M.

Mirzai
A. Gholami
F. Aminifar

J. Electrical Systems 2-1 (2006): 1-12

Regular paper
Failures Analysis and Reliability
Calculation for Power Transformers

JES
Journal of
Electrical
Systems

Failures of transformers in sub-transmission systems not only reduce reliability of

power system but also have significant effects on power quality since one of the
important components of any system quality is reliability of that system. To
enhance utility reliability, failure analysis and its rates, failure origin and physical
damage causes must be studied. This paper describes a case study of the reliability
of sub-transmission transformers (63/20 KV) installed in Mazandaran province,
operated in sub-transmission system. The information obtained from Meandering
Regional Electric Company. The results of study and analysis on 60 substation
including more than 110 transformers installed in sub-transmission system show
that the failure modes of transformers can be represented by Weibull distribution.
Weibull statistics have been widely used and accepted as a successful mathematical
method to predict the remaining life time of any equipment. Useful conclusions are
presented both for power systems operators and manufactures for improving the
reliability of transformers.

Keywords: Failure, Transformer, Reliability.

1. INTRODUCTION
An AC power system is a complex network of many components such as
synchronous generators, power transformers, transmission lines, distribution
network and loads. The operational availability of power transformer is such a
strategic importance for power utility companies. Serious failures in power
transformers owing to insulation breakdown cause considerable financial lose
due to power outage and cost of replacement or repair. Most power utilities
have developed inspection methods and scheme for transformer condition
assessment and they traditionally collect data and information on failure cases
[1]. This information shows that in smaller transformers aging related failures
are dominant. In the medium power rating class, tap changer failures constitute
the highest failure rate. Also In the large transformers insulation coordination
failures are the most common cause in the early service life of transformer [2].
Appropriate maintenance of newer and refurbishment of the older units can
minimize general aging and significantly extend the life time of transformers
[3].Knowledge of the reliability of transformer and other electrical equipment is
Corresponding author: [email protected]
Iran University of Science & Technology (IUST)College of Electrical Engineering P.O. Box # 16844,
Narmak, Tehran, Iran.
Copyright JES 2006 on-line : journal.esrgroups.org/jes

M. Mirzai et al.: Failures Analysis and Reliability Calculation for Power Transformers

an important consideration in the design of power systems. To evaluate the

reliability of a power system, it is necessary to have accurate reliability data on
transformers, together with similar data on other types of electrical
components. The condition of components directly affects the condition of
system with respect to adequacy and security.
Because reliability of power transformers is such an important attribute of a
modern power system as well as its impact on system economy, a good deal of
effort goes into the specification and testing of reliability [4, 5& 6].
The purpose of this paper is outage analysis and description of methods for
estimating the main reliability parameters of power transformers by using of
operational records. The style of information is one of the means of
demonstrating economic justification for spares, redundancy, or improved
maintenance programs.
2. TRANSFORMER FAILURE CAUSES
Transformer problems normally are caused by insulation oil degradation,
overload, thermal stress, humidity in oil/paper and bushing defective and etc.
From the records it may be stated that some of the causes are due to
contaminated oil. This is when they contain moisture or other foreign
substances that are not products of oil oxidation. One or a combination of the
following can cause elevated temperature: excessive load, excessive ambient
temperature, cooling system problems, sludge oil, dark colored exterior paints.
Load and ambient temperature are closely related in their effect on transformer
operating temperature. For a constant transformer load, higher ambient
temperature led to higher operating temperature. A number of cooling system
problems can cause high operating temperature: closed radiator valves, dirty
cooling fins, broken or improperly set cooling fans/pumps, and cooling control
circuit failure [3].
3. DATA PREPARATION
Records of failures reports from 2000 to 2004 on the transformers installed
in Mazandaran province in the north of Iran by Mazandaran Regional Electric
Company. For the purpose of this study data where collected including the
number of outages caused by transformers failure , the outage duration ,
number of units , voltage level , failure cause and restoration time and anything
else.

J. Electrical Systems 2-1 (2006): 1-12

4. GEOGRAPHICAL LOCATION
The nature of Mazandaran province is under the influence of geographical
latitude, Alborz heights, elevation from sea level, and distance from the sea,
southern wildernesses of Turkmenistan, local and regional wind currents, and
diverse vegetation cover. From geographical point of view, Mazandaran
province is divided into two parts i.e. coastal plain and Alborz mountainous
area. Alborz mountain range has surrounded the coastal strip and coastal plains
of the Caspian Sea like a high wall. Due to permanent breeze of the sea and
local winds in southern and eastern coasts of the Caspian Sea, there have been
formed sandy hills that have caused the appearance of a low natural barrier
between the sea and plain. High and low average temperature in these two
weather region are about: 10.9 C in winter, 26 C in summer and the annually
average temperature is 17.7 C. Annually average humidity is about 75.5%.
Figure 1 illustrates the geographical location of the Mazandaran province.
5. FAILURES ANALYSIS
When a failure occurs it is necessary to investigate the causes to improve
production technology and maintenance programs. Some of the most important
causes of failures in power transformers are cited in the first of this study.
Furthermore, transformers must be denergized for some periodical tests and
services. Therefore an outage will occur and it will decrease the component and
ultimately system reliability .So we consider that causes as outages/failures.

MAZANDARAN
PROVINCE

Tehran
Mazandaran
Water

IRAN MAP

Figure 1: Geographical location.

M. Mirzai et al.: Failures Analysis and Reliability Calculation for Power Transformers

In this part, we divide the outage causes into five groups:

1) Periodical test
2) Services
3) Protective operation
4) Insulation problems
5) Others (bushing faults, development, lightning, etc.)
The purpose of this classification is to identify outages/faults that are most
likely to occur in transformers. Table 1 shows the distribution of each failure
mode. Also the details of the failures and corresponding percentages are shown
in figure 2.
Table 1.Classification of outage/failure origins: TTF = Total Time of Fault
(Outage), ATF = Average Time of Fault (Outage).
Description

Failures

Duration

No.

TTF

ATF

Periodical Test

48

12%

80:39:00

1:40:49

Services

56

14%

88:16:00

1:34:34

Protective
Operation

206

50%

264:22:00

1:17:00

Insulation
Problems

16

4%

107:40:00

6:43:45

Others

86

21%

113:23:00

1:19:06

Total

412

100%

654:20:00

Periodical Test
12%
Services
14%

Others
21%
Insulation
Problems
4%

Protective
Operation
49%

Figure 2: The percentage of any type of failures.

Results of table 1 show that protective operation is the highest failure mode,
but average times of the fault is the lowest, due to their causes are temporary
and are removed after a little time. In other hand insulation problems are not
too much but they have the highest average time of faults. The 5th group is
consists of development interruption, bushing faults, tap changer faults,
operator failures, lightning, unknown operation errors and etc. The most
4

J. Electrical Systems 2-1 (2006): 1-12

number of faults of this group are related to Tap Changer failures. The main
causes of the Tap changers failures start from sparking and erosion of the Tap
Changer contacts. Moving parts malfunctioning can also lead to a failure.
Eroded contacts produce sparking and sticking. In off-load Tap Changers long
period without operation, leads to a corrosion and sludge build up which causes
jamming of the moving parts and consequently, a failure of the Tap Changer.
Therefore maintenance and periodical test and services can be reduced this type
of failures.
25

Early Failure
Period

Random Failure
Period

Wear out Failure

Period

Failure rate

20

15

10

0
1

10

15

23

26

30

Time(year)

Figure 3: Component failure rate as a function of operating time, bathtub

curve.
6. BASIC CONCEPTS OF RELIABILITY ANALYSIS
In general, a bathtub curve may be used as an adequate representation of a
transformer failure mode with varying life time. Accelerated life testing has
shown that transformer have a bathtub curved failure rate. It can be divided in
to three failure modes as shown in figure 3. [7, 8]
These modes are:
Early Failures: These occur during the first year of energization and usually are
caused by inherent defects due to poor materials, workmanship or processing
procedures or manufacturers quality control beside installation problems.
Random Failures: These are not associated with early failures. They are
produced by chance or operating conditions such as a failure from switching
surges, lightning and operator faults. This part of curve called useful life time
period and failure rate is nearly constant.
Wear Out Failures: This type is the results of material wear out. Normally, the
wear out mode becomes predominant only after 20 years of operation. This
normal wear out period is followed by an increasing failure rate.

M. Mirzai et al.: Failures Analysis and Reliability Calculation for Power Transformers

Different mathematical models can be used to simulate different failure

modes. Obviously we are most interested in the random failure mode, since it is
not only related to manufacturing quality control but also it is possible related
to the operating conditions of the units. Therefore, calculation and analysis are
concentrated in the random failure mode which period is from 1 to 6 year of
operation. One important methodology is to find a probability distribution
function that can represent the operating time t in the random failure period.
Table 2 summarizes the operating time for the failed transformers (t) , the
number of failures cause to unit outage (mi), the cumulative number (N) and the
relative cumulative frequency (F`(ti)). The histogram of this data is shown in
figure 4.
Table 2. The number and relative cumulative frequency of failures.
t

mi

F`

1<t2

67

67

0.162621359

2<t3

80

147

0.356796117

3<t4

101

248

0.601941748

4<t5

84

332

0.805825243

5<t6

80

412

1.000000000

7. RELIABILITY ANALYSIS
In the first step of reliability analysis, we introduce the Weibull probability
distribution [9].
120

N o . o f failu re s

100
80
60
40
20
0
1<t2

2<t3

3<t4

4<t5

5<t6

Operation time

Figure 4: Number of failures vs. operation time.

J. Electrical Systems 2-1 (2006): 1-12

Therefore we use current data to obtain the specific probability distribution.

Weibull probability distribution is shown as equation (1):

F `(t ) = 1 exp(

tm
)
t0

(1)

In which m and t0 must be determined where t0 is the scale and m is the

shape parameter. For achieving to a linear relation, (ln) function must be used
twice as below:
1
ln ln(
) = m ln t ln t0
1 F `(t )

(2)

or:

Y = mX A

(3)

Where:
1
) , X = ln t and A = ln t0 .
1 F `(t )

Y = ln ln(

Table 3. Shows calculated X and Y for failures data.

Table 3. Calculated X and Y for failures data
t

mi

F`

1<t2

67

67

0.1626

0.6931

-1.7289

2<t3

80

147

0.3568

1.0986

-0.8180

3<t4

101

248

0.6019

1.3863

-0.0821

4<t5

84

332

0.8058

1.6094

0.4941

5<t6

80

412

1.0000

1.7918

1.5272

The operating time t and its corresponding relative commutative F`(ti) are
plotted on a Weibull probability paper as shown in figure 5. From figure 5 the
distribution parameters ( m and t0 ) of Weibull distribution function can be
estimated as follow: m = 2.8076 and t0 = exp(4.01) = 55.14 .
The theoretical probability function of operating time t has the form:

F (t ) = 1 exp(t 2.8076 / 55.14)

(4)
7

M. Mirzai et al.: Failures Analysis and Reliability Calculation for Power Transformers

Theoretical distribution curve F (t ) of failed transformers is shown in figure

6 that t is operating time. To determine the accuracy of this theoretical function,
the new corresponding values for previous data must be recalculate and
compare with basic data. Table 4. Shows F '(t ) , F (t ) and their absolute
deference Dn .
2

Ln(Ln(1/(1-F'(t)))

1.5
1
0.5
0
-0.5
-1
-1.5
-2
0.5

0.7

0.9

1.1

1.3

1.5

1.7

1.9

Ln(t)

Figure 5: The relative commutative function plotted in Weibull distribution paper.

100%
90%
80%
70%
F (t) %

60%
50%
40%
30%
20%
10%
0%
1

10

t(year)

Figure 6: Theoretical distribution curve of operating time of failed transformers.

The maximum difference between F (t ) and F '(t ) is the measure of how

good fit is let the maximum difference be denoted by:

Dn = max | F (t ) F `(t ) |

(5)

Therefore: Dn = 0.0647 .
This deference, which is a random variable which in turn varies with the
sample selected, is then compared with the critical value Dn that called
Kolmogorov-Smirnov criteria, which is defined as:
8

J. Electrical Systems 2-1 (2006): 1-12

P (Dn < Dn ) = 1

(6)

Critical values Dn at various significant levels are tabulated in probability

tables for various values of n . If the observed (Dn < Dn ) the proposed
distribution is acceptable at the selected significance level .
Table 4. F (t ) and F '(t ) and their difference s
t

mi

F`

Dn

1<t2

67

67

0.1626

0.1186

0.0439

2<t3

80

147

0.3568

0.3249

0.0318

3<t4

101

248

0.6019

0.5850

0.0168

4<t5

84

332

0.8058

0.8065

0.0007

5<t6

80

412

1.0000

0.9352

0.0647

For = 0.05 and n > 50 from probability tables we have:

Dn =

1.36

(7)

In this case n = 412 so: Dn = 0.067 and (Dn < Dn )

Therefore the Weibull distribution is verified. The results show that the
studied power transformers are in conformity with the Weibull distribution for
the mathematical failure model.
8. RELIABILITY LEVEL AND FAILURE RATE

The operating time test known as "definite time and tail cut test" was used to
calculate the Mean Time To Failure (MTTF) and the Failure Rate ( Fr ). The
failure rate function % or hazardous function has the form:

m
H (t ) = ( )t m 1
t0

(8)

It can be obtained from the distribution parameters m,t0 of F (t ) . In this case:

H (t ) = (0.0509)t 1.8076
The MTTF or M is equal to exception of operation time and has the form:
9

M. Mirzai et al.: Failures Analysis and Reliability Calculation for Power Transformers

M = (t0 )1/ m (1 + 1/ m)

(9)

Where:

(n) = t n 1e tdt
0

MTTF is calculated and shown as: MTTF=3.71 (year)

For comparison and analysis, other approach is used for calculating the
average MTTF with calculation of Failure rate ( Fr ,av ). In this approach the
average value given as below was used:

Fr ,av =

H (t )dt = T
T

m 1

(10)

/ t0

The results of this approach are tabulated in table 5.

Table 5. H (t ) and Fr ,av of transformers for 5 years
Fr,av

H(t) percentage per year

t=1

t=2

t=3

t=4

t=5

T=5

0.050

0.178

0.370

0.620

0.932

0.332

By considering of the distribution function is an exponential distribution, the

failure rate Fr will be constant and the following formula can be used for
calculating MTTF:

MTTF = 1/ Fr ,av

(11)

Where Fr ,av is shown in Table 5 and therefore: MTTF=3.01 (years).

There is a little difference between two value of MTTF from above
approaches and its reason is the assuming of the constant failure rate in second
approach.
9. CONCLUSION

In this paper, causes of sub-transmission transformer outages have been

classified: Periodical test, Services, Protective operation, Insulation problems
and others (bushing, development, etc.) Also the number and duration of
transformers outage have been investigated and analyzed, according to
following:
10

J. Electrical Systems 2-1 (2006): 1-12

1. Protective operations have the highest number of faults and insulation

problems have longest interrupts duration.
2. The failure mode of sub-transmission transformer can be represented by
Weibull distribution. For most transformers the empirical cumulative
frequency F`(t) versus t on the Weibull probability paper exhibits a
remarkable linearity .
3. The MTTF for sub-transmission transformers installed during 2000-2004 in
Mazandaran province is 3.71 years and the average failure rate ( Fr ,av ) is
0.332%/year.
4. Appropriate maintenance of newer and refurbishment of the older units can
minimize general aging and significantly extended the life of transformers.
Acknowledgement

The authors would like to acknowledge the support of MAZANDARAN

Regional Electric Company, HV Center and electrical engineering department
of Iran University of Science and Technology.
REFERENCES
[1]

M. Minhas, J. Reynders and P. de Clerk, Failures in Power system transformers

and appropriate monitoring techniques, High voltage engineering symposium 2227 august 1999 conference publication no.467, IEE 1999.

[2]

M. Minhas, P. de Clerk and J. Rainders, Failures in large power transformers, In

Proceeding of Cigre International conference on large high voltage electric
system, September 1997, Cairo, Egypt.

[3]

H. Borsi, E. Gochenbach and V. Wasserberg, Life time Extension of Transformers

by On-line Drying, in 19th International Power System Conference (PSC 2004)
Iran-Tehran.

[4]

A. S. Farag, C. Wang and T. C. Chang, Failure analysis of composite dielectric of

power capacitor in distribution system, in IEEE Transaction on Dielectric and
Electrical Insulation, Vol. 5,No 4,August 1998.

[5]

R. Billinton and N. Allan, Reliability assessment of large electric power system,

Kluwer Academic Publisher, Mass. USA, 1988.

[6]

C. C. Fang, B. S. Meng and P. Seng, Reliability Evaluation of transmission and

distribution configuration with duration dependent effect, IEE Vol.136, PT C, no.
2, March 1989.

[7]

R. Billinton and N. Allan, Reliability Evaluation of Engineering Systems Concept

and Techniques, 1993.

[8]

G. Anderes, Probability Concepts in Electric Power Systems, John wiley & sons,
NY, USA, 1990.

11

M. Mirzai et al.: Failures Analysis and Reliability Calculation for Power Transformers

[9]

12

M. Ugur, A. Kuntman and M. Merev, Investigation of the Aging process in

polymeric insulators by using improved Weibull statistics, IEEE Int. sym. on
electrical insulation MA USA, April 7-10, 2002.