The Health Index represents a practical tool that combines the results of operating observations, field inspections, and site and laboratory testing to manage the asset and prioritize investments in capital and maintenance plans.
IntroductionPower transformers have the single highest value of the
equipment installed in high-voltage substations, comprising up to 60% of total investment. An increasing demand for improved financial and technical performance has pushed most power utili-ties to assess the actual condition of their transformers [1]–[3]. To achieve the optimal balance among capital investments, asset maintenance costs, and operating performance, there is a need to provide economic and technical justifications for engineering decisions and capital replacement plans.
The Health Index (HI) represents a practical tool that combines the results of operating observations, field inspections, and site and laboratory testing into an objective and quantitative index, providing the overall health of the asset. Asset HI is a powerful tool for managing assets and identifying investment needs as well as prioritizing investments in capital and maintenance programs [4], [5]. The objective of this paper is to present a condition-based asset management tool that quantifies power transformer degrada-tion and allows for a recommendation regarding the number of power transformers that would likely require replacement within future time horizons. A capital plan for replacement of power transformers is also presented.
Several studies have examined different power transformer condition assessment and life-management techniques. These techniques include measuring or monitoring of dissolved gas, oil or conductor temperature, moisture, oil quality (dielectric strength, acidity, color, and interfacial tension), and partial discharge, as well as frequency response analysis, recovery voltage method, thermal imaging, tap changer tests, and bushing tests [1]–[8]. Such tests are conducted on a routine or condition basis to evaluate the condition of power transformers. However, no method is avail-able to quantify the condition of the asset through combining all available data. This paper describes a practical asset HI calculation method that combines the impact of all available data and also utilizes criteria based on the industry’s common practices. IEC, IEEE, and CIGRE recommendations and Kinectrics’ experience with different utilities are considered in developing the scoring and ranking methods.
Readily Available DataTest plans differ from one utility to the other; however, the
common, well-accepted tests that are useful as diagnostic meth-ods include:
and other accessories • Infraredthermography• Dissolvedgasanalysis(DGA)• Oilqualitytests• Dissipation factor (tanδ) and capacitance measure-
ment
Ali Naderian Jahromi, Ray Piercy, Stephen Cress, Jim R. R. Service, and Wang FanKinectrics Inc.,Transmission and Distribution Technologies,Toronto, ON, Canada
An Approach to Power Transformer Asset Management Using Health Index Key Words: Power transformer, health index, failure, life, age, capital plan
March/April 2009 — Vol. 25, No. 2 21
• Tapchangertestsandinspection(oiltests,contactre-sistance, motor current, and others)
Other than these tests, some useful data are usually available to evaluate the long-term condition of power transformers, such as load history and maintenance data.
Why Health Index?Manufacturers often define the anticipated life of power trans-
formers to be 25 to 40 years. However, some transformers in service are now approaching this age, and a number have reached 60 years old. Nonetheless, failure rates remain low, and there is little evidence that many are at, or near, the end-of-life. In the past, different concepts related to transformer life management have been introduced such as:
In most of the analyses that have been done using these con-cepts, there is an attempt to model the insulation life, mainly the paper insulation. Temperature and DGA are the key factors in this modeling, and other valuable data such as routine test results, maintenance data, and the previous history of the transformer are usually neglected. The purpose of asset condition assessment in this work is to detect and quantify a long-term degradation and to provide a means of quantifying the remaining asset life. This as-sessment includes identifying assets that are at or near to the end-of-life and assets that are at high risk of generalized failure that will require major capital expenditures to replace the assets.
A composite HI is a very useful tool for representing the overall health of a complex asset. HI quantifies equipment condition based on numerous condition criteria that are related to the long-term degradation factors that cumulatively lead to an asset’s end-of-life. HI results differ from maintenance testing or condition-based diagnostics, which emphasize finding defects and deficiencies that need correction or remediation to keep the asset operating during some time period.
Parameters in Health Index FormulationTo assess the overall condition of a power transformer, it is
necessary to include as much data as is available and suitable for a realistic assessment. The development of a condition-based HI
requires an assessment of the relative degree of importance of the different condition factors in determining the health of the asset. Each condition factor is discussed below.
Dissolved Gas AnalysisTheoretically, by means of DGA, it is possible to distinguish
internal faults such as arcing, partial discharge, low-energy spark-ing, severe overloading, and overheating in the insulation system. IEC 60599 [9] provides a coded list of faults detectable by dis-solved gas analysis, and IEEE Standard C57.104 [10] introduces a four-level criterion to classify risks to transformers for continued operation at various combustible gas levels [1]. Practically, DGA data by itself does not always provide sufficient information from which to evaluate the integrity of a transformer. Normal opera-tion will also result in the formation of some gases. Information about the history of a transformer (maintenance, loading practice, previous faults, manufacturer data, and so on) is an integral part of the information required to make an evaluation. In fact, it is possible for some transformers to operate throughout their use-ful life with substantial quantities of combustible gases present. Figure 1 compares the recommended alarm level of hydrocarbon gases from different references including IEEE, IEC, Dornenburg, and Bureau of Reclamation [1], [2]. One distinct difference is the IEEE limit for acetylene, which is an order of magnitude larger than the limit recommended by IEC and others. Moreover, IEEE has the most conservative level for carbon monoxide (350 ppm), which is half of the value recommended by IEC.
Considering the different recommendations, Figure 2 and Table 1 introduce a ranking method developed using the DGA data. The DGA factor is
DGAF =´
=
=
å
å
S W
W
i ii
ii
1
7
1
7
(1)
where Si = 1, 2, 3, 4, 5, or 6, and W
i is the assigned weighting
factor. An initial value for Wi is allocated to be equal to 1 for both
CO and CO2; 3 for CH
4, C
2H
6, and C
2H
4; 5 for C
2H
2; and 2 for H
2.
Figure 1. Comparison of recommended hydrocarbon limits.
22 IEEE Electrical Insulation Magazine
Si is the score of each gas based on Figure 2. If any gas content
exceeds the last limit introduced in Figure 2, the assigned score is 6. For example, if CH
4 is
more than 600 ppm, its score is 6. The
weighting factors can be adjusted according to the current practice of a utility. The rating code starts with A as the best condition to E, which represents the worst situation. This type of coding is employed for the remaining factors. This scoring system is not recommended for use as a diagnostics tool. It tries to give an overall figure of DGA results in a long-term time frame.
Figure 2. Scoring dissolved gases.
Table 1. Transformer Rating Based on DGA Factor.
Rating Code Condition Description
A GoodDAF < 1.2 [AU8: I would left align the last column on DGAF ]
B Acceptable 1.2 ≤ DGAF < 1.5
C Need Caution 1.5 ≤ DGAF < 2
D Poor 2 ≤ DGAF < 3
E Very poor DGAF ≥ 3
Table 2. Standards for Oil Quality Tests.
ParameterASTM recommended by
IEEE[1,10]IEC recommended by
CIGRE[1,11]
Dielectric Breakdown D877, D1816 IEC60156
Water content D1533 IEC 60814
Power Factor D924 IEC247
IFT D971 ISO 6295
Acidity D644, D974 IEC62021
Color D1500 ISO 2049
Table 3. Grading Method for Oil Test Parameters Based on IEEE C57.106-2006.
U ≤ 69 kV69 kV < U< 230 kV 230 kV ≤ U
Score (Si) Wi
Dielectric Strength kV (2 mm gap)
≥45 ≥52 ≥60 1
335–45 47–52 50–60 2
30–35 35–47 40–50 3
≤30 ≤35 ≤40 4
IFTdyne/cm
≥25 ≥30 ≥32 1
220–25 23–30 25–32 2
15–20 18–23 20–25 3
≤15 ≤18 ≤20 4
Acid Number
≤0.05 ≤0.04 ≤0.03 1
1.05–0.1 0.04–1.0 0.03–.07 2
0.1–0.2 1.0–0.15 0.07–.10 3
≥0.2 ≥0.15 ≥0.10 4
Water content(ppm)
≤30 ≤20 ≤15 1
430–35 20–25 15–20 2
35–40 25–30 20–25 3
≥40 ≥30 ≥25 4
Color
≤1.5 1
21.5–2.0 2
2.0–2.5 3
≥2.5 4
Dissipation factor (%)25 oC
≤0.1 1
30.1–0.5 2
0.5–1.0 3
≥1.0 4
March/April 2009 — Vol. 25, No. 2 23
Several classic techniques have been developed for DGA interpretation of power transformers in the past 30 years such as Rogers, Dornenburg, Duval Triangle, and modified Dornenburg [1], [2], [11]. Most of these methods are based on the gas ratio such as C
2H
2/C
2H
4, CH
4/H
2, and C
2H
4/C
2H
6. However, all are
applicable for detecting the fault type rather than determining if a transformer works normally or not, and they are not employed for a long-term investigation. Instead, the rate of gas production is a better analysis to be employed [1], [4]. A reduction of the HI is recommended if the rate of gas increment is more than 30% for three consecutive gas samples, or 20% for five consecutive oil samples.
Oil QualityTable 2 summarizes some of recommended oil test standards
based on the ASTM standard suggested by IEEE and the IEC standard recommended by CIGRE [1], [12], [13]. A combination of electrical, physical, and chemical tests is performed to establish preventive maintenance procedures, avoid premature failure and costly shutdown, and planned maintenance such as oil reclama-tion or replacement.
The latest IEEE guide for acceptance and maintenance of insu-lating oil states that ASTM D1816 (1-mm or 2-mm gap) represents more realistic results than ASTM D877 for power transformers because of the electrode shape [12]. This standard categorizes the
Table 4. Furfural Test Rating or Age Rating Where Test Results Not Available.
Rating CodeFuraldehyde [ppm]
AgeYears
A 0–0.1 Less than 20
B 0.1–0.25 20–40
C 0.25–0.5 40–60
D 0.5–1.0 More than 60
E More than 1.0 —
Table 5. Power Factor Rating.
Rating CodeMaximum Power Factor
[%]
A pfmax < 0.5
B 0.5 ≤ pfmax <0.7
C 0.7 ≤ pfmax <1
D 1 ≤ pfmax <2
E pfmax ≥ 2
Table 6. Tap Changer Standards Differences.
IEC IEEE
Designation OLTC LTC
Tap selection and Arcing control method
Diverter switch Arcing switch
Selector switch Arcing tap switch
Current limiting method Mainly resistor type Resistor and reactor type
Figure 3. LTC normal gas concentration introduced by Weidman-ACTI and WECC gas limits [16], [17] (ppm Log scale).
24 IEEE Electrical Insulation Magazine
test thresholds for each voltage class. A ranking method based on the above references is suggested in Table 3. An oil quality factor (OQF) similar to the DGAF of (1) is developed using the score and the weighting factors of Table 3. The suggested weighting factor in Table 3 is a baseline and can be adjusted according to the utility practice and oil supplier, if necessary. A rating code similar to that in Table 1 is employed to evaluate the calculated OQF using A, B, C, D, and E codes. It is important to note that these values are recommended for continued use of service-aged insulating oil, not for new oil.
FurfuralA furan test is not done on a routine or periodic basis for power
transformers; however, it may be employed as a post-diagnostic technique. Measurement of the furfural content of the oil can be used for a bulk measurement of the degree of polymerization of the paper insulation. IEC 61198 [14] explains the measurement of trace furanic compounds in transformer oil. Furan levels in transformers are typically less than 0.1 ppm. However, CIGRE conducted an extensive study of more than 5000 European trans-formers and found that a significant number have a furan content higher than 1 ppm [15]. This test is recommended when the transformer overheated, or has a high level of carbon monoxide and carbon dioxide. If the transformer is older than 25 years, a furan test is recommended on a periodic basis [1], [2].
If the data are available, the first two columns of Table 4 are employed for the HI calculation. If the transformer oil has been reclaimed or changed, then this test cannot give real information on the paper degradation. In such cases, the age of transformer may be used in the HI calculation using the third column of Table 4. This table does not imply a relation between the furan test and the transformer age; however, the relation between furan and the degree of polymerization for Kraft paper has been developed that is correlated to the remaining life of paper [3].
CIGRE SC A2 and the Insulation Life Subcommittee of the IEEE/PES Transformer committee have recently worked to cor-relate furan analysis to transformer insulation condition. The problem becomes one of determining how the furan values should be interpreted, the correlation to degree of polymerization (DP), normal aging versus fault data, and remaining life.
The recent results of collected furan analysis of over 30,000 data points from 12 countries by the Insulation Life Subcommit-tee of IEEE does not match with the traditional method proposed in Table 4. The gathered data show a wide variability in the data with no firm correlation between furans and age or predictable life. This committee believes that more data from North American utilities, testing companies, and users are needed before a techni-cal paper or guide could be developed. CIGRE SC A2 recently came to the same conclusion. Because work continues on this, the proposed method in Table 4 may be subject to change or receive a low weighting factor in the HI formulation.
Power Factor The power factor or dissipation factor measurement is an
important source of data to monitor transformer and bushing
Figure 4. Scoring dissolved gases of different LTCs based on thresholds (ppm log scale).
March/April 2009 — Vol. 25, No. 2 25
contact resistance, temperature, motor current, acoustic signal, relay timing, maintenance data, and the history of the LTC.
Some factors are not incorporated directly into the HI cal-culation of LTC. For example, the number of operations is not directly meaningful if a maintenance schedule is followed care-fully and the contacts are cleaned and replaced accordingly. The two main factors employed for HI calculation are DGA and oil quality tests.
The presence of certain levels of gas in LTC oil is normal because arcing occurs when an LTC operates. The concentration of DGA in a LTC depends on a number of variables, including mechanism type, breathing type, manufacturer, LTC model, oil brand, operating current, step-voltage of the LTC, and the number of operations. For example, free-breathing LTCs rapidly lose or gain gases, while sealed LTCs retain much of the gases produced. European and North American practices differ in the interpretation of DGA results. Because there is no standard for DGA of LTCs, it is not easy to recommend gas limits for DGA of LTCs. Here, the North American practice for LTCs will be taken into account. Weidmann-ACTI laboratories has developed
Figure 5. Recommendations for insulated conductor hottest temperature and top oil temperature, long-term emergency = 1 to 3 months, short-term emergency = 1/2 to 2 hours.
Figure 6. Individual component condition criteria based on corrective maintenance work orders in five years.
conditions. This test is performed to determine the condition of capacitive insulation between different windings and compart-ments. The measurement of transformer insulation’s capacitance and power factor at voltages up to 10 kV (at 50 or 60 Hz) has long been used both as a routine test and for diagnostic purposes. The tests can be done in the following configurations: high-voltage winding to ground, high- to low-voltage winding, low-voltage winding to ground, high- to tertiary-voltage winding, low- to tertiary-voltage winding, and the tertiary-voltage winding to ground insulation. pf
max is the greatest of all the measured power
factors. Table 5 recommends a ranking method for the power factor of transformers based on literature [1], [2].
Tap ChangerGeneral differences between tap changers used under IEEE
and IEC standards are listed in Table 6 [1]. Depending on the construction, the insulation system of an LTC usually consists of oil, cardboard, fiberglass, or epoxy resin.
There are several types of measurements for assessing the con-dition of LTCs such as: number of operations, DGA, oil quality,
26 IEEE Electrical Insulation Magazine
normal gas concentration levels [1], [16]. The Western Electric-ity Coordinating Council (WECC) has also suggested threshold levels for various LTC configurations [17]. These two resources are compared in Figure 3.
The threshold values are sometimes misleading, and it is suggested that the ratios between gases are also considered. If ethylene exceeds acetylene (except in vacuum type), this is a strong indicator of coking. The arcing-type tap reactive changers should produce acetylene with some heating gases. Resistive tap changers produce high acetylene with less heating gases. A resistive-type LTC will produce very high acetylene if the current
interruption has slowed [17]. Based on limited findings, Figures 4(A)–(D) propose a scoring method for incorporating the DGA analysis results into HI calculations. A DGA factor similar to (1) is used to rank the LTC condition based on the DGA analysis. Table 2 is used to rate the LTC using the calculated DGAF. If any gas content exceeds the last limit of Figure 4, a score of 4 is assigned. If C
2H
4/C
2H
2 ≥ 1, then the portion of HI related to LTC
is multiplied by C2H
2/C
2H
4.
Load HistoryFigure 5 summarizes the recommendations of IEC 354 and
IEEE C57.91-1995-cor. 1-2002 with respect to conductor and oil temperature inside the transformer [18]. The numbers are close, but IEC has a more conservative recommendation for conductor temperature. Moreover, IEC suggests a 1.3 per unit (p.u.) load factor for long-term emergency period and a 1.5 p.u. load factor for a short-term emergency period. The recorded monthly load peaks can be employed to calculate the load history contribution to HI calculations. The load history is categorized according to the five groups listed below:
N0 = Number of S
i/S
B that are lower than 0.6, i = 0,
N1 = Number of S
i/S
B that are between 0.6 and 1, i = 1,
N2 = Number of S
i/S
B that are between 1 and 1.3, i = 2,
N3 = Number of S
i/S
B that are between 1.3 and 1.5, i = 3, and
N4 = Number of S
i/S
B that are greater than 1.5, i = 4,
where Si is the monthly peak load, and S
B is the rated loading of
the transformer. Eq. (2) proposes a linear method of load score calculation,
and Table 7 describes a ranking method of transformer condition using the load history data.
LF =- ´
=
=
å
å
( )40
4
0
4
i N
N
ii
ii
(2)
Maintenance DataA ranking system was developed based on the maintenance
work orders issued in the last five years for the transformer and
Table 8. Overall Condition Based on Trend in Total Corrective Maintenance Work Orders.
Condition Rating Condition Criteria Description
A [Max(last 2 yrs) < 3] OR [increased < 10% over 5 yrs]
B[Max(last 2 yrs) > 3 AND increased > 10% over 5 yrs] OR [Max(last 2 yrs) > 5]
C[Max(last 2 yrs) > 5 AND increased > 30% over 5 yrs] OR [Max(last 2 yrs) > 10]
D[Max(last 2 yrs) > 10 AND increased > 50% over 5 yrs] OR [Max(last 2 yrs) > 15]
E[Max(last 2 yrs) > 15 AND increased > 80% over 5 yrs] OR [Max(last 2 yrs) > 20]
Table 7. Load Factor Rating Codes.
Rating Code Description
A LF ≥ 3.5
B 2.5 ≤ LF < 3.5
C 1.5 ≤ LF < 2.5
D 0.5 ≤ LF < 1.5
E LF ≤ 0.5
Table 9. Ranking of the Turn Ratio Test, Leakage Reactance Test, Core-to-Ground Test, and Winding Resistance Test.
Rating Code
Turn ratio (TR) deviation of actual to declared
[%]
Leakage reactance deviation
[%]
Core-to-ground resistance
[MΩ] Winding resistance
deviation [%]
A ΔTR ≤ 0.1% ΔX < 0.5% R > 1000 ΔR < 1%
B 0.1% < ΔTR ≤0.5% 0.5% ≤ ΔX < 1% 100 ≤ R < 1000 1% ≤ ΔR<2%
C 0.5% < ΔTR ≤ 1% 1% ≤ ΔX < 2% 10 ≤ R < 100 2% ≤ ΔR <3%
D 1% < ΔTR < 2% 2% ≤ ΔX < 3% 1 ≤ R < 10 3% ≤ ΔR<5%
E ΔTR ≥ 2% ΔX ≥ 5% R < 1 ΔR ≥ 5%
March/April 2009 — Vol. 25, No. 2 27
its accessories. Infrared thermography and bushing condition are two important factors in this evaluation. Oil leak, oil level, cool-ing system, gaskets, main tank condition, and grounding are also taken into account. Figure 6 shows the condition criteria based on corrective maintenance work orders in the last five years. If there is no work order in the last five years for any of these factors, the condition rating will be “A.” It is suggested that the rate of increase of work orders be monitored as well. An overall condi-tion factor is introduced to include the rate of maintenance work orders as shown in Table 8. Aside from bushing visual inspection such as oil leak and porcelain or silicon rubber condition, there are recommended tests such as oil tests (DGA, moisture, and so on), power factor tests, and hot collar tests that can be separately
quantified in a manner similar to the transformer oil and power factor test introduced in Tables 1, 2, 3, and 5.
The rest of the tests involved in the HI calculation are sum-marized in Table 9, with their rating factors. Turn ratio test, ex-citation current test, leakage reactance test, core-to-ground test, and winding resistance test are mainly considered as diagnostic tests rather than routine tests, and the related test data may not be available. The scoring system is a combination of the limits in [1] and IEEE Std 62, Part 1 [19].
Health Index Calculation A quantified scoring system can be used to appropriately rep-
resent the transformer health. This involves the following steps:
1) “Deterioration” assessments or scores are converted to health scores in a defined range from “perfect health” to “very poor condition.”
2) Importance weighting is assigned to each factor in a range from “modest importance” to “very high impor-tance.”
3) General deterioration index is formulated by calculating the maximum possible score by summing the multiples of steps 1 and 2 for each factor.
4) The general deterioration index is normalized to a maxi-mum score of 100 based on having a defined acceptable/minimum number of condition criteria available.
5) The dominant factors are normalized to a maximum score of 100.
A calculation of the overall Health Index is performed, where 100% represents excellent health and less than 25% represents “poor” health. Table 10 and Figure 7 provide a summary of the scoring system and a flowchart of the main condition parameters that are used in this study for condition assessment. Totaled scores are used in calculating final HI. For each component, the HI calculation involves dividing its total condition score by its maximum condition score, then multiplying by 100. This step normalizes scores by producing a number from 0 (completely degraded transformer) to 100 (perfect condition). The power transformer is rated against a set of criteria for each condition parameter. The rating (A, B, C, D, E) is converted to a factor be-tween 4 and 0, respectively, called HIF in Table 10. Considering all the discussed parameters and factors, the total HI of a power transformer is proposed as:
HI
HIF HIF
= ´ +=
=
=
=
å
å
å
å60
4
40
4
1
21
1
2122
24
22
24% %
K
K
K
K
j jj
jj
j jj
jj
(3)
Kj and HIF
j are introduced in Table 10. A weighting factor of
40% is assigned to the LTC and 60% to the transformer. This is based on an international survey done by a CIGRÉ working group on failures in large power transformers that found that about 40% of failures were due to LTC [6]. If a utility has the records
Table 10. Health Index Scoring.
#TransformerCondition Criteria K
ConditionRating HIF
1 DGA 10 A,B,C,D, E 4,3,2,1,0
2 Load History 10 A,B,C,D, E 4,3,2,1,0
3 Power Factor 10 A,B,C,D, E 4,3,2,1,0
4 Infra-red 10 A,B,C,D, E 4,3,2,1,0
5 Oil Quality 6 A,B,C,D, E 4,3,2,1,0
6 Overall Condition 8 A,B,C,D, E 4,3,2,1,0
7 Furan or Age 5 A,B,C,D, E 4,3,2,1,0
8 Turns ratio 5 A,B,C,D, E 4,3,2,1,0
9 Leakage reactance 8 A,B,C,D, E 4,3,2,1,0
10 Winding resistance 6 A,B,C,D, E 4,3,2,1,0
11 Core-to-ground 2 A,B,C,D, E 4,3,2,1,0
12 Bushing Condition 5 A,B,C,D, E 4,3,2,1,0
13 Main Tank Corrosion 2 A,B,C,D, E 4,3,2,1,0
14 Cooling Equipment 2 A,B,C,D, E 4,3,2,1,0
15 Oil Tank Corrosion 1 A,B,C,D, E 4,3,2,1,0
16 Foundation 1 A,B,C,D, E 4,3,2,1,0
17 Grounding 1 A,B,C,D, E 4,3,2,1,0
18 Gaskets, seals 1 A,B,C,D, E 4,3,2,1,0
19 Connectors 1 A,B,C,D, E 4,3,2,1,0
20 Oil Leaks 1 A,B,C,D, E 4,3,2,1,0
21 Oil Level 1 A,B,C,D, E 4,3,2,1,0
22 DGA of LTC 6 A,B,C,D, E 4,3,2,1,0
23 LTC Oil Quality 3 A,B,C,D, E 4,3,2,1,0
24 Overall LTC Condition 5 A,B,C,D, E 4,3,2,1,0
28 IEEE Electrical Insulation Magazine
Figure 7. HI calculation flowchart.
March/April 2009 — Vol. 25, No. 2 29
of failure and can calculate the failure rate caused by LTC, it is recommended that the proposed weighting factor is replaced by the calculated number. Figure 8 presents an example of the HI calculation for a large population of power transformers.
Health Index and Probability of FailureTransformer failures can be broadly categorized as electrical,
mechanical, or thermal. The cause of a failure can be internal or external. In addition to failures in the main tank, failures can also occur in the bushings, tap changers, or transformer accessories. According to traditional literature, the failure pattern of a power transformer follows a “bathtub curve” [2]. CIGRE working group
WG12.05 performed an international survey of 13 countries (Eu-rope, North America, and Australia) [6]. Typical failure statistics versus age of three different types of transformers are shown in Figure 9, based on a survey of a large population of transformers [1]. A utility company may experience failure rates that are dif-ferent, so each company should keep accurate records of failures. If actual failure rate data are available, it is recommended these be used to compare with HI rather than the proposed failure rate in Figure 9.
In a simplified manner, one could map the Very Good, Good, Fair, etc., categories to increased probability of failure. Figure 10 shows an example of end-of-life curve versus age. In this example, transformers are assumed to have a median life expectancy of 50 years. The life expectancy curve is estimated through a Weibull distribution. The ratio of failed %, as reflected in the Age Curve (Figure 10) can be mapped to the failure rate versus age (Figure 9) to estimate the probability of failure from the HI, as shown in Figure 11. A simplified summary of such estimation is shown in Table 11.
Capital Plan for Asset ReplacementA common use of HI is to provide justification for a capital
plan for asset replacement at the end-of-life. This is a part of the full capital plan for a utility, which must include capital for system growth, capacity shortfall, obsolescence, and emergency repair, as well as asset replacement due to end-of-life. There are four steps in the development of the capital plan for asset replacement based on HI, as shown in Figure 12.
Step 1: Remove Maintainable Condition ParametersThe HI results provide a snapshot of the condition of the assets.
Figure 8 Example of Health Index versus age for power transformers.
Figure 9 Failure probability versus age [1].
30 IEEE Electrical Insulation Magazine
Figure 10 Ratio of failed % as reflected in the age curve.
Figure 11 Simplified estimating of probability of failure from the Health Index.
In some cases, items that are in poor condition can be returned to good condition by maintenance. In developing a capital plan for asset replacement, the maintainable items can be assumed to be maintained and therefore removed from HI formulation by setting the weight of the associated condition parameter to zero. When the HI formulation is modified to exclude maintainable condition factors, the definitions of the factor values for the remaining con-ditions may require adjustment. Normally, with many condition factors, the Health Index cannot be driven to less than about 0.25. The interpretation of the Health Index to obtain probability of failure and estimate remaining life, therefore, must be based on a Health Index of 0.25, indicating effective end-of-life.
Step 2: Estimate the Probability of Failure and Effective Age of Transformer
The HI can be used to estimate the probability of failure of the transformer in its present condition. Each transformer has a level of remaining strength, both electrical and mechanical, that decreases as its condition deteriorates with age and use. The transformer’s probability of failure depends on whether the stresses in the field exceed the remaining strength. The probability of the stress exceeding the strength is the probability of failure. The relation between probability of failure and HI is defined in Figure 11.
March/April 2009 — Vol. 25, No. 2 31
The “effective age” of the transformer can be determined by relating the probability of failure based on HI to the experienced failure rate for different actual chronological ages. The typical failure rate versus age curve that has been introduced in Figure 9 is again employed. Figure 13 shows a schematic on how to extract the effective age. For example, as shown in this figure, if a 60-year-old transformer had been heavily loaded over its life and now had a HI of 50%, it would have an effective age of 78 years. An alternative is to use the industry standard failure rate curve, which follows a standard bathtub curve and assumes increasing failure rate with an exponential rise, and then fit the exponential to the observed median life and maximum life of the particular transformers in a particular utility.
The failure rate versus age graph can be used to estimate the probability of failure in future years by reading the graph for the years above the effective age. However, the rate at which the degradation of strength, and subsequently HI, occurs needs to be
taken into account. An adjustment must be made to account for assets that are aging more slowly or more quickly than normal. The aging rate is calculated as the effective age divided by the real chronological age. This aging rate assumes that the asset will be in a similar operating environment in the future as it has been in the past. Because this assumption is not necessarily accurate, aging rates that vary widely from the normal rate of one are not reliable. A maximum change of 50% has proven to be practical. For example, the 60-year-old transformer with the effective age of 78 years is aging at 1.3 years per calendar year (78/60). Then the probability of failure in the next year is read from the failure rate versus age chart at the effective age plus 1.3.
Step 3: Calculate the Remaining Life of the Transformer
The remaining life of the transformer can be determined in two ways. The simplest way is to subtract the effective age from the age at which the cumulative probability of failure reaches 99% and divide by the aging rate. Alternatively, the maximum age could be considered to occur slightly earlier, at a particular failure probability that the utility deemed to be the maximum acceptable due to consequences of failure.
The remaining life of the transformer can also be estimated based on a financial risk analysis. This analysis compares the cost of replacing the transformer in any one year, with the “con-sequence cost” of leaving it in service that long. The consequence cost is the probability of failure multiplied by the cost of failure, such as outage cost to customers and environmental cleanup, plus the annual maintenance cost. Each year that the replacement is delayed, the present value of the replacement capital cost is reduced, but the consequence cost increases. At some point, an optimum minimum sum of these two values determines the opti-mal age to replace the transformer. This is the “economic end of life,” where the lowest long-term cost is achieved.
Step 4: Sum the Capital Cost in Each YearReplacement timing of the asset has now been determined
based on the remaining life. The cost of each asset can be de-termined based on the capital, installation, and disposal costs. The costs used should be replacement costs, not original cost. It
Table 11. Health Index Scale.
Health Index Condition Description
Approximate Expected Lifetime
85–100 Very GoodSome aging or minor deterioration of a limited number of components
More than 15 years
70–85 GoodSignificant deterioration of some components
More than 10 years
50–70 Fair
Widespread significant deterioration or serious deterioration of specific components
Up to 10 years
30–50 Poor Widespread serious deterioration Less than 3 years
0–30 Very Poor Extensive serious deterioration At End-of-Life
Figure 12 Capital plan for asset replacement flowchart.
32 IEEE Electrical Insulation Magazine
should be assumed that out-of-date technology will be replaced with the most appropriate present-day technology. In each year, the replacement cost for all the assets reaching end-of-life in that year are summed to produce the total replacement capital plan. An example is shown in Figure 14. Three different types of power transformer are represented by the different colors.
Conclusion The composite HI presented is a very useful tool for represent-
ing the overall health of a complex asset such as a power trans-former. HI quantifies equipment condition based on numerous condition criteria that are related to long-term degradation factors that cumulatively lead to a power transformer’s end-of-life. The
Figure 13 Extracting effective age from the probability of failure.
Figure 14 Example replacement capital plan.
March/April 2009 — Vol. 25, No. 2 33
method’s multi-criteria analysis approach combines the various factors are combined into a condition-based HI. In addition to the regular test data that have been used in the past, a count of corrective maintenance work orders can be used to evaluate the physical health condition of transformers. Some of the important factors include bushing condition, oil leak, tank corrosion, cool-ing system, infrared thermography, grounding, and foundation. The relation between HI and probability of failure was developed based on available data and can be applied to similar analysis ap-plications. HI can effectively be employed to provide justification for a capital plan which includes end-of-life asset replacement.
References[1] ABB Service Handbook for Transformers, 2nded.,Zurich,Switzerland:
[4] T.Hjartarson andS.Otal, “Predicting future asset condition based oncurrent health index andmaintenance level,” presented at 11th IEEEConf.Transmission&DistributionConstruction,OperationandLive-LineMaintenance,Albuquerque,NM,Oct.2006.
[8] K.T.Muthanna,A.Sarkar,K.Das,andK.Waldner,“Transformerinsulationlifeassessment,” IEEE Trans. Power Del.,vol.21,no.1,pp.150–156,Jan.2006.
[9] IEC60599, “Mineral oil-impregnated electrical equipment in service -Guidetotheinterpretationofdissulvedandfreegasesanalysis”,Edition2,1999.
[10] IEEE Guide for the Detection and Determination of Generated Gases in oil-Immersed Transformers and Their Relation to the Serviceability of the Equipment,IEEEStd.C57.104,1978.
[11] M. Duval, “A review of faults detectable by gas-in-oil analysis intransformers,”IEEE Insulation Magazine,vol.18,no.3,pp.8–17,May/Jun.2002.
[12]IEEE Guide for Acceptance and Maintenance of Insulating Oil in Equipment, IEEEStdC57.106-2006,IEEETransformersCommittee,2006.
[19]IEEE Guide for Diagnostics Field Testing of Electric Power Apparatus—Part 1: Oil-Filled Power Transformers, Regulators, and Reactors,IEEEStd62,1995.
Ali Naderian Jahromi (M’06) received his B.Sc. and M.Sc. degrees from Sharif University of Technology, Iran, in 1998 and University of Tehran in 2000, respectively. He received his Ph.D. degree in 2006 after several years’ research at the University of Waterloo and then joined Kinectrics (for-merly Ontario Hydro Research Division) in 2007. His employment experience includes
Iran-Switch Company (1997–1999) in testing of MV switchgears and circuit breakers, Iran-Transfo Company (2000–2001) in de-signing and manufacturing of HV testing transformers, and Iran Power Generation and Transmission Organization (TAVANIR) in substation planning division (2001–2004). His research interests are high-voltage test techniques, diagnostics of power transform-ers, and HV cable field testing and commissioning.
Ray Piercy has been a Principal Engineer in the Distribution Systems group within Kinectrics. He has over twenty-five years’ experience in the analysis, modeling, moni-toring, and testing of distribution systems and components. His major projects have been in long-range planning, loss analysis, cost allocation studies, asset condition assessment and monitoring, end-of-life as-sessment, predictive reliability and power
quality assessment, development of new data acquisition equip-ment, transient recovery voltage modeling and measurement, distribution transformer failure analysis, internal arc testing, and power line carrier communications modeling and testing.
Stephen Cress received his B.A.Sc. degree in electrical engineering from the Univer-sity of Toronto in 1976. He joined Ontario Hydro in 1976 and then started work for Ontario Hydro Research Division in 1978 as a Senior Engineer. He is developer of the computer programs FUSECORD and
TRANSIZE and co-developer of ARCPRO.
He is currently the Manager of Distribution Asset Management in the T&D Business
of Kinectrics Inc. Mr. Cress is a recognized Distribution Power System expert and has over 30 years’ experience in technical investigations, application, analyses, standard development, standard and qualification testing, and research associated with distribution systems and equipment. He is the Canadian Chairman of IEC/SCC SC32A and CSA C254 Committees for Standards on High Voltage Fuses.
34 IEEE Electrical Insulation Magazine
Jim Service has over 30 years of experience related to power systems and, in particular, managing and utilizing large databases of power system data. He has worked in the area of Data Automation, i.e., implementa-tion of complete solutions, through the use of standard or novel methods, to acquire data and process it via spreadsheets or database queries and stored functions. Jim also has wide-ranging experience in data display on
the web, in office applications, asset condition assessment tools, asset condition health indexes, etc. He has developed applied engineering software in such diverse areas as turbo-generator SSFR, robotics, arc hazard analysis, electrical energy market simulation, “smart meter” data analysis, transmission line fault location, and geomagnetically induced current monitoring. In addition to his engineering expertise, Jim is a qualified ISO 9000:2001 and ISO 17025 internal auditor and a Z299.x and N286 QA inspector. Jim is registered as a Professional Engineer in the province of Ontario.
Fan Wang obtained his B. Eng, M. Eng., and Ph.D. degrees from Tsinghua University (China), National University of Singapore (Singapore), and Chalmers University of Technology (Sweden), respectively. He worked as an electrical engineer at South-west Electric Power Design Institute (China) for 5 years, and as a systems engineer at Honeywell Canada for 2 years before join-
ing Kinectrics as an engineer/scientist in 2007. His working and research experiences are in the fields of power system asset management, power quality, protective relays, and power plants/substations engineering design. He is a licensed professional engineer of PEO and member of IEEE.
