642

642
TRANSFORMER RELIABILITY SURVEY
Working Group
A2.37
December 2015
TRANSFORMER
RELIABILITY SURVEY
WG A2.37
Members
S. Tenbohlen, Convenor (DE), J. Jagers, Secretary (ZA),

G. Bastos (BRA), B. Desai (US), B. Diggin (IRE), J. Fuhr (CH), J. Gebauer (DE), M. Krüger (AT),
J. Lapworth (UK), P. Manski (POL), A. Mikulecky (HR), P. Müller (CH), C. Rajotte (CAN),
T. Sakai (JP), Y. Shirasaka (JP), F. Vahidi (DE)
Copyright © 2015
“Ownership of a CIGRE publication, whether in paper form or on electronic support only infers right
of use for personal purposes. Unless explicitly agreed by CIGRE in writing, total or partial
reproduction of the publication and/or transfer to a third party is prohibited other than for personal
use by CIGRE Individual Members or for use within CIGRE Collective Member organisations.
Circulation on any intranet or other company network is forbidden for all persons. As an exception,
CIGRE Collective Members only are allowed to reproduce the publication.”
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”.
ISBN : 978-2-85873-346-0
Transformer Reliability Survey
Acknowledgement
The members of CIGRÉ working group A2.37 first and foremost express their sincere thanks to the
representatives of the 57 utilities that have collected service and failure information and contributed to this
survey.
Our special thanks go to Walter Wasinger and Marshall Clark for compiling information from Australia
/New Zealand utilities.
We also acknowledge the help and assistance from many colleages and associates in carrying out our
tasks:
A. McIntosh, T. Fogelberg, M. Mottier-Favre, O. Glassey, H. Lauper, E. Hill, S. Proulx, M. Hässig, M. Ling,
Li Ing Peh, M. O’Neill, L. Paulhiac, J. Shann, M. H. Jensen, A. Janssen, A. van Schijndel, M. Stach, C. Y.
Choo, P. West, B. S. Andersen, N. Majer, I. Horsey, A. Fisher, P. Gidden, J. Velek, H. Haug, M. Schäfer,
U. Sundermann, G. Ebb, J. Reimer, L. Gunn, P. Onions, P. New, T. Woolley, G. Buckley, L. Eggimann, S.
Wyss, A. Osterholt, M. Jordanoff, A. Kosmata, S. Kumar, J.-F. Boudreau, V. Moorkath.
ISBN : 978-2-85873-346-0
Page 1

Contents
EXECUTIVE SUMMARY ................................................................................................................................... 7
1 INTRODUCTION ................................................................................................................................................................ 10
1.1 Background .................................................................................................................................................................. 10
1.2 Scope and Aim ............................................................................................................................................................ 10
1.3 Structure ........................................................................................................................................................................ 11
2 DEVELOPMENTS IN TRANSFORMER TECHNOLOGY, DESIGN AND MANUFACTURING ................................... 12
2.1 Developments in Technology .................................................................................................................................... 12
2.1.1 Core Material (Magnetic Circuit) ...................................................................................................................... 12
2.1.2 Conductor Material (Windings) ......................................................................................................................... 13
2.1.3 Insulation Materials .............................................................................................................................................. 14
2.1.4 Tap Changers ........................................................................................................................................................ 15
2.1.5 Bushings .................................................................................................................................................................. 16
2.1.6 Condition Monitoring............................................................................................................................................ 17
2.2 Developments in Design Tools .................................................................................................................................. 18
2.3 Developments in Manufacturing ............................................................................................................................... 18
2.3.1 Core Cutting and Annealing ............................................................................................................................... 18
2.3.2 Winding Construction and Conductor Material .............................................................................................. 18
2.3.3 Drying of the Active Part .................................................................................................................................... 19
2.3.4 Clamping Structure ............................................................................................................................................... 19
2.4 Relationships between Reliability and Manufacturing Period ........................................................................... 19
2.4.1 Windings ................................................................................................................................................................ 19
2.4.2 Magnetic Circuit/ Core Reliability .................................................................................................................... 20
2.4.3 Impact of Design Techniques .............................................................................................................................. 20
2.4.4 Tap Changer Reliability ...................................................................................................................................... 20
2.4.5 Bushing Reliability................................................................................................................................................. 21
2.4.6 Cooling System ..................................................................................................................................................... 21
3 DEFINITIONS AND GUIDELINES FOR RELIABILITY SURVEYS .................................................................................... 22
3.1 Guidelines on Reliability Practices .......................................................................................................................... 22
3.2 Reliability Surveys ...................................................................................................................................................... 22
3.3 Definitions ..................................................................................................................................................................... 23
3.3.1 Failure ..................................................................................................................................................................... 23
3.3.2 Reliability ............................................................................................................................................................... 23
3.3.3 Failure Rate ........................................................................................................................................................... 23
3.4 Review of Available Reliability Statistics ............................................................................................................... 24
3.4.1 Constraints.............................................................................................................................................................. 24
3.4.2 Failure Rates .......................................................................................................................................................... 25
3.4.3 Failure Location ..................................................................................................................................................... 28
3.4.4 Failure Mode ......................................................................................................................................................... 28
3.4.5 Failure Causes ....................................................................................................................................................... 28
3.5 Theory of Statistical Analysis .................................................................................................................................... 29
3.5.1 Types of Statistical Data ..................................................................................................................................... 29
Page 2
3.5.2 Basic Renewal Theory .......................................................................................................................................... 30

3.5.3 Weibull Parameter Estimation ........................................................................................................................... 33
3.5.4 Repairable Systems ............................................................................................................................................. 34
4 DESCRIPTION OF EXISTING TRANSFORMER RELIABILITY SURVEYS ....................................................................... 37
4.1 Countrywide Surveys.................................................................................................................................................. 37
4.1.1 Canadian Electricity Association ........................................................................................................................ 37
4.1.2 German FNN-Failure Statistics........................................................................................................................... 41
4.1.3 Japan ...................................................................................................................................................................... 44
4.1.4 National Committee Cigre A2 – Brazil ............................................................................................................ 49
4.1.5 United Kingdom .................................................................................................................................................... 52
4.2 Company Related Surveys ........................................................................................................................................ 55
4.2.1 Utility, ESBI (Ireland) ............................................................................................................................................ 55
4.2.2 Utility, Eskom (South Africa) ................................................................................................................................ 58
4.2.3 Transformer Manufacturer, ABB (Sweden) ...................................................................................................... 61
4.2.4 OLTC Manufacturer - Maschinenfabrik Reinhausen (Germany) .................................................................. 64
4.2.5 Consultancy – EPRI (USA) .................................................................................................................................... 66
5 METHODOLOGY FOR FAILURE DATA COLLECTION ................................................................................................. 73
5.1 Definition of Failure .................................................................................................................................................... 73
5.2 Reliability Questionnaire ........................................................................................................................................... 73
5.4 Data Collection and Limitations................................................................................................................................ 75
6 RESULTS OF PERFORMED RELIABILITY SURVEY .......................................................................................................... 77
6.1 Collected Data ............................................................................................................................................................ 77
6.2 Data Limitations ........................................................................................................................................................... 79
6.3 Failure Rate .................................................................................................................................................................. 79
6.4 Failure Rate Dependent on Transformer Age ....................................................................................................... 81
6.5 Failure Location Analysis ........................................................................................................................................... 86
6.5.1 Transformers .......................................................................................................................................................... 86
6.5.2 Shunt Reactors ....................................................................................................................................................... 91
6.6 Failure Mode Analysis................................................................................................................................................ 92
6.7 Failure Cause Analysis ............................................................................................................................................... 94
6.8 External Effects Analysis ............................................................................................................................................ 97
6.9 Action Analysis ............................................................................................................................................................. 99
6.10 Comparison with Cigré Survey of 1983 ........................................................................................................... 101
6.10.1 Failure Rates .................................................................................................................................................... 101
6.10.2 Failures Classification ..................................................................................................................................... 102
7 CONCLUSION AND RECOMMENDATIONS ............................................................................................................. 108
Collecting, Compiling and Presenting Data in a Standardized Way .................................................................. 109
REFERENCES ........................................................................................................................................................................ 110
ANNEXES ............................................................................................................................................................................. 115
ANNEX A: Questionnaire ............................................................................................................................................... 115
ANNEX B: Definitions from Surveys and International Standards and Guidelines ............................................. 117
ANNEX C: Description of Analysis of Transformer Population Data ..................................................................... 119
Page 3
List of Figures
FIGURE 1: OVERVIEW OF CORE STEEL DEVELOPMENT [6] ........................................................................................................................... 13

FIGURE 2: OVERVIEW OF INCREASE IN TRANSFORMER RATED VOLTAGE AND POWER [6] .............................................................. 14
FIGURE 3:CENSORING OF STATISTICAL DATA ................................................................................................................................................. 30
FIGURE 4: PROBABILITY DENSITY FUNCTION (PDF) [52] ............................................................................................................................... 31
FIGURE 5: FAILURE DISTRIBUTION FUNCTION (CDF) [52].............................................................................................................................. 31
FIGURE 6: SURVIVAL FUNCTION [52] ................................................................................................................................................................. 32
FIGURE 7: THE ‘BATHTUB CURVE’ HAZARD FUNCTION [53] ......................................................................................................................... 33
FIGURE 8: SHAPES OF THE WEIBULL HAZARD FUNCTION [52] ................................................................................................................... 34
FIGURE 9 : DATA ANALYSIS FRAMEWORK [52] .............................................................................................................................................. 36
FIGURE 10: PERCENTAGE OF PRIMARY LOCATION OF DISTURBANCE FOR 110 KV ............................................................................ 43
FIGURE 11: PERCENTAGE OF PRIMARY LOCATION OF DISTURBANCE FOR 220 KV AND 380 KV ................................................... 43
FIGURE 12: NUMBER OF DISTURBANCES PER 100 TRANSFORMERS .......................................................................................................... 44
FIGURE 13: NUMBER OF INSTALLED TRANSFORMERS ................................................................................................................................... 46
FIGURE 14: NUMBER OF FAILURES BY STRUCTURAL COMPONENT ........................................................................................................... 48
FIGURE 15: NUMBER OF FAILURES BY IDENTIFIED CAUSE ............................................................................................................................. 48
FIGURE 16: FAILURES CLASSIFIED ACCORDING TO FAILURE MODE .......................................................................................................... 51
FIGURE 17: FAILURES CLASSIFIED ACCORDING TO COMPONENT INVOLVED ....................................................................................... 51
FIGURE 18: FAILURES CLASSIFIED ACCORDING TO CONSEQUENCES OF FAILURE ............................................................................... 52
FIGURE 19: FAILURE MODE ................................................................................................................................................................................... 54
FIGURE 20: FAILURE CAUSE................................................................................................................................................................................... 54
FIGURE 21: FIELD FAILURE REPORTING PROCESS ........................................................................................................................................... 62
FIGURE 22: FAILURE RECORD SHEET ................................................................................................................................................................... 63
FIGURE 23: TYPICAL AGE PROFILE – IN-SERVICE INDUSTRY-WIDE DATABASE FOR TRANSMISSION SUBSTATION POWER
TRANSFORMERS ............................................................................................................................................................................................. 67
FIGURE 24: IDB DATA CATEGORIES (NOT TO SCALE) ................................................................................................................................... 68
FIGURE 25: SURVIVAL PLOTS FOR NEW VERSUS REPAIRED UNITS—NOTABLE DIFFERENCES IN PERFORMANCE......................... 69
FIGURE 26: THE IDB PROCESS .............................................................................................................................................................................. 70
FIGURE 27: ANALYSIS OF FAILURE LOCATION (1112 OF 1863 RECORDS HAVE FAILED COMPONENT SPECIFIED) .................... 70
FIGURE 28: ANALYSIS OF FAILURE CAUSE (654 OF 1863 RECORDS HAVE FAILURE CAUSE IDENTIFIED)........................................ 71
FIGURE 29: APPLICATION EXAMPLE: YEARLY FAILURES ................................................................................................................................. 72
FIGURE 30: FAILURE RATE DEPENDENT ON VOLTAGE CLASS AND APPLICATION ................................................................................. 80
FIGURE 31: NUMBER OF FAILURES DEPENDENT ON TRANSFORMER AGE ............................................................................................... 82
FIGURE 32: NUMBER OF TRANSFORMERS-YEARS SURVIVING AGE T , (CONSIDERING 11YEAR-FAILURE INTERVAL) ................. 83
FIGURE 33: FAILURE HAZARD RATE (5 YEARS MOVING AVERAGE) ........................................................................................................... 84
FIGURE 34: FAILURE HAZARD AND REPLACEMENT HAZARD FOR UK POPULATION [54] ..................................................................... 85
FIGURE 35: FAILURE LOCATION ANALYSIS BASED ON 675 MAJOR FAILURES FOR U≥100KV ......................................................... 88
FIGURE 36: FAILURE LOCATION ANALYSIS, SUBSTATION TRANSFORMERS, U≥100KV, 536 FAILURES ........................................... 89
Page 4
FIGURE 37: FAILURE LOCATION ANALYSIS, GENERATOR STEP-UP TRANSFORMERS, U≥100KV, 127 FAILURES ........................... 89
FIGURE 38: WINDING FAILURE LOCATION ANALYSIS FOR U≥100KV (LEFT: SUBSTATION TRANSFORMERS RIGHT: GSU) ....... 90
FIGURE 39: FAILURE LOCATION ANALYSIS OF TRANSFORMERS WITH MANUFACTURING PERIOD BEFORE 1980 ...................... 90
FIGURE 40: FAILURE LOCATION ANALYSIS OF TRANSFORMERS WITH MANUFACTURING PERIOD AFTER 1980 ......................... 91
FIGURE 41: FAILURE LOCATION ANALYSIS OF SHUNT REACTORS ............................................................................................................ 92
FIGURE 42: FAILURE MODE ANALYSIS BASED ON 964 MAJOR FAILURES ............................................................................................... 93
FIGURE 43: FAILURE MODE ANALYSIS OF 799 SUBSTATION TRANSFORMERS ...................................................................................... 94
FIGURE 44: FAILURE MODE ANALYSIS OF 165 GENERATOR STEP-UP TRANSFORMERS ...................................................................... 94
FIGURE 45: FAILURE CAUSE ANALYSIS BASED ON 964 MAJOR FAILURES............................................................................................... 95
FIGURE 46: EXTERNAL EFFECTS OF 964 MAJOR FAILURES ........................................................................................................................... 97
FIGURE 47: FAILURE LOCATION WHERE FIRE OR EXPLOSION OCCURRED (126 MAJOR FAILURES)................................................. 98
FIGURE 48: EXTERNAL EFFECTS OF ALL BUSHING FAILURES (115 MAJOR FAILURES) ........................................................................... 98
FIGURE 49: THE ANALYSIS OF THE ACTIONS TAKEN AFTER 964 MAJOR FAILURES .............................................................................. 99
FIGURE 50: FAILURE LOCATION ANALYSIS OF 242 SCRAPPED TRANSFORMERS .................................................................................. 99
FIGURE 51: FAILURE LOCATION ANALYSIS OF 465 REPAIRED TRANSFORMERS ..................................................................................100
FIGURE 52: COMPARISON OF FAILURE LOCATIONS IN ALL TRANSFORMERS ......................................................................................103
FIGURE 53: COMPARISON OF FAILURE LOCATIONS IN POWER STATION TRANSFORMERS ...........................................................103
FIGURE 54: COMPARISON OF FAILURE LOCATIONS IN SUBSTATION TRANSFORMERS ....................................................................104
FIGURE 55: COMPARISON OF FAILURE CAUSES IN ALL TRANSFORMERS ..............................................................................................105
FIGURE 56: COMPARISON OF FAILURE CAUSES IN POWER STATION TRANSFORMERS ...................................................................105
FIGURE 57: COMPARISON OF FAILURE CAUSES IN SUBSTATION TRANSFORMERS ............................................................................106
FIGURE 58: COMPARISON OF FAILURE MODES IN ALL TRANSFORMERS...............................................................................................106
FIGURE 59: COMPARISON OF FAILURE MODES IN POWER STATION TRANSFORMERS ....................................................................107
FIGURE 60: COMPARISON OF FAILURE MODES IN SUBSTATION TRANSFORMERS .............................................................................107
FIGURE 61: SHEET 1 OF QUESTIONNAIRE FOR THE COLLECTION OF POPULATION DATA ..............................................................115
FIGURE 62: SHEET 2 OF QUESTIONNAIRE FOR THE COLLECTION OF FAILURE DATA .........................................................................116
Page 5
List of Tables
TABLE 1: FAILURE RATES FROM SURVEYS FROM 1968 TO 2005 ............................................................................................................... 27

TABLE 2: INVESTIGATED NUMBER OF TRANSFORMERS BANKS 2009 ....................................................................................................... 40
TABLE 3: BREAKDOWN OF COLLECTED DATA DEPENDENT ON VOLTAGE CLASS [57] ........................................................................ 40
TABLE 4: TRANSFORMER BANK ANALYSIS BY SUBCOMPONENT FOR 200 – 299 KV [57] ................................................................. 41
TABLE 5: NUMBER OF INSTALLED TRANSFORMERS VOLTAGES 66KV AND ABOVE .............................................................................. 46
TABLE 6: NUMBER OF FAILURES AND THE RATE OF FAILURES...................................................................................................................... 47
TABLE 7: NUMBER OF TROUBLES AND THE RATE OF TROUBLES ................................................................................................................. 47
TABLE 8: THE DISTRIBUTION OF FAILED COMPONENTS BY VOLTAGE CLASS ......................................................................................... 48
TABLE 9: THE DISTRIBUTION OF IDENTIFIED CAUSE BY VOLTAGE CLASS ................................................................................................. 49
TABLE 10: NUMBER OF TRANSFORMERS AND REACTORS ............................................................................................................................ 50
TABLE 11: COLLECTED FAILURE DATA ACCORDING TO VOLTAGE LEVEL ................................................................................................ 50
TABLE 12: MAJOR UK TRANSFORMERS FAILURES BETWEEN 2003 AND 2009 ....................................................................................... 53
TABLE 13: HV TRANSFORMERS ON THE ESB SYSTEM..................................................................................................................................... 56
TABLE 14: VOLTAGE CLASSES OF TRANSFORMERS ....................................................................................................................................... 56
TABLE 15: TRANSFORMER MVA RATING ........................................................................................................................................................... 56
TABLE 16: TOTAL UNIT YEARS OF TRANSFORMERS........................................................................................................................................ 57
TABLE 17: UNIT YEARS OF TRANSFORMERS ON THE SYSTEM ..................................................................................................................... 57
TABLE 18: MAJOR FAILURES ON HV TRANSFORMERS BETWEEN 1980-2009 ........................................................................................ 57
TABLE 19: FAILURES AND FAULTS ON NETWORKS TRANSFORMERS ........................................................................................................ 57
TABLE 20: FAILURES AND FAULTS ON GSU’S ................................................................................................................................................... 58
TABLE 21: FAILURES AND FAULTS ON AUTOTRANSFORMERS ..................................................................................................................... 58
TABLE 22: CLASSIFICATION INTO FAILURE LOCATIONS, FAILURE CAUSES AND FAILURE MODES [35] ............................................ 60
TABLE 23: INVESTIGATED POPULATION OF TRANSFORMERS ..................................................................................................................... 78
TABLE 24: INVESTIGATED POPULATION OF SUBSTATION TRANSFORMERS ............................................................................................ 79
TABLE 25: INVESTIGATED POPULATION OF GENERATOR STEP-UP TRANSFORMERS ............................................................................ 79
TABLE 26: FAILURE RATES OF SUBSTATION TRANSFORMERS ...................................................................................................................... 80
TABLE 27: FAILURE RATES OF GENERATOR STEP-UP TRANSFORMERS ...................................................................................................... 80
TABLE 28: FAILURE RATES OF COMBINED GROUP OF TRANSFORMERS ................................................................................................... 81
TABLE 29: POPULATION DATA FOR HAZARD CURVE ANALYSIS ................................................................................................................. 83
TABLE 30: FAILURE LOCATION ANALYSIS DEPENDENT ON VOLTAGE CLASS ......................................................................................... 87
TABLE 31: FAILURE LOCATION ANALYSIS OF SHUNT REACTORS ............................................................................................................... 92
TABLE 32: FAILURE MODE ANALYSIS DEPENDENT ON VOLTAGE CLASS .................................................................................................. 93
TABLE 33: FAILURE CAUSE ANALYSIS DEPENDENT ON VOLTAGE CLASS ................................................................................................. 96
TABLE 34: FAILURE RATES DETERMINED FROM THE CIGRÉ 1983 AND A2.37 SURVEYS (ALL AGES) ...............................................101
TABLE 35: COMPARISON OF FAILURE RATES .................................................................................................................................................102
Page 6
EXECUTIVE SUMMARY
In 2008, Cigré Working Group A2.37 Transformer Reliability was formed with the following objectives:
 Reviewing all existing surveys and study different practices (in terms of data collection, compilation,
etc.)
 Conducting a new international survey on transformer failures, and proposing a uniform way of
collecting, compiling and presenting data.
 Compiling and analysing the collected data, and interpreting the results (calculation of failure rates,
classification into failure location, failure causes and failure modes)
The different definitions and information content of the readily available statistics constrained forming a
coherent database from individual sources. Thus, a questionnaire was developed by the Cigré Working Group
A2.37 by which utility failure statistics could be collected in a standardised way. Transformer failure data
could therefore be analysed and interpreted for various types of transformers in terms of failure locations,
failure causes, failure modes, actions, external effects and failure rates in transformers. In contrast to several
public available statistics, the results of this questionnaire can deliver valuable information which can be used
for asset management of a power transformer fleet.
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 56 utilities from 21 countries. The year of
manufacture of the units span from the 1950’s up to 2009, and the reference periods range from 3 to 11
years.
The overall failure rate of substation, GSU and the combined group of transformers were all within 1%. Only
GSU units in the voltage class 300kV to 500kV exceeded 1% failure rate. Although differences in failure rate
were very low, less than 0.5%, the failure rates of GSU units were higher than substation transformers in nearly
all the voltage classes.
INVESTIGATED POPULATION AND FAILURE RATES OF SUBSTATION TRANSFORMERS
HIGHEST SYSTEM VOLTAGE [kV]

POPULATION
INFORMATION
100 ≤ kV < 200 ≤ kV < 300 ≤ kV < 500 ≤ kV <
69 ≤ kV < 100 kV ≥ 700 All
200 300 500 700
Number of Utilities 11 38 31 27 3 4 58
Number of
2,962 10,932 4,272 3,233 434 348 22,181
Transformers
Transformer-Years 15,267 64,718 37,017 25,305 4,774 2,991 150,072
Major Failures 144 280 186 152 27 10 799
FAILURE RATE 0.94% 0.43% 0.50% 0.60% 0.57% 0.33% 0.53%
Page 7
INVESTIGATED POPULATION AND FAILURE RATES OF GENERATOR STEP-UP TRANSFORMERS

POPULATION
INFORMATION 100 ≤ kV < 200 ≤ kV < 300 ≤ kV < 500 ≤ kV <
69 ≤ kV < 100 kV ≥ 700 All
200 300 500 700
Number of
14 320 455 673 167 74 1,703
Transformers
Transformer-Years 153 3,278 4,639 6,740 1,837 740 17,387
Major Failures 0 20 43 89 9 4 165
FAILURE RATE 0.00% 0.61% 0.93% 1.32% 0.49% 0.54% 0.95%
For three groups of substation transformers detailed population data were collected enabling the calculation of
hazard curves. All populations show a low hazard rate and no distinct bathtub curve character. Within two
populations just one failure each occurred in the first 20 years respectively. Both populations showed an
increase of the hazard rate at an age of 30, 40 respectively from almost 0% to a level of app. 1% (5 year
moving average). This level does not further increase until the highest ages. An increasing probability of failure
after a particular age, which would justify an exchange of the transformer, cannot be derived from the
available data. A more likely cause of unreliability is probably damage caused by unusual system events.
Replacement strategies, in which preferably old transformers are replaced, have a biasing effect on the failure
statistics as transformers are not left in service to fail. A possible approach is to estimate on the number of
years a replaced transformer would have remained in service if it would not be replaced. So these statistics
cannot be used directly for lifetime modelling, e. g. Weibull fitting. Because the hazard curve does not show a
distinct increase with time the use of Time Based Maintenance will not be effective for substation transformers.
Therefore maintenance should be planned according to the actual condition.
Windings, tap changer and bushing related failures were the major contributors, followed by lead exit related
failures, irrespective of application or manufacturing period.
GSU transformers had higher contributions of winding and lead exit failures than substation transformers.
Substation transformers on the other hand had higher contributions of tap changer related failures than GSU
transformers probably due by a lower proportion of GSU transformers uses a tap changer. The contribution of
bushing related failures were similar in both transformer applications.
Up to 700kV, the contribution of bushing and lead exit related failures increased with increasing voltage class.
The contribution of tap changer related failures appeared to decrease with increasing voltage level.
Dielectric mode failures were the highest contributor of failure modes, irrespective of transformer application.
Substation transformers also featured higher contributions mechanical type failures, whereas GSU transformers
also featured thermal mode failures.
Design and manufacturing, aging and external short circuits appeared to be major contributors of failure
causes. Due to the large contribution of ‘unknown’ cases in terms of the failure causes, this result should be
treated and interpreted with caution.
Failures originating in the bushings most often lead to severe consequences such as fires and explosions. Due to
their impact, winding failures typically led to a situation where the failed transformer was scrapped. Tap
changer and bushing failures typically resulted in repairs.
Page 8
In order to simplify future surveys the working group recommends using the developed Excel-spreadsheet for
the collection of failure data. Thus statistical data can be collected in a simple and standardized which allows
benchmarking between different transformer populations.
Page 9
1 INTRODUCTION
1.1 Background
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 specifying and buying equipment, when organizing maintenance
and when benchmarking their performance. Statistical analysis of the past failure data can display useful
features with respect to the future failure behavior. Equipment reliability data are also required when assessing
the overall reliability of an electric power system, including studies of the electric energy supply security.
Furthermore, international standards applicable to high voltage equipment are being improved on the basis of
service experience and reliability data [1].
In 1983, Cigré Working Group 12.05 published a report summarizing the results of their analysis of
transformers that failed in the period 1968 to 1978 [2]. 13 countries from 3 different continents took part in
this survey. The authors reported of difficulties to compile and analyse the data of the survey due to incomplete
or incompatible responses. Ten years later, Working Group 12.14 attempted to upgrade this survey but was
unsuccessful due to similar reasons. In the 2000’s Study Committee A2 started a new working group based on
the work done in Study Committee A3 on reliability surveys. The work of A3 was stopped due to the new
international context where the restriction to data access was becoming problematic. Study Committee A2 also
started an Advisory Group on Reliability and a paper on that subject was presented at the Cigré 2006 session.
Many countries have published reliability surveys locally with some being published annually. However, this
knowledge is not shared amongst the international community where most benefit can be drawn.
1.2 Scope and Aim

Around the world, utilities apply different approaches to estimate the actual stage of life of their assets. Two
main methodologies can be distinguished here; bottom-up and top-down analysis. The bottom-up analysis
focuses on the degradation and condition assessment of individual assets. The base for such an analysis are
maintenance and diagnostics reports (e.g. DGA, PD-measurement, FRA, dielectric response), loading history and
aging characteristics obtained through investigations performed on service-aged materials. The top-down
analysis investigates the condition of the whole population by means of analytical tools (e.g. statistical
distributions). In such approach, the information about number and ages of both failed and installed 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. Thus the results can be used to
parameterize Time Based Maintenance strategies. However, both approaches have certain limitations, in
particular imposed by the differences in design and operating regimes. Additionally, a mathematical
incorporation of all degradation mechanisms is constrained by their possible interactions. Moreover, for both
approaches the information necessary for the analysis is either very limited or even unavailable [3].
To date, the only major international survey on large power transformer failures was published in 1983 [2].
The failed units were classified according to voltage level, age and application, and for each class,
corresponding failure rates were calculated as well as classification into failure component, presumed cause
and failure origin. The survey concluded that the average failure rate of transformers may be regarded as 2
% across all voltage categories. Since then, this statistic has become an international benchmark in the
transformer industry for the failure rate performance of transformers. However, this survey was based on
transformer failures occurring in the period 1968 to 1978, and shall be considered with caution if used for
comparison with units designed and manufactured several years after 1978.
Because all worldwide approaches to collect failure data of power transformers failed since 1983, the working
group A2.37 Transformer Reliability was formed in 2008 with the following objectives:
Page 10
 Reviewing all existing survey and study different practices (in terms of data collection, compilation, etc.)
1.3 Structure
In chapter 2, the developments in transformer design, manufacture and technology, as presented in the
literature, are reviewed with the aim of assessing to what extent developments influenced the trends in
reliability (or failure) of transformers identified in chapter 3 and 6.
Chapter 3 gives a general overview of transformer reliability as presented in the literature, with emphasis on
recommendations for reliability practices in the industry, definitions and terminology associated with reliability,
and statistics from available reliability surveys. The application of statistical analysis to transformer data is also
discussed.
Chapter 4 presents existing countrywide and company related surveys, their general information about specific
definitions and methodology of data acquisition and analysis.
In chapter 5 the failure data collection methodology developed by the Working Group A2.37, and the
associated limitations, are described. A uniform way of collecting, compiling and presenting data is proposed.
Chapter 6 presents the results of the conducted failure data survey. The the investigated population and the
analysis in terms of failure rates and failure classifications, are discussed.
Chapter 7 finally summarizes the main points and results of the brochure.
The developed questionnaire, further definitions and the theory of population data analysis are given in the
annexes.
Page 11
2 DEVELOPMENTS IN TRANSFORMER TECHNOLOGY, DESIGN AND

MANUFACTURING
2.1 Developments in Technology
Lucien Gaulard and John Dixon Gibbs first exhibited a device with an open iron core called a 'secondary
generator' in London in 1882, then sold the idea to the Westinghouse company in the United States [4]. They
also exhibited the invention in Turin, Italy in 1884, where it was adopted for an electric lighting system.
However, the efficiency of their open-core bipolar apparatus remained very low. Based on this technology the
first practical usable transformer was built in 1885 [4], and the basic principles of modern transformer design
laid down in the 1920’s and 1930’s [5]. Since then, the basic sub-systems and design features of a transformer
have remained unchanged.
With the development of the computer simulations in the 1990’s the design criteria are based on more precise
calculations of permissible stresses. These techniques offer the advantage of a more secure design process.
In the last few decades (approximately since the 1980’s, based on the publication date of the paper),
transformer technology development has focused on judgement of influence of the operation conditions to the
reliable service of transformers. For an efficient maintenance, diagnostic and monitoring methods using a
modern measuring technique were developed to assess the condition of service aged transformers in the correct
way [6].
2.1.1 CORE MATERIAL (MAGNETIC CIRCUIT)

The development of core material was aimed at reducing core losses, through the use of improved and thinner
grades of core steel [7], and is illustrated in Figure 1.
Cold rolled grain oriented manufacturing of silicon sheet steel was developed in the early 1930’s [8], and the
first commercial quantities introduced in 1939 [5]. During the 1970’s Hi-B steel was introduced [4], and by
1995, this steel type was the norm in the industry [5]. Laser scribed core material was introduced from the
1980’s (Figure 1) as a special material to be used only where the cost of core losses was very high [5].
Especialy for grid transformers which are usually running under 50% load (Europe), the no load losse play a
very important role for the optimal design of the transformer. Although the lowest predicted thickness of steel
was 0.18 millimetres from approximately the 1990’s (Figure 1), the lowest available thickness in the early
2000’s was 0.23 millimetres [5], [6].
Page 12
FIGURE 1: OVERVIEW OF CORE STEEL DEVELOPMENT [6]
2.1.2 CONDUCTOR MATERIAL (WINDINGS)

Improvements in conductor material were driven by the increase in power rating and voltage requirements due
to increased electricity demands (Figure 2). As power ratings increased, low voltage windings using conductors
connected in parallel were used to reduce eddy current losses [4]. As a consequence of the further increase in
rated power the introduction of continuously transposed conductor (CTC, 1960) led to significant reduction of
eddy current losses and circulating currents [6].
Page 13
FIGURE 2: OVERVIEW OF INCREASE IN TRANSFORMER RATED VOLTAGE AND POWER [6]
Using CTC for windings the circulating currents caused by leakage flux linking with parallel connected
conductors are avoided [4]. The transposition ensures that each strand experiences the same overall leakage
flux, as nearly as possible [5]. Continuously transposed conductor is preferably of the epoxy bonded type
(back lack) for greater short circuit strength [7].
According to Seay [9], the IEEE C57 standards of 1975 resulted in new winding materials and mechanical
designs.
2.1.3 INSULATION MATERIALS

Cellulose (in different forms) and mineral oil based insulation systems have remained the most widely used, and
according to Baehr [6] and Metha [10] this is mainly because of their high reliability and comparatively low
prices. The available paper insulation technologies (in 1998, based on publication date of reference) include
Kraft paper, creped paper, highly extensible paper, thermally upgraded paper and diamond dotted paper
for insulation of conductors and pressboard for the barrier isulating system [5].
Page 14
The development of high density pre-compressed pressboards (main insulation) has improved short circuit
withstand [10].
No time periods were provided for these developments.
2.1.4 TAP CHANGERS

Tap Changers are designed to transfer electrical power from one tap winding to the adjacent one. To fulfil this
function two different configurations are used:
• In tank type: The tap changer is installed in the transformer main tank with a separate oil
compartment. Within the separate oil compartment operates the diverter switch or a selector switch
(combined diverter and selector). The switching principle is normally the high speed resistor type
OLTC
• Compartment type: The tap changer is mounted in an external compartment. This external
compartment might be connected to the main tank via a barrier board. The external compartment is
usually either one compartment (selector and diverter in one compartment) or two compartments
(separated diverter or selector). For the use of reactor type tap changers the compartment type is
usually used. With resistor type tap changers this type is only used in special applications (e.g. arc
furnace transformers – The separate compartment gives easier access to the components for
maintenance)
The high-speed resistor type OLTC based on the 1926 patent by Dr. Jansen which uses the switching principle
of the diverter switch and tap selector, is the most widely used principle worldwide; in conventional OLTCs an
arcing contact system with mineral oil as switching medium is used. The transition impedance is carried out with
ohmic resistors. With this principle, the switched current and the recovery voltage are in phase which lightens
the quenching of the arc in the current zero [11].
The tap changer technology in the 1950s and 1960s of the last century used laminated paper oil
compartments and cast material in the mechanic of the diverter. The tap selector contacts were made of copper
or brass. With larger power ratings in the end of the 1960s the tap selectors were redesigned. The use of the
copper or brass contacts was improved with a silver plating or silver coating of the contacts. This helped to
avoid contact heating and eliminated contact problems on the tap selector.
In the 1970s the diverter switch technology switched from former cast material into a mechanic innovation using
sheet metal parts and the use of plastics. Starting from the 1980s diverter switch oil compartments were made
out of Glass Fibre Reinforced Plastics (GFRP) instead of the previously used laminated paper. This new material
had the advantage of not being hydroscopic.
Vacuum tap changer technology was first introduced in North America by Westinghouse in the 1960’s for
reactive type tap changers. Then, the development of the vacuum tap changer technology for resistive type tap
changers began in the mid 1980’s, followed by dry type OLTCs in the mid 1990’s and in-tank types in the
year 2000 [12], [13], [14]. The vacuum tap changer technology offers a big maintenance advantage since it
eliminates arc quenching in oil associated with traditional oil tap changers. With the vacuum tap changer
technology, the arc quenching takes place in the vacuum interrupters, resulting in cleaner diverter oil and
reduced maintenance activities on the tap changer.
Because maintenance requires intrusive work to be done on the tap changer, outages are required to perform
the work; the vacuum technology extends the maintenance intervals resulting in a reduction in outages.
Variations of the vacuum tap changer technology include retrofitting of the diverter switch (vacuum technology)
into an existing oil type selector [12].
Page 15
2.1.5 BUSHINGS
There are two main types of bushings used most frequently in the transformer technology: ceramic (porcelain)
bushings, which nowadays are dominant at distribution voltages, and condenser type bushings, used for the
past 50 years as the only choice for higher voltages. They are produced in three types of technology: RBP
(resin bonded paper), OIP (oil impregnated paper), and RIP (resin impregnated paper).
Early bushings were dry insulated, and produced from a resin coated paper, referred to as resin bonded
paper bushings [15]. These bushings were wound in an uncontrolled workshop environment and as a result had
a high partial discharge and dissipation factor.
From the 1940’s, oil impregnated paper bushings with porcelain as the external insulator were introduced.
During the 1990’s, resin impregnated paper bushings with a silicone rubber insulator were introduced to
eliminate the problems experienced with oil impregnated paper types with porcelain insulators. The
advantages of silicone rubber over porcelain include the following: maintenance free since cleaning of the
insulator in polluted areas is not required, self-extinguishing properties, no projection of dangerous broken
pieces in case of failure, and hydrophobicity [16].
Condenser bodies of RBP and RIP bushings are solid products processed by turning. They mechanically adhere
firmly and tightly to the flange so in this manner and with their integrity, they separate transformer oil from the
surrounding medium. Therefore, the lower envelope is not necessary, because the body itself fulfils its task. In
the case of condenser body breakdown the integrity of the body and its sealing effect on the flange is usually
preserved well enough to prevent the oil from leaking from the transformer, but, nevertheless, in a certain
percentage of failures, leakage does occur. This, then, causes transformer fires because the oil leaks right onto
the glowing hot bushing parts, heated due to the breakdown.
In the case of OIP bushing, the situation is essentially different. There is no sealing effect of the condenser body
to the flange, so in the case of a fracture of both lower and upper envelopes, oil leaks from the transformer,
often leading to fires [17]. In the case of the upper envelope fracture, oil will not leak from the transformer
because the lower envelope is fixed to the flange and the sealing effect is preserved. (In some older versions
of OIP bushings, the sealing effect was assured by the axial force, so the fracture of at least one envelope
would cause oil leakage from the transformer).
Upper oil – air bushings envelopes contain sheds to ensure satisfactory creepage distance and are made of
porcelain or composite materials, with silicone sheds, or, most recently, silicone sheds are applied directly on the
RIP body. In OIP bushings, the space between the condenser body and the upper envelope is filled with oil, and
in RBP bushings, it is filled with insulation liquid. In RIP bushings, this space is filled with oil or, most recently with
insulation foam for the completely dry construction, or the space does not exist if silicone sheds are applied
directly onto the body. The porcelain upper envelopes are durable but breakable. They usually burst during
bushing breakdown and are sensitive to vandalism. Their hydrophobicity is reduced in the polluted atmosphere.
Composite upper envelopes, on the other hand, are mechanically tougher, more resistant to vandalism, they do
not burst and their hydrophobicity is better (because of silicone sheds), but they consider to be less durable
than the porcelain ones. RBP and, especially, RIP bushings can operate for some time even if the upper
envelope breaks. Due to greater toughness of the condenser bodies, RBP and, especially, RIP bushings have
generally better seismic characteristics than OIP bushings. RBP and RIP bushings can withstand temperatures up
to 120 °C, whereas OIP bushings are resistant up to 105 °C. (Nowadays, to reach better withstand
temperature of RIP bushings paper may be replaced with synthetic material or glass [18]. OIP and RIP bushings
have a very low partial discharge (PD), regularly several pC at test voltages. RIP bushings are sensitive to the
presence of PD because they have no possibility of regeneration that OIP ones have. Concerning PD, RBP
bushings have essentially poorer characteristics. Their PD reaches several hundred pC at test voltages and it
can be even a hundred at operating voltage. The reason is that they always contain some air, so this
Page 16
technology is nowadays considered obsolete. Capacitance and tanδ for OIP and RIP bushings are permanent
parameters until a disturbance occurs, making them very favourable for condition diagnostics. RBP bushings
gradually increase capacitance during operation (even by ten or more %) due to oil impregnation and this can
mask their defects.
2.1.6 CONDITION MONITORING

Transformer outages have considerable economic impact on the operation of an electrical network. Therefore it
is the aim to ensure an accurate assessment of the transformer condition. Techniques that allow diagnosing the
integrity through non-intrusive tests can be used to optimise the maintenance effort and to ensure maximum
availability and reliability. With the increasing average age of the transformer population there is an
increasing need to know the internal condition. For this purpose on- and off-line methods and systems have
been developed in recent years. On-line monitoring can be used continuously during the operation of
transformers and offers in that way a possibility to record different relevant stresses which can affect the
lifetime. The automatic evaluation of these data allows the early detection of an oncoming fault. In comparison
to this, off-line methods require disconnecting the transformer from the power network and are mainly used
during scheduled inspections or when the transformer is already suspicious [19].
Approximately since the 1990’s transformer on-line monitoring systems have been developed to such an extent
that the condition of all major parts of the transformer can be assessed. These techniques cannot prevent failure
of transformers, but they can prevent the costly repercussions associated with transformer failures by allowing
the user to take corrective action during the operating life. These advantages range from the ability to know
the overload capability of the transformer to detailed monitoring of faults in the active part and accessories
such as bushings and tapchangers. The cost and complexity of monitoring systems varies widely and the
application of all the available techniques to all transformers from new is not financially justified. Instead a mix
and match approach is more likely to be appropriate with the customer able to choose which type of
monitoring is most appropriate, depending on the importance and health of the transformer. For example, it
may be decided that monitoring cooler operation is important from new, but monitoring partial discharge is
only worthwhile when partial discharges have been detected by using routine dissolved gas analysis [20].
In the simplest case, transformer monitoring means the measurement of single parameters or the realization of a
simple model, e.g. gas in oil monitoring or thermal monitoring based on load current and temperatures. This
kind of monitoring makes it possible, to notify, if a threshold value is reached. The user will react upon reaching
a threshold value. Early warning for oncoming faults and maintenance management however are in this case
very limited. The next level of complexity is the transformer monitoring system. This kind of monitoring is the
basic monitoring of a single subcomponent of the transformer, like standalone OLTC monitoring or bushing
monitoring. Such systems observe more than one parameter and use simple models, in order to assess the health
of the supervised components. The correlation of data across transformer subcomponents is difficult, even if
several transformer subcomponents are monitored, because the data are distributed over different devices. The
correlation of measured data within the subcomponent however is possible. The foresighted detection of
abnormal conditions is possible in many cases. The interpretation of the obtained data is realized as status
information about the monitored transformer subcomponent. A comprehensive and interactive transformer
condition monitoring system integrates all relevant main components of the power transformer within only one
system. The correlation of data across transformers is possible, because all data are located in only one
database. The database and all algorithms are located on only one IED. The correlation of all data of the
transformer or even of several transformers is possible. The interpretation of the data delivers health
information of the transformer and all of its components. The expert system and its diagnosis functions support
the user to take the right decisions regarding foresighted operation and maintenance of the power transformer
[21].
Page 17
The experience gained from on-line monitoring of power transformers is increasing steadily. There is
nevertheless still a lack on how to integrate the information obtained by the on-line monitoring into the actions
taken onto the service of the transformer. The combination of on-line monitoring and off-line diagnosis provides
a powerful tool for the complete and economic assessment of transformer condition. The supplementary
information obtained by the off-line diagnostic after the detection of an abnormal condition is a worthful
information to be integrated into future condition monitoring systems. Compared to the number of power
transformers in operation, the number of installed monitoring system is quite low. Therefore the number of cases
where monitoring prevented major failures is not very big. Nevertheless several cases are reported, e. g.
detection of problems within the insulation system by online gas sensors, detection of bushing and tap changer
problems [22], [23].
Off-line diagnosis provides furthermore powerful tools which allow supplementary information on the
transformer condition, which so far cannot be integrated in an on-line monitoring system reliably and cost
effectively; among these methods are for example Frequency Response Analysis (FRA) for the detection of
winding deformations [24] and dielectric response measurements to assess the moisture content of the insulation
system [25].
2.2 Developments in Design Tools

Since the 1960’s, numerical methods (finite difference) or analogue methods (resistance networks or
electrolytical tank) were used to solve complex three dimensional partial differential equations governing the
electric, magnetic, thermal and hydraulic fields [26].
Computer aided design, employing techniques such as finite element analysis and draughting techniques, was
introduced in the late 1980’s to speed up design work, and to achieve more compact designs with the
improvement of space factor through the effective use of active materials (copper, iron and insulation) [4], [26].
These techniques allowed the design engineer to simulate the transformer operation under abnormal conditions,
and to determine how materials and constructional features would react to abnormal service conditions.
2.3 Developments in Manufacturing

There are two basic types of transformers categorised by their winding/core configuration namely shell type
and core type. The limbs of core type transformers are concentrically surrounded by the main windings. In a
shell-type transformer the flux-return paths of the core are external to and enclose the windings [8].
2.3.1 CORE CUTTING AND ANNEALING

Up to the 1970’s, the annealing process was used whereby the cut core steel laminations were passed through
an annealing furnace to reduce the additional loss due to mechanical stress introduced during the cutting
process. Burrs introduced during the cutting and slitting process were also ground away, and the sheets
reinsulated to replace insulating coating removed during the burr grinding operation [4].
The annealing process has since been eliminated with the use of modern cutting and slitting machines, and the
introduction of Hi-B steel, because the steel was easier to handle. The additional insulating coating on the core
steel sheets was also no longer needed because of the high surface resistance of the Hi-B steel. From the late
1980’s to the early 1990’s (based on the publication date of the reference), cutting and slitting machines
employed high quality silicon carbide, that was aimed at improving the quality of the core edge and reducing
operating costs [4].
2.3.2 WINDING CONSTRUCTION AND CONDUCTOR MATERIAL

Most manufacturers adopted disc type windings to reduce costs; they also require less labour to complete, but
require additional techniques (application of electrical interleaving or inter-shielding) to provide adequate
lightning impulse strength at the higher voltages [4].
Page 18
The introduction of vertical winding machines allowed windings to be manufactured with less handling by the
operator [26].
Measures implemented to achieve greater mechanical strength were hardening procedures during the
manufacturing process of the copper conductor strands, which resulted in controlled increase of the yield point
of copper conductor [6].
2.3.3 DRYING OF THE ACTIVE PART

Vapour phase drying was introduced in the 1940’s. It has the advantage that it reduces the drying time by a
factor of 3 to 6, depending on the size of the transformer, and it provides a more uniform dryness though all
parts of the insulation [4]. Since the 1950’s, major improvements have been made in oil drying plants, and it
became possible (from the late 1980’s to early 1990’s, based on the publication date of the reference) to
impregnate a transformer in the factory in a vacuum below 0.03 millibar, using hot transformer oil with a water
content below 2 parts per million in an air content below 2%, and with no fibrous particles in suspension [4].
2.3.4 CLAMPING STRUCTURE

To account for the loss in clamping during the drying process, manufacturers applied an external clamping
force. This was originally done by hand by clamping down the winding by a force applied through tie rods
between clamping rings. Techniques now being used by manufacturers include the use of large hydraulic
presses to clamp down the winding, and the development of isostatic clamping techniques where the winding is
subjected to a constant clamping force throughout the drying cycle [4].
Cores were originally constructed using bolted frames to clamp the laminations together. Bolts have since been
eliminated, first on core limbs, and then the yokes, from the 1950’s to the 1970’s. The laminations are (since the
late 1980’s or early 1990’s based on the publication date of the reference) held together by the hoop stress
of the windings, by fibre glass or metal banding (with an insulated insert), or by pinching the yoke laminations
between external clamps [4].
2.4 Relationships between Reliability and Manufacturing Period

The surveyed failure statistics in chapter 3 indicated no change in terms of the major contributors of failure with
manufacturing period. Due to the constraints also discussed in chapter 3, in particular the definition of the
manufacturing period of the failed transformers in the surveyed literature, it was difficult to draw any
meaningful inference on the relationship between reliability (failure patterns) and manufacturing periods.
However, Bossi [2] provided the most complete account of pre 1978 manufactured transformer failures.
2.4.1 WINDINGS
The exponential increase in rated power and voltage of transformers is directly related to power system
expansion. According to Seay [9], pre 1970 manufactured transformers did not perform well under increasing
short circuit stresses resulting from power system expansion. The impact of short circuit forces on a transformer
can usually be seen in the windings.
Failure statistics and developments supporting the impact of power system expansion on windings in this period
include:
 An increasing failure rate by year from the late 1960’s up to the 1970’s [2]
 Oscillating phenomena at regulating windings due to the lightning Impulse -> later the ZnO protection
of regulating windings was introduced.
 Failures classification indicated that winding related failures were among the major contributor of
failure location, and mechanical and dielectric type failures as failure modes; these are characteristic
of the impact of short circuit forces on the windings and their insulation (winding movement or
deformation, and premature aging of the winding insulation under repetitive short circuit forces).
Page 19
 The need for increased short circuit strength was acknowledged with the introduction of continuously
transposed conductors (CTC) during the 1960’s.
 With the increase in rated power and voltage, larger sized transformers were also required, and
design engineers had to take the transportation limits (rail, shipping and road transportation) into
consideration in the complete design of the transformer [27]. With optimisation of the transformer
design using computer aided design techniques, internal clearances could be reduced, resulting in lower
oil volumes and a reduction in insulation material; this has since led to dielectric and thermal mode
failures becoming more prevalent [9]. This observation is evident in the surveyed studies in chapter 3,
where dielectric mode failures featured as a major contributor of failure modes, irrespective of
manufacturing period.
2.4.2 MAGNETIC CIRCUIT/ CORE RELIABILITY

The constant development in core technology (material) had a positive impact on reliability and this is evident in
the results from the surveyed failure statistics, where core related problems featured as a minor contributor of
failure, irrespective of manufacturing period.
Pre 1970 core problems related to design and manufacturing of the core, in particular to the understanding of
leakage flux, also affected other parts of the transformer such as the windings. Significant developments in
core design and manufacturing were only implemented after the 1980’s, when problems related to leakage
flux were understood [26]. Larger transformers (GSU and transmission transformers) designed and built in the
1960’s were failing from the 1980’s, due to circulating currents in the unforeseen parallel circuits formed in the
continuously transposed conductor. This suggests that contributions of winding and core related failures of
earlier transformers could have been differently classified if failure had initially been identified as being
related to the core.
There are still some cases where the design did not consider the superposition of leakage flux from windings
with the magnetic field of high current conductors. Such faulty design can only be proved during service, where
either hot-spots in the core or on the tank appear. Such hot spots caused by the superposition of different
magnetic fields can be eliminated only by the reconstruction of the high current leads.
2.4.3 IMPACT OF DESIGN TECHNIQUES

Design related problems were a major contributor of failure in pre 1978 manufactured transformers
irrespective of application [2]. Analysis of the failure rate with age over the period 1955 to 1977 indicated a
high early life failure rate, indicative of design or manufacturing related problems [28]. Allan [26] attributed
the failures of large transformers from the 1960’s, to the inability to investigate the effects of abnormal service
conditions during the design phase. This suggests that numerical design techniques used at the time were not
effective.
However, Kogan [29] demonstrated a contradicting opinion regarding the effectiveness of pre 1980 design
techniques. Lower early life failure rates were observed in higher voltage units manufactured prior to 1977,
which have been attributed to the fact that manufacturers built more margins into the designs, by testing units at
higher lightning and induced test levels than required by the specification.
2.4.4 TAP CHANGER RELIABILITY

Tap changer related failures featured among the major contributors in pre 1978 manufactured substation
transformers [2], and the surveyed literature also indicated that these were observed irrespective of
manufacturing period. Failure statistics and developments in this period that could have contributed to poor tap
changer reliability are:
 Maintenance related problems were identified as a failure cause in pre 1978 manufactured substation
and autotransformers; although they were not related to any specific location in the transformer.
Page 20
 The tap changer technology used during this period was the oil type tap changer. With switching of the
diverter contacts in oil type tap changers, carbon particles are produced as by products in the diverter
oil. Tap changer diverters are usually designed to operate under these conditions, but in extreme cases
the carbon particles may reach levels that are detrimental to the tap changer diverter, causing
flashovers. As a result of the contamination and consequences, oil tap changers have to be maintained
regularly. According to manufacturer specifications, tap changers can either be maintained based on a
certain number of tap operations reached or a certain number of years from commissioning or last
maintenance, whichever of the two comes first.
 Tap changer maintenance involves intrusive work with human intervention, typically involving removal of
the diverter switches, cleaning of the diverter and insulating material, check of the arcing contacts and
replacement of the diverter oil. With human intervention and the nature of the tap changer
maintenance, it is possible that foreign particles or items may find their way into the diverter tank.
Switching operations take place in the diverter tank, and should there be foreign items, these could
result in flashovers, compromising the tap changer operation.
 One of the most common reasons for failures on tap changers manufactured pre 1978 were the non
silver plated selector contacts that led to contact heating and thermal defect of these contacts.
2.4.5 BUSHING RELIABILITY

Bushing reliability in the pre 1980 period could be attributed to the technology being used, which was the oil
impregnated paper type with porcelain insulator. The porcelain insulator in particular has the following
weaknesses:
 Pollutants and moisture deposits can contaminate the surface of the porcelain bushing insulators, causing
flashovers along the insulator, or between phases.
 According to Horning [30], flashovers of an insulator occur when most of the surface is covered with a
wet contaminant layer of low resistivity. As the flashover process develops, tracking may occur which
can cause cracks or punctures to develop on the insulator surface. This allows moisture ingress into the
condenser body which causes it to deteriorate. Saline pollution can cause abnormal corrosion of bushing
flanges and fittings, leading to hermetic seal deterioration that also causes atmospheric moisture
ingress, allowing the bushing condition to degrade with time. Industrial pollution produces irregular
surface voltage grading, which stresses the high voltage insulation, promoting flashovers along the
bushing or between phases, and insulation damage.
 Due to this problem, the surface of porcelain bushing insulators has to be maintained regularly in areas
where bushings are exposed to these conditions. The maintenance activities on bushings involve external
cleaning (hand wiping, washing with specialised equipment), or treating of the porcelain insulator with
dielectric compounds.
 Although no intrusive work is done on the bushing, handling of the bushing can incur damage to the
porcelain, or the condition of existing cracks and punctures may be worsened. This allows moisture
ingress into the condenser body which causes it to deteriorate.
According to Sokolov [31] approximately 80% of bushing failures in the period 1986 to 1996 (the pre
1986 build) occurred after 10 to 12 years in service, where design deficiencies and aging were typically
involved. The contribution of pre 1980 manufactured bushings is not known, so it is not known whether the
problems described are representative of the pre or post 1980 build. If the pre 1980 build was the major
contributor, this could suggest that the design techniques used at the time were not sufficient.
2.4.6 COOLING SYSTEM

According to Doble Engineering [32], known weaknesses less likely to cause failure in pre 1970 transformers
included corrosion in radiators and piping, and wear and failure of pumps and fans.
Page 21
3 DEFINITIONS AND GUIDELINES FOR RELIABILITY SURVEYS

This chapter will provide an overview of top-down approaches to transformer reliability as presented in the
literature, with emphasis on reliability practices in the industry, definitions and terminology associated with
reliability, and statistics from available reliability surveys.
3.1 Guidelines on Reliability Practices

The IEEE developed a guide that addresses the collection and reporting of transformer failures by establishing
a databases and statistical analysis of data through the calculation of failure rates [33]. The guide
distinguished between reliability improvement and reliability measurement. Determining failure rates was
considered as reliability measurement, whereas the classification into failure location and causes should be
pursued from a standpoint of reliability improvement.
In 2000, the Cigré study committee on transformers, Study Committee A2, set up a Reliability Advisory Group
to consider and recommend how Cigré could and should contribute to activities related to transformer reliability
[34]. The work of this advisory group has been summarised and recommendations made. Recommendations
focussed on three areas for future activities, and included (i) reliability of new transformers, (ii) reliability
during mid-life, and (iii) risk management at the end of life.
Cigré WG A2.18 gave recommendations on failure identification by developing the following [35]:
 Standard failure reporting form, which allows for all relevant information to be gathered.
 Failure code, which classifies the failures by the location, the nature of the failure and the cause of the
failure
 Failure identity card, which can produce a complete description in a standardised format.
3.2 Reliability Surveys

There is limited literature available in the public domain discussing failure statistics of transformers. Internal
surveys are being conducted by certain countries and utilities, and a number of well established databases
exist from countries such as Australia, Brazil, Canada, Germany and Japan. The results or methodology of these
surveys are, however, not made available publicly due to confidentiality and competition between utilities.
Another possible reason for the scarce information could be the fact that the industry started referring to the
results of the international survey by Bossi [2] because of its thoroughness, instead of conducting individual
surveys.
Measures of reliability from the surveyed literature included the calculation of failure rates, failures
classification into failure locations, causes, and modes, and the application of statistical analysis to determine
reliability with age. Bossi [2] presented the results of an international survey of transformer failures from 13
countries that calculated failure rates, and implemented a failures classification. The results of this survey have
become an international benchmark in the transformer industry for comparing transformer reliability statistics.
The aforementioned measures of reliability with accompanying definitions of failure were not available in all
the surveyed studies. Where definitions were available, they differed between the studies. Definitions from
various surveys and international standards are summarised in Appendix B.
Although this is not mentioned in the surveyed studies, cases where the studies were limited to certain measures
of reliability could be due to unavailability of data. Factors influencing the unavailability of data could be:
 Lack of skills or time to conduct proper failure analysis to determine failure locations, causes and
modes.
 With the impact of aging on power system infrastructure, changing economic climate and increased
demand for electricity, utilities are forced to invest wisely in new infrastructure. One of the requirements
Page 22
of detailed asset management decisions is sufficient data on the failures history of the installed
population. Because most utilities are only now faced with this problem, the need for the collection and
record keeping of such data in earlier times may not have been appreciated.
3.3 Definitions
The most relevant definitions from Appendix B are discussed below.
3.3.1 FAILURE
The majority of the surveyed studies did not provide consistent failure definitions. In order to have a true
comparison between the failure data, it would be required that the failure definitions are similar. This ensures
that comparisons are based on failures (events) occurring under the same conditions.
The definition of failure and associated terms, and how they are applied depends on the environment in which
it was developed and is being used. The systems operator’s focus would be on the impact on the system,
ranking failure in terms of system reliability, whereas the plant specialist would rank it in terms of what
remedial action would be required to restore equipment functionality. A clear example provided by [34] is that
of transformers being removed in scheduled outages. The system operator would not consider this as a failure
since it would not have an effect on the system reliability, whereas the plant specialist would consider it as a
failure. Another example is that of insurance companies that could be using insurance claims as a benchmark,
where the definition of failure and its severity could be dictated by the value of the claim.
The definitions in IEC [36] and IEEE [33] are described as being very broad, whereas Bossi’s [2] definition is
more restrictive, in that it considers only problems that require the unit to be removed from service to be
repaired [35]. Further examples of restrictive definitions include those of Kogan [29] and Higgins [37]. Despite
the difference between having a broader or restrictive definition, both types allow further breakdown of
failures into levels of failure severity, or outage type and times.
3.3.2 RELIABILITY
According to the IEEE [33], stating reliability without mentioning the period of time involved makes it incomplete.
Reliability was therefore defined as ‘the probability that a transformer will perform its specified function under
specified conditions for a specific period of time’. This definition is in line with that of the IEC 60050-191 [36]
where reliability is defined as ‘the ability of an item to perform a required function under given conditions for
a given time interval’.
Cigré WG A2.18 [35] defined reliability as ‘the probability that the equipment will remain in service without a
failure occurring’, excluding the connotation of a specified time interval.
3.3.3 FAILURE RATE

Failure rate has been the most utilised measure of reliability because it is based on the count of the number of
failures which should be easily obtainable. Where failure rates were available from the available studies, they
were calculated and trended by year of failure, and over a period of time (number of years).
The failure rate of a single population is defined as:
n1 + n 2 + ... + ni
λ= ⋅ 100% (3.1)
( N 1 + N 2 + ... + N i ) ⋅ T
ni = Number of failures in i-th year

Ni = Number of transformers operating in the i-th year
T = Reference Period (normally one year)
Page 23
It is also possible to calculate the combined failure rate of different populations (e.g. different utilities). In this
case it is postulated that the number of operating transformers is constant during the reference period.
n1 + n2 + ... + ni
λ= ⋅ 100% (3.2)
N1 ⋅ T1 + N 2 ⋅ T2 + ... + N i ⋅ Ti
ni = Number of failures by i-th population

Ni = Number of transformers of i-th population
T = Reference Period of i-th population
The hazard function gives the instantaneous failure rate with time [36], [38], [39], [40]. A condition in IEC
60050-191 requires that only the failures of non-repaired items be considered when using the expression for
instantaneous failure rate. A non-repaired item is an item which is not repaired after failure.
The average failure rate over an interval can then be determined from the integral of the hazard function over
that interval, divided by the width of the interval [36], [39], [40].
International standards claimed that failure rate should only be calculated for forced outages IEEE [33]. Failure
rate was calculated as the ratio of the number failures with forced outages of a given population over a given
period of time, to the number of accumulated service years for all transformers in that population, over the
same period of time. [33] and [41] expressed the numerator of the failure rate definition in terms of forced
outages only, whereas [2] calculated failure rate for scheduled and forced outages. Both [2] and IEEE [33]
require the immediate removal of the transformer from service. [2] also requires that the removal from service
take place within 30 minutes. The definitions of scheduled outage and defect in both references are the same.
Failures with scheduled outages occur when a transformer is deliberately taken out of service at a selected
time. Defects arise where imperfection or the lack of performance may be corrected or eliminated without
taking the transformer out of service.
Elerath [40] further distinguished between annual failure rate and annualised failure rate. The annual failure
rate is similar to the average failure rate but the interval is one year. The annualised failure rate is an estimate
of the annual failure rate based on extrapolating beyond the period of time that data was collected, and
normalising to one year. Based on this, the failure rate as determined in [2], Kachler [41] and IEEE [33] is an
average annual failure rate, taken over a number of years, instead of one year.
3.4 Review of Available Reliability Statistics
3.4.1 CONSTRAINTS
Without clear definitions, as well as adequate descriptions of the transformer application, voltage level, failure
period and manufacturing period, making comparisons to identify trends becomes difficult. Most of the
surveyed literature adhered to this constraint. In summary, the following factors constrained the comparison of
the available statistics:
1. Limited publicly available reliability survey results.
2. The majority of the surveyed studies did not provide failure definitions. In order to have a true
comparison between the failure data, it would be required that the failure definitions are similar.
3. Where failure rates were given, the majority of the sources did not provide the formula used for
calculating failure rate.
4. Further, the majority of the studies did not provide the manufacturing period of the failed units. In most
cases only the failure period was given, and an attempt was made to determine the manufacturing
Page 24
period from the failure period. However, this introduced another constraint. For example, if the failure
period was from 2000 to 2005, then the manufacturing period would be pre 2005, meaning that this
could include older and younger units, i.e. pre and post 1980. If the spread of the older versus younger
units is not known, it becomes difficult to assess which group’s failures actually dominated the sample or
if they had equal contributions. The same reasoning applies to the failures classification per
manufacturing period, where a specific failure location cause or mode could dominate a sample.
5. The surveyed studies are from different countries with varying transformer design specifications,
maintenance philosophies, condition monitoring practices and operating conditions. These factors could
have an impact on outcomes of comparisons, however the extent to how much these factors’ impact
affects the results of the surveys is not known.
6. The surveyed studies did not use the same categorisation of transformers according to application. Bossi
[2] referred to power station, substation and autotransformers, whereas others referred to GSU,
transmission and distribution transformers. Although this is not mentioned in Bossi [2], power station
transformers include GSUs, station and unit transformers. Station transformers are used to supply
auxiliary equipment in the power station, and unit transformers’ primary winding is directly connected
to the generator output [7]. In most cases, the difference between these units is the highest system
voltage each connects to. In the case of units housed in substations, higher voltage units are more
generally referred to as transmission units, and the lower voltages referred to as distribution units.
Transmission and distribution transformers are housed in substations, and they are therefore also
substation transformers. Transmission and distribution transformers can also be two winding, three
winding or autotransformers.
7. In the surveyed studies, the classification of failures was done according to failed location, failure cause
or failure mode. Some of the surveys contained data with mixed classification where failure causes,
and/or failure locations, and/or failure modes were used together [32], [42], [43]. This suggests that
there is not consensus in the industry regarding the use and application of failure classification
terminology.
The above-mentioned constraints were acknowledged, and attempts at comparisons have therefore been made
with caution.
3.4.2 FAILURE RATES

Bossi [2] investigated more than 1000 failures from 13 countries, in transformers not older than 20 years, and
failing in the period 1968 to 1978. This was a very comprehensive study where failure rates were calculated
in terms of transformer application, voltage level, and outage type. Forced outages and the combination of
forced and scheduled outages were considered.
The study concluded that for the units installed on systems operating at voltages lower than 700kV, the
average failure rate may be stated as 2%. In this survey there was no distinction made between minor and
major failures. Since then, this statistic has become an international benchmark in the transformer industry for
transformer failure rate performance. Considering the failure period, the year of manufacture of these units
would be up to and including 1978. This would make the failure rate of 2% inapplicable to transformers
manufactured after this period. Bossi [2] also concluded that failure rates increase with increasing voltage level.
The failure rates of variously rated transformers from utilities in different countries spanning from 1968 to
2005 is shown in Table 1. With the exception of Bossi [2] and Fleeman [44], none of the sources in Table 1
provided definitions of failure or formulas for failure rate calculation. The failure rate calculations of Bossi [2]
and Fleeman [44] were based on the same formula.
Page 25
Failure rates per annum across different manufacturing periods spanning from pre 1978 to pre 2005 varied,
but remained within the 2% benchmark, assuming the failure rate calculations have been based on Bossi [2]. At
2.9%, only the failure rate of 765kV pre 1986 manufactured GSUs exceeded the 2% industry benchmark. In a
way, this result could have been expected since the 2% failure rate was based on units up to 700kV, and
therefore excludes the 765kV units. This could suggest that the 2% failure rate determined by Bossi [2] was
valid across all manufacturing periods.
Only two studies showed clear distinctions between the failure rates according to manufacturing period, and
both applied to single phase GSUs. The outcomes of the surveys contradict each other: Kogan [29] reported
that the older designs (post 1977) were more reliable than later versions, whereas Hall [45] reported that
failure rates made older designs (pre 1974) less reliable than their younger counterparts.
Kogan [29], Wagenaar [46], and Fleeman [44] considered 345kV and 765kV GSU and autotransformers in
the United States, which were manufactured pre and post 1977, and pre and post 1986. In the initial study,
Kogan [29] observed that post 1977 manufactured 765kV units had higher failure rates than their pre 1977
counterparts. The lower failure rates of pre 1977 units have been attributed to the fact that manufacturers built
more margins into the designs, by testing units at higher lightning and induced test levels than required by the
specification. In light of the higher failure rates, an initiative was undertaken to improve dielectric specification
of these transformers and this was enforced in 1986 [42] [44]. Since then, studies have shown lower failure
rates in post 1986 manufactured units, compared to pre 1986, which varied from 1.3% to 2.9% before the
implementation of the new specification.
In a study of the reliability of single phase generator transformer sets in the United Kingdom, the failure rate
for units failing in the period 1974 to 1994, post 1974 build, was reported to be 1.2% [45]. This was an
improvement in failure rate which stemmed from initiatives in terms of specification, design and manufacture, to
enhance the reliability of the transformer fleet, after the poor experience with three phase units during 1966 to
1974, the pre 1974 build. Since 1995, the failure rate of single phase transformers has shown further
improvement to 0.6%.
This demonstrates that changes in the specification allowed differences between manufacturing periods to be
observed in the failure rate. Although these two studies are not sufficient to draw a thorough assessment of
reliability, they could suggest significance in behaviour of transformers prior to and after the 1970’s, and
transformer application.
Page 26
TABLE 1: FAILURE RATES FROM SURVEYS FROM 1968 TO 2005
APPLICATION/ MANUFACTURING
FAILURE FAILURE
SURVEY SOURCE
CLASSIFICATION PERIOD PERIOD RATE (%)
Cigré International All voltage levels

1968 - 1978 Pre 1978 2 [2]
Survey (60kV-700kV)
United Kingdom All voltage levels Pre 1987 Pre 1987 <2 [26]
Generator Step-Up,
United Kingdom 1974 - 1995 Pre 1995* 1.2 [45]
Major failures
Generator Step-Up &

ZTZ Service
Transmission Rated 2000 - 2005 Pre 2005 1-2 [28]
Database, Ukraine
power ≥ 100MVA,
US-NGRID, United Distribution 115kV,

- - 0.35 - 0.8 [47]
States 69kV & <69kV
Hydro Quebec, All voltage categories,

- - < 0.5 [48]
Canada major failures
American Electric
345kV & 765kV Pre 1986 Pre 1986 * 1.3 - 2.9 [44]
Power
American Electric
345kV & 765kV Post 1986 Post 1986 * 0.35 - 1.35 [44]
Power
Australia & New

Costly failures Pre 1996 Pre 1996 0.4 [49]
Zealand
* Indicates studies where the manufacturing period was given.
Failure Rate Dependent on Age

Two types of failure rate have been mentioned: a failure rate for a group of transformers over a time period,
and a time dependent failure rate given by the hazard function. Bossi [2] determined the failure rate by age
group, and applied the same failure rate calculation as used for a group of transformers. Later references
used the hazard function for determining the failure rate with age [28], [29], [42], [47], [48]. The failure
probability with age was determined from the cumulative distribution function and the probability density
function.
According to Lapworth [34], although determining the reliability with age statistical analysis techniques (hazard
function, probability density function, and survival function) provides a more realistic model, it is debatable how
relevant a statistical evaluation of transformer failures will be, since they are so dependent on random external
events and internal weaknesses. Lapworth further argues that a more useful approach would be to determine
the likely onset of unreliability on an individual basis.
Results from the literature included the following:
The failure rate of pre 1978 manufactured units in the higher voltage ranges above 100kV tended to increase
with age and those below 100kV decreased with age according Bossi [2].
According Sokolov the failure rate of pre 2005 manufactured GSUs from 220kV decreased with age [28].
Jongen [50] indicated an increasing failure rate with age for units from 50kV to 150kV.
Page 27
Sokolov further indicated a decreasing failure rate with age in pre 1965 manufactured transformers, which
after design improvements, led to an increasing failure trend with age [28], [43].
Tap changer and winding related failures also indicated an increasing failure rate with age [51].
3.4.3 FAILURE LOCATION

Tap changer related failures were the largest contributors in failure rates of substation transformers
manufactured up to 1978 [2]. GSU failure statistics of units manufactured up to 2005 were largely winding
related, where failures developed in the winding minor insulation [42]. Contributions to failure from the tap
changer were most prominent in post 1974 manufactured GSUs in the United Kingdom [45]. In pre 1978
manufactured substation transformers with an on-load tap changer, the tap changer was the major contributor
followed by winding related failures [2]. In units without an on-load tap changer, failure of the windings, and
the tank and dielectric fluid were most prominent. In autotransformers, windings and other accessories
dominated.
The failures of pre 2005 manufactured transmission transformers above 100MVA occurred mostly in the
bushings and major insulation [42]. The evaluation of a group of 167MVA transformers, also manufactured pre
2005, showed a similar distribution, with the inclusion of minor insulation and winding insulation failures. Core
related failures did not feature as a major contributor in any of the surveyed studies, and was a minor
contributor in pre 1978 manufactured transformers, irrespective of transformer application [2].
3.4.4 FAILURE MODE

In Bossi [2], failure origin is equivalent to that of failure mode. For the pre 1978 transformers, dielectric mode
failures were more prominent in units with an on-load tap changer in the group of power station and
autotransformers, and mechanical in substation units, for the combination of forced and scheduled outages.
When considering units without an on-load tap changer, the results are significantly different, with mechanical
and dielectric mode failures most prominent in power station and autotransformers, and substation units,
respectively.
The ZTZ Service database considered 108 major failures of units from 9 different manufacturers rated
100MVA and above, all manufactured up to 2005 [42]. Most GSUs (of which 56% were manufactured after
1975) failed due to dielectric mode failures originating in the windings insulation. Thermal mode failures have
shown an increasing trend where they resulted in overheating of leads and connections.
Kogan [29] reported increased dielectric mode failures in GSU and autotransformers in the range 345kV to
765kV, in the absence of any system disturbances such as lightning, elevated system voltage, line or switching
faults, all of which were manufactured after 1977. In 765/345kV autotransformers problems were nearly all
caused by dielectric failure, and mainly concentrated in the first nine years of life. Failures at later ages were
of the mechanical type.
Dielectric mode failures featured as major contributors in most of the surveyed studies, irrespective of
manufacturing period and transformer application [2], [29], [42].
3.4.5 FAILURE CAUSES

According to Sokolov [42], three main causes of failure may be considered when the withstand strength of the
transformer with respect to one of its key properties is exceeded by operating stresses. These are:
 Originally insufficient safety margin due to underestimation of operational stresses (poor specification),
design deficiencies, manufacturing weaknesses, or material defects.
 Operation stresses exceed specified quantities (unusual event, operational error).
Page 28
 Critical deterioration of safety margin, including inadequate maintenance, low quality repair or
refurbishment. With years, one can expect gain in weight of this third cause, which automatically
increases the influence of initial design margin, as well as operational stresses.
From a manufacturer’s perspective, similar views are shared where transformer failures can be grouped under
three main headings, as shown by Allan [51]:
 Failures due to weak specification, design deficiencies, manufacturing weaknesses or material defects,
 Failures due to system disturbances, operational factors, or interactions between the transformers and
other equipment on the system,
 Failures which result from maintenance operations, repairs or refurbishment that has or have not been
undertaken.
According to Bossi [2] pre 1978 manufactured units, in all three groups of transformer application, failed
mostly due to design and manufacture related problems.
Not many of the literature sources provided a breakdown of failure causes with failure locations. According
Sokolov [31], approximately 80% of bushing failures in the period 1986 to 1996 (the pre 1986 build)
occurred after 10 to 12 years in service, where design deficiencies and aging were typically involved.
Design and manufacturing related problems were identified as major contributors in most of the surveyed
studies, irrespective of manufacturing period and transformer application [2], [47].
3.5 Theory of Statistical Analysis
3.5.1 TYPES OF STATISTICAL DATA

Proper data analysis depends on the type of data. Nelson described two main data types: complete data and
censored data [38].
Complete data arise where the value of each sample unit is observed (Figure 3a). Incomplete data can be
truncated data or censored data. When we know that an event occured before or after a certain moment, we
say that data is censored (left or right respectively) (see (Figure 3f). When we do not know that an event
occured before a certain moment, we say that data is left truncated.
Nelson classified censored data into 6 data types as right censored, left censored, singly censored, multiply
censored, time censored, and failure censored:
 Right Censored Data: Some units are un-failed and their failure times are known to only to be beyond
their present running times. Un-failed units are called suspended unit, run-outs or removals. (Figure 3b,c)
 Left Censored Data: Units known to have failed before certain time/date, whose failure time is not
known. (Figure 3d)
 Singly Censored Data: Arises when units are started on test together and the data analysed before all
units fail. Singly censored data can be left or right censored. (Figure 3b,c)
 Multiply Censored Data: Data with differing running times intermixed with failure times; where units go
into service at different times and have different running times when the data is recorded. (Figure 3e)
 Time Censored Data: Singly and multiply censored data may be time censored. Singly censored data is
time censored if the censoring time is fixed. Multiply censored data is time censored if the running times
differ from the failure times. (Figure 3b)
 Failure Censored Data: Singly and multiply censored data may be failure censored. Singly censored
data is failure censored if the test is stopped when a specified number of failures occur with the time to
that fixed failures being random. Multiply censored data is failure censored if the running times equal
the failure times. (Figure 3c)
Page 29
FIGURE 3:CENSORING OF STATISTICAL DATA
3.5.2 BASIC RENEWAL THEORY

Renewal theory is the theory that explains failure situations where use based preventative actions lead to
complete restoration, leaving the failed item in a good-as-new condition after repair [52]. The theory can be
approached from various angles using different mathematical theory bases; however it is often approached
through the use of statistical distributions [38]. The one most often used is the Weibull distribution.
The five functions that are of primary importance in Renewal Theory are the reliability functions: Probability
Density Function (pdf), Cumulative Distribution Function (cdf), the Survival (Reliability) Function and the Hazard
Function.
Probability Density Function (PDF)

The probability density function (PDF), f(t), gives the probability of failure occurring at any specific time. Its
units are failure/item-time. Thus, at any point it gives the probability of a failure occurring during the following
time [52].
The probability density function is defined as:
1 ∆n(t ) (3.3)
f (t ) = ⋅
N ∆t
∆n(t) = Number of failures in time interval [t, t+∆t]

∆t = Length of time interval
N = Original Population
It is the mathematical model for the population histogram containing relative frequencies which must be greater
than or equal to zero, and must sum up to unity.
Page 30
If t represents age at failure of a device, then a point on the probability density function may be interpreted
as:
 percentage that failed at a certain age,
 probability of failure at a certain age,
 the mode of the distribution, i.e. the age at which failure is most likely to occur, is given by the maximum
point on the probability density function.
FIGURE 4: PROBABILITY DENSITY FUNCTION (PDF) [52]
The relative frequency of failure distribution is required for estimation of PDF. Therefore, the number of failures
(frequency) per age interval 𝑡 𝑘 is normalized dividing by the total number of items (transformers). These values
represent the histogram for failures distribution.
Failure Distribution Function (CDF)

The failure distribution function is a cumulative distribution function (CDF) and gives the cumulative probability of
failure. It thus represents the probability that a failure has occurred on or before a certain time.
The failure distribution function is defined as:
F (t ) =
∑ n (t )
i (3.4)
N
Σni(t) = Number of failures up to time t

N = Original Population
The failure distribution function is the integral of the probability density function f(t),
t
F (t ) = ∫ f (t )dt (3.5)
0
FIGURE 5: FAILURE DISTRIBUTION FUNCTION (CDF) [52]
Page 31
Survival (or Reliability) Function

The survival function R(t) is defined as:
R (t ) = 1 − F (t ) (3.6)
The survival (reliability) function is thus the complement of the failure distribution function. It gives the
probability of survival up to any specific time t.
If t represents age at failure of a device, then a point on the survival/reliability function may be interpreted as:
 The probability of surviving beyond a certain age t is R(t),

 Percentage of that survived at least until age t,
 The survivorship at age t is R(t).
FIGURE 6: SURVIVAL FUNCTION [52]
Hazard Function
The hazard function is a measure of proneness to failure as a function of age. It is defined as:
f (t ) 1 ∆n(t ) (3.7)
h(t ) = = ⋅
R (t ) N (t ) ∆t

N(t) = Population surviving at time t
∆t·h(t) is the probability of an apparatus of age t to fail in the time interval from t to t+∆t.
A decreasing hazard function during the early life of a product corresponds to infant mortality. This often
indicates that the product is poorly designed or suffers from manufacturing and installation defects. The second
region spans the working life of the component. During this period a constant and fairly low risk of failure is
experienced. The failures are of random nature. An increasing hazard function during later life of a product is
said to correspond to ageing behaviour due to wear-out failure. Such failure rate behaviour often indicates
that failures are due to the product wearing out. [38], [52]
A hazard function showing a decreasing failure rate in the early life and an increasing failure rate in later life
is referred to as a bathtub curve by reliability engineers, as shown in Figure 7. The flat part of the bathtub
curve is called useful life.
In case the hazard curve does not show an increase with time the use of Time Based Maintenance will not be
effective. In these cases the only recommended strategies are Condition Based Maintenance or Corrective
Maintenance [52].
Page 32
FIGURE 7: THE ‘BATHTUB CURVE’ HAZARD FUNCTION [53]
3.5.3 WEIBULL PARAMETER ESTIMATION

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
determines 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 [3]. With the help of the Weibull-
Grid the characteristic life and a certain probability of failure of components can be read. One reason for its
popularity is that it has a great variety of shapes. This makes it extremely flexible in fitting data, and it
empirically fits many kinds of data. The two-parameter Weibull propability density function is as follows [38]:
𝛽
𝑓(𝑡) = �𝛼𝛽� 𝑡𝛽−1 𝑒𝑒𝑒�−(𝑡⁄𝛼 )𝛽 � ; t>0 (3.8)
For the hazard curve h(t) the two-parameter Weibull fitting function is [38]:
β -1
 β  t 
h(t) =  . 
α α  (3.9)
The parameter 𝛽 is called the shape parameter (or slope of the CDF) and is positive. 𝛽 is a dimensionless
number and determines the shape of density function. The parameter 𝛼 is called the “characteristic life”, since it
is always the 100 × (1 − 𝑒 −1 ) ≅ 63.2𝑡ℎ percentile. 𝛼 has the same unit as t, for example, year.
If 𝛽 = 1, the hazard rate will not depend on time. For values of 𝛽 larger than 1 the failure probability will
increase with time. In case 𝛽 larger than 2 a gradually increasing failure probability will occur. So an age limit
for the operation of the components would be advisable. If 𝛽 ≈ 3.7, the hazard rate will follow the normal
distribution.
Page 33
h(t)
FIGURE 8: SHAPES OF THE WEIBULL HAZARD FUNCTION [52]
3.5.4 REPAIRABLE SYSTEMS

According to Ascher & Feingold, a repairable system is a system which after failing to perform at least one of
its required functions, can be restored to performing all of its required functions by any other method, other
than replacement of the entire system. A non-repairable system is a system which is discarded the first time that
it ceases to perform satisfactorily. [54]
Although repairable systems (e.g. automobiles, personal computers, etc.) are common, the techniques for
analysing repairable systems are not as widespread as those for non repairable systems. This situation often
leads to incorrect analysis techniques due to confusion between the hazard rate and rate of occurrence of
failures.
Repairable Systems Theory
Renewal theory assumes that repair resulted in the component or system being restored to the as-good-as-new
condition. Repairable systems theory on the other hand assumes that reliability degradation takes place over
time and repair as-bad-as-old or the minimal repair condition, i.e the condition it was just before failure. This
does not imply that the system will immediately fail again, but that it was not restored to the orginal (new)
condition. [52]
Repairable Systems Theory makes use of regression to model the long term life trend of a system. A system is
made up of many components each using its own failure characteristics which can be modelled using Renewal
Theory. But the system itself, if it degenerates over time must be modelled using regression techniques. If the
system does not degenerate over time, Renewal Theory can be used to model the failures of the system. One of
the models that can be used for the purpose of modelling the failure rates (rate of occurrence of failure) is the
Non-Homogeneous Poisson Process models.
Data Trend Testing
Deciding which theory to use, requires the use of a data trend test to determine if the system or component
being analysed is experiencing reliability growth (lives between failures are getting shorter) or reliability
degradation (lives between failures are getting shorter). If the system is experiencing reliabilility growth or
reliability degradation, then the individual lives cannot be thrown into the same hat (distribution) for analysis.
That is, renewal theory cannot be used, and repairable systems theory is required. [52]
The Laplace trend test is data trend test which has been proven to one of the best trend tests for trend in
failure analysis, and is given by the following equation:
Page 34
𝑛−1
∑𝑖=1 𝑇𝑖 𝑇𝑛
−
𝐿= 𝑛 − 1 2
1�
1 2
𝑇𝑛 � �
12(𝑛 − 1)
Where:
𝑛 = number of failures that have occurred
𝑡1 , 𝑡2 , 𝑡3 ………………….𝑡𝑛 = times at which failure occurred
𝑇1 , 𝑇2 , 𝑇3 ………………….𝑇𝑛 = interarrival times to failure
𝑇𝑘 = ∑𝑘𝑖=1 𝑡𝑖 The Laplace quantity when calculated for a failure data set without a trend will produce a value
with an average of 0 and standard deviation of 1.
• If the value of 𝐿 lies between -1 and +1, the data set has no inherent trend and renewal theory can
be used for the analysis.
• If the value 𝐿 is smaller than -2, the data set shows a definite inclination towards reliability growth and
repairable systems modelling can be used for the trend analysis.
• If the value of 𝐿 is larger than +2 the data set shows a definite inclination towards reliability
degradation and the repairable systems theory can be used for analysis.
• If the value of 𝐿 is between -1 and -2, or between +1 and +2, we are in a grey area where,
statistically speaking, both renewal theory and repairable systems theory can be used.
The decision process for deciding which theory to use is shown in the figure below. [52]
Page 35
FIGURE 9 : DATA ANALYSIS FRAMEWORK [52]
Page 36
4 DESCRIPTION OF EXISTING TRANSFORMER RELIABILITY SURVEYS

In this chapter, existing country wide surveys and company surveys are discussed, in particular general
information about specific definitions and the methodology of data acquisition and analysis.
4.1 Countrywide Surveys

Country wide surveys are a systematic collection of data on the availability and disturbances of the electrical
power supply, with emphasis on the frequency, duration and extent of the interruptions. Detailed statistics
about the failure location in the respective equipment, failure cause or mode and repair activities are in most
cases not included. The benefit of these statistics with respect to asset management is therefore limited.
The surveys from Canada, Germany, Japan and Brazil are presented and discussed in this section. The failure
data as presented in the surveys are based on data from multiple utilities in each country. Each survey consists
of five major sections, as follows :
 General Information
 Information on data contributors and the scope of the equipment dealt with in the survey.
 Specific definitions and methodology of data acquisition and analysis
 Investigated population of transformers
 Summary / example of the reliability data
4.1.1 CANADIAN ELECTRICITY ASSOCIATION
General Information
The Canadian Electricity Association (CEA), founded in 1891, is the national forum and voice of the evolving
electricity business in Canada. CEA contributes to the regional, national and international success of its members
through the delivery of quality value-added services.
In 1975, CEA adopted a proposal to create a facility for the centralized collection, processing and reporting
of reliability and outage statistics for electrical generation, transmission and distribution equipment. To
coordinate the development of this Equipment Reliability Information System (ERIS), CEA constituted the
Consultative Committee on Outage Statistics (CCOS). [55], [56], [57]
In 1978, the transmission stage of the information system was implemented when Canadian utilities began
supplying data on transmission equipment in accordance with the Instruction Manual for Reporting Component
Forced Outages of Transmission Equipment. This report is published annually and is based on data of the
period of the last 5 years.
Contributors and Scope

The data used in preparing this report are submitted by the following utilities [56]:
 ATCO Electric
 Britisch Columbia Transmission Corporation (BCTC)
 Churchill Falls (Labrador) Corp.
 EPCOR
 Hydro One
 Hydro-Québec (Trans Énergie)
 Manitoba Hydro
 Newfoundland & Labrador Hydro
 Sask Power
Page 37
Each contributor receives an individual utility report containing the same tables shown in the Appendices but
based on the contributor's own data. Also contained in the Individual Utility Report are statistics for each and
every contributing utility.
The survey covers transmission equipment (transmission lines, transformer banks, circuit breakers, cables) in
Canada with an operating voltage of 60 kV and above and also includes those elements associated with
transmission systems such as synchronous and static compensators, and shunt reactors and capacitors on the
tertiaries of transformers of 60 kV and above. It does not cover equipment while in commissioning, nor SF6
equipment (except circuit breakers) nor DC equipment.
Specific Definitions and Methodology of Data Acquisition and Analysis

All the definitions are based on [56]:
Transmission Equipment
All equipment with an operating voltage of 60 kV and above and includes those elements associated with
transmission systems such as synchronous and static compensators and also shunt reactors and capacitors on the
tertiaries of transformers of 60 kV and above.
Major Component
A unit of Transmission Equipment, including all the associated auxiliaries that make it a functional entity within a
power system.
Subcomponent
The constituent components of a Major Component and includes those external elements which are associated
with it.
Terminal
A transmission line end or cable end which is equipped with primary protection.
A Unit of Transmission Equipment
A three-phase installation whether this installation is made up of one three-phase element or three single-phase
elements.
Primary Cause
The reason to which one can attribute the outage or malfunction of a Major Component.
Component Forced Outage
The automatic or emergency removal of a Major Component directly caused by defective equipment, adverse
weather, adverse environment, system condition, human element or foreign interference. Recording is not done
for the case of healthy Major Components which are removed from service as a result of cascading system
events or as a result of the outage (or malfunction) of some other Major Component and recording is not done
for manual removal of a component from service where that removal may be delayed more than thirty minutes
to allow load transfer or other operations.
Common Mode Outage
An event where more than one component forced outage results from a single primary cause and where the
outages are not consequences of each other.
Failure Mode
A description of the type of fault (or malfunction) which the system sustains as a result of a Component Forced
Outage.
Page 38
Sustained Forced Outage

A transmission line related forced outage the duration of which is one minute or more. It does, therefore, not
include automatic reclosure events.
Transient Forced Outage
A transmission line forced outage the duration of which is less than one minute and is, therefore, recorded as
zero. It covers only automatic reclosure events.
Replacement Time
The elapsed time required to replace the Major Components from stock or other location in the network.
Forced Unavailable Time
The elapsed time required to restore the Major Component to service or to repair it in the case where it has
been replaced. All events have a Forced Unavailable Time of minimum one minute with the exception of
Automatic Reclosures on Transmission Lines which are reported as "zero" forced unavailable time.
Kilometer Years (km. a)
The summation of the product of the length in kilometers and the period duration in years for the transmission
lines or cables under consideration.
Terminal Years (a)
The summation of the product of the number of terminals and the period duration in years, for the transmission
lines or cables under consideration. A terminal refers to a transmission line or cable end which is equipped with
primary protection.
Component Years (a)
The summation of the product of the number of units of a Major Component and the period duration in years,
for the major component under consideration.
Number of Outages
The number of major component related forced outages which involved the indicated subcomponent or primary
cause.
Frequency (per 100 km. a)
The Number of Outages divided by Kilometer Years which are in turn divided by 100.
Frequency (per a)
The Number of Outages divided by Terminal Years or Component Years.
Total Time (h)
The sum of the forced unavailable times (in hours) of Major Component related forced outages involving the
indicated subcomponent or primary cause. Forced unavailable time is the elapsed time required to completely
restore a Major Component to service.
Mean Duration (h)
The Total Time divided by the Number of Outages.
Median Duration (h)
The time at which fifty percent of the forced unavailable times are greater than this value and fifty percent are
less.
Unavailability (%)
The product of Frequency and Mean Duration in years expressed as a percentage.
Page 39
Mean Operating Position(h)

The mean time that the operating position was out of service. This time differs from Mean Duration for those
forced outages which resulted in the Major Component being removed from service and replaced with another.
Investigated Population of Transformers

The tables are presented as stated in [57]. The inventory covered by this survey is given in Table 2:
TABLE 2: INVESTIGATED NUMBER OF TRANSFORMERS BANKS 2009
Voltage Classification Transformer Banks Shunt Reactor Banks

60 –109 kV 1,029 60
110 – 149 kV 1,805 8
150 – 199 kV 124 1
200 – 299 kV 1,029 20
300 – 399 kV 279 15
500 – 599 kV 104 68
600 – 799 kV 314 302
Example of the Reliability Data

The collection form of failure data is represented as following [57]. It has to be regarded that the statistics
covers outages, e.g. minor and major failures.
TABLE 3: BREAKDOWN OF COLLECTED DATA DEPENDENT ON VOLTAGE CLASS [57]
Page 40
TABLE 4: TRANSFORMER BANK ANALYSIS BY SUBCOMPONENT FOR 200 – 299 kV [57]
4.1.2 GERMAN FNN-FAILURE STATISTICS
General information
In Germany, official annual statistics are compiled by the VDE-FNN (formerly VDN and VDEW). They present
the analysis of disturbances from participating utilities in the country. All disturbances are recorded in a
standardized way, a proven procedure that has been used for many years in Germany [58], [59], [60]. There
are two main objectives of this survey.
1) Systematic collection of data on disturbances of primary components of the electrical grid.
2) Systematic collection of data on availability of the electrical power supply.
The reliability of supply can objectively be “measured” by means of appropriate parameters. This enables
utilities to compare their own figures with the national figures. The data can be evaluated according to the
definitions of the UNIPEDE [60]. Thus comparable conclusions regarding the international availability can be
drawn by means of parameters, such as interruption frequency, supply unavailability and interruption duration.
Page 41
Methodology of Data Acquisition and Analysis

To make it easier to evaluate the statistics normally no words or texts are used for the data collection. The
statistics uses exactly defined index lists or figures for automatic data processing. The acquisition of data is
supported by dedicated software which supports the complete and errorless input of indices by a plausibility
check. The data can be transferred by secure internet connection (Verband der Netzbetreiber VDN,
Berichtsjahr 2004).
In 2004 the statistics was adapted according to the new requirements of a liberalized energy market. In order
to get reliable and comparable data the objective was to reduce the room for interpretation of some figures.
Thus the index lists were modified. Furthermore the amount of collected data was reduced to achieve a higher
acceptance of the utilities to participate and deliver data. Main achievement of the statistics is now the
compilation of information about disturbances and availability. So main figures are frequency, duration and
extent of interruptions. Detailled information about the failure, location within the primary equipment, mode or
cause are not included.
The parameters are classified in different groups:
 Organizational information: identification of network operator, failure event, etc.

 Network description: voltage level, earthing of neutral, type (overhead line/cable)
 Data on outage occurrences: date and time of occurrence, occasion, type of failure (e.g. earth fault,
short circuit)
 Data on failure location: type of equipment, mode of failure (e.g. automatic or manual tripping, auto
reclosure, malfunction of protection device)
 Data on supply interruptions: begin and end of interruption, interrupted active power flow, energy not
delivered
 Furthermore quantitative network data, such as number of substations, transformers, switchgear, circuits,
are collected in order to calculate relative figures.
Thus the data collected in the FNN disturbance and availability statistics give the answer to a lot of questions,
but there are numerous technical problems left, needing even more data material. Especially the definition and
use of modern asset management procedures causes a wide and varied number of questions which cannot be
answered by the FNN-statistics. It is not possible to extend the standardized disturbance descriptions without an
unjustifiable waste of time and money. Therefore detailed statistics about the failure location in the respective
equipment, failure cause or mode and repair activities are not included. The individual requirements are
difficult to agree and normally these data are strictly confidential.
Result of FNN-Statistic
In 2004 the disturbance statistics recorded about 50 % of the length of the medium and high voltage grid and
about 60 % in the area of highest voltage (220 kV and 380 kV) (Verband der Netzbetreiber VDN,
Berichtsjahr 2004). In terms of numbers 3894 pieces of 110 kV transformers and 499 pieces of 220 and 380
kV transformers are included. The current statistic of 2010 covers 100% of the transmission grid (220 kV and
380 kV), more than 90% of the 110 kV grid and around 70% of the medium voltage grid.
In Figure 10and Figure 11 the percentage of primary location of disturbance is shown. It can be seen that only
11% of the recorded disturbances are related to transformers.
Page 42
Unknown
Overhead Line
Cable
Substation
Transformer
others
FIGURE 10: PERCENTAGE OF PRIMARY LOCATION OF DISTURBANCE FOR 110 KV
Unknown
Overhead Line
Cable
Substation
Transformer
others
FIGURE 11: PERCENTAGE OF PRIMARY LOCATION OF DISTURBANCE FOR 220 KV AND 380 KV
Figure 12 shows the the disturbance rates of transformers dependent on voltage level with and without failures.
In 2004 the failure rate of the transormers is 0.31 % or the 110 kV units and 0.64% for the 220 kV and 380
kV units.
Page 43
Without failure
With failure
Disturbances
per 100
transformers
< 110 kV 110 kV 220 kV, 380 kV

FIGURE 12: NUMBER OF DISTURBANCES PER 100 TRANSFORMERS
4.1.3 JAPAN
General Information
In the circumstance of deregulated power market trendily, Japanese power utilities have been more strongly
required to use power facilities as long as possible up to their life with avoiding forced outages. Utilities would
like to obtain the efficient maintenance roles based on the actual data which could be analyzed for
preparation of the preventive maintenance and/or the replacement program. Therefore wider scope
investigations were executed to collect various kinds of basic information and failures/troubles on power
facilities including power transformers.
Finally, these collected data and recommended maintenance methods were finalized and compiled into the
maintenance guidance book, through farther information exchanges, evaluations and discussions in the
committee [61], [62].
The executive body of the committee is the Electric Technology Research Association of Japan that mainly
consists of following organizations:
 11 power utilities
 Electrical Manufactures
 Universities
 The Federation of Electrical Power Companies
 Central Research Institute of Electric Power Industry
 The Japan Electrical Manufactures’ Association
Contributors and Scope

The data used in this report are based on the mentioned survey carried out by the following utilities:
 Hokkaido Electric Power Company

 Tohoku Electric Power Company
 Tokyo Electric Power Company
Page 44
 Chubu Electric Power Company

 Hokuriku Electric Power Company
 The Kansai Electric Power Company
 The Chugoku Electric Power Company
 Shikoku Electric Power Company
 Kyushu Electric Power Company
 Electric Power Development Company
Objects of the investigation were the main primary equipment used in power utilities’ substations as follows:
 Power Transformers
 On Load Tap Changers
 Shunt Reactors
 Shunt Capacitors
 Gas Insulated Switchgears
 Gas Circuit Breakers
 Vacuum Circuit Breakers
 Disconnecting Switches
 Instrument Transformers (Voltage Transformers)
 Instrument Transformers (Current Transformers)
 Surge Arresters
 Buses
 Bushings
Specific Definitions and Methodologies of Data Acquisition and Analysis

Data sheet tables used for the survey were discussed, evaluated and determined by the mentioned committee,
because items collected would contain manifold particulars.
Utilities as contributors completed data table with using their data bases involving technical specifications,
maintenance and operational records, failure and trouble records. Collected data were initially reviewed and
put together into simplex statistical formats. In order to evaluate data from various kind view points, collected
data were categorized by criteria having many kind of axis. Furthermore, for obtaining preferable diagnostic
method, prevention maintenance and/or optimal replacement scheme, the committee analyzed individual data
that had significant and/or common issues on maintaining power facilities.
Object Period and Scope

Object periods are divided into two stages, the first one is 6 years from 1992 to 1997 and the second one is 4
years 1998 to 2001. The scope of power transformers in this paper are the operation voltage that is 66kV
and above. The types of transformers are oil immersed and gas filled.
Investigated Items
Investigated items consist of 4 major items as number of installed equipment in the category, failures data,
troubles data and inspection criteria data. The events are classified into failures or troubles.
Data are classified in details as follows:
 Transition number of installed transformers in the each category such as use for
distribution/transmission, oil immersed/gas filled and used period
 Failure data in the each category such as annual number of event, categorized by the structural
component, categorized by the identified cause, categorized by the appearance aspect and
categorized by the finding status.
Page 45
 Trouble data in the each category such as annual number of event, categorized by the structural
component, categorized by the identified cause, categorized by the appearance aspect and
categorized by the finding status.
 Inspection criteria such as frequency of periodical maintenance work and its items carried out during
inspections.
Failure mode : Failure mode is defined as that basically equipment is damaged seriously with disturbance of
operation and put into out of services in long time by automatically or emergency manual switching off.
Trouble mode : Trouble mode is defined as that basically equipment is able to still operate and
countermeasure will be taken in scheduled outage within short period if necessary.
Identified cause : Identified cause is that phenomenon, process, or action is determined as a direct influence on
transformers’ failure or trouble.
Number of Installed Transformers Annually

Number of installed transformers up to 77kV and above owned by 10 utilities in Japan is shown in Table 5 and
Figure 13. The total number of gas filled transformers is 161 banks.
TABLE 5: NUMBER OF INSTALLED TRANSFORMERS VOLTAGES 66KV AND ABOVE
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
up to 77kV 11,450 11,735 11,944 12,166 12,334 12,531 12,525 12,669 12,756 12,831
110 to 154kV 851 872 884 886 891 897 905 920 930 932
187 to 275kV 654 682 697 715 731 741 755 784 789 804
500kV 206 224 230 246 262 276 268 273 275 279
Total 13,161 13,513 13,755 14,013 14,218 14,445 14,453 14,646 14,750 14,846
16,000
14,000
12,000
Number of 10,000 500kV
transformers 8,000 187 to 275kV
6,000
4,000 110 to 154kV
2,000 up to 77kV
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
FIGURE 13: NUMBER OF INSTALLED TRANSFORMERS
Number of Failure and the Rate of Failures

Number of failures and rate of failures in % are shown in Table 6. The average of the rate of failures during
10 years is 0.09%. However, actually almost 30% of failures are caused by malfunction of deteriorated
mechanical relays because any forced out is categorized into failure due to automatic shutdown, even if
transformer is not damaged. It has to be regarded that more then 77% of the failures happend in the voltage
class up to 77 kV. Transformers with voltages above 100 kV showed a failure rate of 0.16%.
Page 46
TABLE 6: NUMBER OF FAILURES AND THE RATE OF FAILURES
Year 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Total
up to 77kV 15 11 20 10 12 10 3 11 7 3 102
110 to 154kV 2 0 1 1 3 1 0 0 4 3 15
187 to 275kV 0 0 0 1 0 4 0 2 2 2 11
500kV 1 0 0 0 0 1 1 0 1 0 4
Total 18 11 21 12 15 16 4 13 14 8 132
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Average
up to 77kV 0.13 0.09 0.17 0.08 0.10 0.08 0.02 0.09 0.05 0.02 0.08
110 to 154kV 0.24 0.00 0.11 0.11 0.34 0.11 0.00 0.00 0.43 0.32 0.17
187 to 275kV 0.00 0.00 0.00 0.14 0.00 0.54 0.00 0.26 0.25 0.25 0.15
500kV 0.49 0.00 0.00 0.00 0.00 0.36 0.37 0.00 0.36 0.00 0.16
Total 0.14 0.08 0.15 0.09 0.11 0.11 0.03 0.09 0.09 0.05 0.09
Number of Troubles and the Rate of Troubles

Number of troubles and rate of troubles in (%) are shown in Table 7. The average of the rate of troubles
during 10 years is 2.43%.
TABLE 7: NUMBER OF TROUBLES AND THE RATE OF TROUBLES
Year 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Total
up to 77kV 252 278 283 274 302 283 202 218 210 163 2,465
110 to 154kV 43 45 48 37 47 45 62 64 56 47 494
187 to 275kV 30 32 46 28 38 35 41 41 53 41 385
500kV 5 4 14 6 13 16 8 8 3 10 87
Total 330 359 391 345 400 379 313 331 322 261 3,431
Year 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Average
up to 77kV 2.20 2.37 2.37 2.25 2.45 2.26 1.61 1.72 1.65 1.27 2.01
110 to 154kV 5.05 5.16 5.43 4.18 5.27 5.02 6.85 6.96 6.02 5.04 5.50
187 to 275kV 4.59 4.69 6.60 3.92 5.20 4.72 5.43 5.23 6.72 5.10 5.22
500kV 2.43 1.79 6.09 2.44 4.96 5.80 2.99 2.93 1.09 3.58 3.41
Total 2.51 2.66 2.84 2.46 2.81 2.62 2.17 2.26 2.18 1.76 2.43
Number of Failure by the Structural Component

Number of failure by the structural component during 1992 to 2001 is shown in Figure 14. HV coil, bushing and
tap changer take significant number of failures in the main components of transformer. For the reference,
causes of mechanical relay operations were almost malfunction by deterioration of mechanical switches.
Page 47
Number of Failures 35
30
25
20
15
10
5
0
HV Coil
CT Wiring
LV Coil
Bushing &
Others
Tap Coil
Mechanical
Changer
Conductor
Insulator
Tap
Relay
FIGURE 14: NUMBER OF FAILURES BY STRUCTURAL COMPONENT
The distribution of failed components by voltage class is shown in Table 8. Since the largest population of
transformers is up to 77kV rating, necessarily this voltage class account for large rate.
TABLE 8: THE DISTRIBUTION OF FAILED COMPONENTS BY VOLTAGE CLASS
Failed Mechanical Bushing & HV Tap CT & Tap LV

Conductor Others Total
Component Relay Insulator Coil Changer Wiring Coil Coil
up to 77kV 26 12 13 8 10 7 8 3 15 102
110 to 154kV 2 2 1 6 0 0 1 1 2 15
187 to 275kV 2 4 3 1 0 0 0 0 1 11
500kV 1 1 1 0 0 1 0 0 0 4
Total 31 19 18 15 10 8 9 4 18 132
Number of Failures by Identified Cause

The number of failures by the identified cause during 1992 to 2001 is shown in Figure 15. The total number of
failures is 133 within the period of 10 years.
Disasters
6 Incomplete Manufacturing
9
16 53 Incomplete Installation
25
23 Inadequate Maintenance
Deterioration
Unknown
FIGURE 15: NUMBER OF FAILURES BY IDENTIFIED CAUSE
Page 48
The distribution of identified cause by voltage class is shown in Table 9. The largest number by identified
causes is inevitable disasters such as lightning stroke, typhoon and earthquake which are typical Japanese
natural circumstances. Since the largest population of transformers is up to 77kV rating, necessarily this voltage
class account for large rate.
TABLE 9: THE DISTRIBUTION OF IDENTIFIED CAUSE BY VOLTAGE CLASS
Identified Incomplete Incomplete Inadequate

Disasters Deterioration Unknown Total
Cause Manufacturing Installation Maintenance
up to 77kV 41 18 20 12 4 7 102
110 to 154kV 6 2 2 2 2 1 15
187 to 275kV 3 2 3 2 0 1 11
500kV 3 1 0 0 0 0 4
Total 53 23 25 16 6 9 132
4.1.4 NATIONAL COMMITTEE CIGRE A2 – BRAZIL

After the deregulation of Brazil electrical market ABRATE, the Brazilian Power Transmission Utilities Association,
created a working group to elaborate a statistics of transformers performance. All utilities members of this
association are requested to inform transformers failure data. As this association represent only transmission
utilities, no step up transformers data are collected. A report is published each two years, although few utilities
agreed to share their data and information.
In 2007 the national committee of Cigré Brazil SC A2 start up a working group to perform a reliability survey
in an attempt to collect data from transformers and reactors installed in Brazil. Ten utilities, one research
institute and 4 manufacturers accepted the invitation to take part of the WG and elaborate a survey to be
filled up by the Brazilian utilities.
The scope of survey covers step-up transformers, transformers, autotransformers and reactors installed on
transmission system. The survey adopted the following premises: equipment shall be manufactured after 1980,
the highest voltage level shall be greater or equal to 138 kV without any limitation of power, and the failures
informed shall have occurred after 1997.
The objective of the Cigre survey is to provide useful information to utilities optimize specifications, maintenance
procedures, assist the Design Review and optimize spare parts policy.
In addition, from June 2008 the National Electrical Energy Agency started collecting heavy penalties for
unavailability of transformers and it is critical that utilities may have the means to compare the performance of
their equipment with other installed and in operation
The Working Group completes the analysis of the data collected by survey and a report is being prepared. A
resume of the results of Cigre Brazil Survey is presented below. [63]
Definitions
Failure : It's the end of condition (ability) or the impossibility of equipment performs its required function,
leading to their unavailability in operational conditions not programmed.
Defect : Any abnormality detected in a equipment that doesn’t required to take out of operation, but that
affects the degree of reliability or performance specified or expected for this equipment.
Forced Outage : Shutdown of equipment in non-programmed, resulting from a failure or an emergency
condition that requires automatic or manual shutdown of the equipment to avoid risk to life or damage to
equipment or other unintended consequences for the electric system
Scheduled Outage: Shutdown of equipment programmed in accordance with the operation procedures.
Page 49
Survey
Data surveys of transformer performance have been unsuccessful in Brazil and abroad. The difficulties in
development of such statistics were reported by Lapworth [34]. Among the reasons for this difficulty are the
difficulties to characterize failures and defects, to establish the best method of calculation and the restriction
imposed by the utilities to provide their data and information in a competitive environment. To minimize these
difficulties the Cigre Brazil SC A2 Working Group decided to use the same criteria and method used in CIGRE
1982 survey, preparing a questionnaire with direct questions and limited to the information needed to
calculate the failure rate. Also all information received is considered confidential and any direct or indirect
reference to the source of collected data will not be disclosed.
The questionnaire based on the Cigre WG 12-05 survey questionnaire was prepared and sent to Brazilian
utilities. Thirteen of the most important utilities answered the survey and data from 3198 transformers and
reactors was received, which represent more than 70% of the installed capacity in Brazil. [63]
A breakdown of the collected data (population) according to voltage level is given in TABLE 10.
TABLE 10: NUMBER OF TRANSFORMERS AND REACTORS
The failure rate is calculated as follows :

∑𝑛
𝑖=0 𝑛𝑛
𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝑅𝑅𝑅𝑅 = 100 𝑥 ∑𝑛
(%) where,
𝑖=0 𝑁𝑁
ni - number os transformer that failed in the ith year

Ni – number of transformers in service during the ith year
The calculated failure rates dependent on voltage range are given in Table 11 for the time interval between
2004 and 2009.
TABLE 11: COLLECTED FAILURE DATA ACCORDING TO VOLTAGE LEVEL
Results from the survey also included the failure mode, the components involved and the consequences of the
failure. The results are shown in the figures below.
Page 50
Dielectric
Chemical
Mechanical
Thermal
Unknown
FIGURE 16: FAILURES CLASSIFIED ACCORDING TO FAILURE MODE
Windings Insulation between Windings

Insulation Winding Earth Leads
Bushings Core
Tank Cooling unit
On-Load Tap Changer Off-load Tap Changer
Other
FIGURE 17: FAILURES CLASSIFIED ACCORDING TO COMPONENT INVOLVED
Page 51
Leakage
Explosion
None
Othen
Fire
FIGURE 18: FAILURES CLASSIFIED ACCORDING TO CONSEQUENCES OF FAILURE
4.1.5 UNITED KINGDOM

A review of major failures of large generator and transmission transformers in the UK between 2003 and
2009 has been carried out.
Background Information
The Privatisation of the UK Electricity Supply Industry in the early 1990’s resulted in the former Central
Electricity Generating Board, the nationalised industry responsible for generation and transmission in England
and Wales, being broken up and replaced by a number of generation and supply companies and a
transmission company, National Grid, all privately owned, whose primary reporting responsibilities are
financial to their shareholders. The industry is overseen and regulated by OFGEM (Office of Gas and
Electricity Markets) whose main roles are to encourage competition where this exists and to act to counter the
exploitation of any monopolistic situations. Apart from the standard requirement to report equipment problems
with health & safety implications to the relevant authorities, there is no centralised and formalised requirement
or system for reporting equipment failures, but major incidents causing loss or disruption of supply are
investigated by OFGEM and companies are expected to have adequate asset management procedures in
place, particularly those who charge their customers for the utilisation of their assets used to deliver energy.
Therefore every electricity company would be expected to keep information on major equipment failures, but
there is no standardised format for this and this information is not usually available to outside parties.
Transformer Populations
The survey covered large generator and transmission transformers (>100 MVA) which are nearly all connected
to the 275 and 400 kV UK grid system.
There are approximately 800 transmission units of voltages from 400/275 to 275/33 kV and ratings from
100 to 1,000 MVA, mainly auto-transformers. Most of these are 40 or more years old. In view of the low
failure rates for transmission transformers it was not considered worthwhile splitting these units into further
subgroups.
The majority of generator transformers are also 40 or more years old, at power stations built before
Privatisation: either three phase units of 340 to 735 MVA (subgroup G1) or single phase banks of 800 MVA
built to a particularly demanding specification (subgroup G2). The other subgroup (G3) are significantly
younger three phase units of transformers at power stations built since Privatisation.
Page 52
Major UK Transformer Failures

A list of 32 major transformer failures occurring between 2003 and 2009 has been compiled and details
supplied according to the A2.37 format. This included in-service failures that required the transformer to be
replaced, and also transformers removed from service before failure in service because of concerns about
condition, about one third of the total. Calculated failure rates are given in the Table below.
TABLE 12: MAJOR UK TRANSFORMERS FAILURES BETWEEN 2003 AND 2009
Transformer type Subgroup Population Number of Failure rate,

failures % p.a.
GSU G1 62 11 2.5
G2 28 3 1.5
G3 114 4 0.5
Transmission T 800 14 0.3
Failure Types and Causes

A third of the total number of failure cases involved transformers removed from service before failure in
service. In most cases the reason was evidence of an internal fault from increasing main tank gas levels. Most
cases involved service aged transformers (more than 30 years old), but three relatively new transformers had
to be removed because of gassing faults, and only one of these, involving an overheated selector contact on an
in-tank tap-changer, was returned to service.
The majority (two thirds) of in-service failures were due to inter-turn failures. Most of these were for service
aged (over 35 years old) units, presumably due to aged insulation. However, three of these in-service inter-turn
failures were for relatively new (less than 20 years old) transformers, two of which involved corrosive sulphur
deposition. Two generator transformers failed in service with inter-turn faults for unknown reasons while being
used to back-feed the station.
Three in-service failures were mechanical in nature, all involving collapse of the tap winding. Two of these
immediately followed short circuits within tap windings due to tap-changer flashover faults. The other in-service
mechanical failure also involved a collapse of the tap winding, but this is thought to have been initiated by a
through fault some years before. A fourth mechanical failure was discovered by diagnostic testing during an
outage and was also thought to have been caused by an external through fault some years before. Therefore,
of the 12% of all failures that were mechanical in nature, half of these were immediate failures involving a
collapse of the tap winding following flashovers within the tap-changer, while the other half were thought to
have been caused by external through faults some years before.
There were only four in-service dielectric failures, two of which were attributed to internal over-voltages arising
from part winding resonances set up when the transformers were energised remotely. Another dielectric failure
for some unknown reason resulted in fire on a large quadrature booster. The other dielectric failure is believed
to have resulted because of incorrect installation of a replacement bushing.
The distribution of failure type and cause are as shown in the following figures.
Page 53
14
4
Dielectric
Inter-turn failure
Gassing
9 Mechanical
FIGURE 19: FAILURE MODE
Aged insulation
1
Unknown
1 1
2 8
Manufacturing
2
Over voltage
3 Tap-changer fault
Through fault
3 7
Winding thermal fault
4
Core circulating current
Corrosive sulphur
Faulty maintenance
FIGURE 20: FAILURE CAUSE
Age at Failure
Perhaps the most obvious fact to emerge from the analysis of this data was that by far the majority of failures
occurred in the 36-40 years age band, for both generator and transmission transformers. This, coupled with the
fact that the majority of failures were inter-turn failures due to aged insulation gives concerns about the life
expectancy of transformers.
However, since the incidence of failure at this age simply reflects the greater number of transformers of this
age on the system, this does not necessarily reflect an increased failure rate with age. In fact, a detailed
hazard rate analysis for UK transmission units has failed to reveal any significant increasing trend up to 50
years of age.
Page 54
4.2 Company Related Surveys

This section describes failure data surveys, which are done by specific companys. Here the approaches of
utilities, manufacturers and consulting firms are presented.
4.2.1 UTILITY, ESBI (IRELAND)
General Information
ESB International (ESBI) is a wholly owned subsidiary of Ireland's Electricity Supply Board (ESB), a vertically
integrated utility that has a number of ring fenced divisions operating independently in the Single Electricity
Market. Unlike ESBI, our parent company ESB, is a semi-state body with a break-even mandate.
ESB International (ESBI) is responsible for the non-regulated businesses in the ESB group and is unique in the
sense that it operates across all areas of the Energy Value Chain i.e. it originates, outsources, constructs,
manages, operates and maintains energy assets in Ireland and overseas.
Fault Recording
Network technicians in the field report all major and minor faults on Networks transformers (110/38, 20, 10kV)
to the transformer maintenance team by filling out a Green Amber Red sheet. This records all details of the
fault.
 The National Control Centre contacts the on call engineer regarding all transmission transformer
(>110kV) faults.
 If further investigation is warranted test engineers from the transformer maintenance team are
dispatched to investigate the fault.
 Transformer specialists review all transformer test data from electrical tests, visual inspections and
protection fault records. The component (winding, core, tap changer etc.) which caused the fault is
identified. The origin of the fault (mechanical, chemical, dielectric, thermal) is established and a
conclusion is reached regarding what caused the fault (design, manufacturer, material, transport,
incorrect erection, material, lightning, overload).
 The manufacturer is contacted if necessary to carry out or advise on any repair that may be required.
In some cases the unit is retired due to financial reasons where it is not cost effective to repair the unit.
 The transformer is then repaired or replaced.
 Finally a fault file is completed by the transformer specialists and kept on a designated database.
 When the component which caused the fault is identified all similar components on other transformers
are assessed during maintenance and condition assessments to determine if they are developing the
same failure mechanisms and if a type fault exists.
Failures
A failure of a transformer and reactor is defined as a major or minor fault which causes a forced or unforced
outage of the transformer. Major failures constitute events that require the transformer to be retired or
removed from site for repair. Minor failures can be repaired on site.
 Design and manufacturing can lead to failures of windings, magnetic circuits, tap changers, bushings,
accessories, terminals and dielectric fluids. Some of these defects can cause instantaneous failures but
other may take years to cause a failure.
 The same failures can occur if the operation and maintenance is not carried out regularly and
effectively. Incorrect operation such as overloading can lead to reduced life.
 Condition assessment and on line monitoring is the most effective way of ensuring that the transformer
operates correctly and that its predicted life expectancy is maintained. The condition of all of the
components in the transformer can be assessed through different diagnostic methods.
Page 55
 All major failures which resulted in a forced outage and necessitated a large repair in the
manufacturer’s factory (e.g. winding replacement, core reconstruction) or in the transformer being
retired have been part of this survey between 1980-2009.
 There are a large number of minor failures some of which have been identified here. However not all
minor faults or saves are recorded as they are picked up during condition assessments, as a result of on
line monitoring or routine maintenance. They would however constitute a substantial number of faults as
we would typically carry out 30-40 condition assessments per year. These faults appear in
transformers while it is energised or being tested and if they are not corrected they will develop into
failures. They can therefore not be classed as failures.
 The failure rate is approximately 0.2-0.5% over a 29 year period between 1980 -2009. These are
failures that had to be removed and repaired or retired. Most failures were due to materials, design or
manufacturing defects.
Data
TABLE 13: HV TRANSFORMERS ON THE ESB SYSTEM
Highest Voltage GSU (two winding) Networks Auto transformer

Winding (kV) transformer (three (two winding)
winding)
110 21 200 0
220 16 21 24
400 3 0 5
TABLE 14: VOLTAGE CLASSES OF TRANSFORMERS
GSU (kV) Networks (kV) Auto transformer (kV)
110/10 110/38/10 220/110
220/20 110/20,10/10 400/220
400/20 220/110/10
TABLE 15: TRANSFORMER MVA RATING
GSU (MVA) Networks (MVA) Autotransformer (MVA)
30-420 20-250 125-500
Page 56
TABLE 16: TOTAL UNIT YEARS OF TRANSFORMERS
Winding Voltage GSU Networks Auto transformer
110 108 1076 0
220 34 38 128
400 0 0 6
TABLE 17: UNIT YEARS OF TRANSFORMERS ON THE SYSTEM
Age yrs GSU Network Auto transformer

(no of transformers)
0-5 0 28 (8) 11 (3)
>5-10 9 (1) 590 (68) 96 (13)
>10-20 151 (12) 496 (35) 27 (2)
>20-30 292 (11) 731 (35) 232 (9)
>30-40 354 (10) 1137 (36) 77 (2)
>40-60 301 (8) 1270 (32) 0
TABLE 18: MAJOR FAILURES ON HV TRANSFORMERS BETWEEN 1980-2009
Year Number of failures Transformer type Number of

failures
1980-1990 3 GSU 4
1990-2000 8 Network 12
2000-2009 7 Auto 2
TABLE 19: FAILURES AND FAULTS ON NETWORKS TRANSFORMERS
Component Major failures, minor faults and saves
Winding 20%
Core 4%
OLTC 42%
Bushing 16%
Accessories 18%
Page 57
TABLE 20: FAILURES AND FAULTS ON GSU’S
Winding 30%
Core 0%
OLTC 30%
Bushing 30%
Accessories 10%
TABLE 21: FAILURES AND FAULTS ON AUTOTRANSFORMERS
Winding 40%
Core 20%
OLTC 0%
Bushing 0%
Accessories 40%
4.2.2 UTILITY, ESKOM (SOUTH AFRICA)

The approach to reliability evaluation of Eskom’s transformer fleet, as carried out within the Research
department, is discussed in this section. The approach has been adopted since 2008.
The results of the reliability evaluation for variously rated transformers are typically presented as failure rates
and the percentage contributions of failure locations, causes and modes, by transformer application (generator
step-up, unit and substation transformers), voltage level, outage type and age groups. Typical results have
been published in [64], [65] and [66].
Definitions
Reliability associated terminology are defined as follows:
 Failure: The termination of the ability of an item to perform a required function.

 Failure with Forced Outage: According to IEEE C57.117-1986 as: Failure of an equipment that requires
its immediate removal from the system. This is accomplished either automatically by the operation of
protection equipment or as soon as switching operations can be performed.
 Failure with Scheduled Outage: According to IEEE C57.117-1986 as: Failure for which the removal of the
equipment from service can be deferred to some more convenient occasion, but still requires a change
to a planned outage programme.
 Severe Failure: Failures which necessitated the removal of the transformer for repair or replacement.
 Intermediate Failures: Failures which were possible to repair onsite, but usually required intrusive work
on the transformer itself.
 Failure Rate: Determined as the ratio of the number of transformer failures over a period, to the number
of transformers in service in the same period.
Page 58
Data Collection and Classification

Records of the failure incidents, and failure information to populate a failures database, are in the form of
failure investigation and repair reports, as well as information obtained from internal asset databases.
The classification of failures are to large extent based on a standard failure reporting form developed and
presented in Cigré WG A2.18 [35], in which information was grouped into three categories: the equipment
description, operational history and description of failure.
Under the description of failure, the failure location, failure cause and failure mode classifications were based
on predetermined classifiers (refer to [35]) provided in Cigré WG A2.18, where:
 Failure location referred to the location in the transformer where the failure was initiated.
 Failure cause referred to cause of failure in the location where the failure was initiated.
 Failure mode referred to the ‘nature of failure’ as used in Cigré WG A2.18.
Page 59
Failure Location Description Failure Causes Description Failure Modes Description
Windings Inherent Inadequate specification Dielectric Partial Discharge

Between windings deficiency Inadequate design Tracking
Winding Connections Tap Leads Inherent material defect Flashover
To Bushings Improper factory assembly Open circuit
Clamping Inherited Improper site assembly Electrical Short circuit

Mechanical Structure Coil Blocking deficiency Improper maintenance Poor joint
Lead Support Improper repair Poor contact
Major Improper adjustment General overheating
Thermal
Insulation Minor Overload Localised hotspot
Material - Liquid, Gas Load removal Moisture
Tank System event Over-voltage Contamination Particles

Physical chemistry
Selector Resonance Gas
Page 60
Diverter Short circuit Corrosion
Tap Changer
Drive Motor & Couplings Vandalism Bending
External event
Control System Impact of external object Breaking
Bushings Porcelain Lightning Mechanical Displacement
Core High Ambient Loosening
Helmet Low Ambient Vibration
Draw Lead Environmental Rain
Magnetic Circuit Water ingress
Wind
Seismic
Improper application
TABLE 22: CLASSIFICATION INTO FAILURE LOCATIONS, FAILURE CAUSES AND FAILURE MODES [35]
Abnormal deterioration
4.2.3 TRANSFORMER MANUFACTURER, ABB (SWEDEN)

This chapter details how a transformer manufacturer manages reporting of field failures, reviews the activity to
resolve these and implements improvement activity to prevent them. This process will result in identifying areas
for improvement in the processes and also the gaps in guidelines, method provisions and work instructions. It
provides an excellent foundation for employing field performance feedback to the processes of producing
transformers and that helps in continuous improvement activity which eliminates wasteful effort and enhances
the operational excellence goal.
The statistic which can be drawn from the field data shall not be presented as general MTBF or Failure
frequencies. The base for the statistics most likely will not be complete and is futile to compare between
different manufacturers.
Definitions
• Field Failure: Transformers that fail during operation with the customer. A failure is any unplanned
outage caused by the transformer or its accessories. An indication which requires the tank to be opened
shall also be regarded as a field failure.
• Metrics DB : Transformer Metrics Database within Lotus Notes where field failures are recorded.
• 4Q : 4 Quadrant Approach to improvements

• Transformer : Can be used generically and include Reactors, Shell type, HVDC and special transformers
also.
• Unit Report : Information in the Metrics DB, such as Serial Number, Rating, Voltage, Customer, etc
applicable to one transformer.
Field Failure Reporting and handling

Field failure reporting is very important, because the data provides opportunities to learn, analyze and
prevent recurrence. This would help to continuously improve the quality of products.
Reporting of field failures follows a simple process as shown in Figure 21
Page 61
From local plant analysis to global experience
FIGURE 21: FIELD FAILURE REPORTING PROCESS

NOTE: THE TIME SPAN BETWEEN 1 AND 2 WILL DEPEND ON CUSTOMER, REPAIR URGENCY, ETC.
Reporting of Field Failures is mandatory. A field failure must be reported into the Metrics DB by the Plant
manager or his assigned representative within one week of learning of the event. The reporting speed is
critical. The mail functionality associated with the database will automatically send notifications to all people
relevant to the failure.
The Metrics DB contains Unit Reports on most transformers manufactured by ABB. The reporting process is as
follows:
1. Open the Metrics DB and select Field Failures. Select “All units per delivery year” (or “All units per Serial
number”)
2. Click to open it then click on the “Report a Field Failure for this unit” button.
3. The “Field Failure Record Sheet” GUI opens. Complete the input data.
Page 62
The picture below gives a quick reference.
Transformer Units can be found in field

failure view e.g. ‘All units per deliver
year’
To report the failure: Select the unit and
click ‘Edit document’
Local Analysis of the Failure

The Plant Manager owning the issue will constitute a team of local experts to analyze the failure and provide
possible root causes to the Power Field Failure expert team. If this action cannot be completed within 3 weeks
of learning of the failure the Plant manager shall contact the Power field failure expert team and agree on the
time schedule for providing the local analysis report. The local analysis team can use the 4Q method with
simple tools like Cause and Effect Diagram, Pareto Diagram to pinpoint possible causes of the failure.
Generally, all units which fail while still under warranty must be thoroughly analyzed and the failure cause
determined. Units which are no longer under warranty shall be dealt with on a case by case basis.
FIGURE 22: FAILURE RECORD SHEET
Page 63
Field Failure Review Process

Once every 3 months, the Power Field Failure expert team and a team of Design and Production Engineers
from plants shall analyse all the failures. This activity shall bring forth actions to prevent the recurrence of
failures and provide feedback to the various processes.
The above team’s main task is to go through all the failures and provide actions that need to be implemented in
various processes to prevent the failures. The actions can include among others, changing
• Work methods
• Method Provisions
• Work Instructions
• Safety Requirements
• Design Criteria
The actions can be piloted in a few plants before rolling it out as necessary. The Power Field Failure expert team
holds the responsibility to decide this. In the event of an “immediate alert” being necessary, the expert team shall
organise to issue the same.
The Plants that receive a “process change alert” shall implement these actions to prevent further failures in the
test area or field. The Power Field Failure expert team shall close the field failure report when all plants have
confirmed the implementation of changes at their plants.
The full report of progress of various field failures will be available in the database. The Power Quality
manager shall review this every quarterly. The review shall focus on timely reporting and feedback action in
the plants and the effect of process changes in reducing the field failures. They may suggest changes to the
reporting and analysis process as necessary.
4.2.4 OLTC MANUFACTURER - MASCHINENFABRIK REINHAUSEN (GERMANY)

Starting with the introduction of the IBM HOST system in MR’s production in the mid 70’s MR systematically
recorded statistic data regarding MR products in service in an ERP system. Up to this time the product
surveillance was recorded manually in paper documents. With the introduction of the SAP System in 2005 the
data collection was adopted to the new requirements. Nowadays the product surveillance includes service
records, failure statistics and complaints regarding either logistical or technical issues.
Methodology of Data Collections

All incoming information with the reference to certain equipment is recorded with this equipment. The content of
the information includes the following topics:
 Service Records: Data are collected from service records of MR service personnel and service partners.
These data are collected and stored with the relevant equipment to have a complete history of the
equipment. The structure of the report includes measured data, transformer data e.g. rated voltages,
rated power, substation information, organizational information, operating conditions, a checklist for the
service personnel and a text component for remarks.
 Incoming E-mail data regarding service offers or orders, complaints, failure data or informational data.
E- Mail has become the most common communication method nowadays.
 Telephone calls, faxes regarding service offers or orders, complaints, failure data or informational
data.
With the exception of the own service reports the information doesn’t come in a structured way that can be
used for a statistic. To evaluate the statistics defined figures have to be used for the data collection. This job is
done in MR’s service dept. by the order processing teams. Up to the introduction of the SAP system there were
different tools to be used for service records and for product surveillance data. With the introduction of SAP all
Page 64
incoming operating data are recorded in one tool called the “service notification”. This tool is used as platform
to store all in- and outgoing information regarding existing equipment. The product surveillance data of the
pre- SAP system were transferred to SAP with a data transfer routine. All paper documents regarding still
existing equipment were scanned and linked to the corresponding SAP equipment file.
Classification of Data Collections

The information is classified in different groups:
 Organizational information: end-user, transformer manufacturer, contact person, etc.

 Equipment description: type of equipment, serial number, nos. of operations. This information in
connection with an integrated ERP system is the link to the detailed equipment structure with all its
information.
 Subject coding: The two step coding that systematically classifies the information
• Notification code: complaint/failure or planned modification/service
• Sub codes to the notification code e.g. technical, logistical complaint failure etc.
• In the event of a technical complaint/failure a “detail” mask opens for the coding of: failure
location, mode of failure, cause of the failure, etc. Additionally activities e.g. service jobs
repairs, investigations are controlled with this tool
A standard routine in the system is available to select and display statistic data. For regular reports add – on
programs were created to transfer the data to Excel and create reports.
Reliability Statistics Based on the Collected Data

For the reliability statistic we follow the IEEE Guide for Reporting Failure Data for Power Transformers and
Shunt Reactors on Electric Utility Power Systems. This Guide includes network-, interconnection-, Distribution,
HVDC-, generator- (or unit-, step- up) transformers but not rectifier- or arc furnace transformers.
All OLTC failures on the above mentioned transformers are considered that caused a forced outage of more
than 1 (one) day and required inspection and/or repair measures with respect to the tap-changer regardless
of the fact if the tap-changer was the cause of the failure or if the tap-changer failed due to an overstress
exceeding the guaranteed properties.
Definitions
Definition of parameters for the reliability evaluation:
The recordings for deliveries of on-load tap-changers start in 1980. To cover the missing information of the
date of commissioning an assumption is used. We deduct the deliveries of the last 6 months of the last statistical
year for the time delivery to commissioning. This results in the following calculation:Operating unit years end of
2011 = (nos. of units commissioned 1980 x 31,5 service years)+(nos. of units commissioned 1981 x 30,5
service years)+…+(nos. of units commissioned 2010 x 1,5 service years) )+(nos. of units commissioned 2011 x
0,5 service years)
Failure rates for LTC's: nF = number of units failed
FR = Failure rate: (no. of units failed/ operating unit years) x 100%
MTBF = Meantime Between Failures: 1 / FR
Calculation method:
Failure rates for LTC's: nF = number of units failed
FR = Failure rate: (no. of units failed/ operating unit years) x 100%
Page 65
MTBF = Meantime Between Failures: 1 / FR

Calculation method:
4.2.5 CONSULTANCY – EPRI (USA)

Managing fleets of aging power transformers is a critical issue for utility companies striving to maintain
reliability and control costs. Predicting the failure rate of transformers nearing the end of their design life is
challenging, making typical asset management activities such as repair and replacement decisions more
difficult. Generic transformer reliability data may not be adequate to inform such decisions, and data from a
single company may not be extensive or diverse enough for useful statistical analysis. To help utilities better
predict transformer service life, EPRI created the Transformer Industry-Wide Database (IDB). The IDB pools
transformer operating and failure data from supporting utilities in order to assemble a statistically valid
population that includes a variety of power transformers [67].
To acquire additional performance data to support risk-based transformer management, EPRI and member
utilities continue developing an industry-wide database (IDB) for power transformers. The transformer IDB
captures in-service and failure data from many sources in a common format for data mining and statistical
analysis. Data is obtained on in-service transformers and those removed from service due to failure or
retirement. The data are historically accurate and include failure mode, operational and repair history, and
equipment design information.
An important application of IDB data is to develop hazard functions through parametric analysis. This involves
fitting a model to the data in order to mathematically describe the transformer aging and wear-out process
over time. Because transformers have different designs, different components, and fail for different reasons,
analysis groups of similar transformers must be properly selected to develop failure models appropriate for
each group. Data must also be assigned to the correct group.
An innovative approach is used to separate transformer populations into subsets of similar units with shared
characteristics and behavior. Statistical analysis focusing on those subsets yields insights into transformer aging,
failures, and projections of future performance that support maintenance repair and replacement decisions,
and asset management decisions to minimize lifecycle costs of equipment replacement and maintenance,
including failure costs.
An advanced risk-based approach to managing transformer assets uses historical performance data to group
and rank transformers. This decision support foundation performs four key steps: 1) assessing existing
performance, 2) specifying required performance, 3) projecting future performance, and 4) understanding how
to bridge gaps.
This approach requires equipment failure models and hazard rates based on the best available data. Generic
reliability data is inadequate to meet decision support requirements. A given utility’s transformer data is
unlikely to be sufficiently extensive or represent statistically diverse transformer population subsets.
The IDB provides a broad-based repository of transformer performance data as a key part of a decision
support foundation. With proper care and analysis, this data can provide information about the past
Page 66
performance of equipment groups (e.g. substation transformers) and subgroups (e.g. 345 kV auto transformers)
and the factors that influence that performance (e.g. voltage, manufacturer). With enough data, projections can
be made about future fleet performance (e.g. expected number of failures), and both past and projected
future performance information can be useful for operations, maintenance, and asset management decisions.
Transformer Demographics and Life Expectancy

Many substation transformers were installed in the 1960s and 1970s and are approaching the end of their
nominal design lives. Figure 23 shows the age profile for over 7,000 units in a particular subset of in-service
transformers contained in the transformer IDB. Clearly depicted is the “asset wall” in the 35 to 45 year age
bracket. This IDB data is aggregated from eight utilities with a variety of sizes and service territories and is
thought to be representative of the general industry in North America.
FIGURE 23: TYPICAL AGE PROFILE – IN-SERVICE INDUSTRY-WIDE DATABASE FOR TRANSMISSION SUBSTATION POWER TRANSFORMERS
Like other equipment, transformers may follow a familiar bathtub curve—an initially high rate of infant
mortality failures, followed by a relatively low and constant failure rate during a long service life, then an
increase in wear-out failures with impending end of life.
One application of the IDB is to assess whether this curve accurately describes historical transformer
performance. If the bathtub curve applies to transformer life, what are the parameters of the curve—
especially the wear-out portion of the curve? Do the curve parameters change with different transformer
makes, models, vintages, and applications? Answering these questions is more important than ever as
transformer fleets age and high replacement costs and uncertain lead times put more pressure on asset
managers striving to meet high reliability standards.
Transformer Data Categories

The IDB includes population and failure data sets based on transformer type, make, model, application, and
age. There are two broad classes of transformer data included: in service units (population data) and units
permanently removed from service due to failure or retirement. For the purpose of IDB, failure is the
termination of a transformer’s ability to perform its functions with acceptable risk without major repair. This
includes failure with a forced outage—requiring the transformer’s immediate removal from service by relay
operation or emergency switching—and failure with a scheduled outage in which the transformer is removed
from service at a selected time (for example, due to unacceptably high dissolved gas levels).
Page 67
FIGURE 24: IDB DATA CATEGORIES (NOT TO SCALE)
In-service population data includes
• Transformer name plate information

• Serial number or unique identifier
• In-service or manufactured date
• Application
• Type, e.g., auto or non-auto transformer
• Previous rewind (yes/no, date)
• Service location (utility, substation, transformer position)
Removed from service data includes the preceding information, plus the date and the reason the transformer
was removed from service.
Data Processing
An important application of IDB data is to develop hazard functions (hazard rates) through parametric
analysis. This involves fitting a model to the data in order to mathematically describe the transformer aging
and wear-out process over time. Since transformers have different designs, different components, and fail for
different reasons, analysis groups of similar transformers must be properly selected to develop failure models
appropriate for each group. Data must also be assigned to the correct group—auto transformers must be
grouped with auto transformers and rewound units grouped with rewound units. Grouping transformer
populations into subsets of similar units with shared characteristics and behavior is essential to enable an
“apples to apples” analysis. The need for such groupings is illustrated by the notable differences in survival
performance between new and repaired units and between auto and non-auto units (Figure 25) demonstrated
by a particular subset of the IDB.
Page 68
FIGURE 25: SURVIVAL PLOTS FOR NEW VERSUS REPAIRED UNITS—NOTABLE DIFFERENCES IN PERFORMANCE
Failure data details are also desirable in defining subgroupings, including failure location (main body, load tap
changer, bushing) and cause (internal or external to the transformer system). In general, utility-supplied data
requires review and cleansing before it is suitable for inclusion in the IDB.
Modeling Objectives and Data Analysis

Because the models under initial development focus on the wear-out portion of the bathtub curve, we are
primarily concerned only with failures that could be wear-out related. Therefore, main body failures initiated
by external events such as a stuck breaker or mis-operating relay are identified and analyzed separately. To
model wear-out, only failures beyond the expected age for inception of wear-out are used for analysis. Load
tap changers and bushings can be expected to wear-out at different rates from each other and from the main
body and therefore also are analyzed separately, even if their failure results in failure of the transformer.
The analysis begins with a search for the best wear-out models to fit one utility’s data. A second utility’s data is
analyzed to find its best models. The two sets of data are compared to assess similarity. If the two utilities’
models are similar enough based on standard statistical tests, the two utility data sets are combined and the
best models are found for the aggregated data. A third utility is then analyzed and compare to the
aggregated models. The process continues, building a larger aggregated data set. As the set gets larger, the
models’ confidence bounds can be expected to decrease.
To estimate parameters of a model, the challenge is to find the parameters that make that model fit the data
best. To compare among models, we find models which fit the data best, and decide if one or more models fit
sufficiently better than the rest. Maximum likelihood estimation (MLE) and likelihood ratio (LRT) tests are used.
MLE reflects goodness-of-fit metrics based on the likelihood (probability) of seeing the data given a particular
model. In this case this means both the general form of the model and specific model parameter values. The
maximum likelihood estimates of the parameters are those values that make the observed data most likely to
have happened. LRT compares two nested models, testing whether the nesting parameters of the more complex
model differ significantly from their null values. LRT tests whether the extra goodness of fit to the data is worth
the added complexity of the additional parameters. IDB statistical analysis is complicated by the large degree
of truncation in the aggregated data and the different censor dates associated with the different utility data
sets.
Project Status and Preliminary Results

Data collection for the IDB began in 2006, and the database now contains records on more than 20,000
power transformers. Several thousand more records are in various stages of review for inclusion. Figure 26
presents an overview of the IDB process. Extensive interaction with the supplying utility is often required to
assure sufficient data quality and proper descriptive classification.
Page 69
FIGURE 26: THE IDB PROCESS
FIGURE 27: ANALYSIS OF FAILURE LOCATION (1112 OF 1863 RECORDS HAVE FAILED COMPONENT SPECIFIED)
Page 70
FIGURE 28: ANALYSIS OF FAILURE CAUSE (654 OF 1863 RECORDS HAVE FAILURE CAUSE IDENTIFIED)
One goal of the IDB work is to develop appropriate hazard rates for transformer subsets of interest. The
hazard functions can be convoluted with the corresponding in-service population to provide forecasts of
anticipated failures. In Figure 29, an application example for a set of transformers from a particular utility
provides the probability distribution of the number of failures in the next year based on a hazard rate
determined from IDB analysis. Also provided are 95% confidence bounds on these probabilities. These results
were computed using the appropriate hazard function and the transformer set demographic data. For
example, the probability of having two failures in the next year is about 0.27. The black bars are the upper
and lower 95% confidence bounds on the individual probabilities. There is essentially 0% chance of having
greater than nine failures. Such calculations can provide information useful for asset management and planning.
Page 71
FIGURE 29: APPLICATION EXAMPLE: YEARLY FAILURES
The development of an industry-wide database for power transformers is an ongoing effort. Through the work
undertaken to date, we have positively addressed several fundamental project issues:
1. Does the potential value of an IDB justify the project?
2. Is there sufficient, useable data available?
3. Can data be collected, organized and reviewed with a reasonable level of effort?
4. Are available statistical techniques appropriate for analysis of the collected data?
5. Can hazard models be constructed with necessary confidence limits from the data?
EPRI’s transformer IDB is an ongoing development and the insights, underlying methodology, approach and
findings continue to be fine-tuned, enhanced and evolve as new data-sets are added and existing data
reviewed. Nonetheless there are some valuable accomplishments based on work to-date:
1. Established a value consensus and assembled critical data mass
2. Developed practical data model
3. Developed data validation processes
4. Tested and identified applicable analysis techniques and methodologies
5. Produced promising preliminary results
a. Descriptive failure statistics
b. Aggregated data from different utilities
c. Hazard rates that show age dependent wear-out
d. Application of results to project anticipated number of failures
The Transformer IDB will provide utilities valuable insights and information to support maintenance repair and
replacement decisions, and asset management decisions to minimize lifecycle costs of equipment replacement
and maintenance, including failure costs.
Page 72
5 METHODOLOGY FOR FAILURE DATA COLLECTION

In this chapter, the data collection methodology developed by the working group is presented and discussed. A
uniform way of collecting, compiling and presenting data is proposed.
5.1 Definition of Failure

Based on the poor data collection experience of previous working groups the decision was taken to limit the
data collection to major failures of transformers and reactors operating at 60kV and above.
A major failure was defined as any situation which required the transformer to be removed from service
for a period longer than 7 days for investigation, remedial work or replacement. The necessary repairs
should have involved major remedial work, often requiring the transformer to be removed from its
installation site and returned to the factory. A major failure would require at least the opening of the
transformer or the tap changer tank, or an exchange of the bushings. A reliable indication that the
transformer condition prevents its safe operation is considered a major failure, if remedial work (longer
than 7 days) was required for restoring it to the initial service capability. In some cases also failures were
assigned as major, if remedial work was shorter than 7 days and extensive work with oil processing had
to be done (e.g. exchange of bushing).
5.2 Reliability Questionnaire

An Excel spreadsheet tool consisting of two major sections was developed to collect data in a standardized
way and in accordance with the definition of major failure in section 5.1. The spreadsheets are given as
appendix A and can also be downloaded at [68].
The first section of the questionnaire requests general information about the utility and the population of the
operating transformers for the indicated failure period. Population information requested included:
 Transformer application:
- Substation – Distribution
- Substation – Transmission
- Power Station – Generator Step-Up,
- Power Station – Unit Transformer
- Shunt Reactor
- Other
 Type:
- Two Winding
- Two Winding with Tertiary
- Autotransformer
- Autotransformer with Tertiary
 Number of phases in tank

- 1-Phase
- 3-Phases
 Voltage Ratios dependent on application and highest system voltage
 Rated power dependent on application and highest system voltage
 Typical loading dependent on application and highest system voltage
 Typical manufacturing period dependent on application and highest system voltage
Information on failure definitions and failure data collection was also requested. Because knowledge of the
age distribution of the transformer population allows determining the hazard function, this information would
Page 73
have been of great interest for the evaluation of the failure data. Nevertheless, the request of age distribution
to data suppliers, often a complicated task, could have resulted on a refusal to contribute to the survey. Thus
the working group decided not to collect these data.
The second section captures the transformer failure data grouped into four categories as follows:
 Identification of the unit: application, type, construction type, year of manufacture.

 Features of the unit: rated power, rated voltage, number of phases, cooling system, type of oil, tap
changer, tap changer arrangement, oil preservation system, over voltage protection.
 Detail of occurrence: year of failure, service years to failure, loading immediately prior to failure.
 Consequences of failure: external effects, failure location, service years of failed bushings (if location is
bushings), failure mode, failure cause, action taken, and detection mode.
A pull-down menu for each field suggests the possible answers and ensures that the answers are given in a
standardized way.
5.3 Classification of Failures
Failures were classified into failure location, failure causes and failure modes.
Failure location referred to the primary location (component) in the transformer where the failure was initiated,
and was classified as:
 Windings (HV, MV, LV, or Tapping)

 Lead Exits (HV, MV or LV)
 Bushings (HV, MV, LV, or Neutral)
 Tap Changer
 Core and Magnetic Circuit
 Magnetic Shunts
 Insulation (Phase to Phase, Winding to Ground, HV-Winding to LV-Winding, Turn to Turn) Electrical
Screen
 Cooling Unit
 Tank
 Current Transformer
 Unknown
Failure cause referred to the cause of failure in the primary location where the failure was initiated, and was
defined as the circumstances during design, manufacture or application that led to the failure. Failure causes
were classified as:
 Design, Manufacturing or Materials

 Transportation, Handling, or Storage
 Loss of Clamping Pressure
 Installation on Site
 Improper Maintenance
 Improper Repair
 Abnormal Overload
 Overvoltage
 Overheating
 Lightning
 External Short Circuit
 Repetitive Through Faults
Page 74
 Seismic Disturbances
 Improper Application
 External Pollution
 Loss of Cooling
 Vandalism
 Abnormal Deterioration
 Aging
 Collateral Damage
 Unknown
According to IEEE [69], failure mode refers to the manner in which a failure occurred, and was categorised as
electrical, thermal, and mechanical and contamination. The description of the ‘nature of failure’ as used in Cigré
WG A2.18 [35] is similar to that of failure mode in IEEE [69]. The nature of failure and failure mode will thus
be considered as being equivalent, referring to the mode/nature of failure in the location where the failure
was initiated. Thus failure mode was classified as:
 Dielectric (Partial Discharge, Tracking, or Flashover)

 Electrical (Open Circuit, Short Circuit, Poor Joint, Poor Contact)
 Thermal (General Overheating, Localised Hotspot)
 Physical Chemistry (Contamination (Moisture, Particles, Gas), Corrosion)
 Mechanical (Bending, Breaking, Displacement, Loosening, Vibration)
 Unknown
5.4 Data Collection and Limitations

The methodology of failure data collection was done in a similar way as performed by a Cigre Working
Group A3.06 [70]. Each participating utility was requested to complete the Excel questionnaire given in annex
A. The completion of the questionnaire was supported and monitored by the members of the working group.
Especially it was taken care to ensure that all major failures in the indicated time period were included. All
responses were compiled into a database. In order to achieve a data security and anonymity, the failure data
from each source was made anonymous by labelling with a code based on the geographical location and a
sequence number. Information about the transformer manufacturer was not collected.
The objectives and constraints of a failure data survey are perfectly described by Cigre Working Group
A3.06 [70]. Therefore the relevant text is cited here again.
“The questionnaire may not have been fully understood by respondents resulting in certain, failure details being
classified as unknown. Not all asset owners responded to the questionnaire. The working group recognizes that
imperfections exist in the data gathered from utilities around the world but they all realize that some good, real
operational information is much better than assuming critical electrical equipment will continue to operate in its
present state far into the distant future.
Too often, utilities are tempted to believe the data presented is of high precision and that a single metric such as
annual failure frequency can be used to make important financial and technical decisions, such is not the case. It
was recognized by the working group that analysis of the survey data is not an absolutely exact activity and
statistical tools had to be utilized to not only obtain the most meaningful information from the data but to also
show a range of likely outcomes suggested by the data. To provide the reader with the most valuable survey
summary information, the working group has distilled the survey responses into a range of meaningful statistical
values, not just one single that can easily be mischaracterized as highly precise.
Carrying out a reliability survey with the objective and scope as of the present study is indeed a major undertaking.
As pointed out above, utilities need good failure statistics for several purposes. However, many of them are to a
Page 75
lesser extent willing to take the job of collecting and contributing data from their own operations to collaborative
enquiries. Hence, getting utilities participate is a major challenge, and presumably becoming even more difficult in
the future as competition increases and less tangible engagements are under greater pressure. In practice it turned
out that having working group members who either were employed by the utility or had good personal relations to
individuals in utilities in their own country was necessary. Sending out invitation letters and in other ways
(repeatedly) contacting utilities in countries not represented in the working group gave very little in return. Thus for
obtaining a broad participation it is necessary to have the survey organized by a rather large group of committed
individuals from many countries.
Content and wording of the questionnaires are extremely important. Anything that has the slightest chance of being
misunderstood will, to a surprisingly large degree, really be misunderstood. Moreover, it is very important to get
the right trade-off between asking too many and too complicated questions, and just collecting information that are
obvious and thus of limited value. Some of those who complete the forms are not necessarily transformer experts,
and asking too difficult questions, seeking information not ready available or using questionnaires in a foreign
language, may easily cause them to give up. Input data should be numbers, dates or given by selecting one or
more alternatives from an available list. If open-ended questions are used, time consuming manual handling and
interpretation of the returned questionnaires become necessary. This is recommendable only for very special topics
and circumstances. It may be advisable to run a small trial or test of the questionnaires before they are sent out to
the wider target group. Such an exercise may disclose inconsistencies, poor wording, questions that are easily
misinterpreted and other imperfections that can easily be corrected.
Finally, it is important to have realistic expectations about the quality of the results that can be obtained from
reliability surveys. Equipment technologies and vintages, substation layouts and other matters can differ
significantly from one country to the other. Moreover, different countries have different traditions with regard to
how they install, operate, maintain, repair and replace their equipment. Consequently, information collected about
populations and failures will never be homogeneous, complete, fully consistent and without flaws. At least some
underreporting of failures, in particular the minor ones, is unavoidable. When analyzing and evaluating service
experience data these shortcomings should be kept in mind and dealt with. However, the bigger picture should
remain clear: Although failure statistics obtained from extensive international surveys have flaws it is still of great
value to the power engineering community. There are no alternative ways of obtaining such information; pure
guesswork is not really an option.” [70]
Page 76
6 RESULTS OF PERFORMED RELIABILITY SURVEY

The results of the questionnaire are presented and analysed in this chapter. The collection of data was done
between 2010 and 2012.
6.1 Collected Data

The working group collected 964 major failures which occurred in the period 1996 to 2010, within a total
manufacture of the units span from the 1950’s up to 2009. The length of reference periods range from 3 to 11
years. The reference period is the period during which the failures occurred, as defined in the reliability
questionnaire.
The participating utilities with their contributed population in transformer-years, is shown in Table 23. Because
the number of operational transformers was only provided for one year, the total number of transformer-years
(population per utility) was calculated under the assumption that the number of transformers in operation was
constant during the reference period. The number of transformers was multiplied with the length of the
reference period in years to obtain an estimate of the total number transformer-years. The utilities normally
didn’t provide data for the full investigated failure period. Therefore the amount of transformers operated by
the respective utility cannot be deduced from the amount of transformer-years given in Table 23.
At 150,072 transformer-years the investigated population of substation transformers was considerably high
(Table 24). This value was almost four times higher than the population in Bossi [2] with 40,547 transformer-
years. The population of generator step-up transformers was considerably lower (17,387 transformer-years)
(see Table 25).
The questionnaire also collected population and failure data of shunt reactors. However, the reported
population of 1350 shunt reactors (13927 shunt reactor-years) was small and originated mainly from two
countries. Therefore a detailed analysis of shunt reactor failures could not be performed. The number of
reported failures was in total 31, which resulted in a failure rate of 0.22%. The failure location of these
failures is analyzed in chapter 6.5.2.
The collected data consisted predominantly of core type transformers. The results presented in this chapter
therefore is applies to core type type transformers.
HVDC transformers failure or population data was not collected. Information on HVDC transformer reliability
can be found in [71].
Page 77
TABLE 23: INVESTIGATED POPULATION OF TRANSFORMERS

Transformer-
Country / Utility
Years
Germany 1 3,350
Germany 2 4,730
Germany 3 1,848
Germany 4 160
Germany 5 85
Germany 6 1,990
Germany 7 1,040
Austria 1 1,793
Austria 2 407
Austria 3 1,221
Austria 4 704
Austria 5 616
Austria 6 1,408
Austria 7 561
Suisse 1 561
Suisse 2 114
Suisse 3 407
Suisse 4 825
Suisse 5 462
Suisse 6 1,771
Czech 1 555
Spain 1 8,745
Croatia 1 950
UK 1 6,993
UK 2 1,582
UK 3 44
UK 4 2,628
UK 5 1,683
The Netherlands 1 1,947
Italy 1 5,670
Ireland 1 3,190
France 1 1,368
France 2 7,416
Denmark 1 781
Denmark 2 440
Poland 1 2436
Hungary 1 1285
Europe Total 71766 (42.86%)
New Zealand 1 275
New Zealand 2 286
New Zealand 3 341
New Zealand 4 3,399
Australia 1 1,309
Australia 2 3,289
Australia 3 88
Australia 4 2,244
Australia 5 594
Australia 6 88
Australia 7 77
Australia 8 1,309
Australia 9 2,475
Australia Total 15774 (9.42%)
Brazil 1 29,920 (17.87%)
India 1 7,996 (4.77%)
Korea 1 17,784 (10.62%)
USA 1 3,146 (1.88%)
Canada 1 15,210 (9.08%)
South Africa 1 5863 (3.50%)
Page 78
TABLE 24: INVESTIGATED POPULATION OF SUBSTATION TRANSFORMERS
POPULATION HIGHEST SYSTEM VOLTAGE [kV]

INFORMATION
69 ≤ kV < 100 100 ≤ kV < 200 200 ≤ kV < 300 300 ≤ kV < 500 500 ≤ kV < 700 kV ≥ 700 All
Number of
2,962 10,932 4,272 3,233 434 348 22,181
Transformers
Transformer-Years 15,267 64,718 37,017 25,305 4,774 2,991 150,072
TABLE 25: INVESTIGATED POPULATION OF GENERATOR STEP-UP TRANSFORMERS
POPULATION HIGHEST SYSTEM VOLTAGE [kV]

INFORMATION
Number of
14 320 455 673 167 74 1,703
Transformers
Transformer-Years 153 3,278 4,639 6,740 1,837 740 17,387
6.2 Data Limitations

Component replacement, in particular bushings and tap changers, was not taken into account due to the lack of
information. The age at failure therefore reflects that of the transformer at failure, and not that of the
component at failure. As the age of the unfailed units was not collected with the questionnaire, only failure
rates for a group of transformers can be calculated instead of hazard curves.
Failure data of some national statistics, e.g. Canada and Japan, couldn’t be used, because of different failure
definitions.
Where the classification into failure locations, failure causes and modes could not be determined, the failures
were left unclassified. Furthermore it has to be regarded that mode and cause of a failure often could not be
determined or are uncertain.
6.3 Failure Rate

The failure rates of combined populations were calculated according to equation 3.2. Because the number of
operational transformers was only provided for one year, the total number of transformer-years (population
per utility) was calculated under the assumption that the number of transformers in operation was constant
during the reference period. The failure rates according to the voltage category for the substation and
generator step-up transformers, as well as the combined group of transformers, are given in Table 26 to Table
28. It is important to note that the number of failures of generator step-up units and of units in voltages classes
above 500 kV, as well as the population of these two categories, was low. The calculated failure rates should
thus be considered with caution.
The overall failure rate of substation, generator step-up and the combined group of transformers were all
within 1%. Only generator step units in the voltage class 300kV to 500kV exceeded 1% failure rate.
Although differences in failure rate were very low, less than 0.5%, the failure rates of generator step-up units
were higher than substation transformers in nearly all the voltage categories.
Page 79
1,60
Substation Transformers
1,40
Generator Step-Up Units
1,20
Combined Group of Transformers
Failure Rate in %
1,00
0,80
0,60
0,40
0,20
0,00
69 ≤ kV < 100
100 ≤ kV < 200
200 ≤ kV < 300
300 ≤ kV < 500
500 ≤ kV < 700
kV ≥ 700
All
FIGURE 30: FAILURE RATE DEPENDENT ON VOLTAGE CLASS AND APPLICATION
TABLE 26: FAILURE RATES OF SUBSTATION TRANSFORMERS
FAILURES &
POPULATION
INFORMATION 69 ≤ kV < 100 100 ≤ kV < 200 200 ≤ kV < 300 300 ≤ kV < 500 500 ≤ kV < 700 kV ≥ 700 All
Major Failures 144 280 186 152 27 10 799

Transformer-
Years 15,267 64,718 37,017 25,305 4,774 2,991 150,072
FAILURE RATE 0.94% 0.43% 0.50% 0.60% 0.57% 0.33% 0.53%
TABLE 27: FAILURE RATES OF GENERATOR STEP-UP TRANSFORMERS
FAILURES &
POPULATION
Major Failures 0 20 43 89 9 4 165

Transformer-
153 3,278 4,639 6,740 1,837 740 17,387
Years
FAILURE RATE 0.00% 0.61% 0.93% 1.32% 0.49% 0.54% 0.95%
Page 80
TABLE 28: FAILURE RATES OF COMBINED GROUP OF TRANSFORMERS
FAILURES &
POPULATION
Major Failures 144 300 229 241 36 14 964

Transformer-
15,420 67,996 41,656 32,045 6,611 3,731 167,459
Years
FAILURE RATE 0.93% 0.44% 0.55% 0.75% 0.54% 0.37% 0.57%
6.4 Failure Rate Dependent on Transformer Age

As discussed in section 3.5.1, proper statistical analysis of data requires an understanding of the data types
being dealt with.
The collected data is a combination of units in failures with repair, units in complete failure and the surviving
population, i.e. those units that did not fail. Complete failure resulted in scrapping of the transformer and can
occur only once. Failures with repair required repair or replacement of components or subsystems, to return the
transformer to service. A transformer could have multiple failures with repair, but knowledge of failures outside
the evaluation period is not known for the collected data.
Because the evaluation period spanned from 2000 to 2010 (refer to evaluation period of the analysis
presented later in this section) censoring and truncation as discussed in section 3.5.1 would apply to the
collected data (failures and surviving population) referred to as the sample, as follows:
• The failure times of ‘failures with repair’ which occurred before 2000, are not known. Only the failure
times beyond 2000 (truncation time) is included in the sample, therefore the sample is left truncated.
• Because the sample contains transformers which survived to the end of the evaluation period, up to
2010, the sample is also right censored. The times at which the surviving transformers in the sample exit
the evaluation period, are referred to as the right censoring times.
According to Nelson [38], multiply censored data consists of failure times intermixed with running times that can
result from the removal of units from use before failure or collection of data while units are still running. The
running times are referred to as censoring times, and can be left or right censored. Since the collected data
contained failure times, and right censoring times, it suggests that the data could be classified as Multiply
Censored Data, if the failures with repair were excluded from the sample.The treatment of multiply censored
data is addressed in Nelson [38] where non parametric plotting techniques such as probability plotting and
hazard plotting are discussed for determining the cumulative distribution function and the hazard function of
such data, respectively.
Applying Renewal Theory
By definition, a power transformer is a system which consists of several repairable components and subsystems
(windings, tap changers, etc.) and should therefore be treated as a repairable system. However, in the event of
a major failure of a transformer, the concerned component or subsystem is normally fully exchanged and
renewed. As shown in Figure 49 and Figure 50 in case of winding failures, which is the dominant failure
location, often a complete new rewinding is done or in many cases also scrapping of the transformers. Assuming
the system could be 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 could be applied. This approach is
applied in this section for a restricted part of the collected failure data.
The hazard function, also called momentary probability of failure, gives the probability that the component will
fail at a certain life. One can read the real hazard (risk) of failure directly from the hazard function. This
Page 81
information is valuable for asset managers, because it aids decision making for the proper maintenance
strategy. As discussed in chapter 2, the hazard function is computed using the following formula:
1 ∆n(t )
h(t ) = ⋅
N (t ) ∆t

Figure 31 depicts the total number of major failures of transformers (voltage classes between 100 kV and 500
kV) by age in the failure interval. No significant trends, in terms of number of failures with age, can be
observed, because information about age distribution is missing. To determine the hazard rate function, the age
distribution of the population is required. For the sake of simplicity, an inquiry into the age distribution of the
population was not included into the developed questionnaire. As a result, the hazard curve could not be
calculated for the full data set.
35
100kV<=U<200
30
200kV<=U<300
25 300kV<=U<500
Number of Failure
20
15
10
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Age
FIGURE 31: NUMBER OF FAILURES DEPENDENT ON TRANSFORMER AGE
(100 KV – 500 KV, TRANSFORMERS, WORLDWIDE, 770 FAILURES)
To overcome the problem of missing age distribution of the full investigated population, the failure behavior of
three substation transformer populations for which the age distribution was known, was investigated. Group A
consisted of substation transformers of two European utilities [72]. Populations B and C originated from New
Zealand and Canada respectively. Table 29 shows the population data of these groups.
Page 82
Table 29: Population Data for Hazard Curve Analysis
Transformer Population Voltage No. of No. of Failure Rate

Class (kV)
Transformer- Major
Years Failures
Group A 110≤U≤380 6510 35 0,54
Group B 110≤U≤220 4686 24 0,51
Group C 120≤U≤315 11950 28 0,23
The data collection for the three groups spans from 2000 to 2010. Therefore the number of units surviving age
t, N(t), have to be adapted to this 11 year failure interval. This procedure is explained in ANNEX C:
Description of Analysis of Transformer Population Data. The resulting transformer years of transformers
surviving time t, N(t)∙Δt, is shown in Figure 32. The high service experience of population A in the age interval
between 20 and 40 corresponds to the extension of the transmission grid in Europe between 1970 and 1980.
Consequently this age group is represented in the failure interval between 2000 and 2010 quite well (Figure
32). The comparison of the three population groups reveals the different installation times of the transformers.
450
400 Population A (110kV-380kV)
350 Population B (110kV-220kV)

Population C (120kV-315kV)
Transformer Years
300
250
200
150
100
50
0
1 6 11 16 21 26 31 36 41 46 51 56 61
Age
FIGURE 32: NUMBER OF TRANSFORMERS-YEARS SURVIVING AGE t ,
(CONSIDERING 11YEAR-FAILURE INTERVAL)
Page 83
10,0% Population A (110kV-380kV)

9,0% 1 failure
Population B (110kV-220kV)
5 TrYears
8,0%
Population C (120kV- 315kV)
1 failure
7,0%
Hazard rate %
6,0%
1 failure
5,0%
4,0%
3,0%
2,0%
1,0%
0,0%
1 6 11 16 21 26 31 36 41 46 51 56 61
Age
FIGURE 33: FAILURE HAZARD RATE (5 YEARS MOVING AVERAGE)
Figure 33 shows the hazard rates of the three population groups. In order to identify trends also the 5-year
moving average of hazard rate curves is shown. Because only failures between 2000 and 2010 were
evaluated, these calculated hazard rate functions does not give any information on early failure rates of older
transformer designs (e. g. before 1980) and ageing problems of newer designs (e. g. after 1990).
All populations show a low hazard rate and no distinct bathtub curve character (compare Figure 7). Within
population A and C just one failure each occurred in the first 20 years at an age of 2 years and 9 years
respectively. Both population showed an increase of the hazard rate at an age of 30, 40 respectively from 0%
to a level of app. 1% (5 years moving average). This level does not increase further until the highest ages with
the exception of two outliers at an age of 48 (Pop. A, 20%) (57 (Pop. C, 4,3%) resp.). These can be attributed
to uncertain statistical data as the operational experience tends to zero (compare Figure 32), e. g. one failure
within 5 transformer years at age 48.
The hazard curve of population B appeared to suggest a relatively constant probability of failure across all
ages. Also here the outliers at age 55 (2 failures within 39 TransformerYears) and 60 (1 failure within 17
TransformerYears) are due to low operational experience (compare Figure 32).
Thus a significantly increasing probability of failure after a particular age cannot be derived from the
available data. The hazard curve for substation transformers does not seem to show a distinct wear-out
characteristic. Although it is often assumed that an increasing probability of failure after a particular age is
due to the effects of component ageing, i.e. oil or paper ageing for transformers, for transformers, a more
likely cause 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 [34].
Substation transformers are normally not so heavily loaded which could make ageing pronounced. Failures due
to ageing play a minor role and are masked by random or external failure reasons and early replacement of
transformers. Furthermore, a transformer consists of several subsystems, with their own individual ageing
characteristics and failure modes. Because major failures occur in all of the subsystems (see chapter 6.5), the
Page 84
hazard curve can be considered as the result of competing failure modes, which have presumably diverse
characteristics. Therefore no distinct increase of the failure probability in later life can be observed in the
hazard curve; but only the behaviour of subsystems with random failures.
A second reason for the absence of a distinct ageing characteristic of the hazard curve is the replacement
strategy. Utilities often use replacement strategies, in which preferably old transformers are replaced, which do
not anymore fulfill the operational requirements (e.g. losses, noises, regulating range). Thus potential failures of
replaced units cannot contribute to the hazard curve. For the present analysis, information about preventive
replacements in the past was not available. The availability of this information could have a profound biasing
effect on the failure statistics, as transformers are not left in service to fail, so that these statistics cannot be
used directly for lifetime modelling.
In [73] the failure analysis of the transformers used on the National Grid system in the UK is given (see Figure
34). “It shows the historical hazard rate (conditional probability) for both failure (blue line) and replacement (pink
line) against age for the transformer population. Figure 34 shows that although replacement becomes more likely
as the transformer gets older, due to the operation of the condition based replacement scheme, the probability of
failure does not. This shows that a condition based replacement scheme does appear to be capable of controlling
the failure rate of an ageing population, at least at the early stages, however the operation of the scheme also
means that the historical failure hazard rate curve will not be a useful predictor of transformer lifetimes. The fact
that the failure hazard rate (blue line) is essentially independent of age shows that these failures are occurring
randomly. This is borne out by an analysis of the causes of these failures that are only age related in a very few
cases. Historical failure data shows that failures in service are not showing an increasing trend with age, but this is
due to an active condition based replacement programme that is replacing older transformers even though age is
not directly considered as a factor in the condition assessment process.
This means that traditional approaches to building a statistical end-of-life model cannot be used. Analysis of the
insulation of transformers removed from service for any reason indicates a very wide range of condition, some
samples show severe thermal ageing and it is clear that age-related failures can be expected if replacement is not
carried out, other samples show little ageing and for these transformers it appears that very long lifetimes might be
expected if other ageing mechanisms do not become apparent.”
FIGURE 34: FAILURE HAZARD AND REPLACEMENT HAZARD FOR UK POPULATION [54]
Because the hazard curve does not show a distinct increase with time the use of Time Based Maintenance (e. g.
exchange of substation transformers at a certain age) is not an effective asset management strategy.
Maintenance should be planned according to the actual condition.
Page 85
Generator step-up units are often heavier loaded near their nominal rating, so a more pronounced ageing
characteristic could be expected. The number of GSU failures collected in this survey was much lower than the
substation transformers and there was no information about the age distribution of the GSU population
available. Thus a calculation of a hazard curve was not possible and consequently the impact of aging on the
hazard curve could not be assessed. The homogeneity of the GSU data, e.g. operational characteristics,
voltage level, manufacturer and commissioning time, were also not as complete as with substation transformers.
Fitting the hazard curve allows its extrapolation in time beyond the age of the oldest assets in the population.
Most renewal failure data sets encountered in the maintenance environment can be fitted lifetime modeling
distributions such as the Weibull distribution. However, in case of power transformers, the occurrence of a
failure event is more or less a random event. Furthermore, the advance of transformer maintenance and
replacement strategies limits any age-dependent increase of failure rate of power transformers. Lifetime
modelling techniques such as Weibull fitting are therefore not suitable for modelling transformer failure data. A
possible approach is to estimate on the number of years a replaced transformer would have remained in
service if it would not be replaced [74].
6.5 Failure Location Analysis
6.5.1 TRANSFORMERS
The failure data of the full population were analysed as a function of the primary location (component) in the
transformer where the failure was initiated. In Table 30 the failure location analysis is shown according to
voltage class for substation and generator step-up transformers. The number of failures classified as “unknown”
location is given in the last line of Table 30. They were not taken into account in the percentage values and in
the following analysis.
Winding, tap changer and bushing related failures were the major contributors, followed by lead exit related
failures.
Failures originating in the windings were the major contributor (89%) in units with voltages lower than 100 kV.
The contribution in this voltage class was based on the data of three utilities, which is non homogeneous. The
contribution of winding failures was also the highest in this voltage class, compared to the other voltage classes.
This could be attributed to the fact that transformers in this voltage class are designed and operated
differently from units in the higher voltages classes.
Up to 700kV, the contribution of bushing related failures increased with increasing voltage class. Lead exit
related failures exhibited the same increasing trend, across all voltage classes. The contribution of tap changer
related failures appeared to decrease with increasing voltage level, but it doesn’t take into account the
possibility that at higher voltage level, less transformers are equipped with tap changers.
Page 86
TABLE 30: FAILURE LOCATION ANALYSIS DEPENDENT ON VOLTAGE CLASS

Failure Location
All
(without
69 ≤ kV < 100 100 ≤ kV < 200 200 ≤ kV < 300 300 ≤ kV < 500 500 ≤ kV < 700 kV ≥ 700 “unknown”)
(122) (246) (199) (201) (18) (11) (797)
Winding 109 89.3% 107 43.5% 75 37.7% 79 39.3% 4 22.2% 4 36.4% 378 47.4%
Lead Exit 1 0.8% 7 2.9% 16 8.0% 20 10.0% 2 11.1% 4 36.4% 50 6.3%
Insulation 2 1.6% 4 1.6% 6 3.0% 6 3.0% 0 0 1 9.1% 19 2.4%
Electrical screen 0 0 0 0 0 0 3 1.5% 0 0 0 0 3 0.4%
Bushing 0 0 32 13.0% 33 16.6% 44 21.9% 5 27.8% 1 9.1% 115 14.4%
Core and
7 5.7% 7 2.9% 10 5.0% 4 2.0% 2 11.1% 0 0% 30 3.8%
magnetic circuit
Flux Shunts 0 0% 0 0% 0 0% 3 1.5% 0 0 0 0% 3 0.4%
Tank 0 0% 1 0.4% 0 0% 2 1.0% 0 0 1 9.1% 4 0.5%
Cooling unit 0 0% 3 1.2% 1 0.5% 2 1.0% 2 11.1% 0 0% 8 1.0%
Tap Changer 3 2.5% 84 34.2% 57 28.6% 38 18.9% 3 16.7% 0 0% 185 23.2%
CT 0 0% 1 0.4% 1 0.5% 0 0% 0 0 0 0% 2 0.3%
Number of
unknown failure 22 15.3% 54 18% 30 13.1% 40 16.6% 18 50% 3 21.4% 167 17.3%
locations
Figure 35 shows the failure locations for all failures with voltage class 100kV and above, and excluding
failures classified as unknown.
Failures were predominantly winding (40%) and tap changer (27%) related.
The contribution of tap changer related failures decreased significantly in comparison with the statistics from
1983 given in [2]. The lower failure rate of OLTC can be attributed to the development of better contact
materials (use of silver plated contacts) see respective chapter 2.1.5. 95% of the failed substation transformers
and 86% of the failed generator step-up units were equipped with a tap changer.
Bushings (14%), lead exits (6%) and core (4%) were minor contributors.
Page 87
CT; 0,30%
Tap Changer;
26,96% HV Winding;
21,19%
Cooling unit; 1,19%
Tank; 0,59%
MV Winding; 4,44%
Flux Shunts; 0,44%
Core and magnetic LV Winding; 11,41%

circuit; 3,41%
LV Bushings; 0,74% Tapping Winding;

HV Bushings;
2,81%
14,07%
MV Bushings; 2,22% HV Lead Exit; 5,04%
Electrical Screen; MV Lead Exit; 1,33%
Winding to Winding 0,44%
Isolation; 0,74% Winding to Ground Phase to Phase LV Lead Exit; 0,89%
Isolation; 1,19% Isolation; 0,59%
FIGURE 35: FAILURE LOCATION ANALYSIS BASED ON 675 MAJOR FAILURES FOR U≥100KV
Figure 36 and Figure 37 depict the failure location dependent on application in units with voltages 100kV and
above. Substation transformers below 100kV were excluded because of their different failure location
behavior. GSUs without tap changers were also excluded from the analysis in order to prevent
misinterpretation of the data.
Winding related failures were the largest contributor in both transformer applications. GSU transformers had a
higher contribution of winding and lead exit failures than substation transformers. Substation transformers on
the other hand had a higher contribution of tap changer related failures than GSU transformers. The
contributions of bushing related failures were similar in both transformer applications.
The analysis of the winding failures dependent HV, MV, LV and tap winding, as depicted in Figure 38, shows
that more than 50% of the major failures originate from the HV winding.
Page 88
CT; 0,37% HV Winding;

19,40%
Tap Changer;
31,16%
MV Winding;
5,60%
Cooling unit; LV Winding; 9,33%

1,12%
Tapping Winding;
Tank; 0,75%
3,36%
Flux Shunts; HV Lead Exit;
0,37% LV Bushings; 3,17%
Core and 0,37% MV Lead Exit;
magnetic circuit; 1,68%
HV Bushings; LV Lead Exit;
2,43%
13,99% 1,12%
Winding to Winding to
MV Bushings; Electrical Screen; Winding Isolation; Ground Isolation; Phase to Phase
2,80% 0,56% 0,37% 1,31% Isolation; 0,75%
FIGURE 36: FAILURE LOCATION ANALYSIS, SUBSTATION TRANSFORMERS, U≥100KV, 536 FAILURES
Cooling Tap Changer;

unit; 11,81%
Flux Shunts;
1,57%
0,79% HV Winding;
Core and 28,35%
magnetic
circuit; 6,30%
LV Bushings;
2,36%
HV Bushings;
14,17%
Winding to LV Winding;
Winding 18,90%
Isolation; HV Lead Exit; Tapping
2,36% 12,60% Winding; 0,79%
FIGURE 37: FAILURE LOCATION ANALYSIS, GENERATOR STEP-UP TRANSFORMERS, U≥100KV, 127 FAILURES
Page 89
Tapping Winding to
Winding Winding
9% Isolation;
HV Tapping
Winding 4,69%
LV Winding;
51% 1,56%
Winding
25%
LV
Winding;
37,50% HV
MV Winding;
Winding 56,25%
15%
FIGURE 38: WINDING FAILURE LOCATION ANALYSIS FOR U≥100KV (LEFT: SUBSTATION TRANSFORMERS RIGHT: GSU)
To assess the contributions of failure location with manufacturing period, the major failures were classified into
two groups representing transformers manufactures before and after 1980. It is important to note, that since
the failures occurred in the period 2000 to 2010, the pre 1980 manufactured group of transformers consisted
mainly of older units.
Winding related failures featured as the major contributor, followed by tap changers and bushings,
irrespective of manufacturing period.
CT
0,30%
Tap Changer Winding
23,12% 42,94%
Cooling unit
0,90%
Core and
magnetic circuit
4,20%
Bushings
15,92%
Electrical Screen
Insulation Lead Exit
0,90%
2,70% 9,01%
FIGURE 39: FAILURE LOCATION ANALYSIS OF TRANSFORMERS WITH MANUFACTURING PERIOD BEFORE 1980
(333 FAILURES, U≥100KV)
Page 90
CT
0,29%
Tap Changer Winding
30,70% 36,84%
Cooling unit
1,46% Tank
1,17%
Flux Shunts
0,88% Lead Exit
Core and
5,56%
magnetic Insulation
Bushings
circuit 2,34%
18,13%
2,63%
FIGURE 40: FAILURE LOCATION ANALYSIS OF TRANSFORMERS WITH MANUFACTURING PERIOD AFTER 1980
(342 FAILURES, U≥100KV)
6.5.2 SHUNT REACTORS

The number of reported failures of shunt reactors was small and originated mainly from two countries.
Nevertheles an analysis of the failure location was performed and is shown in Table 31.
Bushing and core related failures were the major contributors. This could be attributed to the fact that shunt
reactors are severely exposed to vibrations, which can lead to bushing and care failures.
Page 91
TABLE 31: FAILURE LOCATION ANALYSIS OF SHUNT REACTORS
FAILURE LOCATION NUMBER OF FAILURE PERCENTAGE OF FAILURE
WINDING 4 12,90%
LEAD EXIT 2 6,45%
ISOLATION 2 6,45%
BUSHINGS 10 32,26%
CORE AND MAGNETIC

CIRCUIT 10 32,26%
OTHER 3 9,68%
SUM 31 100,00%
Other Winding
10% 13%
Lead Exit
7%
Isolation
6%
Core and
magnetic circuit Bushings
32% 32%
FIGURE 41: FAILURE LOCATION ANALYSIS OF SHUNT REACTORS
6.6 Failure Mode Analysis

Information about the failure mode was also collected in the questionnaire. The failure mode describes the
nature of the failure illustrating what actually happened when the failure occurred. The definitions of the failure
modes are according to [35]. Dielectric failure means partial discharge, tracking and flashover. Electrical
failure means open circuit, short circuit, poor joint, poor contact, ground deterioration, floating potential.
The failure mode analysis of substation and generator step up transformers is shown in Figure 42, and
according to voltage class shown in Table 32. Dielectric mode failures were the most prominent, followed by
mechanical and electrical type failures. Dielectric mode failures were also the major contributor in all the
voltages classes, with the highest contributions occurring in the 69kV to 100kV, and the 500kV to 700kV
voltage classes.
Figure 43 and Figure 44 show the failure mode according to transformer application. Substation transformers
had higher contributions of dielectric and mechanical type failures, whereas GSU transformers had higher
Page 92
contributions of thermal and dielectric mode failures. This can be explained by the normally higher loading of
GSUs compared to substation transformers.
TABLE 32: FAILURE MODE ANALYSIS DEPENDENT ON VOLTAGE CLASS

Failure Mode
(144) (300) (229) (241) (36) (14) (964)
Dielectric 70.14% 25.33% 36.68% 24.89% 72.22% 42.86% 36.62%
Electrical 12.50% 19.33% 15.72% 17.84% 0 28.57% 16.49%
Thermal 0.69% 12.00% 10.04% 18.25% 2.78% 0.00% 10.89%
Physical Chemistry 0.00% 3.66% 4.37% 4.56% 2.78% 0.00% 3.32%
Mechanical 6.25% 27.33% 22.71% 18.25% 11.11% 7.14% 20.02%
Unknown 10.42% 12.33% 10.48% 16.18% 11.11% 21.43% 12.66%
Unknown;
12,66%
Dielectric;
36,62%
Mechanical;
20,02%
Physical
chemistry;
3,32%
Thermal; Electrical;
10,89% 16,49%
FIGURE 42: FAILURE MODE ANALYSIS BASED ON 964 MAJOR FAILURES
Page 93
Unknown
12,14% Dielectric
38,30%
Mechanical
22,15%
Physical
Electrical
chemistry Thermal
18,02%
2,88% 6,51%
FIGURE 43: FAILURE MODE ANALYSIS OF 799 SUBSTATION TRANSFORMERS
Unknown
15,15% Dielectric
28,48%
Mechanical
9,70%
Physical
chemistry
5,45%
Thermal Electrical
32,12% 9,09%
FIGURE 44: FAILURE MODE ANALYSIS OF 165 GENERATOR STEP-UP TRANSFORMERS
6.7 Failure Cause Analysis

Determining the root cause of failure can be a difficult and extensive task. Failure can often be attributed to
different failure causes, and in some cases the causes are unknown; 29% of the analysed major failures were
due to unknown causes. The results presented in this section should therefore be treated and interpreted with
caution.
Due to the restriction on the failure period from 2000 to 2010, information about aging problems of newer
transformers or design problems of older transformers cannot be derived from this analysis.
Design and manufacturing (20%), aging (12%) and external short circuits appeared to be major contributors.
Page 94
Lightning and overvoltage contribute just by app. 3% and seem to be almost negligible as failure cause. This
could be attributed to the fact that almost all transformers in the investigated population were protected by
surge arresters. Furthermore, the existence of overvoltages is unknown, because they are normally not
measured. Thus, failures that were actually caused by "Lightning" or "Overvoltages" may have been
interpreted as having a different root cause, e.g. "Unknown" which is the second largest root cause. This
consideration is supported by the fact that most failures were inter-turn of dielectric nature, which could be
explained by resonant internal overvoltages against which surge arresters do not provide adequate protection.
In order to get a deeper insight also the installation of monitoring systems for the measurement of transient
overvoltages may be beneficial.
External short- Design

circuit 9,96%
11,62%
Manufacturing
Aging 9,96%
12,34%
Improper repair
6,02%
Other reasons
4,88%
Unknown
29,05% Material
3,73%
Improper
maintenance
3,22%
Corrosive Sulphur Abnormal
0,21% Lightning Deterioration
External Pollution 2,18% 2,49%
Improper Loss of cooling 0,52% Installation on-site
application 0,21% Loss of clamping Repetitive through 0,83%
0,21%
Collateral Damage Overheating pressure Overvoltage faults
0,31% 0,31% 0,41% 0,62% 0,83%
FIGURE 45: FAILURE CAUSE ANALYSIS BASED ON 964 MAJOR FAILURES
Page 95
TABLE 33: FAILURE CAUSE ANALYSIS DEPENDENT ON VOLTAGE CLASS

Failure Cause
(144) (300) (229) (241) (36) (14) (964)
Improper Specification 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Design 0.69% 8.67% 13.10% 15.77% 2.78% 0.00% 9.96%
Manufacturing 10.42% 11.33% 9.61% 9.54% 0.00% 14.29% 9.96%
Material 0.69% 5.00% 3.93% 4.56% 0.00% 0.00% 3.73%
Transport, Handling or
0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Storage
Loss of clamping
0.00% 1.00% 0.44% 0.00% 0.00% 0.00% 0.41%
pressure
Installation on-site 0.00% 1.67% 0.44% 0.83% 0.00% 0.00% 0.83%
Improper maintenance 0.69% 2.33% 6.55% 3.32% 0.00% 0.00% 3.22%
Improper repair 14.58% 7.67% 3.06% 2.90% 0.00% 0.00% 6.02%
Abnormal Overload 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Overvoltage 0.00% 0.33% 0.87% 1.24% 0.00% 0.00% 0.62%
Overheating 0.00% 1.00% 0.00% 0.00% 0.00% 0.00% 0.31%
Lightning 1.39% 2.33% 2.18% 2.90% 0.00% 0.00% 2.18%
External short-circuit 25.69% 10.33% 10.04% 8.71% 0.00% 0.00% 11.62%
Repetitive through
2.08% 1.00% 0.44% 0.41% 0.00% 0.00% 0.83%
faults
Seismic Disturbances 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Improper application 0.00% 0.00% 0.87% 0.00% 0.00% 0.00% 0.21%
External Pollution 0.00% 0.33% 1.31% 0.41% 0.00% 0.00% 0.52%
Loss of cooling 0.00% 0.00% 0.44% 0.41% 0.00% 0.00% 0.21%
Vandalism 0.69% 0.00% 0.00% 0.00% 0.00% 0.00% 0.10%
Abnormal Deterioration 0.00% 1.33% 3.06% 3.32% 0.00% 35.71% 2.49%
Aging 15.28% 14.67% 13.54% 9.13% 0.00% 0.00% 12.34%
Collateral Damage 0.00% 0.67% 0.44% 0.00% 0.00% 0.00% 0.31%
Corrosive Sulphur 0.00% 0.33% 0.00% 0.41% 0.00% 0.00% 0.21%
Other reasons 10.42% 3.33% 5.68% 3.32% 0.00% 7.14% 4.88%
Unknown 17.36% 26.67% 24.02% 32.78% 97.22% 42.86% 29.05%
Page 96
6.8 External Effects Analysis

Most of the major failures did not result in any external effects (76.5%) (see Figure 46). The most problematic
situations after a major failure were fires and explosions. 7.1% of failures lead to fire while 5.9% of external
effects were related to explosions or bursts.
Because external effects are in most cases associated with significant financial and economic implications, the
originating failure locations for fire and explosion related failures were analyzed, and is shown in Figure 47.
Failures originating in the bushings most often lead to severe consequences.
Figure 48 shows the external effects after a bushing failure occurred. Almost 50% of these failures lead to fire
and explosions.
Collateral
Damages;
Others; 4,88%
1,24%
Fire; 7,16%
Explosion,
Burst; 5,91%
Leakages;
4,25%
None; 76,56%
FIGURE 46: EXTERNAL EFFECTS OF 964 MAJOR FAILURES
Page 97
HV Winding; LV Winding;
9,52% 5,56%
Unknown;
Tapping
23,02%
Winding;
2,38%
HV Lead Exit;
5,56%
Phase to Phase
Tap Changer; Isolation;
13,49% 1,59%
Winding to
Ground
Core and Isolation;
magnetic 0,79%
circuit; 0,79%
LV Bushings; HV Bushings;
MV Bushings; 30,16%
1,59%
5,56%
FIGURE 47: FAILURE LOCATION WHERE FIRE OR EXPLOSION OCCURRED (126 MAJOR FAILURES)
Others; 7,83%
Collateral
Damages;
0,87% None; 45,22%
Fire; 30,43%
Explosion, Leakages;
Burst; 10,43% 5,22%
FIGURE 48: EXTERNAL EFFECTS OF ALL BUSHING FAILURES (115 MAJOR FAILURES)
Page 98
6.9 Action Analysis

The actions taken after a major failure are shown in Figure 49. The majority of the units were repaired either
onsite or in a workshop, and more than a third were scrapped upon failures.
Figure 50 and Figure 51 shows the failure locations of scrapped and repaired transformers, respectively. Due
to their impact, winding failures normally lead to a situation where the failed transformer was scrapped. Tap
changer and bushing failures are normally resulted in repairs.
Onsite Repair < 1

week; 7,37%
Unknown; 13,80%
Onsite Repair > 1
week; 12,97%
Onsite Repair > 1
month; 2,28%
Repair in
Scrapping; 31,74% workshop; 31,85%
FIGURE 49: THE ANALYSIS OF THE ACTIONS TAKEN AFTER 964 MAJOR FAILURES
Tap Changer
Core and 12,40%
magnetic circuit
2,89%
Bushings
10,33%
Insulation
2,48%
Lead Exit
7,02%
Winding
64,88%
FIGURE 50: FAILURE LOCATION ANALYSIS OF 242 SCRAPPED TRANSFORMERS
Page 99
CT
0,43%
Tap Changer Winding

26,24% 42,37%
Cooling unit
1,51%
Tank
0,86%
Flux Shunts
0,65%
Core and
magnetic
circuit
4,09% Lead Exit
Bushings 6,45%
Electrical Screen
14,62% Insulation
0,65%
2,15%
FIGURE 51: FAILURE LOCATION ANALYSIS OF 465 REPAIRED TRANSFORMERS
Page 100
6.10 Comparison with Cigré Survey of 1983
6.10.1 FAILURE RATES

The following factors constrained a comparison between the Cigré 1983 and A2.37 survey failure rates (refer
to Table 34 for calculated failure rates from each survey):
 Definition of failure used by Bossi and the current survey differs. Bossi defined failure as the lack of
performance by a transformer of its required functions so that the unit must be taken out of service to
be repaired. The current survey defined failure as any situation which requires the equipment to be
removed from service for a period longer than 7 days for investigation, remedial work or replacement.
 The Bossi survey considered failures and transformer populations up to 20 years, and manufactured be
up to 1978, only. The current survey had no restriction on age at failure and included all years of
manufacture.
 In order to compare the results of both surveys the age distribution of the population would be required
to determine the population up to 20 years. Unfortunately the age distribution of the transformer
populations was only available for 6 utilities. It was therefore not possible to determine the failure rate
of units up to 20 years, or the failure rate according to age, for the full investigated population.
TABLE 34: FAILURE RATES DETERMINED FROM THE CIGRÉ 1983 AND A2.37 SURVEYS (ALL AGES)
Cigré Survey 1983 Cigré A2.37

APPLICATION &
VOLTAGE GROUPS Unit- Failure Unit- Failure
Failures Failures
Years Rate Years Rate
60 ≤ kV < 100 15868 322 2.00% 15420 144 0.93%
Power Station TRFS 988 15 1.50% 153 0 0.00%
Substation & Auto TRFS 14880 307 2.10% 15267 144 0.94%
100 ≤ kV < 300 24927 496 2.00% 109652 529 0.48%
300 ≤ kV < 700 6234 182 2.90% 38656 277 0.72%
60 ≤ kV < 700 47029 1000 2.10% 163728 950 0.58%
Despite the constraints, an attempt was made to perform a failure rate comparison of both surveys based on
populations up to 20 years. Using the failure and population data of just six utilities from the A2.37 survey,
where the age distribution of the populations was available, the comparison was based on the following:
 The 1983 data consisted of failures and population up to 20 years and manufactured up to 1978;
referred to as pre1978. The recent data of the 6 utilities was therefore grouped into units
manufactured post 1978 and age up to 20 years.
 The failures in the 1983 survey were grouped according to downtime after failure. Three categories
were shown: range of downtime ≤ 1 day, range of downtime 1 to 30 days, and range of downtime ≤
30 days. To minimize the impact of the difference in failure definition between the two surveys, the
Page 101
number of failures with downtime >30 days were estimated and classified as major failure, as per the
A2.37 definition, since this is typically the range of downtime of such failures.
The estimated failure rates of 1983 survey and the 6 utilities from the A2.37 survey are given in Table 35. The
failure rate was less than 1% in both surveys. At 0.4%, the failure rate of the post 1978 units was lower than
pre 1978 units.
TABLE 35: COMPARISON OF FAILURE RATES
POPULATION
SOURCE FAILURES FAILURE RATE
(TRF YEARS)
Pre 1978 : Cigré Survey
47029 386 0.8%
1983
Post 1978 : Cigré A2.37 (6
12805 50 0.4%
utilities)
6.10.2 FAILURES CLASSIFICATION

Using the estimated number of failures of transformers with downtime >30 days from the Bossi 1983 survey,
the failures classification (failure location, causes and modes) of major failures could be estimated and
compared with the A2.37 survey. The data of all transformers manufactured after 1978 and up to 20 years,
of the entire A2. 37 survey was used in the failures classification.
According to email communication with Bossi, all scrapped units were excluded from the failures classification in
Bossi 1983. In order to compare the data the A2.37 data included units up to 20 years, manufactured after
1978 and excluded units that were scrapped.
Bossi 1983 classified transformer application into power station, substation and autotransformers. The Cigré
A2.37 data were classified into distribution and transmission substation transformers. The substation and
autotransformers of the 1983 survey were grouped together as substation transformers, for comparison with
the A2.37 data.
Bossi also confirmed that bushings and winding exit leads were classified under ‘terminals’ in Bossi, 1983.
The comparison of failure locations by transformer application and the combined group of transformers, for the
1983 and A2.37 surveys are given in Figure 52 to Figure 54.
Winding, tap changer and terminal related failures were the major contributors in both surveys for the
combined group of transformers, and substation transformers.
In the A2.37 survey, the contribution of terminal and tap changer related failures were higher, and winding
related failures lower, irrespective of transformer application.
Page 102
Tap changer
Other accessories
Tank and dielectric

fluid
Terminals
Magnetic Circuit
Windings
0% 10% 20% 30% 40% 50% 60% 70%
Bossi 1983 Cigre A2.37
FIGURE 52: COMPARISON OF FAILURE LOCATIONS IN ALL TRANSFORMERS
Tap changer
Other accessories
Tank and dielectric

fluid
Terminals
Magnetic Circuit
Windings
0% 10% 20% 30% 40% 50% 60%
FIGURE 53: COMPARISON OF FAILURE LOCATIONS IN POWER STATION TRANSFORMERS
Page 103
Tap changer
Other accessories
Tank and dielectric fluid
Terminals
Magnetic circuit
Windings
0% 10% 20% 30% 40% 50% 60% 70%

FIGURE 54: COMPARISON OF FAILURE LOCATIONS IN SUBSTATION TRANSFORMERS
The comparison of failure causes by transformer application and the combined group of transformers, for the
1983 and A2.37 surveys are given in Figure 55 to Figure 57.
Failures appeared to be predominantly attributed to design, manufacturing and material related problems, in
both surveys, and irrespective of application.
The contribution of external short circuits and improper repair also featured among the major contributors in the
A2.37 survey for substation transformers and the combined group. Improper repair was not included in the
classification of the 1983 survey.
Failures due to lightning had lower contributions in the A2.37 survey, irrespective of application.
Page 104
Poor Oil Management

Overvoltage
Other Reasons
Improper Repair
Improper Application
Aging
Abnormal Deterioration
External short circuit
Lightning
Overfluxing
Abnormal overload
Incorrect maintenance
Incorrect erection in site
Transport or Storage
Material
Manufacture
Design
0% 5% 10% 15% 20% 25% 30%
FIGURE 55: COMPARISON OF FAILURE CAUSES IN ALL TRANSFORMERS
Overvoltage
Loss of cooling
Lightning
Overfluxing
Abnormal overload
Material
Manufacture
Design
0% 5% 10% 15% 20% 25% 30% 35% 40% 45%
FIGURE 56: COMPARISON OF FAILURE CAUSES IN POWER STATION TRANSFORMERS
Page 105
Poor Oil Management

Improper Repair
Improper Application
Aging
Other reasons
Lightning
Overfluxing
Abnormal overload
Incorrect erection in site
Material
Manufacture
Design
0% 5% 10% 15% 20% 25% 30%
FIGURE 57: COMPARISON OF FAILURE CAUSES IN SUBSTATION TRANSFORMERS
The comparison of failure modes by transformer application and the combined group of transformers, for the
1983 and A2.37 surveys are given in Figure 58, Figure 59 and Figure 60.
Dielectric mode failures were the major contributor in both surveys, irrespective of transformer application.
In power station transformers, the contribution of thermal mode failures was higher in the A2.37 survey,
whereas mechanical mode failures were lower.
Substation transformers also featured mechanical mode failures as a major contributor; contributions were
similar in both surveys.
Electrical
Chemical
Thermal
Dielectric
Mechanical
0% 10% 20% 30% 40% 50% 60%

FIGURE 58: COMPARISON OF FAILURE MODES IN ALL TRANSFORMERS
Page 106
Electrical
Chemical
Thermal
Dielectric
Mechanical
0% 10% 20% 30% 40% 50% 60%

FIGURE 59: COMPARISON OF FAILURE MODES IN POWER STATION TRANSFORMERS
Electrical
Chemical
Thermal
Dielectric
Mechanical
0% 10% 20% 30% 40% 50% 60%

FIGURE 60: COMPARISON OF FAILURE MODES IN SUBSTATION TRANSFORMERS
Page 107
7 CONCLUSION AND RECOMMENDATIONS

To date, the only major international survey on large power transformer failures was published in 1983. Since
then, the results of this survey have become an international benchmark in the transformer industry for the
performance of transformers, in particular the failure rate. In 2008, Cigré Working Group A2.37 Transformer
Reliability was formed with the following objectives:
 Reviewing all existing surveys and study different practices (in terms of data collection, compilation,
etc.)
After the description of developments in transformer design, manufacture and technology the definitions and
guidelines for reliability surveys are explained. Existing transformer reliability surveys were differentiated into
countrywide and company related surveys. The working group reviewed the methodology of surveys from 5
countries and 5 company surveys of different utilities, manufacturers and a consultancy. The main objective of
the national surveys is the systematic collection of data on the availability and disturbances of the electrical
power supply, with emphasis on the frequency, duration and extent of the interruptions. Detailed statistics
about the failure location in the respective equipment, the failure cause or mode and repair activities are
normally not included, therefore limiting its benefit with respect to asset management. In contrast to this, internal
company surveys offer the advantage of a statistic dedicated to the performance of the individual equipment,
where valuable information for asset management can be obtained. Initial working group discussions
concentrated on analysing these readily available statistics, but the different definitions and information content
constrained forming a coherent database from individual sources.
Thus it was agreed that the scope needed to be broadened to allow comparison with the failure statistic of
1983 survey. A questionnaire was therefore developed to collect utility failure statistics in a standardised way.
Besides information about the population under investigation, failure data was collected for various groups of
transformers in terms of the failure locations, failure causes, failure modes, actions, external effects and others
parameters. In this brochure the data collection methodology developed by Working Group A2.37 is
described, and an uniform way of collecting, compiling and presenting data is proposed.
An Excel spreadsheet tool consisting of two major sections (population data, failure data) was developed to
collect data in a standardized way and in accordance with the definition of major failure. A major failure was
defined as any situation which required the transformer to be removed from service for a period longer than 7
days for investigation, remedial work or replacement. A major failure would require at least the opening of the
transformer or the tap changer tank, or an exchange of the bushings. Pull-down menus for each field in the
Excel spreadsheet suggest possible answers and ensure that the answers are given in a standardized way.
The working group has collected 964 major failures which occurred in the period 1996 to 2010, within a total
manufacture of the units span from the 1950’s up to 2009.
The overall failure rate of substation, GSU and the combined group of transformers were all within 1%. Only
GSU units in the voltage class 300kV to 500kV exceeded 1% failure rate. Although differences in failure rate
were very low, less than 0.5%, the failure rates of GSU units were higher than substation transformers in nearly
all the voltage classes.
For three groups of substation transformers detailed population data were collected enabling the calculation of
hazard curves. All populations show a low hazard rate and no distinct bathtub curve character. Within two
populations just one failure each occurred in the first 20 years respectively. Both populations showed an
Page 108
increase of the hazard rate at an age of 30, 40 respectively from almost 0% to a level of app. 1% (5 year
moving average). This level does not further increase until the highest ages. An increasing probability of failure
after a particular age, which would justify an exchange of the transformer cannot be derived from the
available data. A more likely cause of unreliability is probably damage caused by unusual system events.
Replacement strategies, in which preferably old transformers are replaced, have a biasing effect on the failure
statistics as transformers are not left in service to fail. A possible approach is to estimate on the number of
years a replaced transformer would have remained in service if it would not be replaced. So these statistics
cannot be used directly for lifetime modelling, e. g. Weibull fitting. Because the hazard curve does not show a
distinct increase with time the use of Time Based Maintenance will not be effective for substationr transformers.
Therefore maintenance should be planned according to the actual condition.
Winding related failures appeared to be the largest contributor of major failures, irrespective of transformer
application or manufacturing period, and due to their impact typically led to a situation where the failed
transformer was scrapped. GSU transformers had higher contributions of winding and lead exit failures than
substation transformers. Substation transformers on the other hand had higher contributions of tap changer
related failures than GSU transformers. The contribution of bushing related failures were similar in both
transformer applications. Failures originating in the bushings most often lead to severe consequences such as
fires and explosions.
Dielectric mode failures were among the most prominent, irrespective transformer application and voltage
class. Substation transformers were also more prone to mechanical type failures, and GSU transformers to
thermal mode failures. This can be explained by the normally higher loading of GSUs compared to substation
transformers.
Design and manufacturing, aging and external short circuits appeared to be major contributors. However,
determining the root cause of failure can be a difficult and extensive task. Failure can often be attributed to
different failure causes, and in some cases the causes are unknown. This result should therefore be treated and
interpreted with caution.
The majority of the units were repaired either onsite or in a workshop, and more than a third were scrapped
upon failure. Winding failures mostly lead to the transformer being scrapped upon failure, whereas tap
changer and bushing failures mostly resulted in repairs.
A comparison with the Bossi 1983 survey [2] was constrained by three factors: different failure definitions, age
restriction of the failures, and the availability of the age distribution. Despite the constraints, an attempt was
made to perform a failure rate and failures classification comparison. The failure rate was less than 1% in both
surveys. At 0.4%, the failure rate of the post 1978 units was lower than pre 1978 units. The failures
classification comparison appeared to suggest similarities in terms of the major contributors of failure in both
surveys. Winding, tap changer and terminal related failures were the major contributors in both surveys, for the
combined group of transformers, and substation transformers. Failures appeared to be predominantly
attributed to design, manufacturing and material related problems, in both surveys, and irrespective of
application. Dielectric mode failures were the major contributor in both surveys, irrespective of transformer
application.
Collecting, Compiling and Presenting Data in a Standardized Way

In order to simplify future surveys the working group recommends using the developed Excel-spreadsheet for
the collection of failure data [68]. Thus statistical data can be collected in a simple and standardized which
allows benchmarking between different transformer populations. The individual recording of the failure should
be done in the Excel-sheet directly after failure cause analysis.
Page 109
REFERENCES
[1] CIGRE WG A3.06, „Final Report of the 2004 - 2007 International Enquiry on Reliability of High
Voltage Equipment,“ Brochure 509, Paris, 2012.
[2] A. Bossi, J. Dind, J. Frisson, U. Khoudiakov, H. Light and e. al, "An international survey on failures in
large power transformers in service," Electra, no. 88, pp. 21-48, 1983.
[3] L. Chmura, Life-cycle assessment of high-voltage assets using statistical tools, PhD thesis
Technical University Delft: ISBN 978-94-6182-396-0, 2014.
[4] D. Allan, „Power transformers - the second century,“ Power Engineering Journal, pp. 5-14, January
1991.
[5] M. Heathcote, The J&P transformer book - A practical technology of the power transformer, 12th
Edition: Reed Educational and Professional Publishing Ltd, 1998.
[6] R. Baehr, „Transformer technology state-of-the-art and trends of future development,“ Electra, pp.
13-19, October 2001.
[7] S. Kulkarni und S. Khaparde, Transformer engineering – design and practice, New York: Marcel-
Dekker, 2004.
[8] A. Franklin and D. Franklin, The J&P transformer book - A practical technology of the power
transformer, 11th Edition, Butterworth & Co. Ltd, 1983.
[9] D. Seay, „Transformer fleet management practices,“ in Proceedings of the 73rd Annual International
Conference of Doble Clients, Boston, 2006.
[10] S. Metha and H. Jin Sin, "Power transformers technology review and assessments," in Status
Report of Cigré SC A2 Advisory Group – Transformer Technology, Paris, 2006.
[11] A. Krämer, On-load Tap Changers for Power Transformers, Maschinenfabrik Reinhausen, 2000.
[12] ABB, „ABB Components, 2008: On-load tap-changers, type UC and VUC Technical Guide,“ pp.
Document No. 1ZSE 5492-105 en, Rev. 7, 2008.
[13] Maschinenfabrik Reinhausen, „FirstVACUTAP® VRC 700 delivered,“ [Online]. Available:

http://www.reinhausen.com/desktopdefault.aspx/tabid-195/573_read-520.
[14] D. Dohnal and B. Kurth, "Vacuum switching, a well proven technology has found its way into
resistance-type load tap changers," in Transmission and Distribution Conference and Exposition,
Atlanta, GA, 2001.
[15] L. Jonsson, „Reliability and service characteristics on ABB tap changers and bushings,“ Trafotech
Seventh International Conference on Transformers, India, January, 2006.
[16] ABB, „ABB Components, 2005: Silicone rubber – Product information,“ 2005.
[17] A. Mikulecky, „International Colloquium Transformer Research and Asset Management,“ Cavtat,
Croatia, 2009.
Page 110
[18] U. Krüsi, D. Egger, A. Dais and J. Czyzewski, "New paper-free insulation technology for dry high-
voltage condenser bushings," in 22nd International Conference on Electricity Distribution,
Stockholm, 2013.
[19] S. Tenbohlen, D. Uhde, J. Poittevin, U. Sundermann, P. Werle, H. Borsi and H. Matthes, "Enhanced
Diagnosis of Power Transformers using On- and Off-line Methods: Results, Examples and Future
Trends," in CIGRE Session, A2-204, Paris, 2000.
[20] P. Jarman, B.-O. Stenestam, I. Hunter, A. Garnitschnig, E. Mortensen, K. Viereck, R. Kuppuswamy,

S. Tenbohlen, S. Meijer, A. Wilson, M. Chauke, M. Judd and J. Vancotthem, "Recommendations for
Condition Monitoring and Condition Assessment Facilities for Transformers," CIGRE, Brochure 343,
WG A2.27, Paris, 2008.
[21] M. Rösner, J. Weenen and X. Hao, "Benefits of Comprehensive and Interactive online Monitoring
and Expert Systems for Power Transformers," in CMD, Beijing, China, 2008.
[22] S. Tenbohlen, T. Stirl, G. Bastos, J. Baldauf, P. Mayer, M. Stach, B. Breitenbauch and R. Huber,
"Experience-based Evaluation of Economic Benefits of On-line Monitoring Systems for Power
Transformers," in Cigre Session, 12-110, Paris, 2002.
[23] T. Stirl, R. Skrzypek, S. Tenbohlen and R. Vilaithong, "On-line Condition Monitoring and Diagnosis
for Power Transformers their Bushings, TapChanger and Insulation System," in International
Conference on Condition Monitoring and Diagnosis CMD, Changwon, Korea, 2006.
[24] P. Picher and e. al., "Mechanical Condition Assessment of Transformer Windings using Frequency
Response Analysis (FRA)," Cigre, Brochure 342, WG A2.26, Paris, 2008.
[25] S. Gubanski and e. al., "Dielectric Response Diagnoses for Transformer Windings," Brochure 414,
WG D1.01 (TF 14), Paris, 2010.
[26] D. Allan and T. Harrison, "The argument for transformer replacement," in Proceedings of the IEE
Conference, No.273, 1987.
[27] R. Bruce, H. Royal and J. Price, "Factors influencing the design of large high-voltage power
transformers," Transactions of the South African Institute of Electrical Engineers, no. 59, pp. 1-27,
January 1968.
[28] V. Sokolov, „Reliability problems with large power transformers and shunt reactors – typical failure
modes and failure causes,“ in Cigré SC A2 (Transformers) Colloquium, Moscow, 2005.
[29] V. Kogan, J. Fleeman, J. Provanzana and C. Shih, "Failure analysis of EHV transformers," IEEE
Transactions on Power Delivery, Vol. 3, No2, April 1988.
[30] M. Horning, J. Kelly and S. Myers, "Transformer maintenance guide," in 2nd Edition, Transformer
Maintenance Institute, Division of S.D. Meyers, 2001.
[31] V. Sokolov and B. Vanin, "Evaluation and identification of typical defects and failure modes in 110kV
-750kV bushings," in Proceedings of the 64th Annual International Conference of Doble Clients,
Boston, 1997.
[32] Doble Engineering, „Transformer technology - Design group impact on failure,“ in Eskom/Doble
Seminar Presentation 4.2, August, 2006.
Page 111
[33] IEEE, „IEEE C57.117-1986: IEEE guide on reporting failure data for power transformers and shunt
reactors on electric utility power systems, IEEE,“ 1986.
[34] J. Lapworth, „Transformer reliability surveys,“ in Cigré Session, A2-114, Paris, 2006.
[35] CIGRE WG A2.18, „Life management techniques for power transformer,“ Brochure 227, Paris, 2003.
[36] IEC, „IEC 60050-191 International electrotechnical vocabulary Chapter 191 -Dependability and
quality of service, IEC Standard,“ 1990.
[37] T. Higgins, „The difference online condition monitoring could have made,“ in Cigré SC12
(Transformers) Colloquium, June, Dublin, 2001.
[38] W. Nelson, Applied Life Data Analysis, Hoboken, NJ, USA: John Wiley & Sons, Inc., 1982.
[39] P. Tobias and D. Trinidade, "Applied reliability," Van Nostrand Reinhold Company, 1986.
[40] J. Elerath, „AFR: Problems of definition, calculation and measurement in a commercial

environment,“ in Proceedings of the Annual Reliability and Maintainability Symposium, 2000.
[41] A. Kachler, „Unique transformer failure statistics - An important contribution to economics of

transformer management,“ SC12 (Transformers) Cigré Colloquium, Dublin, June, 2001.
[42] V. Sokolov, S. Tsurpal and A. Drobyshevski, "Reliability problems with large power transformers and
shunt reactors - typical failure modes and failure causes," in Cigré SC A2 (Transformers)
Colloquium, Moscow, 2005.
[43] V. Sokolov, "Transformer reliability - typical failure causes," in TechCon® Asia Pacific, 2005.
[44] J. Fleeman and L. Wagenaar, "Transformer specifications that works - exceeding standards," in
Presentation by American Electric Power to IEEE/PES Transformers committee, October, 2002.
[45] A. Hall, „Engineered reliability - UK experience of 400kV 800MVA generator transformers,“ in Cigré
Session, Paris, 2006.
[46] L. Wagenaar, J. Schneider and J. Fleeman, "EHV transformer dielectric specification

improvements," IEEE Transactions on power delivery, Vols. 9, No. 1, January 1994.
[47] P. Prout, A. Wilson and M. Locarno, "Factors affecting the lifetime of distribution transformers," in
Proceedings of the 70th Annual International Conference of Doble Clients, Boston, 2003.
[48] M. Foata, C. Rajotte and A. Jolicoeur, "On-load tap changer reliability and maintenance strategy," in
Cigré Session, A2-102, Paris, 2006-1.
[49] P. Austin, "Transformer economic issues in Australia & New Zealand," in Cigré SC 12
(Transformers) Colloquium, Dublin, June 2001.
[50] R. Jongen, P. Morshuis, J. Smit, A. Janssen and E. Gulski, "Failure data of power transformers as
input for statistical analysis, T6-207," XVth International Symposium on High Voltage Engineering,
Slovenia, August, 2007.
[51] D. Allan and A. White, "Transformer design for high reliability," in IEEE Second international
Page 112
conference on the reliability of transmission and distribution equipment, March 1995.
[52] J. Coetzee, Maintenance, Victoria, BC, Canada: Trafford Publishing, 2004.
[53] Wikipedia, „Bathtub curve,“ Feb. 2014. [Online]. Available: en.wikipedia.org/wiki/Bathtub_curve.
[54] H. Ascher und H. Feingold, Repairable Systems Reliability, University of California, 1984.
[55] Canadian Electricity Association, „CEA - Analytics Reports,“ December 2014. [Online]. Available:
http://www.electricity.ca/resources/analytics/analytics-reports.php.
[56] Canadian Electricity Association, „Forced Outage Performance of Transmission Equipment,“

Montreal, Canada, 2003.
[57] Canadian Electricity Association, „Forced Outage Performance of Transmission Equipment,“ CEA,
Canada, 2009.
[58] FNN, „Versorgungszuverlässigkeit,“ [Online]. Available:

www.vde.com/de/fnn/arbeitsgebiete/versorgungsqualitaet/Seiten/versorgungszuverlaessigkeit.aspx.
[Zugriff am December 2014].
[59] VDN, „Störungs- und Verfügbarkeitsstatistik,“ Verband der Netzbetreiber VDN, Berlin, 2004.
[60] „UNIPEDE: Availability of Supply Indices,“ Distribution Study Committee, Group of Experts 50.05
DISQUAL, Paris, 1997.
[61] Electric Technology Research Association, „Advanced and Efficient Technology for Substation
Equipment Maintenance,“ Vol.50, No.2, Japan, October, 1994.
[62] Electric Technology Research Association, "Rationalized Technology for Substation Equipment
Maintenance," Vol.56, No.2, Japan, 2000.
[63] G. Bastos, M. A. Marin and e. al., "AVALIAÇÃO DO DESEMPENHO DE TRANSFORMADORES DE

POTÊNCIA E REATORES NO SISTEMA ELÉTRICO BRASILEIRO," Cigre Brazil, GTA2-02, 2014.
[64] J. Jagers, L. Geldenhuis and C. Gaunt, "Large Power Transformer Reliability Improvement in
ESKOM Distribution," in 19th International Conference on Electricity Distribution CIRED, Vienna,
2007.
[65] J. Jagers, J. Khosa, P. DeKlerk and C. Gaunt, "Transformer Reliability and Condition Assessment in
a South African Utility," in 15th International Symposium on High Voltage Engineering, Proceedings
of the ISH, Ljubljana, Slovenia, 2007.
[66] J. Jagers and S. Tenbohlen, "Evaluation of Transformer Reliability Data Based on National and
Utility Statistics," in 16th International Symposium on High Voltage Engineering, Proceedings of the
ISH, Cape Town, South Africa, 2009.
[67] B. Desai and M. Lebow, "Needed: ASAP Approach," Power and Energy Magazine, IEEE, vol.8,
no.6, pp. 53-60, November 2010.
[68] CIGRE WG A2.37, “Questionnaire "Transformer Reliability Survey",” July 2015. [Online]. Available:
http://a2.cigre.org/content/download/64536/3071663/version/1/file/CIGRE_A2-
Page 113
37_Reliability_Questionnaire_x20140814x.xlsx. [Accessed 1 December 2015].
[69] IEEE, „IEEE C57.125: IEEE guide for failure investigation, documentation, and analysis for power
transformers and shunt reactors, IEEE,“ 1991.
[70] CIGRE WG A3.06, „Final Report of the 2004 - 2007 International Enquiry on Reliability of High
Voltage Equipment,“ Brochure 509, Paris, 2012.
[71] CIGRE WG B4.04, „Protocol for Reporting the Operational Peformance of HVDC Transmission
System,“ Brochure 590, Paris, 2014.
[72] F. Vahidi and S. Tenbohlen, "Statistical Failure Analysis of European Substation Transformers," in 6.
ETG-Fachtagung Diagnostik Elektrischer Betriebsmittel, Berlin, 2014.
[73] P. Jarman, R. Hooton, L. Walker, Q. Zhong, T. Ishak und Z. Wang, „Transformer life prediction using
data from units removed from service and thermal modelling,“ in Cigre-Session, A2-212, Paris,
2010.
[74] P. Picher, J. Boudreau, A. Manga, C. Rajotte, C. Tardif, G. Bizier, N. d. Gaetano, D. Garon, B.

Girard, J. Hamel und S. Proulx, „Use of Health Index and Reliability Data for Transformer Condition
Assessment and Fleet Ranking,“ in Paper A2-101, Cigre Session, Paris, 2014.
[75] A. Singh, „Power Transformers in New Zealand: A Review of Current Fleet Age, Asset Management
Strategies and Procurement Practices,“ in EEA Conference & Exhibition, Auckland, 2014.
Page 114
ANNEXES
ANNEX A: Questionnaire
FIGURE 61: SHEET 1 OF QUESTIONNAIRE FOR THE COLLECTION OF POPULATION DATA
Page 115
FIGURE 62: SHEET 2 OF QUESTIONNAIRE FOR THE COLLECTION OF FAILURE DATA
Page 116
ANNEX B: Definitions from Surveys and International Standards and Guidelines

SOURCE TYPE DEFINITION
The termination of the ability of an item to perform a

[IEC, 1990] Failure
required function. [36]
Lack of performance by a transformer of its required

[Bossi, 1983] Failure functions so that the unit must be taken out of service
to be repaired. [2]
The termination of the ability of a transformer to

[IEEE, 1986] Failure
perform its specific function. [33]
[Cigré WG A2.18, Any situation which requires the equipment to be

Failure
2003] removed from service for investigation. [35]
Any forced outage of a transformer due to its failure

in service. Trouble which requires transformer to be
returned to a factory for repair, or which requires
[Kogan, 1988] Failure
extensive field repair. Transportation damage and
minor troubles which may require an equipment
outage are not considered as failures. [29]
Can be repaired quickly in situ. The resulting outage

[Higgins, 2001] Minor Failure
typically would be less than one month. [37]
Must be repaired (if this is possible) off-site, usually at

manufacturers' works. The resulting outage typically
[Higgins, 2001] Major Failure would be measured in months. Failures in which the
transformer is destroyed or must be retired are also
major failures. [37]
Failure with
Failure for which the transformer can deliberately be
[Bossi, 1983] scheduled
taken out of service at a selected time. [2]
outage
Failure with Failure for which the transformer must be taken out of
[Bossi, 1983]
forced outage service immediately (within 30 minutes). [2]
Failure of a transformer that requires its immediate

Failure with removal from service. This is accomplished
[IEEE, 1986]
forced outage automatically or as soon as switching operations can
be performed. [33]
Page 117
SOURCE TYPE DEFINITION
Failure with
Failure for which a transformer must be taken
[IEEE, 1986] scheduled
deliberately out of service at a selected time. [33]
outage
Failure of an equipment that requires its immediate

removal from the system. This is accomplished either
[Cigré, WG A2.18, Failure with
automatically by the operation of protection
2003] forced outage
equipment or as soon as switching operations can be
performed. [35]
Failure for which the removal of the equipment from

Failure with
[Cigré, WG A2.18, service can be deferred to some more convenient
scheduled
2003] occasion, but still requires a change to planned outage
outage
programme. [35]
The state of an item characterised by inability to

perform a required function, excluding the inability
[IEC, 1990] Fault
during preventive maintenance or other planned
actions, or due to the lack of external sources. [36]
[Cigré, WG A2.18,
Fault Any deterioration beyond normal wear or aging. [35]
2003]
Imperfection or partial lack of performance that can

[IEEE, 1986] Defect be corrected without taking the transformer out of
service. [33]
Imperfection in the state of a transformer or partial

lack of performance that may be eliminated without
[Bossi, 1983] Defect
taking the transformer out of service, or on the occasion
of ordinary and detailed servicing. [2]
[Cigré, WG A2.18, Any non conformance to normal condition requiring

Defect
2003] some investigative or remedial action. [35]
The ratio of the number of transformers that failed

[Bossi, 1983] Failure Rate over a certain number of years, to the number of
transformers in service during those years. [2]
The ratio of the number ‘failures with forced outages’

of a given population over a given period of time, to
[IEEE, 1986] Failure Rate the number of accumulated service years for all
transformers in that population over the same period
of time. [33]
The ratio of the number of failures or number of units

failed from N units in service, to the number of service
[Kachler, 2001] Failure Rate
years accumulated with N units in service, where N is
the number of units in service in a 10 year period. [41]
Page 118
ANNEX C: Description of Analysis of Transformer Population Data

The age distribution of population data and failures are required for calculation of hazard rates:
1 ∆n(t )
h(t ) = ⋅
N (t ) ∆t

In this investigation the hazard rate has to be calculated using failures documented in the period between
2000 and 2010. So, the operational experience of transformers must be calculated regarding this period, for
instance a transformer installed in 2005 can just contribute an operational experience with age of 0, 1, 2 , 3 ,4
and 5 in the defined period 2000-2010. The procedure for calculation of hazard rate is introduced by the
following example:
In table A the number of acquired transformers is given for each year. Table B shows the number of failures
dependent on age of the transformer which occurred in the interval between 2000 and 2010 (failure interval).
TABLE A: ACQUISITION YEAR TABLE B: NUMBER OF FAILURES DEPENDENT ON AGE
Acquisition year Number of Transformers Age Number of Failures

2010 a1 1 n1
2009 a2 2 n2
2008 a3 3 n3
2007 a4 4 n4
2006 a5 5 n5
2005 a6 6 n6
2004 a7 7 n7
2003 a8 8 n8
2002 a9 9 n9
2001 a10 10 n10
2000 a11 11 n11
1999 a12 12 n12
1998 a13 13 n13
1997 a14 14 n14
1996 a15 15 n15
1995 a16 16 n16
1994 a17 17 n17
1993 a18 18 n18
1992 a19 19 n19
1991 a20 20 n20
1990 a21 21 n21
The hazard value for the first operation year is calculated by division of the number of first year failures n1 by
the number of transformers which had an operational experience in the first year in the given failure interval
N(1). So, just the transformers acquired from 2000 to 2010 can have a first year operation experience in the
investigated period T (2000-2010). The values of N(1) and n(1) will be as follows:
N(1)=a1+a2+…....+a11, n(1)=n1,
which leads to a hazard rate in the first year of : h(1) = n1 / N(1)

In the same manner, the hazard values for operation years bigger than 1 can be calculated using the given
values in following table:
Page 119
e
Number of transformers with t years
Number of failures at age t, f(t)
operational experience, N(t)
N(1)=a1+a2+…....+a11 n(1)=n1
N(2)=a2+a3+…....+a12 n(2)=n2
N(3)=a3+a4+…….+a13 n(3)=n3
N(4)=a4+a5+…….+a14 n(4)=n4
N(5)=a5+a6+…….+a15 n(5)=n5
N(6)=a6+a7+…….+a16 n(6)=n6
N(7)=a7+a8+…...+a17 n(7)=n7
N(8)=a8+a9+…...+a18 n(8)=n8
N(9)=a9+a10+….+a19 n(9)=n9
N(10)=a10+a11+...+a20 n(10)=n10
N(11)=a11+a12+...+a21 n(11)=n11
N(12)=a12+a13+…+a21 n(12)=n12
N(13)=a13+a14+...+a21 n(13)=n13
N(14)=a14+a15+...+a21 n(14)=n14
N(15)=a15+a16+...+a21 n(15)=n15
N(16)=a16+a17+...+a21 n(16)=n16
N(17)=a17+a18+...+a21 n(17)=n18
N(18)=a18+a19+…+a21 n(18)=n18
N(19)=a19+a20+a21 n(19)=n19
N(20)=a20+a21 n(20)=n20
N(21)=a21 n(21)=n21
Page 120

642

Uploaded by

Copyright:

Available Formats

642

Uploaded by

Document Information

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

642

Uploaded by

Copyright:

Available Formats

642

TRANSFORMER RELIABILITY SURVEY

S. Tenbohlen, Convenor (DE), J. Jagers, Secretary (ZA),

Transformer Reliability Survey

3.5.2 Basic Renewal Theory .......................................................................................................................................... 30

FIGURE 1: OVERVIEW OF CORE STEEL DEVELOPMENT [6] ........................................................................................................................... 13

TABLE 1: FAILURE RATES FROM SURVEYS FROM 1968 TO 2005 ............................................................................................................... 27

HIGHEST SYSTEM VOLTAGE [kV]

Transformer-Years 15,267 64,718 37,017 25,305 4,774 2,991 150,072

Major Failures 144 280 186 152 27 10 799

FAILURE RATE 0.94% 0.43% 0.50% 0.60% 0.57% 0.33% 0.53%

INVESTIGATED POPULATION AND FAILURE RATES OF GENERATOR STEP-UP TRANSFORMERS

HIGHEST SYSTEM VOLTAGE [kV]

Transformer-Years 153 3,278 4,639 6,740 1,837 740 17,387

Major Failures 0 20 43 89 9 4 165

FAILURE RATE 0.00% 0.61% 0.93% 1.32% 0.49% 0.54% 0.95%

1.2 Scope and Aim

2 DEVELOPMENTS IN TRANSFORMER TECHNOLOGY, DESIGN AND

2.1.1 CORE MATERIAL (MAGNETIC CIRCUIT)

FIGURE 1: OVERVIEW OF CORE STEEL DEVELOPMENT [6]

2.1.2 CONDUCTOR MATERIAL (WINDINGS)

FIGURE 2: OVERVIEW OF INCREASE IN TRANSFORMER RATED VOLTAGE AND POWER [6]

2.1.3 INSULATION MATERIALS

2.1.4 TAP CHANGERS

2.1.6 CONDITION MONITORING

2.2 Developments in Design Tools

2.3 Developments in Manufacturing

2.3.1 CORE CUTTING AND ANNEALING

2.3.2 WINDING CONSTRUCTION AND CONDUCTOR MATERIAL

2.3.3 DRYING OF THE ACTIVE PART

2.3.4 CLAMPING STRUCTURE

2.4 Relationships between Reliability and Manufacturing Period

2.4.2 MAGNETIC CIRCUIT/ CORE RELIABILITY

2.4.3 IMPACT OF DESIGN TECHNIQUES

2.4.4 TAP CHANGER RELIABILITY

2.4.5 BUSHING RELIABILITY

2.4.6 COOLING SYSTEM

3 DEFINITIONS AND GUIDELINES FOR RELIABILITY SURVEYS

3.1 Guidelines on Reliability Practices

3.2 Reliability Surveys

3.3.3 FAILURE RATE

ni = Number of failures in i-th year

ni = Number of failures by i-th population

3.4 Review of Available Reliability Statistics

3.4.2 FAILURE RATES

TABLE 1: FAILURE RATES FROM SURVEYS FROM 1968 TO 2005

Cigré International All voltage levels

Generator Step-Up &

US-NGRID, United Distribution 115kV,

Hydro Quebec, All voltage categories,

Australia & New

* Indicates studies where the manufacturing period was given.

Failure Rate Dependent on Age

3.4.3 FAILURE LOCATION

3.4.4 FAILURE MODE

3.4.5 FAILURE CAUSES

3.5 Theory of Statistical Analysis

3.5.1 TYPES OF STATISTICAL DATA

FIGURE 3:CENSORING OF STATISTICAL DATA

3.5.2 BASIC RENEWAL THEORY