT
WG B3.24
Members
M. Reuters, Convenor DE T. Linn CH
D. Templeton, Secretary UK J. Lopez-Roldan AU
M. Beck DE J-F. Penning FR
M. Bélec CA I. Ring IE
E. Dullni DE S. Rowland UK
S. Hoek AT T. Saida JP
J-R. Jung KR C. Xiao CN
S. Kyrkjeeide NO X. Zhang CN
Z. Lefter HU
Corresponding Members
N. Andersen DK
D. Dorgan IE
P. Glaubitz DE
R. Kurte DE
P. Prieur FR
K. Tsuboi JP
Corresponding Members
S. Xiao CN
Copyright © 2017
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in
any form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on
their internal intranet or other company network provided access is restricted to their own employees. No part of this
publication may be reproduced or utilized without permission from CIGRE”.
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent
permitted by law”.
WG XX.XXpany network provided access is restricted to their own employees. No part of this publication may be
reproduced or utilized without permission from CIGRE”.
Disclaimer notice ISBN : 978-2-85873-377-4

“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
BENEFITS OF PD DIAGNOSIS ON GIS CONDITION ASSESSMENT
BENEFITS OF PD DIAGNOSIS
ON GIS CONDITION
ASSESSMENT
Table of Contents
EXECUTIVE SUMMARY ........................................................................................................................................3
INTRODUCTION ....................................................................................................................................................4
1 TYPES OF PD DIAGNOSIS AND THEIR ADVANTAGE FOR USERS ............................................................5

1.1 Technical methods for PD diagnosis ...........................................................................................................5
1.1.1 UHF method ............................................................................................................................................5
1.1.2 Acoustic method ......................................................................................................................................6
1.1.3 Conventional method ...............................................................................................................................8
1.2 Other methods ...........................................................................................................................................9
1.3 Combined methods ....................................................................................................................................9
1.4 Time based strategies of PD diagnosis ..................................................................................................... 10
2 RELIABILITY OF PD DIAGNOSIS AND ITS INFLUENCE ON RISK ASSESSMENT .................................... 12

2.1 Introduction .............................................................................................................................................. 12
2.2 Probability analysis................................................................................................................................... 13
2.3 Consequence analysis ............................................................................................................................. 13
2.4 Impact of reliability of PD diagnosis on risk assessment............................................................................ 15
2.5 Approach for determination of an ‘asset risk index’ ................................................................................... 16
3 BENEFIT ANALYSIS FOR INVESTMENT IN PD DIAGNOSIS SYSTEMS .................................................... 19

3.1 Introduction .............................................................................................................................................. 19
3.2 Decision paths for investments ................................................................................................................. 20
3.3 Cost drivers for different time based strategies of PD diagnosis ................................................................ 21
3.4 Calculation examples for different PD diagnosis and intervention scenarios .............................................. 22
3.5 Spare part policies optimized by PD diagnosis .......................................................................................... 25
3.6 Comparing effectiveness of PD diagnosis against other diagnostic techniques e.g. SF6 analysis............... 25
4 SURVEY OF CASE STUDIES....................................................................................................................... 27
4.1 Example from Australia – Periodic insulation condition monitoring of hybrid switchgear (MTS) .................. 27
4.2 Example 1 from China – Periodic PD diagnosis on 500 kV MTS with uncertain PD defect identification .... 29
4.3 Example 2 from China – PD diagnosis on 500 kV MTS with identified PD defect....................................... 30
4.4 Example from Hungary – Retrofit of a 275 kV HV GIS with a continuous PD diagnosis system ................. 32
4.5 Example from Ireland – ESBI’s PD condition monitoring strategy for HV GIS ............................................ 34
4.6 Example from Japan – Reasons in CEPCO for discontinuance of on-line PD diagnosis in HV GIS............ 39
4.7 Example from Korea – GIS condition assessment on a 154 kV substation of a semiconductor factory....... 40
4.8 Example from Brazil – Return of experience from ITAIPU DAM power plant.............................................. 45
4.9 Example from various countries – Reliable PD diagnosis for condition assessment .................................. 47
DEFINITIONS .......................................................................................................................................................49
Page 1
REFERENCES ..................................................................................................................................................... 50
ANNEX ................................................................................................................................................................. 52
Page 2
EXECUTIVE SUMMARY
The application of PD (partial discharge) measurement to diagnose HV (high-voltage) equipment has been around
for many years. There are many methods to measure and monitor PD with different expectations, and each giving
different experiences to the end user. There is also much controversy over the value of PD diagnosis. This can
originate from bad experience, misunderstanding of results, or simply poor PD equipment implementation. The true
value of a PD diagnosis can be lost in all the hearsay.
CIGRÉ Working Group B3.24 addresses these points and summaries the benefit of PD diagnosis for HV GIS (gas-
insulated switchgear). This summary is based on the experience of GIS manufacturers, GIS users, PD diagnosis
equipment suppliers, and the expertise of test service companies.
In chapter 1, commonly used methods for PD measurement are evaluated along with the advantages and
disadvantages for each method, and their specific application to GIS. Terms used to describe PD testing can vary
from country to country and company to company. All PD diagnosis terms and strategies are therefore defined.
Chapter 2 addresses the reliability of PD diagnosis results and its influence on a risk assessment of substation
equipment as the condition assessment is the ultimate goal. For this, PD data must be processed by some means
to enable an evaluation of the performance of the equipment. This evaluation process and its key factors are
reviewed, and combinations of different reliabilities of PD diagnosis and probabilities of failures were compared. A
consideration of possible actions and possible costs related to this comparison emphasize the paramount
importance of reliable results of a PD diagnosis to minimize costs coming from repair and maintenance efforts. The
lower the reliability, the higher the decision will be to wait with possible consequences of the occurrence of a fault.
PD diagnosis will add cost to the operation of the GIS, comprising capital outlay cost and operational cost. These
costs can be considerable, and the true costs of the PD diagnosis over the life of the GIS needs to enable the cost
benefit evaluation to be carried out. In chapter 3, the results of the Working Group discussion are evaluated and
assessed, and some scenarios for applications of PD diagnosis systems are shown to illustrate possible benefits.
This includes also the opposite viewpoint, where no PD diagnosis is employed. An alternative to PD diagnosis for
condition assessment can be to stock a suitable range of spares and to be prepared to replace damaged parts in
the event of a failure. A general optimization of spare part policies based on the experience with PD diagnosis
systems has not been reported and could also not be developed by the Working Group.
A decision to intervene based on PD diagnosis will be the outcome of the cost benefit analysis where the PD
diagnosis is compared to the risk of failure and cost of repair in GIS. There may be many other factors beyond
direct costs of PD diagnosis systems and repairs of switchgear. All factors are needed to make a profound decision
on whether or not are diagnosing the GIS for PD. An overall cost benefit analysis process is proposed.
In chapter 4 a list of case studies identifies the direct and indirect benefits of PD diagnosis for specific customers
and sites.
Working Group B3.24 acknowledges and expresses their thanks to all involved companies of the WG members
that have allowed their representatives time, travel and encouragement to contribute to the realization of this
technical brochure.
Page 3
INTRODUCTION
Condition assessment of high voltage gas-insulated switchgear (HV GIS) has become important with economic
factors driving users to extend the service life of equipment, to optimise maintenance and to prevent unexpected
outages. To support these objectives, partial discharge (PD) diagnosis systems have been proposed to provide
information during commissioning and operation for supporting decision-makers. Continuous or periodic PD
measurements can provide information on the condition of HV GIS. However, the application of these methods and
the interpretation of the results are not yet standardised (e.g. non-conventional PD measurements), and the related
recommendations to manage PD defects derived from the results may vary. Accordingly the benefit of PD
diagnosis is sometimes not tangible to users.
A strong motivation to conduct this work was to analyse what kind of benefit can be created by the different
available PD diagnosis systems in condition assessment of HV GIS. Previous concerns indicated that PD diagnosis
often does not deliver reasonable guidance, even if PD defects are detected without doubt. All systems should
support the user’s risk assessment at the end of a diagnosis process how to handle detected PD with the aim to
prevent failures and optimize maintenance. In some user’s organisations implementation and operational
procedures for PD diagnosis systems are less or not fully developed, limiting the ability of the organisation’s
operational and asset management teams to take full advantage of the diagnosis systems.
Some key factors influencing the user’s decision to purchase a PD diagnosis system are the probability of detecting
an insulation defect by PD diagnosis, the proportion of insulating defects to all possible defects, the criticality of the
substation, the voltage rating of the GIS and of course the costs of the PD diagnosis system.
Costs to employ a PD diagnosis system may be significantly different from each equipment manufacturer, and can
vary depending on the complexity of the substation and data distribution to higher level systems. In this context
often efforts for service contracts (e.g. for PD data evaluation, training, handling of false alarms), PD diagnosis
system operation strategy (which require also resource allocation at user), and upgrade or replacement of soft- and
hardware are considered.
Page 4
1 TYPES OF PD DIAGNOSIS AND THEIR ADVANTAGE FOR USERS

The broad methodology behind how the benefits of PD diagnosis are evaluated and used, can be applied to many
systems in the network or other assets. There may be specific differences that are not directly applicable when
methods other than UHF (ultra-high frequency), acoustic and conventional PD detection methods are used. A
comparison of their benefits should only be used if there is no effective difference on the measurement method
applied. Therefore a direct comparison of methods is difficult, if one method provides diagnosis results another
method is not able to give. So in this chapter strengths and weaknesses of different PD diagnosis methods are
summarized related to relevant quantities.
1.1 Technical methods for PD diagnosis

For the purposes of this technical brochure, the evaluated methods for measuring PD on HV GIS (rated voltage
above 52 kV) are limited to UHF and acoustic for on-line measurements and UHF, acoustic and conventional for
off-line measurements. The types of PD defects discussed in this context are ‘mobile particles’, ‘floating elements’,
‘protrusions’, ‘particle on insulation’, and ‘void’ [4].
1.1.1 UHF method

When PD activity occurs, electromagnetic waves in the UHF range incidentally propagate inside the GIS enclosure,
attenuated by epoxy-resin insulators, view port windows, and the geometry of the enclosure and conductors, before
being detected by sensors in a range from several 100 Mega-Hertz (MHz) up to some Giga-Hertz (GHz).
Table 1 displays a basic overview of general strengths and weaknesses of the UHF method.
METHOD STRENGTHS WEAKNESSES

UHF high sensitivity when using sensitive to external noise
internal PD sensors signals (increased level of
possibility to be retrofitted interference)
(occasionally with lower cannot be calibrated in terms of
sensitivity) pC (minimum detection
allows continuous PD sensitivity can be ensured)
monitoring long-term behaviour of
localization of PD defects electronic components
possible not always applicable to all
same life expectancy of all types types of GIS
of UHF sensors compared to
GIS
Table 1 : Comparison of strengths and weaknesses of the UHF method
In the following, strengths and weaknesses of the most common PD sensor technologies in terms of operational
conditions (continuous vs. periodic), localization of PD defect (new vs. existing installation, refurbishment), and
noise discrimination are summarized.
Besides different PD sensor technologies also different PD signal detection approaches have been developed by
many PD diagnosis system manufacturers. This topic is not covered in the scope of this brochure.
Sensor technology STRENGTHS WEAKNESSES

Loop-shaped metallic electrode able to detect a PD signal with a very few GIS designs have
(internal, embedded in an high S/N (signal-to-noise) ratio adopted this kind of technology
epoxy-resin insulator) due to direct coupling of EM impossible to be retrofitted
(electromagnetic) waves
multiple sensor ports possible,
one per epoxy-resin insulator at
intervals of a few meters in order
to optimize the detection
Page 5
Sensor technology STRENGTHS WEAKNESSES

sensitivity
additional measurement ports
can be added afterwards
Disc / conical shaped monopole able to detect a PD signal with a in case of retrofit additional GIS
sensors (internal) high S/N ratio due to direct openings required
coupling of EM waves might need a protector against
possibility to be used as a high voltage coupling (inductive
voltage sensor for certain or resistive 50 or short-circuit)
designs
design is optimized to achieve
highest sensitivity
Sensor on viewport or easy to retrofit, if viewports are limited to GIS types equipped
inspection window (external) available in sufficient quantities with viewports (and of sufficient
and suitable locations size to achieve sufficient
no outage or gas handling for sensitivity)
mounting required highly susceptible to noise
useful for periodic, continuous interference requiring extra
and temporary continuous PD shielding
diagnosis customized design required in
most cases (sometimes
additional pre-amplification
required)
Barrier sensor (external, e.g. easy to retrofit not applicable to all GIS designs
“Coaxial-to-waveguide no outage or gas handling for susceptible to noise interference
antenna”) mounting required lower sensitivity (can be
periodic PD measurements easy compensated by more sensors)
to apply with portable equipment may require customized sensor
simple localization possible (e.g. design
by moving sensors around the
barrier and from one barrier to
the next)
sensors generally very robust
additional measurement points
can be added afterwards
Shield type sensor (internal) depending on the sensor design very difficult to be retrofitted
it is possible to achieve a
comparable sensitivity as
internal disc shaped monopole
sensors
less susceptible to noise
interference
existing structure and coupling
surfaces can be used for signal
coupling
Table 2 : Comparison of strengths and weaknesses of most common

PD sensor technologies for the UHF method
1.1.2 Acoustic method

It is widely accepted that both during commissioning and in-service testing of HV GIS the acoustic method is
suitable for PD diagnosis, though quantitative results in terms of pC cannot be obtained. It is advantageous that
several kinds of PD defects can be detected, located, and evaluated. Moreover the acoustic method for PD
Page 6
diagnosis is non-invasive, exhibits a high immunity against disturbances, and can be easily retrofitted into existing
GIS [2], [5].
However sufficient detection sensitivity can only be achieved for PD defects in the immediate vicinity of the acoustic
sensor. Thus, a comprehensive PD diagnosis of a substation with the acoustic method can be very time
consuming, when moving sensors along the GIS external surface or quite expensive in case of a continuous
monitoring approach due to the necessity to install many acoustic sensors. Thus the acoustic method is commonly
not applicable for continuous monitoring for PD diagnosis [2].
The signal spectrum which can be detected by an acoustic sensor ranges from kHz to MHz. It is influenced by the
kind of the sound source and the absorption behaviour of the signal transmission path both in the gaseous and
solid insulation and the metallic material of the enclosure. Concerning the gaseous part of the absorption, the SF 6
gas pressure has a considerable impact on the signal absorption in dependence of the frequency spectrum of the
acoustic signal (Figure 1).
Figure 1 : Acoustic signal absorption as function of SF 6 gas pressure [6]
On the other hand, the coupling of the acoustic signal from the SF 6 gas to the enclosure is improved with higher
gas pressure [5]. In case of the PD defect “protrusion on HV conductor” the better coupling results in enhanced
detection sensitivity at higher SF6 gas pressures (Figure 2).
Figure 2 : Acoustic signal amplitude as a function of SF 6 gas pressure [7]
Basically, the sound can be generated by pressure-waves from discharges, mechanical impact of particles on the
GIS enclosure, and mechanical movements between metallic parts [2]. For evaluation of acoustic signals
Page 7
originating from PD defects, comprehensive methods are available to distinguish between the different types of PD
defects. In case of mobile particles the acoustic method is able to detect even small particles of 1 mm to 2 mm
length with a reasonable signal-to-noise ratio. For PD defects like ‘floating elements’ generally an intensive
acoustic signal with a periodicity of 100 Hz or 120 Hz was found depending on the power-frequency of the
operating voltage [5], [6].
The acoustic method also offers comparatively good opportunities to locate a PD defect by moving acoustic
sensors while comparing signal amplitudes. Moreover with increasing distance between PD sensor and PD source,
the acoustic signal appears more smoothed out regarding the acoustic wave front due to enhanced dispersion of
the signal propagation. Thus it can be possible to differentiate this case against a signal with a short rising acoustic
wave front, which would correspond to a short distance [2].
Table 3 summarizes strengths and weaknesses of the acoustic method.

Acoustic able to be applied on in-service
sensitivity is highly dependent
GIS on the PD defect and its location
occasionally less sensitive to
time consuming in case of
noise interference complete GIS survey
simple to use for localization of
occasionally high amount of
a PD defect manual work
comparatively cheap and easy
sensitive to rain in case of
to carry equipment outdoor GIS, to mechanical
impact, and to magnetostriction
good applicability in case of
access to gas compartments(resulting from inductive voltage
non-intrusive transformers)
not applicable for continuous PD
diagnosis as per common
practice
cannot be calibrated in terms of
pC
Table 3 : Comparison of strengths and weaknesses of the acoustic method
1.1.3 Conventional method

The conventional method to perform PD diagnosis is described in IEC 60270 [8]. In this context only a very brief
summary of strengths and weaknesses of the conventional method is given in Table 4.

Conventional can be calibrated in pC very difficult to apply at in-
according to [8] service HV GIS and at voltage
restricted bandwidth levels above 300 kV
a coupling capacitor is
a complete GIS section can be
surveyed by one system necessary for quantitative
measurements
invasive (in terms of gas
compartments when an
encapsulated coupling capacitor
is used)
susceptible to noise interference
on-site
no options for PD defect
localization
Table 4 : Comparison of strengths and weaknesses of the conventional method
Exemplary illustrations showing test set-up’s installed in a factory (Figure 3) and on-site (Figure 4) are displayed
below.
Page 8
Figure 3 : Test set-up for PD measurement with the conventional method in a factory
Figure 4 : Test set-up for PD measurement with the conventional method on-site [22]
In Figure 3 the GIS under test is energized by a HV test transformer. A HV capacitor is coupled with an inductor
and constitutes a high frequency filter. Indeed, it allows measuring high frequency voltages that are induced by the
apparent currents generated by PD. This filter function prevents HV power frequency to damage the PD
measurement unit. The GIS under test and the HV test-set can be identified in Figure 4 by the different colours.
The HV test transformer is preferably integrated inside the GIS enclosure in order to avoid any air insulated
connection between the transformer and the GIS. An air insulated bushing could generate corona discharges in air
with such a magnitude that any PD activity inside the GIS would most probably be masked.
1.2 Other methods

This sub-chapter gives a short description on other methods able to measure PD. They have in common a very
limited applicability on HV GIS, neither during commissioning tests nor in service. Transient earth voltage (TEV)
detection might be applied on in-service GIS, but is mainly used to diagnose MV (medium voltage) GIS. The
method is comparatively cheap and uses easy-to-carry equipment and non-intrusive sensors. PD detection using
Rogowski coils or optical sensors has not been widely applied to perform PD diagnosis on HV GIS according to the
knowledge of the Working Group. Older as well as recent references - e.g. [6], [9] - indicate the basic principle.
However, no recent information presents application cases of optical PD diagnosis on in-service HV GIS.
1.3 Combined methods

Considering advantages and drawbacks of the various methods described above, a combination can be an
interesting solution by associating the advantages of several methods. For example, the conventional method can
be used in combination with the UHF method. Thus, both PD level in pC and the location of a PD defect can be
diagnosed. The drawback of a combination is the need of several measuring equipment. Also, a combination of
methods can help to discriminate noise and enhance localization possibilities.
Page 9
1.4 Time based strategies of PD diagnosis

Before using or even installing any PD diagnosis system, the user has to decide if measurements are to be
conducted periodically or continuously. The latter is often referred to as monitoring. Based on the WG experience,
it is sometimes meaningful to perform a so-called temporary continuous PD diagnosis, which offers an intermediate
approach. A comparison of three different time based strategies of PD diagnosis is shown in Table 5 and illustrated
in Figure 5.
DESCRIPTION ADVANTAGES DRAWBACKS TIME SCALE RELATIVE COST

PERIODIC representing a flexible use no PD 1 day lowest initial costs
snapshot in time diagnosis … (depends on
on a limited between two 2 days strategic role of
number of campaigns the substation)
simultaneously risk of non-
connected detection of
locations intermittent
PD
high level of
human
expertise
required
TEMPORARY representing the flexible use specific 1 week in the mid-level
CONTINUOUS development on tracking of algorithms … (the GIS section
a few connected every are required 1 year selected for PD
locations for a discharge to eliminate diagnosis might
limited time occurring in noise be chosen
time window interference strategically)
time history expert
trend system
analysis software is
alarm required
generation if
occurring in
time window
remote
control
CONTINUOUS representing the tracking of specific GIS life time highest initial
development for every algorithms costs (depends on
all locations discharge are required strategic role of
without time time history to eliminate the substation)
limitation trend noise
analysis interference
alarm expert
generation system
remote software is
control required
Table 5 : Overview of different time based strategies of PD diagnosis
Page 10
a) b) c)
Figure 5 : Illustration of a) periodic, b) temporary continuous and c) continuous PD

diagnosis procedures
A periodically applied PD diagnosis system normally uses portable equipment. During a measurement campaign
(during GIS commissioning or for periodic scanning), already installed sensors are successively connected to the
portable PD diagnosis equipment.
In a temporary continuous procedure, measurements are made continuously, but with limited duration using
transportable PD diagnosis equipment. During the measurement campaign, a limited number of sensors are
connected to the PD diagnosis equipment for a certain time. This kind of campaign can be made for example,
when abnormal PD signals have been detected during a former periodic application. Thus the diagnosis can be
focused on some sensors for a longer time in order to monitor PD trends, and to assess the severity of a PD defect.
Continuous procedure measurements are conducted with a permanently installed diagnosis system, connected to
every sensor of the substation.
Finally, all described methods in chapter 1 have to cope with noise interference. Noise which is emitted by sources
outside the GIS might be detected and interpreted as PD signal. Therefore enhanced noise discrimination methods
may be applied, which can vary for different equipment. The following is a not exhaustive list:
Hardware filtering,
Frequency band scanning,
Noise interference gating by use of separate sensors,
Phase coincidence filterung,
Star diagram,
Frequency vs. time diagram,
Combined detection methods.
Page 11
2 RELIABILITY OF PD DIAGNOSIS AND ITS INFLUENCE ON RISK ASSESSMENT

2.1 Introduction
PD diagnosis is a means of assessing the condition of substation equipment either in order to prolong maintenance
periods or to decide for intervention if PD activity is detected.
For a well-founded risk assessment of the criticality of the defect and the subsequent decision, the results from PD
diagnosis need to be reliable. Otherwise, it is difficult, if not impossible, to arrive at a sound decision. This fact has
been addressed frequently in the past [4], [6].
The weighting of information required for a comprehensive risk assessment is difficult, and depends on the different
experiences of the users. To lower the uncertainty of a risk assessment it might be helpful to concentrate on a very
few well choosen key parameters in determining the risk. As key factors the following can act:
PD defect identification, location and trend analysis,
Past operational experience with PD defects in HV GIS.
From practical Working Group experience it is known that sometimes it is not possible to identify the location of the
detected PD source with absolute certainty due to a wide variation of reasons (some can possibly be influenced by
the user of the PD diagnosis system, others not) when performing any kind of PD diagnosis.
In terms of location possibilities of PD defects, different approaches are available even under the consideration of
reduced investment costs for PD diagnosis equipment (e.g. less internal UHF sensors and smaller scope of
hardware of a continous PD diagnosis system) [20], [21]. The latter would be supported if some requirements are
covered:
Experience with PD detection sensitivity in similar or same GIS configurations,
Procedures to assess UHF signal attenuation characteristics for GIS sections or even complete GIS,
Possibility to apply temporary external UHF sensors for signal injection purposes or on-demand locating.
PD diagnosis requires certain expertise that can be provided by either in-house experts, external experts, or expert
systems (software). Comprehensive knowledge about GIS design, substation layout and historical failure behaviour
is essential to increase reliability of PD diagnosis as much as possible.
Finally, the consequence of a fault in the substation plays an important role. A substation with high importance will
require perhaps immediate intervention, whereas a substation with redundancy may allow waiting.
In CIGRÉ Technical Brochure 525 [4] an analysis of critical PD defects in HV switchgear is provided and examples
of failure probabilities of typical defects are given. The technical brochure also provides a calculus to obtain the
probability of occurrence of a failure in dependence of several factors including the confidence (reliability) of the PD
diagnosis depending on the expertise of the observer, the known technical impact of an identified defect, and the
known criticality range of an identified defect. This consideration is the first part of a risk assessment (Figure 6) and
briefly repeated in the probability analysis in chapter 2.2.
Page 12
Figure 6 : Flow chart of risk assessment procedure [4]
The second part of the risk assessment is the evaluation of consequences of a possible failure, which is only briefly
mentioned in [4], but will be considered in more detail in the following sections.
2.2 Probability analysis

The first step of a risk assessment is to perform a sensitive PD measurement to detect a potential defect. Any kind
of PD pattern detected in the GIS needs a careful evaluation with respect to the type and location of the PD defect.
Together with other essential information from laboratory measurements, manufacturer's experience, design
aspects, inception and extinction voltages of PD defects, switching operations during PD activity, load of GIS, small
deviations in operating voltage, trend analysis of the PD activity, PD defects can be identified and their criticality
estimated. The highest success rate for such an assessment is given when the PD activity can be related to a
known critical location or component.
In general, it will be possible to relate a PD pattern to a PD defect. However, some uncertainty in the assessment
will remain which can be reduced e.g. by proper evaluation of the amplitude of the signals, the development of the
signal starting from first observation, the location of the PD defect (if feasible), and the known history and failure
statistics of the particular switchgear under observation.
CIGRÉ Technical Brochure 525 “Risk assessment on defects in GIS based on PD diagnostics” provides details on
the assessment of criticality of a measured PD pattern.
The probability that the PD defect will develop into a failure (a voltage breakdown and subsequent internal arc fault)
is an important input parameter for the risk assessment. This information bears the highest uncertainty, since it is a
prediction for the future. Unless there is sound experience on the development of similar PD defects only a trend
analysis (temporary continuous or continuous) would provide an indication as reliable as possible. Thus the
decision to intervene depends on the confidence in the PD diagnosis result, which itself depends on the sensitivity
and reliability of the PD detection technique.
2.3 Consequence analysis

The consequences of a fault created by a PD defect are determined by the criticality of the substation and the
operation mode of the switchgear. If a failure happens because the defect was not fixed, considerable outage costs
might occur among them the following:
Repair costs of the affected component and possible replacement of further damaged components,
Page 13
Costs of non-delivered energy depending on the duration of outage and affected consumers,
Liability costs due to third party damage,
Penalties coming from customer contracts.
These costs depend on the importance of the substation, the build-in of redundancy in the network, public image of
the company, possible personnel injuries and environmental aspects.
On the other hand, the shutdown of a substation and the search for the failure defect and subsequent removal of
the defect also creates costs which have to be considered. In addition, there is a probability that the defect is not
found so that either the switchgear is taken back into service without any replacement - with the possibility of a later
fault - or the switchgear is precautionary refurbished to a large extend with high costs. The following paragraphs try
to give guidance for a decision based on possible consequences.
Figure 7 displays costs or expenses over the age of the switchgear in a substation for the case some PD activity
with associated failure probability P is detected.
Figure 7 : Exemplary scenario for possible intervention based on PD diagnosis
The different costs are related to:
Ex: Costs for repair and exchange of damaged parts of the switchgear after a fault produced by the PD
defect. These corrective costs normally increase slightly in time, since spare parts become more
expensive.
E0: Costs for a planned preventive repair of the switchgear to remove the PD defect. These costs may
considerably vary in height depending on the kind of necessary repair. The lowest costs are achieved
when the search for the PD defect was successful and only the associated switchgear part is replaced
or repaired, respectively. This is the ideal case of ‘condition based maintenance’. If the search for the
PD defect was not successful, many more switchgear parts need to be precautionary refurbished.
Historical experience shows that the overall costs can increase considerably, if during such a repair
additional irregularities are introduced into the HV GIS, which later develop into potential failures. In
this case, costs are similar to that of a ‘time based maintenance’, where repairs and refurbishments
are carried out according to experience and replacement plans. Since it is not known before whether
the search for the PD defect will be successful or not, the higher costs for “time-based maintenance”
need to be applied.
ExP: Is the product of the costs Ex times the probability of failure P. For the probability of failure, a bath tub
curve is assumed (see Figure 8), which implies that possible defects of the particular switchgear and
possible consequences are known from history. Depending on the failure probability, the curve ExP
can have different shapes e.g. may increase at later times or not.
Page 14
Three different time ranges are recognized, which are named as:
Time range I (infant mortality phase),
Time range II (stabilized/normal phase),
Time range III (end-of-life/wear-out phase) 1.
In dependence of the intersections of the preventive cost curve E0 with the curve ExP, the recommended action is
different for the three different time ranges I to III. It depends on whether the costs to remove the PD defect are
higher or lower than the costs for repair of a fault multiplied with the probability of occurrence of that fault.
When ExP value exceeds E0 then intervention is recommended. And if ExP value is less than E0 then intervention
can be postponed.
According to Figure 7, one would recommend in time range I to search for the PD defect and repair the switchgear.
The intervention will depend on the type of detected PD defect. If it is a ‘high risk’ defect with a high probability of
failure it will require urgent intervention, otherwise the outage for repair can be planned on a longer time scale. If
the occurrence of the PD activity is in time range II, one would not intervene, but observe the PD activity. In range
III, one would eventually replace the complete switchgear or parts of it, since it has anyway approached end of life.
Figure 8 : Figure 5-48 taken from [20]
2.4 Impact of reliability of PD diagnosis on risk assessment

Irrespective of the consequence of a fault for a risk assessment, possible intervention after the detection of PD in
switchgear may be influenced by the reliability of the PD diagnosis result, which is explained below:
1) Most reliable, since the PD defect has been classified and located,
2) Reliable, since the PD defect has been either classified or located with uncertainty,
3) Unreliable, since the PD defect has been neither classified nor located.
In order to derive scenarios for possible actions the expected failure probability associated with the PD defect
needs to be estimated:
a) The total failure probability is high,
b) The total failure probability is medium,
c) The total failure probability is low.
1
The increasing part of the curve ExP is not always appearing or obvious.
Page 15
Besides the classification and localization of the PD defect as given above, knowledge of the estimated time to
failure would be useful in order to plan an outage. Since this parameter is most often not assessable, the time to
repair is rather determined by the criticality of the substation, the technical and economical boundaries and the
expected failure probability.
Table 6 summarizes three possible actions in dependence of the reliability of PD diagnosis (1 – 3) and probability
of failure (a – c), and provides an assessment of costs.
POSSIBLE ACTION
no action (with risk of
Timing of repair As soon as possible Scheduled maintenance
failure)
Scenarios in terms of
1c
reliability of diagnosis 1b
1a 2b, 2c
result (1-3) and 2a
3
probability of failure (a-c)
Low High
Medium
(defective part) (defective part +
Spare parts (defective part +
medium, if search was + damaged parts +
+ express costs)
not successful express costs)
POSSIBLE COST
X + Y2 days
X+Y1 days X days
(repair or exchange of
Manpower (repair or exchange of (repair or exchange of
defective + damaged
defective part) defective part)
parts + cleaning)
Medium Lower Higher
Outage
(short-term) (scheduled) (unexpected)
Project management
Medium Minor Medium
(of repair works)
Environmental impacts
Minor Minor Medium
(SF6 to atmosphere)
Table 6 : Actions vs. reliability of PD diagnosis
X: costs for repair and/or exchange of a defective part in days
Y1: costs for acceleration/express due to non-availability of manpower (standard basis)
Y2: costs for additional efforts in terms of extra work on damaged parts regarding repair/exchange and
cleaning
Where the PD defect and its location cannot be identified (i.e. case 3), it seems advisable to continue operation,
and wait on a possible failure with subsequent repair.
The conclusion is that a reliable result of a PD diagnosis system is essential for an action plan focused on
minimization of costs coming from repair and maintenance. The lower the reliability is, the higher the decision will
be to wait with the possible consequence of the occurrence of a fault. In [23] additional information for conducting
risk assessments can be found.
2.5 Approach for determination of an ‘asset risk index’

Risk assessment leading to intervention is always linked to reliable results of a PD diagnosis, expertise and
decision making process available on discovering PD signals. A first PD activity notification will result in a PD
alarm. Subsequently, a competent PD diagnosis system user is provided with early information which can help with
the decision making process, but other considerations can be also necessary like cost to repair, spare part
availability, importance of the asset and/or outage time. Figure 9, Figure 10 and Figure 11 show a possible
approach to classify an ‘asset risk index’ with ‘component factors’ as well as ‘network factors’ as important inputs.
Page 16
Figure 9 : Component factor for ‘asset risk index’ classification
Figure 10 : Network factor for ‘asset risk index’ classification
Page 17
Figure 11 : PD notification procedure and ‘asset risk index’ assessment
Page 18
3 BENEFIT ANALYSIS FOR INVESTMENT IN PD DIAGNOSIS SYSTEMS

3.1 Introduction
The owner or user of a GIS substation will on occasion face the decision of purchasing and installing a PD
diagnosis system when ordering a new substation or when refurbishing existing equipment. Cost versus benefit
scenarios vary due to many factors ranging from the value of the asset to the operational strategy of the system
owner and user and the type of PD diagnosis system applied (e.g. periodic or continuous). PD diagnosis systems
are one of the very few ways the dielectric integrity of the GIS insulation can be assessed in real time without any
operational hindrance, and able to provide predictive warnings or change of conditions.
Considerations associated with PD diagnosis systems are the costs, system reliability, and the user’s operational
strategy. Sometimes it is obvious that implementation and operating procedures for PD diagnosis systems within
the user’s operational teams is poorly applied, especially when training levels are comparably low. This can be
exacerbated by the organizational culture and structure at utilities. There are cases known that PD diagnosis
systems are installed, but sometimes remain without user interaction for data evaluation until a system error or
false alarm occurs.
Costs to employ a PD diagnosis system vary with manufacturer, regional market, substation size, installation,
commissioning and training costs. The cost to monitor a typical GIS substation can vary greatly depending on the
complexity of the system and data distribution to higher level systems. Once the PD diagnosis system is handed
over, the costs do not cease but continue for system operation and user benefit. The following aspects should be
considered as additional cost factors by the user:
Service contracts (e.g. PD data evaluation, education and training, handling of false alarms),
PD system operation strategy (resource allocation at user),
Upgrade/replacement of system.
It is also important to note that a PD diagnosis system - in particular the electronic components and software -
would need replacement or even sometimes upgrading during the life of the GIS substation.
It has to be evaluated which benefits the user can achieve from employing PD diagnosis systems to monitor the
quality of the GIS operation. The ultimate goal must be fault/failure avoidance to reduce power failures, significant
operational down time, asset repairs/replacements and possible fines or lost revenue. To achieve fault/failure
avoidance, the user expects early detection with a clear direction on PD defect type and location followed by expert
analysis. Therefore, a PD diagnosis system must provide data for classification, localization, trend evaluation and
alarming as a minimum for users to determine the failure risk.
Employing a PD diagnosis system may also be seen as a proactive approach in increasing the reliability of the
electrical network, which could be favorably acknowledged by investors, customers and regulators. When installed
in new HV GIS substations, such system can already be used during the HV commissioning test and initial
energization period of the substation to detect and eliminate quality issues. Of course, they should be applied to get
a base line for later condition assessment.
With respect to the measuring technique, the benefit of PD diagnosis systems depends on the selection and
positioning of PD sensors. To achieve the maximum benefit, proven signal capturing methods such as UHF and/or
acoustic measurements should be preferably selected. PD sensors are positioned according to experience,
available positions on the GIS and calculations using the PD system thresholds and GIS attenuation figures for
each compartment. The benefit depends on the balance found between detection sensitivity and PD sensor
quantities.
Page 19
Other factors associated with the GIS also determine criteria to employ a PD system include the following:
Importance of the GIS in the network,
Known/previous defects from historical data or specific HV testing/energization,
Budget for investments.
Very often it is concluded that monetizing of values in a risk assessment is difficult or even impossible since
consequences and costs cannot be linked appropriately. But also very often this hurdle can be taken with
reasonable effort, though not always on first sight. The sources for required data may be indirect ones (e.g.
insurance costs, fines, clean-up costs after environmental impacts), but basically these costs are available perhaps
with higher uncertainty [10].
3.2 Decision paths for investments

The following section summarizes the experience collected by the Working Group concerning the different paths
investment decisions for PD diagnosis systems can take. In Figure 12 three different decision paths are shown.
Figure 12 : Relations for investment decision in PD diagnosis system
They describe the most common reasons for the decision to invest in a PD diagnosis system. The reason can be
any one, or a combination of any of these three paths:
Policy: This path is determined by policies, which may be mandatory, existing within the company or by
legislation. Many utilities have policies to ensure that best practices are applied within the company.
These policies can be internally or externally driven. Internal policies are only provided by the company,
and may be influenced by historical/previous operating experience with certain type of GIS design and/or
expert knowledge levels. External policies may arise from governmental legislation, insurance companies,
or companies relying on the availability on the GIS (such as a nuclear plant or semiconductor factory).
Technical: This path is driven by the technical evaluation within the company. Usually it is initiated by some
existing problem, and the PD diagnosis system is supposed to aid or assist in the technical issue. The
purchase of a PD diagnosis system may also help in optimizing maintenance schedules where
investigations can be carried out along with routine maintenance and preventing a separate outage. The
decision is based on an existing/current or historical problem with a certain site or equipment family.
Page 20
Economic: This path is taken when there is a financial gain in installing a PD diagnosis system. There may
be many reasons such as: reduced insurance premiums, primary fault or unscheduled penalty avoidance,
life extension (or delayed investment), reduced repair costs (by saving the costs of damage caused by a
fault).
In the evaluation of the scope and investment costs of a PD diagnosis system, the equipment should provide a
minimum quantity of operational features, but additional capabilities like software with expert systems for feature
extraction and localizing options should also be considered to increase the probability of a reliable result of a PD
diagnosis. This decision is part of the technical route. The final decision may be a result of all three paths. Every
customer situation is unique, and the individual situation must be assessed before a decision can be made.
3.3 Cost drivers for different time based strategies of PD

diagnosis
Generally internal PD sensors are installed in the easiest and most effective way during the initial process of
manufacturing and installation of the GIS. In this case, ports, windows or other enclosure elements which are
suitable to include PD sensors can be added already in the original GIS design. Because of the compactness of
modern GIS, retrofit solutions can be more difficult and are probably a compromise, e.g. in terms of detection
sensitivity. PD sensors often cannot be placed in the optimal location because of the lack of suitable enclosure
ports. For this reason the effectiviness of a retrofitted PD diagnosis system usually is lower, but still can fulfil the
basic requirements as well. This situation is presented later in chapter 4.4.
In Table 7 an approach was made to visualize possible cost drivers when purchasing, installing and operating any
kind of PD diagnosis system in a new GIS or as retrofit into an existing GIS installation. It is not linked to any
specific diagnosis method or device, and situations can vary between users. The ranking ranges from low to high
and represents an indication only, and not a 1:1 represention of real costs.
Table 7 : Possible cost drivers for different PD diagnosis system implementation

scenarios
A review of recent public data revealing the percentage of application of PD diagnosis systems for GIS condition
assessment as function of the rated voltage of the GIS can be found in Figure 13, taken from [19].
Page 21
Figure 13 : Table 6-13 taken from [19] showing ‘PD measurement application – number
of responses within individual voltage classes’
Based on the responses it was evaluated by CIGRÉ Working Group A3.06 that PD measurements are used mainly
as ‘diagnostic tool at present’ (periodic PD diagnosis system), but it is assumed it will be used more frequently in
the future as ‘monitoring tool’ (continuous PD diagnosis system). CIGRÉ Working Group A3.06 emphasizes that
present monitoring practices are probably influenced by difficulties with the PD system application (high uncertainty
on the diagnosis results) and in terms of retrofit purposes on GIS.
3.4 Calculation examples for different PD diagnosis and

intervention scenarios
In the past, few publications deal with cost analysis of PD diagnosis [17] and proposals how to evaluate the
benefits gained from different kinds of PD diagnosis systems. These proposals ranged from quite general figures to
estimates considering different approaches for PD diagnosis and their benefit for condition assessment of HV GIS
[18]. Periodic PD diagnosis using the UHF method and internal PD sensors was proposed as the most effective in
terms of costs and benefit for the detection of the most abundant PD defect ‘mobile particles’. In the early 1990’s it
was stated first that the costs for continuous PD monitoring ranged from 7 %…12 % of the bay cost, later the range
was modified to 3.5 %...4 % [18]. Continuous PD diagnosis systems using the UHF method should be only
considered if economic and technical justification is given, and a benefit can be generated to the user. Finally the
latter keeps undetermind [18].
In a total cost analysis when a PD diagnosis system is to be purchased, it is important to identify clearly the aim of
the PD diagnosis. The main goal of using a PD diagnosis system is obviously to prevent a failure of the switchgear
or substation, or at least to reduce the in-service failure rate. An in-service fault will result in significant repair costs,
and may also generate significant energy disruption costs which can excced the repair costs, especially when the
outage of the substation lasts for weeks or even months.
It is also important that the purchasing costs of the PD diagnosis system take into account main aspects of the
system including initial purchasing, training, data acquisition and evaluation, maintenance and operational costs.
This is to ensure realistic benefit considerations are gained from the system once purchased.
An example of the total cost associated with a PD diagnosis system comprising 60 UHF sensors is described
below:
Equipment cost (sensors, acquisition unit and software): 350 k$
Installation and commissioning cost: 65 k$
On-site training cost: 25 k$
Communication link cost: 5 k$
Page 22
Consulting/expertise cost (per year): approx. 10 k$
Maintenance cost (per year): approx. 10 k$
Amortization cost (per year): 30 k$
Cost of external UHF sensor (per unit): approx. 1 k$
Typical electronic components of the used equipment are expected to have live times of around 15 years. This
would amount to a total of 1.1 M$ for the investment, use and maintenance of a continuous PD diagnosis system in
the course of 15 years.
Additionally, the following example resumes the cost associated with a failure. In this specific case, a major phase
to ground fault occurred on a 315 kV GIS, which resulted in expenses of up to 1.2 M$:
Utility's labour cost: 400 k$
Manufacturer's labour cost: 120 k$
Replacement parts: 700 k$
These expenses did not include lost income (k$) related to the outage, that lasted 3 months in that specific case. A
main factor in the determination of failure costs can be penalties for loss-of-energy generation or costs for
importing/transfering energy, which could be more than 10 times higher.
Another example considers the costs of a continuous PD diagnosis system using the UHF method applied to HV
GIS in a large DC grid connecting system. The costs given below are based on rough estimations, and do not claim
to be complete.
Costs for a continuous UHF PD diagnosis system:
For equipping a DC platform with two independent PD monitoring systems using the UHF method the following
expenses are assumed for a 30 years deployment:
Installation ready-for-use of 2 independent PD diagnosis systems (one for 155-kV-GIS and one for 420-kV-GIS)
incl. all corresponding secondary equipment (installation in the yard):
approx. 1.5 – 2 M€
Costs for service (accumulated) and Maintenance (including relevant site visits):
approx. 300 – 400 k€
Consultant (accumulated) / expertise: approx. 300 – 400 k€
Depreciation (accumulated): approx. 1 M€
Therefore all investment, use and maintenance costs amount up to 4 M€.
On the other hand the costs for a major failure leading to an unplanned outage have to be calculated. In this
example only the costs for the regulated (by law) compensation for a failure which will lead to an outage of
240 hours (10 days) are considered. This outage duration should reflect the reality quite well, since the utility in this
case claimed to have all spare parts available. Any additional costs (e.g. for repair) are not considered, since they
turn out to be much lower than the outage compensation costs.
Failure in 155-kV-GIS (wind farm feeder):
100 MW (e.g. one wind farm feeder) * (190 €/MW*h) * 240 h * 0,75 (reduction factor for not 100% wind
power delivery) = 3,42 M€.
Failure in 420-kV-GIS:
Page 23
800 MW * 190 €/MWh * 240 h * 0,75 = 27,36 M€ for a failure in the busbar
or
400 MW * 190 €/MWh * 240 h * 0,75 = 13,68 M€ for a failure in one of the redundant bays
In the following four scenarios a)…d) are described explaining the development of outage compensation costs with
regard to the accuracy of the results of a PD diagnosis on the 155-kV-GIS. Any additional costs (e. g. for repair) are
only roughly estimated.
PD diagnosis system gives a clear indication and localization of a PD defect and a major failure can be
prevented by timely intervention:
The repair will be done during a regular outage so no outage compensation costs have to be regarded.
Costs compensation for 48 h outage: 0 M€
Costs of repair (only worktime): 0.1 M€
Sum: 0.1 M€
PD diagnosis system detects a false signal and gives an alarm:
It is assumed that after some additional investigations the GIS is de-energized and opened as soon as
possible in order to remove the cause of the detected PD signals. Spare parts are available so the outage
time is estimated to be 48 h. During the repair works no irregularity is found and therefore concluded, that
the signal was a false alarm.
Costs compensation for 48 h outage: 0.7 M€
Costs of additional investigations: 0.02 M€
Costs of repair (only worktime): 0.1 M€
Sum: 0.82 M€ per false alarm
PD diagnosis system gives an unreliable result:
After further investigations, the decision is made not to open the GIS, but to observe the PD signal further:
Costs of further investigations 0.02 M€
Sum: 0.02 M€
PD diagnosis system gives no alarm and a major failure occurs afterwards:
Spare parts are available therefore the time to repair is estimated to be 10 days. The costs for repair are
higher than for the repair of a PD defect because of additionally damaged parts and cleaning due to
decomposition products:
Costs compensation for 240 h outage: 3.42 M€
Costs of repair: 0.2 M€
Sum: 3.62 M€
In the cost benefit analysis a reduced insurance premium due to the installation of a PD diagnosis system is not
considered, however, can increase the benefit. A higher benefit is also possible if maintenance strategies are
adapted to the outcome of the PD diagnosis system. The investment costs of the PD diagnosis systems are given
to 1.5 M€…2.0 M€ and have to be considered for all scenarios described below.
Page 24
In summary it can be pointed out, that there is no hard line which gives a clear investment decision for or against
any PD diagnosis system. The most important issue is the reliability of the PD diagnosis result. This means the
more trusted the results of the PD diagnosis system are, the higher the benefit of the investment in a PD diagnosis
system will be. In cases where legal requirements or the criticality of the substation are of much higher importance,
investment costs might not play a role at all.
3.5 Spare part policies optimized by PD diagnosis

In chapter 2 some hints are already given concerning spare parts. If PD activity is detected and intervention
decided due to a risk assessment, spare parts are needed do perform the repair. Probably, it is not feasible to just
purchase the required spare parts on the spot, since delivery times may be several months. Also, there is always
some uncertainty in the determination of the kind and location of PD defect. Therefore, much more spare parts for
other replacements need to be available in order to refurbish the switchgear occasionally with other parts, which
have a known failure probability. Consequently, the user probably needs to purchase and keep the same spare
parts on-site as with time based maintenance.
Condition based maintenance as such might save some spare parts, for example, if a circuit-breaker is only
maintained after a certain number of operations or fault interruptions instead always after a certain time interval.
Since nowadays, switchgear often is maintenance-free for 20 years, a condition based maintenance could hardly
save spare parts, since one set of these spare parts should always be on stock in order.
While the initial intent in the scope of this work was to include an evaluation of the merits of PD diagnosis on
optimizing spart part and redundancy policies, it could not be properly assessed due to limited Working Group
knowledge of this topic.
3.6 Comparing effectiveness of PD diagnosis against other

diagnostic techniques e.g. S F 6 analysis
In GIS, different types of electrical activity besides the normal switching duties may occur: partial discharges,
sparks and arcs. Partial discharges have their origin in ‘voids’ and ‘cracks’ of solid insulating material, in protrusions
emitting electrons (corona discharge) or in moving particles giving rise to a temporary gas discharge. Sparks arise
during opening and closing of disconnect switches, whereas arcs of varying intensity occur during load switching
and short-circuit and fault current interruption [14].
Except for partial discharges inside solid material, all these electrical activities involve the insulating gas so that SF 6
molecules are decomposed and ionized. By reaction with moisture (H2O) and different wall materials (in particular
the oxygen in metal oxides), a range of by-products can be created from the decomposition molecules. Experience
has shown that several long-life by-products i.e. SOF2, SO2F2 and SO2 remain, which are the most abundant
substances after electrical activity. In particular, SO2 can easily be detected by sampling the insulating gas into a
detection tube sensitive for SO2. Other detection tubes are available, which are in general sensitive to acidity, and
therefore cover SO2, HF and SO2F2. Lately, some portable analyzers came into the market, which may allow
analyzing by-products on-site periodically.
The probability of detecting decomposition products or by-products is determined by several physical processes as
follows:
the formation rate of by-products as function of the arc or discharge energy,
the transport speed (diffusion or convection) of by-products from the arc or discharge location to the walls of
the compartment, where the detection device is positioned,
the reaction rate of by-products in particular with water molecules, which can be in the order of 10 to 100
days in the case of SO2, but also much shorter for other molecules,
the capture rate of by-products in adsorber materials installed in the GIS depending on the properties and
the location of the adsorber material.
Page 25
In compartments of GIS such as bus-bar and bus-duct sections, which do not contain load switching equipment,
but only bushings, earthing switches and instrument transformers, any level of decomposition products would be a
cause for concern, independently of whether the compartment includes adsorber or not. In this case, a check of the
by-product concentration SO2, for example, would provide additional information to an alarm of a PD detector. The
gas analysis could differentiate corona activity from internal partial discharges in solid material. However, the
location of the activity cannot be determined by gas analysis. Here, the different PD detection methods have a
much better capability. So far a correlation between results from a SF 6 analysis and the kind of a PD defect has
been evaluated only with the help of laboratory test set-up’s with defined conditions. Corresponding results from a
cross correlation with a SF6 analysis performed under on-site conditions are not reported.
In gas compartments, which contain a switching device i.e. a load switch or circuit-breaker, for example, there will
always be some background level of by-products, surely more if no adsorber is installed. This background level is
normally of no concern. In order to distinguish between unusual electrical activity and normal switching activity, the
time delay between the last switching operation and gas sampling instant could be evaluated e.g. by sampling the
gas again after some time. In compartments fitted with adsorber capable of trapping SF 6 decomposition products,
the level of by-products will always reduce over time. If the level does not decrease then it might indicate discharge
activity or possible overheating, though it could also indicate an adsorber ceasing to function correctly. Because of
these inferring processes, it is extremely difficult, if not impossible to detect an incipient failure in a compartment
with switching device. In this case, gas analysis in parallel to PD detection will not provide useful information.
The Working Group is aware of pilot installations in some countries on SF 6 analysis on-line monitoring equipment
used for HV GIS condition assessment. But it is the general understanding of the WG that this kind of diagnosis is
very vague in the outcome.
Page 26
4 SURVEY OF CASE STUDIES

Users have reported a range of experiences with PD diagnosis giving both positive and negative results for
successful condition assessment as described in the following cases:
4.1 Exam ple from Australia – Periodic insulation condition

monitoring of h ybrid switchgear (MTS)
Since 1999, Powerlink Queensland in Australia has constructed a significant number of new substations using the
hybrid GIS switchgear concept (MTS - Mixed Technologies Switchgear) Figure 14. Nearly 60 bays at 275 kV and
330 kV, and over 150 bays at 110 kV and 132 kV are in service [11].
Figure 14 : Photo showing a MTS
Depending on its operating voltage level and importance in the network the commissioning of the substation
require HV testing only for voltage levels 275 kV and above. But practical experience has shown that even if the
MTS is completely tested during the production process and no disassembly for transport is necessary, failures due
to dielectric irregularities have been occurred. Therefore a cost-effective method for condition monitoring of MTS
was applied by use of external window sensors for the UHF method and PD diagnosis at operating voltage. In
Powerlink’s case internal sensors could not be justified economically at the time of initial purchase and to retrofit
them has also been assessed to be too expensive, and moreover require difficult-to-obtain outages. Therefore
external sensors where specially designed to fit in the disconnector inspection windows Figure 15. Alternative
methods for GIS condition assessment (e.g. SF6 analysis) have not been taken into any account.
In this case the PD diagnosis was performed periodically, since a continuous procedure was considered not to be
cost-effective. The time span between periodically site-visits has been assessed depending on the outcome of the
previous PD diagnosis.
Page 27
Figure 15 : Photo showing PD measurements at MTS

a) Diagnosis equipment is installed in the rear of a car, b) On-site overview, (c) and (d)
external sensors placed on the inspection window of each MTS pole
Since the PD measurement in this case was highly affected by noise interference originating mainly from bus-bars
and overhead line connections a high-pass filter was used to improve the signal-to-noise ratio of the PD
measurement. When performing the PD diagnosis steps have been done to confirm the location of the PD defect
inside the switchgear. According to the measuring result shown in Figure 16 the PD defect location was assessed
on pole B. The PD measurements have been done periodically over a period of several months, and the detected
PD signal exhibits consistently.
Figure 16 : UHF spectrum of MTS pole A (green), MTS pole B (blue), and a sensor
placed on the ground close to the MTS unit
After a high level of confidence in the location of the PD defect a convenient outage was planned to conduct an
inspection. This process was strongly influenced by the past experiences with this type of MTS equipment, which
has experienced already dielectric failures. According to data given the ‘risk of failure’ was evaluated to ‘1 failure in
10 years’ which consider a probability of an internal dielectric fault occurring in the complete population of a type of
MTS equipment installed. The economic loss was estimated to be more than 20 times the cost of preventive
maintenance/repair. So it was assessed as ‘significant risk’ requiring a ‘short-medium term intervention’.
The repair procedure can basically be different for MTS and GIS. In this case after the inspection showing a
flashover trace on an epoxy-resin spacer, the spacer was replaced by the utilities maintenance team, and the unit
was back energized again the same day. A subsequent PD measurement does not show any present PD signals.
This experience has not led to a higher investment of the utility in spare parts.
Page 28
4.2 Exam ple 1 from China – Per iodic PD diagnosis on 500 kV MTS
with uncertain PD defect identification
In the course of a periodic on-line PD diagnosis during maintenance in July 2012, PD activities in one unit of a
500 kV MTS (hybrid GIS) have been detected. For a better assessment of the initially observed PD signals, a
continuous PD diagnosis system was installed afterwards for about 1 month. Since the accessibility with external
barrier sensors was not optimal the measuring sensitivity available was limited. The distance between the two
measuring locations at the isolated earthing switches on each side of the circuit breaker gas compartment was
about 8 meters. In Figure 17 till Figure 20 the situation on-site is shown.
Figure 17 : External view of MTS Figure 18 : Position of external

unit UHF sensors on the left and right
side of the insulation of an
isolated earthing switch
Figure 19 : Temporary installation Figure 20 : Barrier sensor –

of continuous PD diagnosis exemplary view
system
Page 29
Some exemplary results received from this PD diagnosis can be summarized as follows:
Signal amplitude and repetition rate were varying over the investigation period with no clear trending to be
extracted,
Acoustic PD diagnosis, SF6 analysis as well as check of moisture content of the SF6 on the MTS unit have
not given any abnormal result,
Location of the detected PD signals have not been able to found out doubtless, so the assessed location
was assumed thereupon,
Type of PD defect was also not assessable.
After involved parties concluded to de-energize the affected equipment, and to first dismantle one module
containing the isolated earthing switch which was assumed to be the origin for the PD signals. The module was re-
tested at the manufacturer’s laboratory with no irregularity found. Afterwards the outdoor bushing on the
transformer side of the MTS unit was also dismantled, since it was considered also as possible source for the PD
signals, and re-tested at the manufacturer’s laboratory with no irregularity observed too. Subsequently, the MTS
unit was re-completed on-site with a new bushing and re-energized. Diagnosis tests as moisture content
measurement and PD diagnosis were performed in this context showing no irregularity.
All of the 500 kV MTS in this area are from the first batch of the hybrid GIS products. These types have also been
faulty in other regions detected by PD diagnosis, which indicates problems in the assembly and process control.
But afterwards this type of equipment operates without any further irregularity, so generally negative experiences
can’t be assessed.
This case can be summarized as follows:
UHF method adopted for PD diagnosis was too sensitive on the one hand side, and superior in detection
capabilities on the other hand side compared to other common methods; faulty PD signals disappeared after
disassembly and thoroughly cleaning, so in unit were small PD defects which did not affect the normal
operation of the equipment,
Unnecessary overhauls can be triggered by a high detection sensitivity of the applied UHF PD diagnosis
system; moreover contamination on the surface of insulators can also lead to misleading UHF signals.
4.3 Exam ple 2 from China – PD diagnosis on 500 kV MTS with

identified PD defect
During the dielectric on-site testing of many units of a 500 kV MTS, several locations have been found with PD
signals. Basically the UHF method was applied, and for locating purposes acoustic PD diagnosis was added.
Finally the result of the PD diagnosis was that PD defects of type ‘particles’ are active in 2 out of 9 phases in the
hybrid GIS. The results are summarized in Figure 21 till Figure 24.
Page 30
Figure 21 : UHF PD signals on phase A Figure 22 : UHF PD signals on

of circuit breaker no. 5011 close to phase B of circuit breaker no.
isolated earthing switch amplitude is 5011 close to isolated earthing
about 40 mV switch
Figure 23 : UHF PD signals on phase C Figure 24 : On-site inspection of

of circuit breaker no. 5011 close to phase C (exemplary shown as
isolated earthing switch amplitude is result on phase A was similar)
about 60 mV
The particles found in the shown locations of phase A and C of circuit breaker no. 5011 have been collected and
analyzed further in a laboratory with the help of SEM (scanning electron microscope) and EDS (energy dispersive
spectrometer). The composition of the impurities includes metal materials such as aluminum, silver, copper,
chromium, and compounds such as fluoride, fiber and epoxy. The longest length of the metallic particles was
diagnosed to about 2.7 mm.
It is suggested that the scope of the inspection should be further expanded. If the phenomenon of abnormal
impurities gathered is generally found all the hybrid GIS equipment of this station should be thoroughly cleaned. A
large-scale disassembly on-site was difficult to implement, so it was recommended to return the equipment to the
factory to carry out this work.
Page 31
4.4 Exam ple from Hungar y – Retrofit of a 275 kV HV GIS with a

continuous PD diagnosis system
Around 6 years after the initial installation of 4 HV GIS in Malaysia one of them (Jalan Pantai 275 kV) should be
retrofitted with a continuous PD diagnosis system. The HV GIS was originally delivered without internal UHF PD
sensors, and comprise 10 circuit breaker bays, double busbar including busbar sectionalizer. At first possible
internal UHF PD sensors has to be proposed and tested. In Figure 25 and Figure 26 the chosen types are shown
as well as the intended flange locations.
Figure 25 : PD sensor type 1 Figure 26 : Possible flange

and 2 location for PD sensors type 1 and
2
The general sensitivity of the proposed UHF PD sensors was measured in a GTEM cell test set-up, and the results
are summarized in Figure 27 and Figure 28.
Figure 27 : Effective height of PD Figure 28 : Effective height of PD

sensor type 1 (180 mm disc sensor type 2 (75 mm disc
diameter) diameter)
It was evaluated that the PD sensor type 1 has 11.6 mm mean effective height, and in the 500-1500 MHz
frequency range the sensitivity curve is in 100 % above 2.0 mm. PD sensor type 2 has 6.8 mm mean effective
height, and in the 500-1500 MHz frequency range the sensitivity curve is in 98 % above 2.0 mm. Finally it was
concluded that the chosen PD sensors design was such as the highest sensitivity possible for the given GIS
geometries.
Page 32
More information regarding PD sensor sensitivity and corresponding parameters like mean effective height can be
found in detail e.g. in [26].
Subsequently the dielectric withstand capability of the GIS modules accommodating the different PD sensors had
to be demonstrated. For that purpose a test set-up (Figure 29) was used. The test result was satisfactory after
passing a test program fulfilling the corresponding IEC standard.
Figure 29 : Test set-up for dielectric withstand verification of GIS modules containing
PD sensors
The positioning of the internal UHF PD sensors was recommended on both sides of the circuit breaker module as
well as at suitable locations at busduct ‘corner modules’ and GIS modules connecting other equipment like HV
cables or HV transformers. In this case due to ‘no free flanges’ any PD sensors were placed between circuit
breaker and busbar module. To overcome a lack of sensitivity in the ‘busbar’ section additional PD sensors were
positioned at each end of the busbar and in the busbar sectionalizer section. Additionally the different outgoing
busduct sections were equipped with PD sensors also. For example, the section between circuit breaker and
transformer bushing (SF6/Oil) module consist of 5 PD sensors per phase. An example is shown in Figure 30.
Page 33
Figure 30 : Exemplary situations for retrofit HV GIS with internal UHF PD sensors
After the installation of all PD sensors to the HV GIS a HV test was performed on-site. The reason for conducting
this test has been some repair works after a transformer fire had damaged some GIS sections. This test would
have been invalid if the only impact on the GIS has been the retrofit of the PD sensors. A test for sensitivity
verification was not reported in this case study.
Finally some experiences were shared regarding the operational behaviour of the PD diagnosis system. Overall the
user has been satisfactory with the system’s performance. Occasionally the results of the PD diagnosis system
triggered maintenances activities on the GIS manufacturer side. In a time span of approx. 3 month 21 different PD
signals in total have been arisen in this context. In 13 cases a further investigation and localization of the PD defect
was recommended. The registered PD signals are reported of the types ‘mobile particles’, ‘floating elements’, and
‘voids’. Further locating activities (e.g. time-of-flight measurement, acoustic PD diagnosis system) have not been
reported as well as how to follow-up with the generated results from the PD diagnosis system.
4.5 Exam ple from Ireland – ES BI’s PD condition monitoring

strategy for HV GIS
GIS Installations:
ESBI (Electricity Supply Board International) installed its first HV GIS (220 kV) in late 1970’s. This GIS is still in
service to this day. It is a dual pressure system with a blast valve for arc extinction in the circuit breakers. This is
the only dual pressure GIS remaining on the ESBI system. Since then there was a continuous growth in
development of GIS stations at 110 kV, 220 kV and one station at 400 kV. The table below summarizes the
installed GIS currently on the system with several more under construction.
Page 34
VOLTAGE LEVEL NUMBER OF INSTALLATIONS

110 kV 40
220 kV 10
400 kV 1
Table 8 : Overview of ESBI’s HV GIS population
Early PD measurements & trials:
During the installation of all early HV GIS there was no decision taken to install sensors to allow for PD continuous
or periodic monitoring. There was also no specific policy on PD measurement using alternative techniques. Several
trials have taken place over the years including the use of UHF ‘window sensors’ and ‘gas barrier spacer sensors’
with a continuous on-line PD monitoring system. The first trial took place on a 7 bay, double bus; 400 kV HV GIS
installation in the mid 1980’s using UHF ‘window sensors’. The installation monitored 2 bays, and was left in place
for several weeks. No activity was detected, and the equipment was removed. However for the following reasons
no decision was taken to install continuous on-line PD monitoring systems:
The costs of the installations and on-going maintenance were considered prohibitive,
The benefits to be derived from continuous on-line PD monitoring were perceived as uncertain,
The faith in the technology was uncertain,
The ‘black-art’ interpretation of the signals meant a heavy dependence on the measuring equipment
supplier.
HV GIS flashovers and subsequent PD measurements:
On the same installation, where the continuous on-line PD monitoring system was trialed, two flashovers occurred
on horizontal gas barrier spacers, which were directly below the busbar and maintenance earthing switches on the
coupler bay. This resulted in half station outages with considerable costs. Particles generated by the contacts of
the disconnector switches and loose corona shields rubbing against conductors fell on to the upturned horizontal
spacers resulting in flashovers. As a result of this, a major modification was made to the HV GIS to remove these
horizontal spacers from the coupler bay and replace them with vertical spacers. This of course required new
enclosures and a major modification.
As a result of these problems, a decision was made to annually survey this HV GIS using an acoustic technique.
The acoustic measuring equipment developer was contracted to visit the site annually to survey for PD. These
periodic surveys were carried out for about 5 years in the late 1980’s and early 1990’s. During this period some
successes were recorded in detecting loose corona shields and some particle generation in one location.
Policy decision to use periodic acoustic technique for PD measurement:
Following the perceived success of the acoustic technique, a decision was taken to evaluate acoustic instruments
commercially available on the market. Two instruments were evaluated including trials and visits to the HV
laboratories of both locations. Subsequently one device was purchased in the mid to late 1990’s. Since then all GIS
installations rated at 110 kV and above are annually surveyed for PD activity using this instrument to this day.
Some successes were recorded over the years, as:
220 kV voltage transformer was removed from service; subsequent investigation revealed a loose clamping
leading to a floating shield, the voltage transformer was repaired,
Particles found in the 400 kV HV GIS; compartments were opened and cleaned without incident.
However some possible PD activity which may have been detected was missed, as:
220 kV HV GIS – disconnector switch:
Page 35
2 flashovers occurred in busbar disconnector compartments; subsequent investigations revealed very

significant particle generation as the disconnector switches were not rated for bus transfer switching;
investigation of other bus disconnector switches revealed the same problem; it is assumed that this issue
may have been picked up with a continuous on-line PD monitoring system.
Figure 31 : Busbar disconnector Figure 32 : Flashover in

switch female fixed contact disconnector switch compartment
400 kV HV GIS – grading capacitors in circuit breaker:
Following an internal inspection and overhaul of 400 kV circuit breakers significant arcing and tracking was
found in one capacitor prior to failure; it is assumed continuous on-line PD monitoring with internal sensors
may have tracked this activity.
Figure 33 : Grading capacitor for Figure 34 : Grading capacitor for

400 kV circuit breaker 400 kV circuit breaker - stripped
Page 36
Figure 35 : Grading capacitor for Figure 36 : Grading capacitor for

400 kV circuit breaker - plates 400 kV circuit breaker - plates
with visible punctures close up
Recent developments – delivery of new HV GIS only with UHF sensors:
Since early 2013 with the placement of a new supply contract for GIS all suppliers for 110 kV, 220 kV and 400 kV
switchgear were requested to provide capacitive or window type sensors integrated into all bays and busbars. The
intention was to provide access points for the later addition of PD diagnosis systems by the customer for life
extension of the switchgear. Integrated sensors were chosen as would allow more direct comparison with baseline
PD measurement performed during the on-site HV test. The use of external acoustic measurement was considered
however, integrated capacitive sensors were deemed less susceptible to external noise sources and fixed location
within the GIS will provide a more comparable measurement against baseline PD fingerprints. As the UHF sensors
are used by the manufacturer during the on-site HV test with PD measurement, the sensors themselves may
therefore be seen as ‘proven’. In one installation following localization of a PD source the sensor itself was found to
be the source due to mishandling during installation in the factory.
Figure 37 : Window type sensor showing scratches from factory installation

discovered during initial on-site PD measurement
Recent developments – several flashovers:
At the time of the design review stage of the new HV GIS supply contract several issues were encountered with
other HV GIS installations at their commissioning and initial energization phases. A 400 kV HV GIS installation in
Turkey suffered flashover several months following energization. Investigations relieved a disconnector that was
incorrectly aligned caused localized heating when in the closed position eventually leading to melting of the
Page 37
contact, displacement of molten material and flashover (Figure 38). It is probable that had active continuous on-line
PD monitoring been employed during the early initial period that the abnormal PD likely to have been generated
may have given warning of such an event.
A 220 kV HV GIS circuit breaker suffered flashover 40 seconds after initial energization. Particles contained within
the arc shield of the interrupter that were not detected during the on-site HV test following final assembly were
thought to have been dislodged during mechanical operation and testing during the commissioning phase (Figure
39).
Figure 38 : 400 kV HV GIS Figure 39 : 220 kV circuit breaker

disconnector moving contact with traces of flashover
showing local heating and
burning
While active PD monitoring may not have detected this particular fault a large concern remains in place over the
existing circuit breakers in service which may also contain particles that have not yet migrated to critical areas of
the circuit breaker to cause flashover. A second circuit breaker pole was removed from service for investigation
based on manufacturing records.
Decision to fit UHF sensors to all new HV GIS:
The initial reason for the decision was taken on life extension of critical assets to be utilized in the future. However,
given the number of manufacturing and assembly related quality issues that have arisen in recent installations, the
migration to more compact types of HV GIS and general concern toward over the economizing of production the
use of UHF sensors is now foreseen as a critical method of ensuring acceptance of factory and site assembly
quality.
Decision to go for periodic PD monitoring using UHF sensors:
At present, due to ongoing quality concerns and the high number of flashover events during on-site HV tests one
manufacturer has been requested to provide active periodic on-line PD monitoring as a provisional quality
assurance measure before handover acceptance. In conjunction with the manufacturer and based on the utilities
experience the bathtub curve of reliability for GIS is highly appropriate. Active periodic on-line PD monitoring of
critical installations in conjunction with increased frequency of periodic on-line PD monitoring is foreseen as a
required quality assurance measure to provide confidence to the user that any unsatisfactorily resolved quality
issues are monitored and early warning provided where possible. This particular quality assurance measure is
being applied to new HV GIS for which there is little service history to base reliability data on.
No decision to implement continuous on-line PD monitoring:
At present an initial three year period of increased monitoring (active periodic on-line PD monitoring) is foreseen. If
during this period further quality issues are discovered then it is likely a longer period will apply, even a continuous
Page 38
on-line PD diagnosis can be applied depending on the severity of the previously received results. However, at
present due to data export capability in remote substations, reliability of the continuous on-line PD diagnosis
systems themselves and compatibility issues with centralized collection and condition monitoring systems, it is not
foreseen to adopt a continuously actively monitored on-line PD diagnosis system. Acoustic PD diagnosis is still
performed in parallel with UHF PD diagnosis to gain complementary results.
Regarding the different investment costs of the different diagnosis devices no specific cost-benefit-analysis has
been performed in advance.
A risk assessment approach implemented at ESBI can be sketched as follows:
Basically the applied risk assessment approach is initially fed by two independent input parameters:
Risk for component to fail (influenced by severity and probability),
Risk to the network (influenced by severity and probability).
The further assessment results then in a 5-level-categorization ranking basically based on the assumed action.
Level 5 means the lowest value for time to action (this can also take e.g. 2 month until next possible outage). An
immediate action (very short-term basis resp. as soon as possible) is very unlikely. The infant mortality phase of HV
GIS is taken generally up to 2 years; but exceptions are possible. It also includes abnormalities during routine and
commissioning test.
4.6 Example from Japan – Reasons in CEPCO for discontinuance

of on-line PD diagnosis in HV GIS
Formerly (around 1990’s) in CEPCO (Chubu Electric Power Company) PD sensors for on-line PD diagnosis had
been installed at the busbars of HV GIS in new main 500 kV and 275 kV substations of certain importance in the
network. However, on-line PD diagnosis has been discontinued due to the following reasons (from the late 1990s
onwards):
Assembly errors during manufacturing process and other faults usually can be detected during the factory
test; the quality can be verified (e.g. voids and the gaps inside the insulation materials can be found) by
conventional PD measurement with sufficient detection sensitivity,
Meanwhile, portable UHF PD and acoustic sensors can be used also on-site for verification tests like site-
acceptance-tests; even if foreign particles due to assembly work exist inside the HV GIS, AE sensors can
detect the foreign particles accurately,
During operation the HV GIS enclosure is tightly closed and no foreign particles can contaminate the HV
GIS,
Furthermore, failure rate of HV GIS has been a substantially decreasing trend due to strict and high quality
control at the Japanese manufacturer’s side.
SURVEY YEAR FAILURE RATE

(10-3 cases/unit-year)
1983-1991 0.450
1992-1997 0.200
1998-2001 0.068
Table 9 : Failure rate of HV GIS in Japan [15]
In contrast to the on-line PD diagnosis in CEPCO in the period from 1995 till 2004 off-line PD diagnosis has been
conducted, which did not give concerns to abnormal dielectric integrity of the investigated HV GIS. In this surveys
acoustic PD diagnosis and PD diagnosis with ‘surface current detectors’ [16] were applied.
Page 39
Finally a PD verification test is conducted only after the installation and the dismantling work when internal
components of the HV GIS were exposed to air.
In addition it can be mentioned that the presented practice is common at Japanese utilities, e.g. Kansai Electric
Power Company and Tokyo Electric Power Company started around 1999/2000 to end the operation of continuous
PD diagnosis systems mainly due to low failure rates of HV GIS during operation, and inappropriate operational
behaviour of the continuous PD diagnosis equipment. But the application of periodic PD diagnosis still exist (e.g.
after maintenance work, after inspection/opening of gas compartments) where these users apply PD diagnosis.
The described practice does not influence the spare part policy of the Japanese utilities in terms of keeping more
spare parts ready to avoid longer outage times. Power companies and manufactures in Japan have made great
efforts for preventive maintenance policies. A lot of case studies about failures were conducted and those results
were basically shared by utilities in order to study, plan and set up sufficient counter measures before to avoid
possible failures. In this sense, many spare parts are not prepared. Some special components such as bushings
(long lead time) may be stocked by utilities as spare, but this is not directly linked to the current practices in PD
diagnosis in Japan.
In case a failure occurred and energy supply is e.g. interrupted it do not exist any regulatory penalties in Japan for
non-supply of electric power, neither in Electric Business Act (law) nor in Electricity Supply Stipulation of each
power utilities.
4.7 Exam ple from Korea – GIS condition assessment on a 154 kV

substation of a semiconductor factory
The semiconductor industry in Korea requires large amounts of reliable energy. Loss of supply or unreliability can
cause significant financial cost. Therefore, substations of a semiconductor factory are managed as critical facilities.
In a 154 kV substation of a semiconductor factory in Korea, an on-line PD monitoring system based on the UHF
method has been installed in 2011 on a 170 kV GIS (installed in 2002) which has been in-service for more than
10 years. The decision to install an on-line PD monitoring system was influenced additionally by a good operational
experience of the main Korean utility as well as some past experiences with failures on similar types of GIS. A kind
of purchase procedure has been set up (Figure 40).
Figure 40 : Decision making process to install an on-line PD monitoring system
An alarm had occurred in the 1 bay of the GIS on the on-line PD monitoring system in October 2012 (Figure 41).
The alarm was kept for 10…20 minutes and cleared. After two hours, the alarm had occurred again and it was kept
for one hour and cleared. This kind of occasional alarms may occur because of PD. However, a possibility that any
external corona discharge may inflow still remains and thus analysis of PD data with trend and phase pattern was
carried out in parallel.
Page 40
Figure 41 : PD Alarms on on-line PD monitoring system (Sensor #157, #170 and #180)
In Figure 42 exemplarily some PD data showing a trend graph are displayed. PD alarms had occasionally occurred
from October 23 to 29. as a repeating creation and clearing of PD alarms. However, the PD alarm was continuously
kept after October 30. Considering the fact that the PD alarm was getting longer and occurred continuously, it was
highly possible that the cause of PD alarm was an internal PD source rather than external noise or corona signals.
a)
b)
c)
d)
Figure 42 : PD alarm trend graph

a) October 23, 2012 PD alarms occurred
b) October 24, 2012 PD alarms occasionally occurred
Page 41
c) October 25 – 29, 2012 no PD alarm occurred

d) October 30, 2012 PD alarms re-occurred and continued
In Figure 43 exemplary PD frequency and phase pattern data of sensor #160 is shown. The left side of the figure is
linked to data in a frequency span ranging from 500 till 1500 MHz, and the right side of the figure displays data in a
phase angle ranging from 0 till 360 . The pink marked section in the frequency spectrum was to mask frequency
ranges of noise interference. Past experience with this kind of PD defect in similar location has influenced the
further procedure. A signal for phase synchronization was provided according to Figure 45.
a)
b)
c)
Figure 43 : PD frequency and phase pattern diagram of sensor #160

a)…c) separate PD measurements within approx. 10 min.
In Figure 42 and Figure 43 above the diagrams display a vertical scaling in pC though the UHF method was
applied. An estimated relation was found in previous investigations (Figure 44) and other references [12] since is
assumed to be more difficult to understand the physical meaning of dBm for less experienced users of a PD
monitoring system. In case of PD defect locationing the PD monitoring system displays the corresponding pC level
considering the attenuation according to the PD location.
Page 42
The general understanding of the Working Group is not to use pC in the context of UHF and acoustic methods for
PD measurement.
Figure 44 : Test set-up for assessing a correlation between pC and dBm signal levels
Figure 45 : Configuration of on-line PD monitoring system for HV GIS
The used on-line PD monitoring system performs a frequency and phase spectrum analysis, not a PRPD pattern
analysis. Basically its functionality is summarized in Figure 46. Some areas within the detection frequency span are
masked by users to eliminate noise signals depending on substation circumstances (e.g. mobile or aircraft
communication frequencies) and user’s experience.
Page 43
Figure 46 : Procedure of PD diagnosis
Finalization of the performed PD diagnosis led to the conclusion, that a PD defect of type ‘floating elements’ was
the cause for the detected PD data. For localization of the PD defect on-site PD measurements were conducted
with portable PD measuring equipment and oscilloscope. The largest signal (approx. -30 dBm ~ -25 dBm) was
measured at Phase S between and . It was determined that the PD defect was located in the section
between and . For a more precise PD locating, the PD signals were measured using an oscilloscope to
evaluate the PD defect location with the ‘time of arrival’ method. It was determined that the location was around
(Figure 47).
Figure 47 : Locating of PD defect
Furthermore no additional impact parameters (e.g. importance of substation, possibility for switching off the
affected section, costs of an outage, estimated costs of failure) have influenced the decision to shut down. The
result of the performed PD diagnosis was with a high level of confidence, and the current load was not high. The
gained confidence in the PD diagnosis result was also established by the expertise level of people from GIS and
PD monitoring system OEM. In cases when the PD monitoring system recorded a PD alarm and subsequent
analysis show that it was not a signal triggered by a PD defect, the expert system of the PD monitoring system was
updated by these data to avoid similar occurrence in the future (Figure 48).
Page 44
Figure 48 : PD diagnosis procedure
Finally the internal inspection exhibited a contamination layer close to a bolt fixing a post insulator to the enclosure
originating from decomposition products generated at that location (Figure 49). The affected bolt was not tightened
correctly.
a) contamination layer on post insulator b) loose bolt on post insulator
Figure 49 : Result of internal inspection showing the detected PD defect
4.8 Example from Brazil – Return of experience from ITAIPU DAM

power plant
ITAIPU DAM is currently the world's largest hydro-electric plant in power generation comprising 20 generator units
and 14 GW of installed capacity. After a test period of approx. 3 years a continuous PD diagnosis system using the
Page 45
UHF method was installed, and in total 462 shield type sensors at earthing switches were available as access
points to the GIS. Later also barrier sensors were available and have been integrated to the PD diagnosis system
[24].
Before this installation a cost-benefit-analysis has shown that only an continuous UHF PD diagnosis system has
sufficient technical performance for the on-site condition at ITAIPU GIS at lower cost levels than all other methods
compared, like conventional, acoustic, and other methods (chemical and optical based) [24].
During the operation of the PD diagnosis system more than 10 cases of PD detection and prevention of developing
to a failure occurred. It was concluded that after 6 cases a benefit has been created when applying the selected PD
diagnosis system as part of a condition based maintenance strategy. This covers cost to repair and investment cost
to the PD diagnosis system itself [24]. In the following 2 cases are described exemplarily to demonstrate the
capabilities of the applied PD diagnosis system.
In the first example mobile particles in a circuit breaker gas compartment were detected, and action has been taken
by the user to open the affected area based on the information given by the PD diagnosis system. After removal of
the PD defect the GIS component has been put back into operation Figure 50.
Figure 50 : PD defect ‘mobile particle’ in circuit breaker gas compartment [24]
The second example describes the detection of a void in a barrier insulator. The PD defect was analyzed further
after localization and removal from the affected gas compartment. The size and position of the void is displayed in
Figure 51.
Figure 51 : PD defect ‘void’ in barrier insulator [24]
Page 46
4.9 Exam ple from various countries – Reliable PD diagnosis for

condition assessment
In a first case description during commissioning and energization works in a 400 kV GIS PD signals started to
appear at the UHF couplers closest to the interfaces between the GIS and several Transformers and Reactors. The
PD signals started within a few hours up to several days after the first energizing of the GIS connections to the
transformers and reactors. The 100-day-history of one typical PD coupler is shown
Figure 52 : 100-day-history of a PD coupler in 400 kV GIS showing PD
From the starting point of the discharges the magnitude of the PD signals steadily increased to a stable level. The
very high level of the PD signals caused saturation of the continuous UHF PD diagnosis system at some PD
couplers. It was necessary to attenuate the signals from the PD couplers to allow the development of the
magnitude of the discharges to be seen. Analysis of the PD signals indicated that most discharges occurred shortly
before the peak of the power frequency cycle. At this time the location of the PD signal could not be identified as it
lay beyond the last available PD coupler therefore time-of-flight measurements to locate the PD source were not
possible. The development of the discharges did not stop. After remaining at a stable level for some time PD
signals started again to increase in magnitude.
Based on current findings it was decided to perform further investigations on-site. All involved parties (GIS
manufacturer, transformer manufacturer, PD diagnosis system manufacturer and user) participated in the on-site
diagnostic measurements. The results clearly showed that a SF 6/Oil bushing was the source of the discharges. The
rapid onset of the PD signals from an initial zero baseline and it’s amplitude trend as well as the position of the
detected void in the critical field area near the inner stress shield suggest that this PD defect would not have
stabilized on the long term, and it would have continued to grow until the insulator failed. Further investigations at
the bushing manufacturer laboratory verified the on-site results. Voids were discovered inside the bushings.
The description of a second case is related to a 400 kV GIS on a nuclear power plant where a failure of the GIS to
take generated power can lead to significant penalties in addition to loss of supply and high repair costs. A
continuous PD diagnosis system analyzing UHF signals was installed to give early warning of possible defects in
the GIS.
Page 47
Figure 53 : 400 kV GIS on nuclear power plant
While in service, the PD diagnosis system generated a PD alarm after detection and analysis of PD signals. These
have been continuous, and on-site investigations to locate the PD source by means of time-of-flight measurements
were conducted. Based on the signal amplitude variation between PD sensors, the general location of the PD
signal was along the busbar connection to a reactor or at the reactor itself. When arriving at site the PD activity had
stopped and remained inactive for a whole day. The PD source could not be located then. Further investigations
via remote control did not show any further activity over days and months after the initial PD alarm.
10 months later a further PD alarm was triggered in the same section of the GIS, and on reviewing the PRPS
display the activity was high both in amplitude (full scale) and number of discharges per cycle. The PD pattern was
obviously generated by a floating element.
Figure 54 : 400 kV GIS on nuclear power plant
After that the PD source was immediately located close to a hatch cover, and the section of the GIS was de-
energized and inspected. At the location of the hatch cover two busbar sections were joined as a ‘palm connection’
with a corona shield surrounding the joint. Upon inspection of the shield, carbon deposits were found on the shield.
On removing the shield to clean and prepare for secure connection to the busbar it was found that the ‘palm
connection’ had a hairline fracture which completely separated when the bolts were removed from the joint. The
fault was due to insufficient jointing subjected to the apparently high level of vibrations initiated by the reactor.
Page 48
DEFINITIONS
Periodic Measurement carried out on demand
Continuous Measurements carried out without interruption (Maybe a third definition describing
approaches in between periodic and continuous is useful? e.g. “temporary continuous”,
“permanent continuous”…)
Diagnostics Process (maybe iterative) to be performed, application of test methods
Diagnosis Result of a process
Off-line Test objects are energized by external voltage sources
On-line Test objects are in service
Monitoring Multiple measurements that allow trend analysis
Failure probability Failure probability is the extent to which a dielectric failure is likely to occur
Risk assessment Risk assessment is the evaluation of identified risks to determine whether action is
required
Condition assessment Condition assessment is the evaluation of identified conditions to determine whether
action is required
PD defect Abnormality in a component that can be detected by PD diagnostics (‘mobile particles’,

‘floating elements’, ‘protrusions’, ‘particle on insulation’, and ‘void’)
Level of confidence Reliability of a PD diagnosis result (to detect, classify and locate a PD defect)
Note: References [1], [4], [10], and [25] have been used for synchronization of some definitions.
Page 49
REFERENCES
[1] CIGRÉ technical brochure 502: High-voltage on-site testing with partial discharge measurement, 2012
[2] Lundgaard L.E., Skyberg B., Schei A., Diessner A.: Method and instrumentation for acoustic diagnoses of GIS,
paper 15-309, CIGRÉ session 2000
[3] Lundgaard L.E.: Particles in GIS – characterization from acoustic signatures, IEEE Transactions on Dielectrics
and Electrical Insulation, vol. 8, no. 6, pp. 1064-1074, 2001
[4] CIGRÉ technical brochure 525: Risk assessment on defects in GIS based on PD diagnostics, 2013
[5] Schei A., Kyrkjeeide S., Larsen V.: Acoustic insulation analyser for periodic condition assessment of gas
insulated substations, IEEE/PES Transmission and Distribution Conference and Exhibition, vol. 2, pp. 919-924,
2002
[6] CIGRÉ working group 15.03: Diagnostic methods for GIS insulating systems, paper 15/23-01, CIGRÉ session
1992
[7] Leijon M., Ming L., Hoff, P.: SF6 gas pressure influence on acoustical signals generated by partial discharges in
GIS, International Symposium on High Voltage Engineering, pp. 197-200, 1991
[8] IEC 60270: High-voltage test techniques – partial discharge measurements, 2000
[9] Ran Z., Ju T., Lei F., Yonggang L., Zhang, X.: Study on relationship between optical signals and charge quantity
of partial discharge under typical insulation defects, Conference on Electrical Insulation and Dielectric Phenomena,
pp. 1209-1212, 2013
[10] CIGRÉ technical brochure 462: Obtaining value from on-line substation condition monitoring, 2011
[11] Lopez-Roldan J., Blundell M., Allan D., Scott R. and Saha T.: Insulation condition monitoring of hybrid
switchgear, paper A3-201, CIGRÉ session 2008
[12] Kato T., Endo F. and Hironaka S.: Sensitive partial discharge monitoring system by UHF method and
calibration technique, CIGRÉ SC 15 Gas Insulated Systems Symposium, pp. 73-76, 2001
[13] CIGRÉ technical brochure 167: User guide for the application of monitoring and diagnostic techniques for
switching equipment for rated voltages of 72.5 kV and above, 2000
[14] CIGRÉ technical brochure 567: SF6 analysis for AIS, GIS and MTS condition assessment, 2014
[15] Electric technology research association (ETRA): Advanced maintenance technology against deterioration of
metal enclosed substation equipment, vol. 61, no. 3, 2006
[16] Emoto K., Higuchi K., Murase H., Koyama H., Wakabayashi S. and Sakakibara T.: A new partial discharge
detecting sensor for gas circuit breakers, International Symposium on High Voltage Engineering, paper 3184, 1997
[17] Petit A.: Diagnostic and maintenance techniques (discussions) – Special report for the second discussion
meeting – Substation components and equipment, CIGRÉ Symposium Berlin, p. 89, 1993
[18] Ali G.: A consultant’s overview of experience with condition monitoring systems worldwide, IEE Colloquium
on HV Measurements, Condition Monitoring and Associated Database Handling Strategies, vol. 9, pp. 1-5, 1998
[19] CIGRÉ technical brochure 514: Final report of the 2004 - 2007 international enquiry on reliability of high
voltage equipment - Part 6 - Gas insulated switchgear (GIS) practices, 2012
[20] CIGRÉ technical brochure 513: Final report of the 2004 - 2007 international enquiry on reliability of high
voltage equipment - Part 5 - Gas insulated switchgear (GIS), 2012
Page 50
[21] Schichler U., Reuter M. and Gorablenkow J.: Partial discharge diagnostics on GIS using UHF and acoustic
method, International Symposium on High Voltage Engineering, paper D-9, 2009
[22] Wei S., Dejun L., Schichler U. and Suermann B.: Performance of GIS epoxy insulators and related tests,
International Conference on High Voltage Engineering and Application, pp. 273-276, 2012
[23] CIGRÉ technical brochure 541: Asset management decision making using different risk assessment
methodologies, 2013
[24] Filho Rodrigues J.G. and Deligi C.: Itaipu GIS PDM system - the need, the choice, implementation and
successfully cases prevented, International Conference on Condition Monitoring and Diagnosis, paper OC1-04,
2014
[25] Kopejtkova D., Kopejtko P. and Spurny P.: Centralized IT system to support HV equipment condition based
maintenance, CIGRÉ Colloquium and Tutorial of CIGRÉ SC B3, Berlin, paper 304, 2007
[26] Judd M. D., Farish O. and Hampton B. F.: The excitation of UHF signals by partial discharges in GIS, IEEE
Transactions on Dielectrics and Electrical Insulation, vol. 3, no. 2, pp. 213-228, 1996
Page 51
ANNEX
Check list resp. addition of estimates to find out if scenarios to create a benefit exist or not:
total costs of a PD diagnosis system (one-time-cost)
cost savings (one-time-cost)
total failure costs (cost per failure)
costs of false PD notification (cost per notification)
costs of true PD notification (cost per notification)
Type of costs to be filled with values/assumptions by

users (typically case dependent)
Costs of PD initial costs

diagnosis system
costs for evaluation of a PD diagnosis

system/costs for specification
equipment cost (sensors, acquisition

system, hardware, software)
costs for data structures and server

installation
costs for basic/enhanced

commissioning
operational cost (incl. costs for

handling/training of false alarms
costs for system operators which could

also include rental fees for diagnosis
equipment; skill-level of system
operator impact the salary; additional
labor costs (dedicated employees,
etc.)
costs for miscellaneous (e. g. power

consumption)
disposal costs (listed as other costs)
costs for hard- and software upgrades
Summation of assumed values A
Saving costs reduction of insurance premium
optimize maintenance costs
life extension/delayed investments
Page 52
Summations of assumed values B
Costs of failure costs of outage (loss of energy

generation/revenue, penalty costs)
cost of repair (manpower, travelling,

spare parts, equipment rental (e. g. HV
test-set))
liability costs
environmental penalty costs
Summation of assumed values C
Costs of PD costs for complementary techniques

notification with no (e. g. for oscilloscope, acoustic
intervention required sensors in case of UHF originally used;
additional measures for improving
noise immunity)
costs for potential outage
costs for further investigations
costs for repair (manpower, travelling,

gas handling, equipment rental (e. g.
HV test-set) when required by policies)
Summation of assumed values D
Costs of PD costs for complementary techniques

notification with (e. g. for oscilloscope, acoustic
rectification work to sensors in case of UHF originally
be planned used)
costs for potential outage
costs for further investigations
costs for repair (manpower, travelling,

gas handling, spare parts, equipment
rental (e. g. HV test-set))
Summation of assumed values E
Note: For applications of this table, see also examples in chapter 3.4. There is a ‘grey’ area between D and E
where PD notification was given, no result was found from PD diagnosis, and the PD was no longer present after
intervention. This can be viewed both negative and positive in terms of a result. The removal of the PD should be
seen as positive.
The total outlay cost for a PD diagnosis system will be ‘A - B’. The cost of a failure must include the replacement of
damaged components, the time and any penalties that apply will be ‘C’. Both D and E assume that the PD
diagnosis system has generated a PD notification, and the associated costs depend on the outcome of the
investigation, on whether it was necessary or not necessary.
Page 53
Generally four typical situations are to be considered for the PD diagnosis investment depending on PD activity.
Scenario 1) No PD notification in the life of the equipment
Net cost = A - B
Scenario 2) PD diagnosis system fails to detect/prevent failure
Net cost = (A - B) + (C)
Scenario 3) PD diagnosis system giving false notification
Net cost = (A - B) + (D)
Scenario 4) PD diagnosis system giving true notification
Net cost = (A - B) + (E - C)
Example 1 (chapter 3.4) <values for 15 years, 50 sensors, all USD>:
A = 1070 k$, B = 0 $, C = 1220 k$, D = 10 k$, E = 132 k$ (estimated at 10 % of flashover repair)
Scenario 1 = 1070 k$
Scenario 2 = 2290 k$ (for one flashover)
Scenario 3 = 1080 k$ (for one notification)
Scenario 4 = -18 k$ (saving for one notification)
Example 2 (chapter 3.4) <values for one PD monitoring system for 155-kV-GIS, all in EUR>:
A = 1350 k€, B = 0 €, C = 3620 k€, D = 712 k€, E = 100 k€
Scenario 1 = 1350 k€
Scenario 2 = 4970 k€ (for one flashover)
Scenario 3 = 2062 k€ (for one notification)
Scenario 4 = -2170 k€ (saving for one notification)
Page 54

674 Benefits of PD Diagnosis On GIS Condition Assessment

Uploaded by

Copyright:

Available Formats

674 Benefits of PD Diagnosis On GIS Condition Assessment

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

674 Benefits of PD Diagnosis On GIS Condition Assessment

Uploaded by

Copyright:

Available Formats

674

Disclaimer notice ISBN : 978-2-85873-377-4

1 TYPES OF PD DIAGNOSIS AND THEIR ADVANTAGE FOR USERS ............................................................5

2 RELIABILITY OF PD DIAGNOSIS AND ITS INFLUENCE ON RISK ASSESSMENT .................................... 12

3 BENEFIT ANALYSIS FOR INVESTMENT IN PD DIAGNOSIS SYSTEMS .................................................... 19

1 TYPES OF PD DIAGNOSIS AND THEIR ADVANTAGE FOR USERS

1.1 Technical methods for PD diagnosis

1.1.1 UHF method

METHOD STRENGTHS WEAKNESSES

Sensor technology STRENGTHS WEAKNESSES

Sensor technology STRENGTHS WEAKNESSES

Table 2 : Comparison of strengths and weaknesses of most common

1.1.2 Acoustic method

Figure 1 : Acoustic signal absorption as function of SF 6 gas pressure [6]

Figure 2 : Acoustic signal amplitude as a function of SF 6 gas pressure [7]

Table 3 summarizes strengths and weaknesses of the acoustic method.

METHOD STRENGTHS WEAKNESSES

1.1.3 Conventional method

METHOD STRENGTHS WEAKNESSES

1.2 Other methods

1.3 Combined methods

1.4 Time based strategies of PD diagnosis

DESCRIPTION ADVANTAGES DRAWBACKS TIME SCALE RELATIVE COST

Figure 5 : Illustration of a) periodic, b) temporary continuous and c) continuous PD

Frequency band scanning,

Noise interference gating by use of separate sensors,

Phase coincidence filterung,

Frequency vs. time diagram,

Combined detection methods.

2 RELIABILITY OF PD DIAGNOSIS AND ITS INFLUENCE ON RISK ASSESSMENT

PD defect identification, location and trend analysis,

Past operational experience with PD defects in HV GIS.

Experience with PD detection sensitivity in similar or same GIS configurations,

Figure 6 : Flow chart of risk assessment procedure [4]

2.2 Probability analysis

2.3 Consequence analysis

Liability costs due to third party damage,

Penalties coming from customer contracts.

Figure 7 : Exemplary scenario for possible intervention based on PD diagnosis

The different costs are related to:

Time range I (infant mortality phase),

Time range II (stabilized/normal phase),

Time range III (end-of-life/wear-out phase) 1.

Figure 8 : Figure 5-48 taken from [20]

2.4 Impact of reliability of PD diagnosis on risk assessment

a) The total failure probability is high,

b) The total failure probability is medium,

c) The total failure probability is low.

X: costs for repair and/or exchange of a defective part in days

Y1: costs for acceleration/express due to non-availability of manpower (standard basis)

2.5 Approach for determination of an ‘asset risk index’

Figure 9 : Component factor for ‘asset risk index’ classification

Figure 10 : Network factor for ‘asset risk index’ classification

Figure 11 : PD notification procedure and ‘asset risk index’ assessment

3 BENEFIT ANALYSIS FOR INVESTMENT IN PD DIAGNOSIS SYSTEMS

PD system operation strategy (resource allocation at user),

Importance of the GIS in the network,

Known/previous defects from historical data or specific HV testing/energization,

Budget for investments.