See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/321660440

Standardized survey of transformer reliability: On behalf of CIGRE WG A2.37

Conference Paper · September 2017

DOI: 10.23919/ISEIM.2017.8166559

CITATIONS READS
8 2,172

3 authors:

Stefan Tenbohlen Janine Jagers

Universität Stuttgart Eskom
378 PUBLICATIONS   3,690 CITATIONS    5 PUBLICATIONS   63 CITATIONS   

SEE PROFILE SEE PROFILE

Farzaneh Vahidi
Universität Stuttgart
17 PUBLICATIONS   69 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

HVDC Transmission View project

VFT in GIS View project

All content following this page was uploaded by Stefan Tenbohlen on 13 December 2017.

The user has requested enhancement of the downloaded file.

 Conference Proceedings of ISEIM 2017

Standardized Survey of Transformer Reliability

on behalf of CIGRE WG A2.37
S. Tenbohlen1*, J. Jagers2, F. Vahidi1,
1
University of Stuttgart, Germany
2
Eskom Research, Testing & Development, South Africa
*E-mail: [email protected]

Abstract-- There is limited literature available in the public following objectives: to review existing surveys and to study
domain discussing failure statistics of transformers. This different practices, to conduct a new international survey on
contribution presents the methodology for a standardized failure transformer failures and to analyse the collected data. This paper
data acquisition developed by the Cigré Working Group A2.37. presents and discusses the working group’s data collection
The working group collected 964 major failures which occurred in methodology as well as the results of the analysis in terms of
the period 1996 to 2010, within a total population of 167,459 failure rates and the classification into failure locations.
transformer years, contributed by 58 utilities from 21 countries.
The overall failure rates of substation transformer were all within II. DATA COLLECTION AND PREPARATION FOR
1%. For three groups of substation transformers, detailed
ANALYSIS
population data were collected enabling the calculation of hazard
curves. All populations show a low hazard rate and no distinct Initial working group discussions concentrated on analyzing
bathtub curve character. An increasing probability of failure after the readily available statistics, but the different definitions and
a particular age, which would justify an exchange of the information content constrained forming a coherent database
transformer, cannot be derived from the available data. from individual sources. It was also agreed that the scope needed
Windings, tap changer and bushing related failures were the to be broadened to allow comparison with the failure statistic of
major contributors, followed by lead exit related failures, 1983 survey. A questionnaire was therefore developed to collect
irrespective of application or manufacturing period. Dielectric utility failure statistics in a standardized way. Besides
mode failures were the most prominent, followed by mechanical information about the population under investigation, failure
and electrical type failures, for substation transformers, whereas data was collected for various groups of transformers in terms of
GSU transformers had higher contributions of thermal and the failure locations, failure causes, failure modes, actions,
dielectric mode failures.
external effects and other parameters.
Keywords: Failure Statistics, Hazard Curve, Questionnaire, A. Definition of Failure
Major Failure, Power Transformer, Reliability Survey Based on the poor data collection experience of previous
working groups, the decision was taken to limit the data
I. INTRODUCTION collection to major failures of transformers and reactors
Accurate information about service experience of high operating at 60 kV and above.
voltage equipment is of significant value for both electric A major failure was defined as any situation which required
utilities and for manufacturers of such equipment. It helps the the transformer to be removed from service for a period longer
manufacturers to improve their products and to provide than 7 days for investigation, remedial work or replacement. The
important inputs for the utilities when buying equipment, necessary repairs should have involved major remedial work,
scheduling maintenance and benchmarking their performance. usually requiring the transformer to be removed from its plinth
Statistical analysis of past failure data can display useful features or would require at least the opening of the transformer or the
with respect to future failure behaviour [1]. tap changer tank, or an exchange of the bushings. A reliable
The first international survey on large power transformer indication that the transformer condition prevents its safe
failures was published in 1983 summarizing the results of the operation is considered a major failure, if remedial work (longer
analysis of transformers that failed in the period 1968 to 1978 than 7 days) was required for restoring it to the initial service
[2]. The survey concluded that the average failure rate of capability. In some cases, also failures were assigned as major,
transformers may be regarded as 2 % across all voltage if remedial work was shorter than 7 days and extensive work
categories. Since then, this statistic has become an international with oil processing had to be done (e.g. exchange of bushings).
benchmark in the transformer industry for the failure rate B. Reliability Questionnaire
performance of transformers.
A questionnaire consisting of two major sections was
There is limited literature available in the public domain developed to collect data, in accordance with the definition of
discussing failure statistics of transformers. Internal surveys are major failure. The spreadsheets can be downloaded at [3]. The
being conducted by certain countries and utilities and a number first section of the questionnaire requested general information
of well-established databases exist from countries such as about the population of the operating transformers for the
Australia, Brazil, Canada, Germany and Japan. The results or indicated failure period. The population information included
methodology of these surveys are however not made available the transformer application, type, number of phases, voltage,
publicly due to confidentiality and competition between utilities. rated power, typical loading, and manufacturing period.
Another possible reason for the scarce information could be the
fact that the industry started referring to the results of the The second section captured the transformer failure data,
international survey by Bossi [2] because of its thoroughness grouped data into 4 categories as follows:
instead of conducting individual surveys. Working group A2.37 • Identification of the unit: application, type,
Transformer Reliability Survey was formed in 2008 with the
 Conference Proceedings of ISEIM 2017

construction type, year of manufacture. TABLE I:

Investigated population and failure rates of substation transformers
• Features of the unit: rated power, nominal voltage,
number of phases, cooling system, type of oil, tap HIGHEST SYSTEM VOLTAGE

POPULATION
INFORMATIO
changer, tap changer arrangement, oil preservation

FAILURES

above 700
system, overvoltage protection.

100 kV˗

200 kV˗

300 kV˗

300 kV˗
69 kV˗

199kV

299kV

499kV

699kV
&

99kV

All
kV
• Detail of occurrence: year of failure, service years to
failure, loading immediately prior to failure.
• Consequences of failure: external effects, failure Number of
11 38 31 27 3 4 58
Utilities
location, service years of failed bushings (if location is Number of
2,962 10,932 4,272 3,233 434 348 22,181
bushings), failure mode, failure cause, action taken, Transformers
Transformer-
and detection mode. Years
15,267 64,718 37,017 25,305 4,774 2,991 150,072
Major Failures 144 280 186 152 27 10 799
C. Data Collection FAILURE
0.94% 0.43% 0.50% 0.60% 0.57% 0.33% 0.53%
RATE
Each participating utility was requested to complete the
Excel spreadsheet questionnaire. The completion of the
questionnaire was supported and monitored by the members of A. Failure Rate Dependent on Age
the working group, in particular, ensuring that all major failures
in the indicated time period, were included. All responses were By definition, a power transformer is a system which
compiled into a database. In order to achieve data security and consists of several repairable components and subsystems
anonymity the failure data from each source was made (windings, tap changers, etc.) and should therefore be treated as
anonymous by labelling with a code based on the geographical a repairable system. However, in the event of a major failure of
location and a sequence number. Information about the a transformer, the concerned component or subsystem is
transformer manufacturer was not collected. normally fully exchanged and renewed. According [4], in case
of winding failures, which is the dominant failure location, a
D. Definition of Failure Rate complete new rewinding is often done or in many cases also
Failure rate was calculated according to the definition in scrapping of the transformers. Assuming the system could be
Bossi [2], which was expressed as: seen to be as-good-as-new after the repair/exchange, it could
practically allow the transformer to be treated as a non-
repairable system and renewal theory can be applied [5].
(1)
The hazard function, also called momentary probability of
with: failure, gives the probability that the component will fail at a
𝑛𝑛𝑖𝑖 : Number of failures in the ith year certain life. One can read the real hazard (risk) of failure directly
𝑁𝑁𝑖𝑖 : Number of transformers operating in the ith year from the hazard function. This information is valuable for asset
𝑇𝑇 : Reference period (normally one year) managers, because it aids in decision making for the proper
maintenance strategy. The hazard function is computed using
III. RESULTS the following formula:
The working group collected 964 major failures which
occurred in the period 1996 to 2010, within a total population of
167,459 transformer years, contributed by 58 utilities from 21 (2)
countries. The year of manufacture of the units span from the where:
1950’s up to 2009 and the reference periods range from 3 to 11 ∆n(t) : Number of failures in time interval [t, t+∆t]
years. Because the number of operational transformers was only ∆t : Length of time interval
provided for one year, the total number of transformer-years 𝑁𝑁(𝑡𝑡) : Population surviving at time t
(population per utility) was calculated under the assumption that
the number of transformers in operation was constant during the To overcome the problem of missing age distribution of the
reference period. The number of transformers was multiplied full investigated population, the failure behavior of three
with the length of the reference period in years to obtain an substation transformer populations for which the age
estimate of the total number of transformer-years. distribution was known, was investigated. Group A consisted of
At 150,072 transformer years the investigated population of substation transformers of two European utilities. Populations B
substation transformers was considerably high (Table I). This and C originated from New Zealand and Canada respectively.
value was almost four times higher than the population in Bossi Table II shows the population data of these groups.
[2] with 40,547 transformer years. The calculated failure rates TABLE II
for substation transformers are also given in Table I. The overall Population data for hazard curve analysis
failure rate is within 1%. Transformer Transformer Major
Voltage Class Failure Rate
Population Years Failures

Group A 110 kV-380 kV 6510 35 0.54

Group B 110 kV-220 kV 4686 24 0.51
Group C 120 kV-315 kV 11950 28 0.23

The data collection for the three groups spans from 2000 to
2010. Therefore the number of units surviving age t, , have
 Conference Proceedings of ISEIM 2017

to be adapted to this 11-year failure interval. In this investigation 10,0%

the hazard rate has to be calculated using failures documented in Population A (110kV-380kV)
9,0%
1 Failure
the period between 2000 and 2010. So, the operational 8,0% 1 Failure
Population B (110kV-220kV) 5 YrYears
experience of transformers must be calculated for this period, for 7,0%
Population C (120kV- 315kV)

Hazard rate %
instance a transformer installed in 2005 can just contribute an 6,0%
operational experience with age of 0, 1, 2, 3, 4 and 5 in the 5,0% 1 Failure
defined period 2000-2010. This procedure is explained in detail 4,0%
in [4]. 3,0%
450 2,0%
Population A 1,0%
400 (110kV-380kV)
Population B 0,0%
350
(110kV-220kV) 1 6 11 16 21 26 31 36 41 46 51 56 61
Transformer-Years

300 Population C Age

(120kV-315kV)
250 Figure 2: Failure hazard rate and its 5 years moving average
200
150 A significantly increasing probability of failure after a
100 particular age cannot be derived from the available data. Thus
50
the hazard curve for substation transformers does not seem to
show a wear-out characteristic. Although it is often assumed that
0
1 6 11 16 21 26 31 36 41 46 51 56 61
an increasing probability of failure after a particular age is due
Age
to the effects of component ageing, i.e. oil or paper ageing for
Figure 1: Number of transformer-years surviving age t, (considering 11 year
transformers, for substation transformers, a more likely cause of
failure interval) unreliability is probably damage caused by unusual system
events, e.g. short circuits, lightning strikes or switching
The resulting transformer-years surviving time t is shown in transients, particularly when transformers have design or
Figure 1. The high service experience of population A in the age manufacturing weaknesses.
interval between 20 and 40 corresponds to the extension of the
Substation transformers are normally not so heavily loaded
transmission grid in Europe between 1970 and 1980.
which could make ageing pronounced. Failures due to ageing
Consequently this age group is represented in the failure interval
play a minor role and are masked by random or external failure
between 2000 and 2010 quite well (Figure 1). The comparison
reasons and early replacement of transformers. Furthermore, a
of the three population groups reveals the different installation
transformer consists of several subsystems, with their own
times of the transformers.
individual ageing characteristics and failure modes. Because
Figure 2 shows the hazard rates of the three population major failures occur in all of the subsystems (see Figure 3), the
groups. In order to identify trends also the 5-year moving hazard curve can be considered as the result of competing failure
average of hazard rate curves is shown. Because only failures modes, which have presumably diverse characteristics.
between 2000 and 2010 were evaluated, these calculated hazard Therefore, no distinct increase of the failure probability in later
rate functions do not give any information on early failure rates life can be observed in the hazard curve but only the behavior of
of older transformer designs (e. g. before 1980) and ageing subsystems with random failures.
problems of newer designs (e. g. after 1990).
A second reason for the absence of a distinct ageing
All populations show a low hazard rate and no distinct characteristic of the hazard curve is the replacement strategy.
bathtub curve characteristic. During the full investigated Utilities often use replacement strategies, in which preferably
lifetime, a constant and fairly low risk of failure was old transformers are replaced, which do not fulfill the
experienced; the failures were of random nature. Within operational requirements anymore (e.g. losses, noises,
population A and C just one failure each occurred in the first 20 regulating range). Thus potential failures of replaced units
years at an age of 2 years and 9 years respectively. Both cannot contribute to the hazard curve. This means that traditional
population showed an increase of the hazard rate at an age of 30, approaches to building a statistical end-of-life model cannot be
40 respectively from 0% to a level of app. 1% (5-year moving used. Because the hazard curve does not show a distinct increase
average). This level does not increase further until the highest with time the use of Time Based Maintenance (e. g. exchange of
ages with the exception of two outliers at an age of 48 (Pop. A, substation transformers at a certain age) is not an effective asset
20%) and 57 (Pop. C, 4.43%). These can be attributed to management strategy. Maintenance should be planned
uncertainty in the statistical data as the operational experience according to the actual condition and not to the age of the
with this high age tends to zero (compare Figure 1), e. g. one transformer.
failure within 5 transformer years at age 48.
B. Failure Location Analysis
The hazard curve of population B appeared to suggest a The failure data of the full population were analyzed as a
relatively constant probability of failure across all ages. Also function of the primary location (component) in the transformer
here the outliers at age 55 (2 failures within 39 transformer where the failure was initiated.
years) and 60 (1 failure within 17 transformer years) are due to
low operational experience (compare Figure 1). Figure 3 shows the failure location analysis according to
transformer application in units with voltages 100 kV and above,
respectively. The number of failures classified as ‘unknown’
location was 17%. These ‘unknown’ cases were not taken into
 Conference Proceedings of ISEIM 2017

account in the percentage values in the table and in the analysis The failure mode analysis according to application is shown
according to application. GSUs without tap changers were also in Figure 4. Dielectric mode failures were the most prominent,
excluded from the analysis in order to prevent misinterpretation followed by mechanical and electrical type failures, for
of the data. substation transformers, whereas GSU-transformers had higher
contributions of thermal and dielectric mode failures.
Up to 700 kV, the contribution of bushing related failures
increased with increasing voltage class. Lead exit related failures
exhibited the same increasing trend, across all voltage classes. IV. CONCLUSION
The contribution of tap changer related failures appeared to A questionnaire was developed by the Cigré Working Group
decrease with increasing voltage level. Winding related failures A2.37 by which utility failure statistics could be collected in a
were the largest contributor in both transformer applications. standardized way. The working group collected 964 major
GSU transformers had a higher contribution of winding (48%) failures which occurred in the period 1996 to 2010. The overall
and lead exit failures (13%) than substation transformers (38% failure rate of substation transformers was within 1%. Only GSU
and 6% resp.). Substation transformers on the other hand had a units in the voltage class 300kV to 500kV exceeded 1% failure
higher contribution of tap changer related failures (31%) than rate. Although differences in failure rate were very low, less than
GSU transformers (12%). The contributions of bushing related 0.5%, the failure rates of GSU-units were higher than substation
failures were similar in both transformer applications. transformers in nearly all the voltage classes.
60,0% For three groups of substation transformers detailed
Substation transformers population data were collected enabling the calculation of
50,0% hazard curves. All populations show a low hazard rate and no
Generator step-up units distinct bathtub curve characteristic. Within two populations,
40,0%
just one failure each occurred in the first 20 years. Both
Failure in %

30,0% populations showed an increase of the hazard rate at the ages of

30 and 40, from almost 0% to a level of app. 1% (5 year moving
20,0% average). This level does not further increase until the highest
ages. An increasing probability of failure after a particular age,
10,0%
which would justify an exchange of the transformer, cannot be
0,0%
derived from the available data.
Windings Lead exit Bushings OLTC other
insulation
Windings, tap changer and bushing related failures were the
major contributors followed by lead exit related failures
Figure 3: Failure location of substation transformers (based on 536 failures)
and generator step-up units (based on 127 failures)
irrespective of application or manufacturing period. Dielectric
mode failures were the highest contributor, irrespective of
C. Failure Mode Analysis transformer application. Substation transformers have also
The failure mode describes the nature of the failure, featured higher contributions of mechanical type failures,
illustrating what actually happened when the failure occurred. whereas GSU-transformers featured thermal mode failures.
The failure modes were classified according to [4], as: In order to simplify future surveys, it is recommended to use
• Dielectric (partial discharge, tracking or flashover); the developed Excel-spreadsheet for the collection of failure
data [3]. Thus statistical data can be collected in a simple and
• Electrical (open circuit, short circuit, poor joint, poor standardized way which allows benchmarking between different
contact); transformer populations.
• Thermal (general overheating, localised hotspot);
• Physical chemistry (contamination (moisture, particles, ACKNOWLEDGEMENTS
gas), corrosion); The authors appreciate the discussions within CIGRÉ
• Mechanical (bending, breaking, displacement, loosening) Working Group A2.37 “Transformer Reliability Survey” and
thank the transformer specialists of the utilities involved in the
60% data collection for their valuable support.
Substation transformers
50%
Generator step-up units REFERENCES
40% [1] CIGRÉ WG A3.06, “Final Report of the 2004 - 2007
Failure in %

30%
International Enquiry on Reliability of High Voltage
Equipment”, Brochure 509, Paris, 2012.
20% [2] A. Bossi, J. Dind, J. Frisson, U. Khoudiakov, H. Light and
10%
e. al, “An international survey on failures in large power
transformers in service,” Electra, 1983.
0% [3] http://a2.cigre.org/Publications/Other-documents
Dielectric Electrical Thermal Physical Mechanical Unknown
chemistry [4] CIGRÉ WG A2.37, “Transformer Reliability Survey”,
Figure 4: Failure mode analysis based on 799 failures of substation and Brochure 642, Paris, 2015.
165 failures of GSU-transformers [5] J. Coetzee, Maintenance, Victoria, BC, Canada: Trafford
Publishing, 2004.

View publication stats