Content uploaded by Farzaneh Vahidi
Author content
All content in this area was uploaded by Farzaneh Vahidi on Feb 10, 2015
Content may be subject to copyright.
Statistical Failure Analysis of European Substation Transformers
Farzaneh Vahidi*, Stefan Tenbohlen
Institut für Energieübertragung und Hochspannungstechnik der Universität Stuttgart, Deutschland
*farzaneh.vahidi@ieh.uni-stuttgart.de
Kurzfassung
Im vorliegenden Beitrag werden die Ausfalldaten von europäischen Netzkupplungstransformatoren analysiert. Basie-
rend auf einer Population von mehr als 45.000 Transformatorjahren und 212 schweren Fehlern, die zwischen 2000 und
2010 aufgezeichnet wurden, lässt sich eine durchschnittliche Fehlerrate von 0,5% angeben. Die ermittelten Lebensdau-
erkurven zeigen ein durch Zufallsausfälle geprägtes Ausfallverhalten. Da kein alterungsbedingter Anstieg der Ausfallra-
te zu erkennen ist, ist der Einsatz einer zeitorientierten Instandhaltungsstrategie für Netzkupplungstransformatoren nicht
ratsam. Als Entstehungsorte für schwere Fehler überwiegen die Wicklungen. Durchführungsfehler führen verhältnismä-
ßig oft zu Schäden, die mit Explosion oder Feuer einhergehen.
Abstract
This contribution addresses the analysis of substation transformer failures in Europe. Based on a transformer population
with more than 45.000 unit-years and 212 major failures over a period of 11 years (2000-2010) a failure rate of app.
0.5% was determined. The derived hazard curves show a constant probability of failure at all ages. Replacement strate-
gies, in which preferebly old transformers are replaced, have a biasing effect on the failure statistics as transformers are
not left in service to fail. Because the hazard curve does not show an increase with time the use of Time Based Mainte-
nance will not be effective for power transformers. Winding related failures appear to be the largest contributor of major
failures, and a significant decrease in tap changer related failures has been observed in comparison with results of the
1983 survey. Bushing failures most often lead to severe consequences like explosion or fire.
1. Introduction
Accurate information about service experience of high
voltage equipment is of significant value for both electric
utilities and for manufacturers of such equipment. It helps
the manufacturers improve their products, and provides
important inputs for the utilities when organizing mainte-
nance and benchmarking their performance [1]. Statistical
analysis of the past failure data can display useful features
with respect to the future failure behavior. Equipment re-
liability data are also required when assessing the overall
reliability of an electric power system, including studies
of the electric energy supply security. Furthermore, inter-
national standards applicable to high voltage equipment
are being improved on the basis of service experience and
reliability data.
Around the world, utilities apply different approaches to
estimate the actual stage of life of their assets [2]. Two
main methodologies can be distinguished here; bottom-up
and top-down analysis. The bottom-up analysis focuses
on the condition assessment of individual assets. The base
for such an analysis are maintenance and diagnostics re-
ports (e. g. DGA, PD-measurement, FRA, dielectric re-
sponse), loading history and aging characteristics ob-
tained through investigations performed on service-aged
materials. The top-down analysis investigates the condi-
tion of the whole population by means of analytical tools
(e.g. statistical distributions). In such approach, the in-
formation about number and ages of both failed and in-
stalled units are essential. Emphasis is put in this case on
economic and strategic life-time assessment. Results of a
top-down analysis are e. g. failure frequency, age of assets
which are most likely to fail. However, both approaches
have certain limitations, in particular imposed by the dif-
ferences in design, operating regimes and maintenance.
Additionally, a mathematical incorporation of all degrada-
tion mechanisms is constrained by their possible interac-
tions. Moreover, for both approaches the information nec-
essary for the analysis is either very limited or even
unavailable.
The failure data analyzed in this contribution are acquired
by means of the reliability questionnaire form of CIGRE
WG A2-37 [3, 4]. Each utility filled this questionnaire
form and all the answers were collected in a database. In
this paper the failure data of European substation trans-
formers with an operating voltage of 110kV, 220kV and
380kV are analyzed.
2. Investigated Population
In this contribution the results of a failure data collection
of 32 utilities from Germany, Austria, Swiss, France,
United Kingdom, Spain, Denmark and Netherlands are
presented based on major failures which occurred be-
tween 2000 and 2010 in substation transformers. This
population of transformers can be regarded as homogene-
ous concerning age distribution, specification, operational
and maintenance conditions.
The survey addresses failure data of substation with oper-
ating voltage between 100 and 500 kV. Table shows the
transformer population investigated in this approach.
Table 1: Investigated population of substation trans-
formers
POPULATION
INFORMATION OF
SUBSTATION UNITS
HIGHEST SYSTEM VOLTAGE [kV]
100 ≤ U < 200
200 ≤ U < 300
300 ≤ U < 500
All
Number of utilities 22 18 14 32
Number of transformers 2775 2124 1214 6113
Number of major failures
90
72
49
211
Transformer-Years
20915
15221
9271
45407
Failure rate p.a.
0.43%
0.47%
0.53%
0.46%
3. Statistical Analysis of Major Fail-
ures
A comparison of different failure surveys is only possible
in case of the same definition for major failure. Here any
situation which requires the equipment to be removed
from service for a period longer than 7 days for investiga-
tion, remedial work or replacement is a major failure. A
major failure requires at least the opening of the tank, in-
cluding the tap changer tank or an exchange of bushings.
Also a reliable indication that the condition of the trans-
former prevents a safe operation should be counted as a
major failure if remedial work (longer than 7 days) is
needed for restoring original service capability (e.g. de-
tection of strong PDs).
3.1 Failure Rate
To determine the failure rate, the following formula is
used [2]:
%100
N
n
i
1i
i
1i
⋅
∑
∑
=
λ
(1)
:
λ
Failure rate p.a.in percentage
:n
i
Number of transformers that failed in the ith year
:N
i
Number of transformers in service during the ith year
For the calculation of failure rates a constant transformer
population was assumed for the investigated failure time
period. The calculated failure rates are given in table 1
dependent on the voltage level. Although there is a slight
increase of failure rate with increasing voltage, an average
failure rate of app. 0.5% can be designated to European
substation transformers.
3.2 Hazard Rate Function
Calculating the failure rate for ever smaller intervals of
time, results in the hazard function. It shows the momen-
tary probability of a failure dependent on the transformer
age. In order to calculate the hazard rate the age distribu-
tion of the full investigated transformer population is re-
quired. Then the hazard function can be computed using
the following formula:
%100
)t(r )t(f
)t(h ⋅=
(2)
:)t(h
Failure hazard rate in percentage
:)t(f
Number of failures at age interval T
:
)t
(
r
Number of transformers in operation and surviving at
age interval T
So, hazard rate is the instantaneous failure rate at age t.
The collected European data did not contain the age dis-
tribution of transformers in operation for all utilities due
to simplicity reasons. Therefore a hazard curve for the full
data set cannot be calculated directly. To overcome the
problem of missing age distribution for the full investigat-
ed population the known population data of three Europe-
an utilities was used as a reference for the full population.
The transformer populations in European utilities have
similar erection times. So, the population data of three
known utilities can be used as a reference to scale the
population data for whole population. Figure 1 shows the
density functions of the population data of the three refer-
ence utilities depending on the voltage class:
Figure 1: Normalized acquisition data of substation trans-
formers of three reference European utilities and its three
year moving average (dashed line)
The high number of installations between 1960 and 1980
corresponds to the extension of the transmission grid.
Using the normalized acquisition data of substation trans-
former in Europe, the number of transformers in operation
and surviving at age interval T (2000-2010, 11years) is
calculated. The resulting age distribution of surviving
transformers is shown in Figure 2.
0
1
2
3
4
5
6
7
8
9
10
2010
2005
2000
1995
1990
1985
1980
1975
1970
1965
1960
Normalized Number of Transformers
in %
100<=U<200
0
1
2
3
4
5
6
7
8
9
10
2010
2005
2000
1995
1990
1985
1980
1975
1970
1965
1960
Normalized Number of Transformers
in %
200<=U<300
0
1
2
3
4
5
6
7
8
9
10
2010
2005
2000
1995
1990
1985
1980
1975
1970
1965
1960
Normalized Number of Transformers
in %
Year
300<=U<500
Figure 2: Number of transformers surviving at time t based
on 11 years interval from 2000 up to 2010
Having the age distribution of the population, the only re-
quired information for calculation of hazard rate is the
transformer age at failure occurrence. This was infor-
mation was taken from the collected failure database.
In figure 3 the age distribution of failed transformers is
presented depending on voltage class:
Figure 3: Age distribution of failed substation transformers
dependent on voltage class
The participating utilities submitted altogether 211 trans-
former major failures which occurred in the time interval
between 2000 and 2011. Main contributors are transform-
ers in the voltage class between 100kV and 200kV with
almost 43%. The transformers in the voltage class be-
tween 200kV and 300kV have a portion of 34% and the
highest voltage class builds 23% of the reported failures.
The hazard rates which are calculated for all voltage clas-
ses between 100 kV and 500 kV are given in Figure 4.
The dashed line describes the 3-year moving average.
With this smoothed curve the tendency of hazard rate is
easier to identify. Because only failures between 2000 and
2010 were evaluated, it has to be regarded that as a matter
of principle this hazard rate function can not deliver in-
formation about early failure rates of older transformer
designs and ageing problems of newer designs.
In principle the hazard curves show a constant probability
of failure at all ages which is an indicator for random fail-
ure behaviour. The values over 45 years are statistical un-
certain, because the failure experience of transformers el-
der than 45 years is insufficient (see Figure 2). Therefore
the increase of the hazard curve should not be attributed
to component ageing. It is usually assumed that the in-
creasing probability of failure after a particular age is due
to the effects of component ageing, i.e. oil or paper ageing
for transformers, but for transformers a more likely cause
of the onset of unreliability is probably damage caused by
unusual system events, e.g. short circuits, lightning strikes
or switching transients, particularly when transformers
have design or manufacturing weaknesses [5]. It has also
to be considered that utilities often use a replacement
strategy, in which preferably old transformers are re-
placed. For this analysis information about preventive re-
placements in the past was not available. This could have
a profound biasing effect on the failure statistics as trans-
formers are not left in service to fail, so that these statis-
tics cannot be used directly for lifetime modeling.
Figure 4: Hazard rate of European substation transformers
dependent on different voltage classes between 2000 and
2010 and its three years moving average (dashed line)
The hazard curve for substation transformers does not
show a distinct wear-out characteristic (Figure 4). Substa-
tion transformers are normally not so heavily loaded.
Therefore ageing is not very pronounced. Failures due to
ageing play a minor role and are masked by random or
external failure reasons and by early replacement of trans-
formers. Furthermore it has to be considered that the
transformer consists of several subsystems, which may
have their own individual ageing characteristics and fail-
ure modes. Because major failures occur in all of the sub-
systems, the hazard curve is a result of competing failure
0
200
400
600
800
1000
1200
1400
0 5 10 15 20 25 30 35 40 45 50
Number of Transformers
Age
100<=U<200
200<=U<300
300<=U<500
010 20 30 40 50
0
2
4
6
8
10
12
14
Age
Num ber of Fai lure
100 <= U< 200
200 <= U< 300
300 <= U< 500
0,00%
0,50%
1,00%
1,50%
2,00%
0 5 10 15 20 25 30 35 40 45 50
Hazard rate in %
100 <=U <200
0,00%
0,50%
1,00%
1,50%
2,00%
0 5 10 15 20 25 30 35 40 45 50
Hazard rate in %
200 <=U <300
0,00%
0,50%
1,00%
1,50%
2,00%
0 5 10 15 20 25 30 35 40 45 50
Hazard rate in %
Age
300 <=U<500
modes, which have diverse characteristics. Thus no in-
crease of the failure probability for higher operational life
can be seen in the hazard curve but the behavior of sys-
tems with random failures. It also has to be regarded that
the operational experience of old power transformers is
limited due to existing replacement strategies, which leads
to statistical insecure data. Because the hazard curve does
not show an increase with time the use of Time Based
Maintenance will not be effective for power transformers.
Maintenance should be planned according to the actual
condition.
Fitting the hazard curve allows its extrapolation in time
beyond the age of the oldest assets in the population.
Many probability distributions can be used to model the
failure distribution. The shape of the hazard curve deter-
mines which continuous distribution can be fitted to the
data. Most renewal failure data sets encountered in the
maintenance environment can be fitted with the Weibull
distribution. But in case of power transformers, the occur-
rence of a failure event is more or less a random occasion.
Furthermore the advance maintenance and replacement
strategies avoid any age-dependent rising of failure rate of
power transformers. So, power transformers failure statis-
tics cannot be used for lifetime modeling, e. g. Weibull
fitting.
Generator step-up units are often heavier loaded near their
nominal rating. So a more pronounced ageing characteris-
tic could be assumed. Unfortunately the collection of
GSU failures is much more difficult as of substation
transformers, because the population is distributed on a
larger amount of companies. Also the homogeneity of the
GSU data, e.g. considering operational characteristic,
voltage level, manufacturer, commissioning time is not as
good as with substation transformers. Therefore a reason-
able analysis of failures dependent on age for GSUs is
practically not possible.
3.3 Failure Location Analysis
The substation transformers failure data are further classi-
fied regarding to primary location of major failures. Fig-
ure 5 shows the failure locations for all failures with the
exception of failures with unknown location. Only 5.2%
of all failures were submitted with unknown location.
Figure 5: Failure location of 200 substation transformers,
voltage class 100kV
U
500kV
As seen in figure 5, major failures originate from several
transformer components. Windings related failures appear
to be the main contributor of major failures with 35%.
The contribution of tap changer (30%) related failures de-
creased in comparison with the statistics from 1983 given
in [6]. The lower failure rate of OLTC can be attributed to
the development of better contact materials (use of silver
plated contacts). 95.3% of the failed substation transform-
ers were equipped with a tap changer. Bushings (17.5%),
lead exits (9.5%) and core (4%) are listed with a minor
percentage as a reason for major failures.
3.4 Failure Mode Analysis
The failure mode describes the nature of the failure illus-
trating what actually happened when the failure occurred.
Dielectric failure means partial discharge, tracking and
flashover. Electrical failure means open circuit, short cir-
cuit, poor joint, poor contact, ground deterioration, float-
ing potential. The definitions of the failure modes are ac-
cording to [7]. There is no single prominent failure mode.
The categories of dielectric and mechanical are with
31.3% and 20.4% the most dominant failure modes (see
Figure 6).
Figure 6: Failure mode analysis of 211 substation trans-
formers failures
3.5 Failure Cause Analysis
The circumstances during design, manufacture or opera-
tion that led to the failure are analyzed. Because it is often
quite difficult and extensive to determine the root cause of
a failure, thus the significance of this analysis is unsure
and the results should be seen with care. Figure 7 shows
the percentages of the various failure causes in substation
transformers.
The determination of the main failure cause is often quite
difficult. Consequently, more than 18% of failures have
unknown causes. Among the different failure causes ex-
ternal short-circuit is with a contribution of 14.2% the
most mentioned one. Astonishingly design and manufac-
turing are mentioned quite often as a failure cause but this
cannot be proven by the low failure rate during the first
30 years of transformers operation. Furthermore, the dis-
tribution of failure causes is very wide. In some cases,
there is more than one reason for the happened major fail-
ures and the fine distinction between various causes
makes a statistical analysis uncertain.
HV Winding
16.5%
MV Winding
5%
LV Winding
11.5%
Tapping
Winding 2%
HV Lead Exit
3.5%
MV Lead Exit
3.5%
LV Lead Exit
2.5%
Isolatio n 1.5% Electrical
Screen
0.5%
HV Bushings
12%
MV Bushings;
5%
LV Bushings
0.5%
Core and
magnetic circuit
4%
Tank 1%
Cooling u nit 1%
Tap Changer;
30%
Dielectric
31.3%
Electrical
18%
Thermal
16.1%
Physical
chemistry
5.2%
Mechanic
al 20.4%
Unknown
9%
Figure 7: Failure cause analysis based on 211 substation
transformers failures
3.6 External Effects
In case of a major failure, it is important to look at the ex-
ternal effects which result from the failure occurrence.
Figure 8 presents, the various external effects which are
caused by the failures. The classification of severe exter-
nal effects was performed with six groups. The statistical
analysis demonstrates that most of the major failures do
not result in any external effect (78.7%). The most prob-
lematic situations after a major failure are fire and explo-
sion. 9.5% of failures lead to fire while 3.3% of external
effects are explosion or burst.
Figure 8: External effects of 211 Substation transformers
major failures
Figure 9: Failure Location of substation transformer where
Fire or Explosion occurred (27 major failures)
Such external effects are always connected with huge
economic consequences. Therefore, the originating loca-
tion for fire and explosion related failures is analyzed in
Figure 9. It can be seen that bushing failures most often
lead to severe consequences.
4. Conclusion
By means of a questionnaire developed by CIGRE work-
ing group A2.37 (Transformer Reliability Survey) major
failures of European substation transformers were ana-
lyzed. Based on a transformer population with more than
45.000 unit-years and 212 major failures a failure rate of
app. 0.5% was determined. The derived hazard curves
show a constant probability of failure at all ages. An in-
creasing probability of failure after a particular age could
not be observed. Replacement strategies, in which pref-
erebly old transformers are replaced, have a biasing effect
on the failure statistics as transformers are not left in ser-
vice to fail. So these statistics cannot be used directly for
lifetime modelling, e. g. Weibull fitting. Because the haz-
ard curve does not show an increase with time the use of
Time Based Maintenance will not be effective for power
transformers. Therefore maintenance should be planned
according to the actual condition.
Winding related failures appear to be the largest contribu-
tor of major failures, and a significant decrease in tap
changer related failures has been observed in comparison
with results of the 1983 survey. Bushing failures most of-
ten lead to severe consequences like explosion or fire.
5. Acknowledgement
The authors appreciate the fruitful discussions within CI-
GRE Working Group A2.37 “Transformer Reliability
Survey” and thank the transformer specialists of the utili-
ties involved in the data collection for their valuable sup-
port.
6. Literature
[1] CIGRE, Final Report of the 2004 - 2007 International
Enquiry on Reliability of High Voltage Equipment,
Brochure 509, Paris, 2012.
[2] L. Chmura, Life-cycle assessment of high-voltage as-
sets using statistical tools, PhD thesis Technical Uni-
versity Delft: ISBN 978-94-6182-396-0, 2014.
[3] Questionnaire of CIGRE WG A2.37 “Transformer
Reliability Survey”, May 2011, http://www.uni-
stuttgart.de/ieh/wga237.html., last accessed: Sept. 01,
2014
[4] S. Tenbohlen, J. Jagers, G. Bastos, B. Diggin, P. Man-
ski, B. Desai, M. Krüger, J. Gebauer, P. Müller, J.
Lapworth, A. Mikulecky, C. Rajotte, T. Sakia, S.
Yukiyasu, Transformer Reliability Survey: Interim
Report, Report WG A2.37, Electra, No. 261, April
2012
[5] J. Lapworth, „Transformer reliability surveys, A2-
114,“ in Cigré Biennial Conference, Paris, 2006
[6] A. Bossi, J. Dind, J. Frisson, U. Khoudiakov, H.
Light und e. al, „An international survey on failures in
large power transformers in service,“ Electra, pp. 21-
48, 1983.
[7] CIGRE, Life Management Techniques for Power
Transformers, Brochure No. 227, June 2003.
Design
11.4%
Manufacturing
12.3%
Material
11.4%
Loss of
clamp ing
pressure
0.5%
Inst allation
on-site
1.4%
Improper
maintenance
5.2%
Improper repair
0.9%
Overvoltage
1.4%
Overheating
1.4%
Lightning
0.5%
External
short-circuit
14.2%
Repetitive
through faults
0.5%
Externa l
Pollution
0.9%
Abnor mal
Deterioration
3.32%
Aging
6.2%
Collat eral
Damage
1.4%
Corrosive
Sulphur
0.5%
Other reasons
8.1%
Unknown
18.5%
None
78.7%
Leakages
5.7%
Explo sion,
Burst
3.32%
Fire
9.5%
Collat eral
Damages
0.5%
Others
2.4%
HV Winding
3.7%
LV Winding
3.7%
Tapping
Winding
7.4%
HV Lead Exit
7.4%
HV Bushings
26%
MV Bushings
18.5%
Tap Changer
22.2%
Unknown
11.1%
