A

AU

a

ARRAA

KACLMT

1

oatmot

rntaadiasqocvle

m

0d

Electric Power Systems Research 80 (2010) 456–464

Contents lists available at ScienceDirect

Electric Power Systems Research

journa l homepage: www.e lsev ier .com/ locate /epsr

sset management techniques for transformers

hmed E.B. Abu-Elanien ∗, M.M.A. Salamaniversity of Waterloo, Waterloo, ON, Canada N2L 3G1

r t i c l e i n f o

rticle history:eceived 10 June 2009eceived in revised form 9 October 2009ccepted 12 October 2009vailable online 5 November 2009

a b s t r a c t

In a deregulated/reformed environment, the electric utilities are under constant pressure for reducingoperating costs, enhancing the reliability of transmission and distribution equipments, and improvingquality of power and services to the customer. Moreover, the risk involved in running the system withoutproper attention to assets integrity in service is quite high. Additionally, the probability of losing any

eywords:sset managementondition monitoringife cycleaintenance

equipment vital to the transmission and distribution system, such as power and distribution transformers,is increasing especially with the aging of power system’s assets. Today the focus of operating the powersystem is changed and efforts are being directed to explore new approaches/techniques of monitoring,diagnosis, condition evaluation, maintenance, life assessment, and possibility of extending the life ofexisting assets. In this paper, a comprehensive illustration of the transformer asset management activitiesis presented. The importance of each activity together with the latest researches done in the area is

ransformers highlighted.

. Introduction

Transformer asset management is generally considered to bene of the most important power system equipment asset man-gements. This is due to the substantial investments in the powerransformers and the importance of the transformers as one of the

ajor factors that affect the system reliability. The un-scheduledutages of the transformers due to unexpected failures are catas-rophic in many cases.

Transformer asset management activities are numerous andesearchers tackle them from different points of view. Mainte-ance plans and condition monitoring techniques are samples ofhe general asset management activities that can be applied tony equipment such as transformer, circuit breaker, high volt-ge capacitor, etc. However, each asset management activity isifferent from equipment to another. For example, condition mon-

toring techniques applied to transformers are different from thosepplied to circuit breakers or high voltage capacitors althoughome of these techniques may have some similarities. Also, oneuantity can be tackled from different asset management pointsf view. For example, transformer hot spot temperature (HST)

an be tackled from transformer condition monitoring point ofiew because it may represent an overloading or serious prob-em inside the transformer, and it can also be tackled from thend of life point of view because the higher the hot spot tem-

∗ Corresponding author. Tel.: +1 519 888 4567x33367; fax: +1 519 746 3077.E-mail addresses: abayoumy@engmail.uwaterloo.ca (A.E.B. Abu-Elanien),

.salama@ece.uwaterloo.ca (M.M.A. Salama).

378-7796/$ – see front matter © 2009 Elsevier B.V. All rights reserved.oi:10.1016/j.epsr.2009.10.008

© 2009 Elsevier B.V. All rights reserved.

perature over the normal value, the shorter the lifetime of theasset.

This paper focuses on the transformer asset management asone of the important power system assets. Fig. 1 shows the trans-former main asset management activities. The transformer assetmanagement can be classified into the following activities:

(1) Condition monitoring (CM) and condition assessment (CA)techniques.

(2) Performing maintenance plans.(3) Aging, health, and end of life assessments.

In the following sections, each activity is discussed in detail.

2. Condition monitoring and condition assessmenttechniques

Transformer CM is concerned with the application and devel-opment of special purpose equipments/methods that are involvedin monitoring a condition of a parameter in a transformer and itsdata acquisition while CA means the development of new tech-niques for analyzing this data to both predict the trends of themonitored transformer and evaluate its current performance. CMfocuses mainly on the detection of incipient faults inside the trans-former that are created from the gradual deterioration. Some of

these incipient faults may be detected during routine maintenance;however, other faults may cause numerous problems before theroutine maintenance cycle. As a result, the ability to have detailedinformation on the state-of-health of the transformer prior to car-rying out maintenance work was unavailable. Also, the diagnosis

A.E.B. Abu-Elanien, M.M.A. Salama / Electric Pow

oum

dpoCacask

tbcTtwflmmabeIs

2

matttftTa

Fig. 1. Transformer asset management activities.

f many incipient faults in the transformer was, in many cases,navailable especially with those faults occurring after the routineaintenance cycle [1–3].The CM has multiple benefits: it reduces the maintenance costs

ue to its ability to detect faults early, limits the probability of com-lete failures, and identifies the root causes of the failure. On thether hand, there are some obstacles during the realization of theM techniques such as extra added cost to the system due to thedded monitoring and communication equipments, increase in theomplexity of the control and communication system, need for newnd high speed processing systems for data processing and deci-ion making, and need for suitable memory storage for data basenowledge.

In order to have information about the state-of-health of theransformer, the monitored data and the incipient faults detectedy the CM system should be analyzed to assess the transformerondition. This assessment is done using the CA of the transformer.ransformer CM can be divided into five main categories: moni-oring the hot spot temperature, monitoring the vibration of theall and winding, monitoring the dissolved gases in the trans-

ormer oil, monitoring the partial discharges (PDs) in the solid andiquid insulations of the transformer, and monitoring the winding

ovement and deformations. In order to have a meaning of theseonitored parameters, the monitored data should be analyzed to

ssess the condition of the transformer. Each CM data category cane assessed using certain CA technique. Fig. 2 shows the main cat-gories of transformer CM and the corresponding CA techniques.n the following sub-sections, each CA technique will be discussedeparately.

.1. Condition assessment by thermal analysis

Thermal analysis of the transformers can provide useful infor-ation about its condition and can be used to detect the inception of

ny fault. Most of the faults cause change in the thermal behavior ofhe transformer. Abnormal conditions can be detected by analyzinghe HST. The most famous abnormal condition of the transformer

hat can be detected by thermal analysis is the overload. Trans-ormer life is affected greatly for a continuous maximum HST morehan 110 ◦C [4]. Predicting the HST can be done by two techniques.he first technique uses the artificial intelligence techniques suchs Artificial Neural Network (ANN) to predict the HST [5]. The sec-

Fig. 2. Transformer condition monitor

er Systems Research 80 (2010) 456–464 457

ond technique develops a thermal model to predict the thermalbehavior of the transformer [6–8].

2.2. Condition assessment by vibration analysis

The usage of the vibration signals in assessing the transformerhealth is a relatively new technique compared with other meth-ods of transformer CA. The transformer vibration consists of corevibrations, winding vibrations, and on load tap changer vibrations[9,10]. These generated vibrations propagate through the trans-former oil until they reach the transformer walls, at which theycan be collected via vibration sensors. The health condition of thecore and windings can be assessed using the vibration signature oftransformer tank [9]. Vibration analysis is a very powerful tool forassessing the health of the on load tap changer [10,11].

2.3. Condition assessment by partial discharge analysis

PDs occur when the electric field strength exceeds dielectricbreakdown strength of a certain localized area, in which an electri-cal discharge or discharges partially bridge the insulation betweenconductors. The dielectric properties of the insulation may beseverely affected if subjected to consistent PD activity over longperiods of time. This may lead to complete failure if the PD activityremains untreated [12]. PD can be detected and measured usingpiezo-electric sensors, optical fiber sensors [13], and Ultra HighFrequency (UHF) sensors [12,14,15]. On site PD measurement isoften affected by strong coupled electromagnetic interference thatincreases the difficulty of extracting PD signals without noise. Themost common methods for PD de-noising are the usage of theWavelet Transform [16,17], the gating method, and the directionalsensing [19]. PD measurement was used extensively for the con-dition assessment of the transformer insulation due to the factthat large numbers of insulation problems start with PD activity[14,18,19].

2.4. Condition assessment by dissolved gas analysis (DGA)

All transformers generate different gases at normal operat-ing temperatures. Nevertheless, the concentration of these gasesincreases in the presence of an abnormality (fault) such as ther-mal, partial discharge, and arcing faults [20]. During internal faults,oil produces gases such as hydrogen (H2), methane (CH4), acety-lene (C2H2), ethylene (C2H4), and ethane (C2H6), while celluloseproduces methane (CH4), hydrogen (H2), carbon monoxide (CO),and carbon dioxide (CO2). Each fault type produces certain gasesfrom the above-mentioned gases [20]. Analyzing transformer oil for

these key gases by chromatography helps to know the fault type,and location [20,21]. Also, laboratories may rely upon defined crit-ical levels of gases, rates of increase in gas level (on a year-by-yearbasis), or one of the ratio methodologies such as Rogers or Dor-nenberg ratios [20,22,23] to asses the condition of oil. However,

ing and assessment techniques.

4 ic Power Systems Research 80 (2010) 456–464

it

dp(Aoth(aadduw

mqiscs[

vstpdos

2

tidieWfiaad

2a

tstMsnCstynp

58 A.E.B. Abu-Elanien, M.M.A. Salama / Electr

nterpretation by the individual gases can become difficult whenhere is more than one fault in the transformer.

Thermal faults such as sustained overloads and high HST pro-uce many gases. Low temperature decomposition of mineral oilroduces relatively large quantities of hydrogen (H2) and methaneCH4), and trace quantities of ethylene (C2H4) and ethane (C2H6).t medium temperatures, the hydrogen concentration exceeds thatf methane, the amount of ethylene increases but still less thanhe amount of ethane. At the upper end of the thermal fault range,ydrogen and ethylene quantities increase and traces of acetyleneC2H2) may be produced [20]. The solid insulation begins to degradet lower temperatures than the oil, therefore, its products are foundt normal operating temperatures in the transformer. The thermalecomposition of cellulose produces carbon monoxide (CO), carbonioxide (CO2), and water vapor. The ratio of CO2/CO is sometimessed as an indicator of the thermal decomposition of cellulose asill be discussed in Section 4.1.1.4.

Low intensity discharges such as partial discharges produceainly hydrogen, with decreasing quantities of methane and trace

uantities of acetylene. The acetylene and ethylene concentrationsncrease as the intensity of the discharge increases. As the inten-ity of the electrical discharge increases and reaches arcing orontinuing discharge proportions. The temperature will increaseignificantly and the quantity of acetylene becomes pronounced20].

These incipient faults affect the reliability of the transformerery much if not detected and treated early. The paper insulationystem may be damaged due to local high temperature hot spots ifhe thermal faults are left untreated. Moreover, the paper insulationroperties decreased notably for sustained PD or arcing faults. Theegradation of the paper insulation can be detected using the ratiof CO2/CO dissolved in transformer oil, which represents the tensiletrength of the paper insulation.

.5. Condition assessment by frequency response analysis

When a transformer is subjected to high through fault currents,he windings are subjected to severe mechanical stresses caus-ng winding movement, deformations, and in some cases severeamage. Deformation results in relative changes to the internal

nductance and capacitance of the winding which can be detectedxternally by frequency response analysis (FRA) method [24].inding damage detection can be accomplished by comparing the

ngerprints of a healthy winding (or the calculated response usingtransformer equivalent circuit) with the fingerprints of a dam-

ged winding. Changes in fingerprints can be used to estimate theegree of winding damage and its location [25,26].

.6. New developments in condition monitoring and conditionssessment

With the development of sensors technology and communica-ion systems, more than one parameter can be monitored in theame time [27]. New online CM and CA systems that monitor morehan one parameter in the transformer are commercially available.

any parameters can be monitored online using these new systemsuch as HST, dissolved gases, and oil temperature. Advanced tech-

ology sensors are used for parameter measurements in these newM systems. All data measured are then collected using data acqui-ition subsystem to be analyzed and to provide interpretation forhe operator. Recently, intelligent systems are used for data anal-sis and interpretation such as multi agent systems [28,29]. Theseew CM systems provide fast and accurate interpretation to anyroblem in the transformer.

Fig. 3. Classification of maintenance activities.

3. Performing maintenance plans

Performing maintenance plans is the second transformer assetmanagement activity. Transformer outage has harmful effects onthe system and can be assumed as one of the most catastrophicoutages, especially for high rating power transformers. Accord-ingly, maintenance of the transformers should be planned carefullyto avoid harmful outages. According to Fig. 3, the maintenancetypes can be classified into corrective maintenance, preventivemaintenance, and reliability centered maintenance. The definitiontogether with the advantages and disadvantages of each mainte-nance type will be illustrated in the next sub-sections.

3.1. Corrective maintenance

Corrective maintenance is designed to perform maintenanceactivity upon occurrence of failure. This type of maintenance is notwidely spread. Corrective maintenance may lead to catastrophicfailures that cannot be maintained and, finally, to losing the asset.This type of maintenance was the main maintenance activity a longtime ago. After the knowledge of the failure consequences which, insome cases, may be catastrophic, this type of maintenance, if used,has been reserved for the types of defects that are not serious andhave no great consequences, such as failure of some accessories. Asa conclusion, the general meaning of the corrective maintenance isperforming maintenance upon failure occurrence. The advantagesand disadvantages of the corrective maintenance are listed below.

Advantages:

(1) It is the least expensive type of maintenance.(2) It saves manpower.(3) It spares the system from un-necessary shutdowns.(4) It performs the maintenance only when it is needed, saving un-

necessary inspections.(5) It is widely understood by the maintenance members.

Disadvantages:

(1) Transformer failure becomes costly to repair and may needexpensive spare parts.

(2) Some transformer failures may be un-repairable if not detectedearly.

(3) Some transformer failures may cause complete shutdown of theproduction line or the power system for long time. This meansloosing revenue which in some cases exceeds the cost of theregular inspection.

3.2. Preventive maintenance

Preventive maintenance aims to prevent the failure from occur-rence. Also, it aims to guarantee long lifetime of the asset. This

A.E.B. Abu-Elanien, M.M.A. Salama / Electric Power Systems Research 80 (2010) 456–464 459

CBM s

cfp

3

eacTumatiib

((

(

(

(

(

3

dcttuontwo

af(d2uirtonm

Fig. 4.

an be achieved by shutting down the equipment regularly to per-orm time based maintenance (TBM) or by installing CM system toerform condition based maintenance (CBM).

.2.1. Time based maintenanceTBM is based on examining and maintaining the transform-

rs according to a time schedule, i.e., performing the inspectionnd the maintenance activities at constant intervals. TBM is theurrent maintenance strategy for many industries and utilities.BM may prevent many failures; however, it may also causennecessary outages, wasting manpower, time, and money if theaintenance interval is too small [30,31]. In addition, unexpected

ccidents may still occur in the intervals between maintenanceasks if the maintenance interval is too large. The general mean-ng of the TBM maintenance is performing maintenance at regularntervals. The advantages and disadvantages of the TBM are listedelow.

Advantages:

1) It is understood by maintenance engineers and technicians.2) It can detect the inception of faults to some extent, if the inspec-

tion interval is reduced.3) It increases the lifecycle of the transformer due to regular

inspections and maintenance.

Disadvantages:

1) It is expensive due to regular un-necessary inspections and thelarge number of the needed maintenance staff.

2) In some cases, TBM is unable to detect faults especially whenthe inspection interval is large.

3) It needs un-necessary shutdowns which add extra cost to themaintenance activity.

.2.2. Condition based maintenanceCBM relays on performing maintenance when the CM system

etects an incipient fault. This incipient fault will change to be aomplete failure if not treated early by the CBM, i.e., a suitable main-enance will be performed upon detection of an incipient fault byhe CM system. By using this technique, the risk of complete fail-re is reduced. CBM lets operators know more about the conditionf a transformer to know clearly when and what maintenance iseeded. A transformer’s historical data such as operation parame-ers, diagnostic tests and the environmental conditions will identifyhich parameter/part should be monitored and the correct method

f monitoring [32].Advanced online monitoring and assessment techniques such

s dissolved gas analysis, partial discharge, furan analysis (FA),requency response analysis, and recovery voltage measurementRVM) play a key role in developing CBM strategies [33,3]. The con-ition monitoring and diagnostic techniques discussed in Sectionare the main core of the CBM. The CBM may depend on contin-

ous, scheduled, or on request CM. The most widely spread CBMs the continuous one. The scheduled or on request CBM aims to

educe the cost of the continuous condition monitoring, which ishe largest problem in the application of the CBM. CBM dependsn monitoring the parameters/parts of the transformer and diag-osing the incipient faults. When an incipient fault is found, theaintenance activity should take place to avoid the complete fail-

ystem.

ure of the equipment. Thus, maintenance is only performed whennecessary.

Fig. 4 shows a block diagram of a CBM system integrated with aCM and CA systems. The first stage of this integrated system is theraw data stage, in which different types of sensors are used to collectthe raw data such as thermal data, partial discharges, vibration sig-nals, gases in oil, etc. The second stage is the pre-conditioning datastage which aims to adjust the input data. This adjustment includesremoving extreme data levels, normalizing the input data if needed,or removing the noise contained in the raw data. With the advancein the sensor technology, this stage may be included in the firststage. The next stage is the extraction of useful information. In thisstage, some useful information can be extracted before the diag-nosis stage. For example, estimation of the hot spot temperature isbased on the top oil temperature, ambient temperature, and loadcurrent. The pre-conditioned and pre-processed data will be usedto assess the condition of the equipment and classify the type offault if any in the CA and fault diagnosis stage. The next stage is theoutput stage, in which the outputs from the fault diagnosis stage,either by Artificial Intelligent (AI) agents or by derived logics, areinterpreted and sent to the targeted maintenance staff. The main-tenance action is taken in the final stage according to the outputsstage [14,34]. The general meaning of the CBM is performing main-tenance only upon request from the CM system. The advantagesand disadvantages of CBM are listed below.

Advantages:

(1) Maintenance is done when it is necessary.(2) Saving costly unnecessary inspections.(3) Saving manpower.(4) Reducing the unnecessary shutdowns of the system.(5) Low possibility of complete failure.

Disadvantages:

(1) Continuous condition monitoring for many parameters isexpensive.

(2) Less understood by maintenance engineers and technicians.(3) It needs fast data communication and manipulation facilities

for successful online monitoring.(4) It needs experienced persons to design the monitoring system,

select the suitable parameters to be monitored, and select thesuitable frequency of data collection.

3.3. Reliability centered maintenance (RCM)

The RCM is a technique initially developed by the commercialairline industry. The fundamental goal of RCM is to preserve thefunction or operation of a system with a reasonable cost [35,36].RCM can be defined as a mix of more than one maintenance strat-egy in an optimized manner in order to reduce the system risk.For a successful RCM plan, the degree of risk of each fault shouldbe identified in order to define the optimum maintenance actions[35,36]. The risk index can be found as follows

risk = probabilty of failure × consequences index (1)

The main items in the implementation of RCM according to(1) are the prioritization of the failure modes according to theirconsequences on the system and modeling the probability of fail-

4 ic Pow

ucefifcafamtmTeo

td

(

((

(

(((

MA5mTuaosps

eptad

4

arieot

60 A.E.B. Abu-Elanien, M.M.A. Salama / Electr

re modes [37,38]. The consequences index of each failure modean be determined by the analysis of the history of failures or byxperience. RCM starts with collecting data about the transformerailures to model the failure modes in a probabilistic form. Thenformation about the consequences of each failure can be collectedrom the past experience of the skilled engineers. The informationollected about consequences of failures together with the prob-bility of each failure are used to calculate the risk index of eachailure. The failure modes that have low risk index are separatednd treated by low cost maintenance method such as correctiveaintenance. The failure modes that have high risk index can be

reated by preventive maintenance such as CBM or TBM with opti-um maintenance interval based on the maintenance cost [39,40].

he possibility of failures still exists in the system with RCM; how-ver, the risks are minimized as high risk failures are not likely toccur.

The general meaning of the RCM maintenance is optimizinghe maintenance plan based on risk analysis. The advantages andisadvantages of RCM are listed below.

Advantages:

1) The cost of the maintenance operation is optimized based onrisk.

2) It reduces the unnecessary shutdowns for low risk failures.3) It saves money paid for unnecessary close timed inspections in

case of TBM.4) It guarantees low possibility of occurrence of high risk failures.

Disadvantages:

1) Less understood by maintenance engineers and technicians.2) Complexity of building the maintenance model.3) Need for large amount of data about failures rates, modes, and

consequences.

RCM can be assumed as the most recent maintenance strategy.ore industries are converting from the regular TBM into RCM.

ccording to [41], routine preventive maintenance is reduced by0% on 11 kV transformers after using RCM. Moreover, the overallaintenance cost is reduced by 30–40% after converting to RCM.

he challenges facing the RCM are the data needed about the fail-re modes and their consequences on both the transformer itselfnd the system. This data includes recorded information from manyperating transformers about their failure modes and failure con-equences. Furthermore, very experienced persons are needed torioritize the consequences of the failure modes on the system andet the consequences indices to be able to calculate the risk.

The main aim of the asset management is to maximize the ben-fits of the asset. The benefits are maximized from the asset byerforming suitable CM techniques and/or performing good main-enance plan to maximize the usage, reducing the outage time,nd increasing the lifetime of the asst. The lifetime issue will beiscussed in the next section.

. Aging, health, and end of life assessments

Equipment aging is a fact of life in power system components. Aspiece of equipment ages, it fails more frequently and needs more

epair time until reaching its end of life [42]. Maintenance activ-ties can extend the life of equipment but become very costly forquipment near their end of life. There are three different conceptsf lifetime for power transformers: physical lifetime, technical life-ime, and economic lifetime [42].

er Systems Research 80 (2010) 456–464

(1) Physical lifetime: A piece of equipment starts to operate fromits brand-new condition until it cannot be used in its normaloperating state and must be retired.

(2) Technical lifetime: A piece of equipment may have to bereplaced due to technical reasons although it may not reachits end of physical lifetime. For example, a new technology isdeveloped for a type of equipment and manufacturers no longerproduce spare parts.

(3) Economic lifetime: A piece of equipment is no longer econom-ically valuable, although it may still be physically used. Thecapital value of any equipment is depreciated every year. Oncethe asset value approaches zero, it reaches the end of its eco-nomic lifetime.

Transformer life management has gained remarkable recogni-tion in the recent years due to the economical and technical reasons[43,44]. Due to the importance of the physical and economic life-times over the technical lifetime, physical and economic lifetimeswill be discussed in some detail in the next subsections.

4.1. Physical aging mechanisms

Most of transformer solid insulations are based on cellulose inthe form of paper. In the presence of heat, oxygen, water and otherchemicals the cellulose molecules go into chemical changes. Thesechanges cause electrical and mechanical degradation of the insu-lation paper [45]. The degradation of the solid insulation can beconsidered as the main reason for transformer end of life [46–49].Abnormal operating conditions such as repetitive overloading forlong times and non-sinusoidal loads can affect the transformeraging. The transformer physical aging mechanisms can be dividedinto intransitive aging and transitive aging [46].

4.1.1. Intransitive agingThe intransitive aging is the ability of the solid insulating

material to withstand the designed stresses such as electrical,mechanical, and thermal stresses with the passage of time [45,46].The ability of the insulating material to withstand the abovemen-tioned stresses remains constant or decreases slightly with thepassage of time until the wear-out of the asset, in which this abil-ity decreases. In what follows, the previous work done to assessthe intransitive aging and the end of life of the transformer will begiven.

4.1.1.1. Degree of polymerization (DP). The DP is the main indicationof the paper health. Paper fibers are composed of cellulose. Glucosemonomer molecules are bonded together by glycosidic bonds toform cellulose. The average length of the cellulose polymer, mea-sured as the average number of glucose monomers in the polymerchins, is referred to as DP. It was approved based on the experienceof previously retired transformers and experimental work that a200 DP or less means the end of life of the solid insulation [47–50].

The DP is a very strong accepted measure to the degradation ofthe solid insulation; however, the measurement of the DP for anytransformer needs a sample of the transformer insulation paperto perform the measurements. The paper sample taken from anyoperating transformer may cause a local damage in the windingsystem or may lead to complete failure of the transformer. Also, theDP measurement is usually done by viscometry method which isnot accurate because it is affected by the surrounding temperature[51].

4.1.1.2. Retained tensile strength. One of the main mechanicalparameters for insulation paper in transformers is tensile strength.As the paper ages, its strength against mechanical forces decreasesespecially against those arises from inrush current or short-circuits

ric Power Systems Research 80 (2010) 456–464 461

[wwiodtlt

4rtdsfs

icfiisIprmffpwfpi

D

caip

D

i

aappodo

4aiLadtCn

A.E.B. Abu-Elanien, M.M.A. Salama / Elect

52,53]. According to [54], the insulation paper reaches its end of lifehen it reaches 50% retention of tensile strength and it can be leftorking until it reaches 40% retention of tensile strength accord-

ng to [55] or it can be left working until it reaches 25% retentionf tensile strength according to [4]. The problem of the insulationamage still exists in the measurements of tensile strength becausehe measurement of the tensile strength needs sample of the insu-ating paper, which may damage the solid insulation system of theransformer.

.1.1.3. Furanic compounds. The needed paper sample for DP andetained tensile strength tests limit their usage practically. Also,he DP is associated with many errors during measurement. Theseifficulties and errors in measuring the DP and retained tensiletrength limited their usage for assessing the health of the trans-ormer despite their reliable decision in assessing the age of theolid insulation and the transformer.

As the paper insulation ages, the polymer chains starts break-ng and generating glucose monomer units that undergoes furtherhemical reaction and becomes one of a family of derivatives of 2-uraldehyde (2FAL) [53] or what is called furanic compounds thatncreases in the transformer oil with the decrease of the DP of thensulating paper [50,51,53,56,57]. These furanic compounds dis-olved in the insulating oil and it can be detected by oil analysis.t is possible to analyze the oil for a number of these furanic com-ounds as parts per billion by weight. Chendong introduces a linearelationship between the total furfural concentration in logarith-ic scale and the degree of polymerization [58]. The relation holds

or non-thermally upgraded paper but it does not apply correctlyor thermally upgraded paper [53]. The amount of the furanic com-ounds corresponding to DP of 200 units is modified in [53], inhich two formulas are proposed to relate the DP to the amount of

uranic compounds in ppb by weight (�g/kg). The first formula isroposed for thermally upgraded paper, in which the total furans

n ppb by weight (�g/kg) are used to calculate the DP, is as follows:

P = log10(total furans) − 4.0355−0.002908

(2)

According to (2), 2844 ppb by weight (�g/kg), total furans con-entration corresponds to DP of 200 units. The second formula ismodification to Chendong formula. It relates the 2FAL, measured

n ppb by weight (�g/kg), with the DP for non-thermally upgradedaper, which is in use these days, as given below:

P = log10(2FAL × 0.88) − 4.51−0.0035

(3)

According to (3), 2FAL concentration corresponds to a DP of 200s 6457 ppb by weight (�g/kg).

Assessment of the transformer aging using furanic compoundsnalysis is growing more these days because this method givescceptable indication of the solid insulation age without using anyaper specimen. However, the absolute correlation of furanic com-ounds to DP varies from transformer to another and it dependsn humidity, operating temperature, type of oil and paper, andesign. Further research on the dependency of furanic compoundsn moisture and temperature is necessary [48].

.1.1.4. Dissolved gas analysis. As a transformer ages, the cellulosend oil degrade. The rate of cellulose and oil degradation is signif-cantly increased in the presence of a fault inside the transformer.ow temperature thermal degradation of cellulose produces CO2

nd high temperature produces CO [56]. High rate of paper degra-ation is estimated when the ethylene concentration increases andhe CO2/CO ratio decreases below a ratio of about 6 [56,58]. At aO2/CO ratio less than 2, the probability of failure increases sig-ificantly [56,59]. However, the scatter of measurements made on

Fig. 5. Charging and return voltage during RVM test [52].

service transformers is so large that no reliable values were foundto assess the transformer end of life. Transformer service lives canonly be estimated to within ±10 years using generation of CO andCO2 phenomenon [59].

4.1.1.5. Recovery voltage measurement. The rate of paper degra-dation depends on several parameters such as pulp composition,thermal upgrading, moisture content, and temperature. The higherthe water contents of the paper, the higher the degradation rate[46,56]. The RVM technique uses the dielectric response to evaluateits condition with respect to moisture content [46].

To perform RVM, first a sample is charged from a high volt-age source for a charging time (tc). Then, the sample is isolatedfrom the HV source and short-circuited for a discharging time (td),where tc > td. At the end of discharging time (td), the short-circuitis removed and the return voltage appearing at the electrodes ismeasured as shown in Fig. 5, where Vr is the maximum value ofthe return voltage, treak is the time at which the maximum returnvoltage is occurred, and Sr is the initial slope of the return voltage.

A polarization spectrum can be obtained by plotting the valueof (Vr) for different values of the charging time (tc). The time atmaximum (Vr) in that polarization spectrum is called the domi-nant time constant. If the condition of the oil–paper insulation ishomogeneous, and the distribution of the temperature, moisture,and aging in the insulation is uniform, the resulting curves willhave one dominant time constant [56], otherwise, it will have manydominant time constants.

4.1.2. Transitive agingIt is the rapid aging of the asset when subjected to abnormal con-

dition [45]. The abnormal conditions may be overloading, supplyingnon-sinusoidal loads, or exposure to higher ambient temperaturethan normal. The main reason for accelerating the end of life ofa transformer under the abovementioned abnormal conditions isthe increase of its HST over normal accepted values. This increasein HST has an effect of reducing insulation life [45,46,60]. Accordingto [4] the HST can be calculated as follows

�HS = �A + ��TO + ��H (4)

where �HS is the temperature of hot spot in ◦C, �A the ambienttemperature in ◦C, ��TO the top oil temperature rise over ambientin ◦C, and ��H is the winding HST rise over top oil in ◦C.

More details for HST calculation can be found in [4]. The value ofthe HST affects the lifetime of the transformer solid insulation. Theincrease of the transformer HST accelerates the end of the trans-former lifetime and vice versa. The average lifetimes of the solid

insulation of the oil immersed transformers based on different endof life criteria are documented in [4].

The relationship between the HST and the transformer life con-sumption is governed by the Arrhenius reaction rate theory which

462 A.E.B. Abu-Elanien, M.M.A. Salama / Electric Power Systems Research 80 (2010) 456–464

e tran

s

p

w

oiaacgt

ga

F

wt(ti

ttctHbolHFa

4

uvbc

Fig. 6. A complete classification of th

tates that [4,54,61–64]:

er unit life = A e(B/�HS+273) (5)

here A and B are empirical constants.These constants (A and B) are based on material characteristics

f the insulation and they are determined such that per unit lifes unity at HST of 110 ◦C. The values of A and B are (9.8 × 10−18)nd (15,000) respectively [4,54,61–64]. The reciprocal of (5) is theging acceleration factor (FAA) which can be used to calculate theonsumed life for a given HST over a given period. FAA has a valuereater than unity for winding hottest spot temperatures greaterhan 110 ◦C and vice versa.

The equivalent life (in hours or days) that will be consumed in aiven time period for the given temperature cycle can be calculateds shown below [4,62,64]:

eq =∑N

n=1FAAn �tn∑N

n=1�tn

(6)

here Feq is the equivalent aging factor for the total time period, nhe index of the time interval, t, N the total number of time intervalsusually 24 h), FAAn the aging acceleration factor for the tempera-ure which exists during the time interval �tn, and �tn is timenterval, hours.

The value of Feq is higher than unity if HST for the day is higherhan 110 ◦C and vice versa. Using the previously mentioned steps,he lost lifetime of the transformer due to the increase in the HSTan be calculated. Generally, transitive aging causes acceleration ofhe transformer end of life which is mainly due to the increase in theST [4,54,61–64]. It is clear that calculating the transformer agingased on the HST is well understood; however, the new researchn this topic focuses on the correct modeling of the transformeroad and the ambient temperature to achieve reliable values of theST and accordingly reliable values of transformer loss of life [62].ig. 6 shows a complete classification of the transformer physicalging mechanisms.

.2. Mechanisms and modeling of economic aging

As the asset is purchased, it losses part of its value every yearntil reaching zero value or its salvage value, where the salvagealue is an estimate of the value of the asset at the time it wille disposed of; it may be zero. The yearly lost part of the assetost is called the depreciation cost. The lifetime of the asset ends

sformer physical aging mechanisms.

when it depreciates to zero or the salvage value [65]. The time-based depreciation has two main types [65–67]: the straight-linedepreciation method and the accelerated depreciation method.

4.2.1. Straight-line depreciationStraight-line depreciation is the simplest and most often used

technique, in which the asset is assumed to lose equal amounts ofits capital cost throughout its lifetime. In other words, the annualdepreciation cost equals the capital cost of the asset minus its sal-vage value divided by number of years of its useful lifetime (usefullifetime is the average lifetime for such type of asset) [65–67].

4.2.2. Accelerated depreciationThe accelerated depreciation assumes the asset to lose higher

amounts of its value during the initial years and less in the lateryears. It includes the reducing balance depreciation and some ofyears digits depreciation (SOY). The reducing balance depreciationprovides a steady declining balance of the depreciation cost over theestimated lifetime of the asset. The most common ways to calcu-late the reducing balance deprecation are 200% and 150% reducingbalance depreciation [65–67]. The annual depreciation cost for year(k) can be calculated as follows

d(k) = p

lifetime× C(k) (7)

where d(k) is the depreciation cost for year (k), p the reducingbalance value (1.5 for 150% and 2 for 200% reducing balance depre-ciation), lifetime the expected lifetime in years, and C(k) is the assetvalue at the beginning of year (k).

The asset value at the beginning of year (k) can be calculated asfollows:

C(k) = C −k−1∑

i=1

d(i) (8)

where C is the asset capital cost, and∑k−1

i=1 d(i) is the accumulateddepreciation of the asset from the installation year until the begin-ning of year (k).

Like the reducing balance depreciation, SOY depreciation alsouses steadily declining periodic amounts. SOY is performed byapplying successively smaller depreciation amounts each year tothe asset value at the beginning of the calculation year. The annualdepreciation for year k using the SOY method can be calculated as

A.E.B. Abu-Elanien, M.M.A. Salama / Electric Pow

f

d

wots

blnamfotttic

$ciswoam

5

lmtCnaAatttact

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

Fig. 7. EUAC for a transformer.

ollows [65–67]

(k) = lifetime − k + 1S

× (C − SV) (9)

here lifetime is the expected lifetime in years and S is the sumf the years digits of the lifetime, for example, If the expected life-ime is 4 years, then, S = 1 + 2 + 3 + 4 = 10, and SV is the transformeralvage value.

It is sometimes economic to retire and replace the equipmentefore its capital value reaches zero and before the end of physical

ifetime rather than continue to face high operating and mainte-ance costs. In this regard, the operating and maintenance costsre taken in the economic end of life decision by calculating theinimum Equivalent Uniform Annual Cost (EUAC) of the trans-

ormer [66,67]. EUAC relays on converting all un-equal annual costsf the transformer into equivalent equal annual cost starting fromhe decision year and ending at each year of the remaining years ofhe transformer useful lifetime. The result is a curve that representshe equivalent equal annual cost of maintaining the transformern service until each year of the expected remaining lifetime. Theurve is like the one shown in Fig. 7.

According to Fig. 7, the transformer owner will pay around7000 for leaving the transformer in service for one year after thealculation date, and he will pay around $5500 each year for leav-ng the transformer in service 2 years after the calculation date ando on. For the case shown in Fig. 7, the minimum EUAC is $1300,hich represents the amount of money paid by the transformer

wner every year for leaving the transformer in service 12 yearsfter the calculation date. According to the minimum EUAC, theost economic end of life is 12 years after the decision date.

. Conclusion

An overview on the transformer asset management was high-ighted in this paper. The general classification of the asset

anagement activities was illustrated. The three major activi-ies of the transformer asset management are the application ofM techniques in the transformer operation, performing mainte-ance plans and investigating new less-cost maintenance methods,nd assessing the health and end of life of the transformer.

detailed explanation of each transformer asset management

ctivity was investigated in this paper. The various CM and CAechniques used to monitor and assess the condition of theransformer were discussed in detail. Transformer maintenanceechniques were highlighted including the advantages and dis-dvantages of each type. The transformer lifetime types werelassified and the different end-of-life criteria were illustrated inhis paper.

[

[

er Systems Research 80 (2010) 456–464 463

References

[1] W. Young, Transformer life management—condition monitoring, in: Proc. IEEColloq. on Transform, Life Manag., London, UK, 1998.

[2] C. Myers, Transformer conditioning monitoring by oil analysis large orsmall; contaminant or catastrophe, in: Proc. 1st IEE Int. Conf. on Power Stn.Maint.—Profitab. through Reliab, Edinburgh, UK, 1998.

[3] M. Arshad, S.M. Islam, A. Khaliq, Power transform. asset manag, in: Proc. Int.Conf. on Power Syst. Technol., PowerCon, 2004.

[4] ANSI/IEEE C57.91, IEEE Guide for Loading Mineral-Oil-Immersed Transformers,1995.

[5] J.K. Pylvänäinen, K. Nousiainen, P. Verho, Studies to utilize loading guides andANN for oil-immersed distribution transformer condition monitoring, IEEETrans. Power Deliv. 22 (2007) 201–207.

[6] W.H. Tang, Q.H. Wu, Z.J. Richardson, Equivalent heat circuit based power trans-former thermal model, IEE Proc. Electr. Power Appl. 149 (2002) 87–92.

[7] J.Q. Feng, P. Sun, W.H. Tang, D.P. Buse, Q.H. Wu, Z. Richardson, J. Fitch, Imple-mentation of a power transformer temperature monitoring system, in: Proc.Int. Conf. on Power Syst. Technol., PowerCon, 2002.

[8] O. Roizman, V. Davydov, Neuro-fuzzy algorithms for power transformers diag-nostic, in: Proc. Int. Conf. on Power Syst. Technol., PowerCon, 2000.

[9] J. Shengchang, S. Ping, L. Yanming, X. Dake, C. Junling, The vibration measuringsystem for monitoring core and winding condition of power transformer, in:Proc. 2001 Int. Symp. on Electr. Insulating Mater, ISEIM, 2001.

10] P. Kang, D. Birtwhistle, Condition assessment of power transformer on-loadtap-changers using wavelet analysis, IEEE Trans. Power Deliv. 16 (2001)394–400.

11] P. Kang, D. Birtwhistle, Condition assessment of power transformer on-loadtap changers using wavelet analysis and self-organizing map: field evaluation,IEEE Trans. Power Deliv. 18 (2001) 78–84.

12] S.M. Strachan, S.D.J. McArthur, M.D. Judd, J.R. McDonald, Incrementalknowledge-based partial discharge diagnosis in oil-filled power transformers,in: Proc. 13th Int. Conf. on Intell. Syst. Appl. to Power Syst, 2005.

13] A. Zargari, T.R. Blackbum, Acoustic detection of partial discharges using non-intrusive optical fiber sensors, in: Proc. IEEE Int. Conf. on Conduct. andBreakdown in Solid Dielectr., Vasteras, Sweden, 1998.

14] S.D.J. McArthur, S.M. Strachan, G. Jahn, The design of a multi-agent trans-former condition monitoring system, IEEE Trans. Power Deliv. 19 (2004) 1845–1852.

15] M.D. Judd, S.D.J. McArthur, J.R. McDonald, O. Farish, Intelligent conditionmonitoring and asset management: partial discharge monitoring for powertransformers, IEE Power Eng. J. 16 (2002) 297–304.

16] H. Zhang, T.R. Blackburn, B.T. Phung, D. Sen, A novel wavelet de-noising methodfor on-site PD measurements on HV cables, in: Proc. 2005 Int. Symp. on Elec-trical Insulating Mater., Kitakyushu, Japan, 2005.

17] H. Mingyou, H. Xie, T.B. Tiong, X. Wu, Study on a spatially selective noise filtra-tion technique for suppressing noises in partial discharge on-line monitoring,in: Proc. the 6th Int. Conf. on Prop. and Appl. of Dielectr. Mater., 2000.

18] P. Werle, A. Akbari, H. Borsi, E. Gockenbach, Partial discharge localization onpower transformers using neural networks combined with sectional windingtransfer function as knowledge base, in: Proc. Int. Symp. on Electr. InsulatingMater., 2001.

19] J.S. Foo, P.S. Gosh, Artificial neural network modeling of partial dischargeparameters for transformer oil diagnosis, in: Proc. Annual Rep. Conf. on Electr.Insulation and Dielectr. Phenom., 2002.

20] IEEE std. C57.104, IEEE Guide for the interpretation of gases generated in oil-immersed transformers, 2008.

21] M. Domun, Condition monitoring of power transformers by oil analysis tech-niques, in: Proc. IEE Colloq. on Cond. Monit. and Remnant Life Assess. in PowerTransform., 1994.

22] B. Pahlavanpour, A. Wilson, Analysis of transformer oil for transformer condi-tion monitoring, Proc. IEE Colloq. Eng. Rev. Liq. Insul. (1997).

23] Y.C. Huang, C.M. Huang, Evolving wavelet networks for power transformercondition monitoring, IEEE Trans. Power Deliv. 17 (2002) 412–416.

24] P.M. Nirgude, B. Gunasekaran, Channakeshava, A.D. Rajkumar, B.P. Singh, Fre-quency response analysis approach for condition monitoring of transformer,in: Proc. Annual Rep. Conf. on Electr. Insulation and Dielectr. Phenom., 2004.

25] N. AI-Khayat, L. Haydock, G. McDowell, Swept frequency response tests for con-dition monitoring of power transformers, in: Proc. Electr. Electron. InsulationConf. and Electr. Manuf. & Coil Winding Conf., 1995.

26] S. Birlasekaran, F. Fetherston, Off/on-line FRA condition monitoring techniquefor power transformer, IEEE Power Eng. Rev. 18 (1999) 54–56.

27] A. Kingsmill, S. Jones, J. Zhu, Application of new condition monitoring technolo-gies in the electricity transmission industry, in: Proc. Sixth Int. Conf. on Electr.Mach. and Syst., ICEMS, Sydney, Australia, 2003.

28] S.D.J. McArthur, E.M. Davidson, V.M. Catterson, A.L. Dimeas, N.D. Hatziar-gyriou, F. Ponci, T. Funabashi, Multi-agent systems for power engineeringapplications—Part I: concepts, approaches, and technical challenges, IEEETrans. Power Syst. 22 (2007) 1743–1752.

29] S.D.J. McArthur, E.M. Davidson, V.M. Catterson, A.L. Dimeas, N.D. Hatziar-

gyriou, F. Ponci, T. Funabashi, Multi-agent systems for power engineeringapplications—part II: technologies, standards, and tools for building multi-agent systems, IEEE Trans. Power Syst. 22 (2007) 1753–1759.

30] O. Wilhelmsson, Evaluation of the introduction of RCM for hydro power gen-erators at Vattenfall Vattenkraft, Master thesis, R. Inst. of Technol., Sweden,2005.

4 ic Pow

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[oil-cooled transformers, in: Proc. Rural Electr. Power Conf., Little Rock, AR, 2001.

64 A.E.B. Abu-Elanien, M.M.A. Salama / Electr

31] A. Setayeshmehr, A. Akbari, H. Borsi, E. Gockenbach, A procedure for diagnosisand condition based maintenance for power transformers, in: Proc. IEEE Int.Symp. on Electr. Insulation, Indianapolis, USA, 2004.

32] Y. Zhu, L. Wu, X. Li, J. Yuan, A transformer condition assessment frameworkbased on data mining, Proc. IEEE Power Eng. Society Gen. Meet. (2005).

33] M. Wang, A.J. Vandermaar, K.D. Srivastava, Review of condition assessment ofpower transformers in service, IEEE Electr. Insulation Mag. 18 (2002) 12–25.

34] J.Q. Feng, C. Ma, W.H. Tang, J.S. Smith, Q.H. Wu, A transformer predictivemaintenance system based on agent-oriented programming, in: Proc. IEEE/PESTransm. and Distrib. Conf. & Exhibition: Asia and Pacific, Dalian, China, 2005.

35] S. Costa, A.F. Brandcio, Optimization of the maintenance of oil filled powertransformers based on reliability, in: Proc. IEE/PES Transm. & Distrib. Conf. &Exposition, Latin America, 2004.

36] M.E. Beehler, Reliability centered maintenance for transmission systems, IEEETrans. Power Deliv. 12 (1997) 1023–1028.

37] B. Handley, M. Redfern, S. White, On load tap-changer conditioned based main-tenance, IEE Proc. Gener. Transm. Distrib. 148 (2001) 296–300.

38] G. Theil, Parameter evaluation for extended Markov models applied tocondition- and reliability-centered maintenance planning strategies, in: Proc.9th Int. Conf. on Probab. Methods Applied to Power Syst., KTH, Stockholm,Sweden, 2006.

39] N. Liu, W. Gao, K. Tan, L. Wang, G. Liang, W. Li, Decision on maintenance periodfor power transformer based on monte carlo simulation of reliability, in: Proc.the 7th Int. Conf. on Prop. and Appl. of Dielectr. Mater., Nagoya, June, 2003.

40] I.P. Siqueira, Software requirements for reliability-centered maintenance appli-cation, in: Proc. 9th Int. Conf. on Probab. Methods Applied to Power Syst., KTH,Stockholm, Sweden, June, 2006.

41] A.K.S. Jardine, A.H.C. Tsang, Maintenance, Replacement, and Reliability: Theoryand Applications, Boca Raton, Florida, 2006.

42] W. Li, E. Vaahedi, P. Choudhury, Power system equipment aging, IEEE PowerEnergy Mag. 4 (May/June) (2006) 52–58.

43] M. Wang, A.J. Vandermaar, K.D. Srivastava, Review of condition assessment ofpower transformers in service, IEEE Electr. Insul. Mag. 18 (2002) 12–25.

44] CIGRÉ Working Group 05, An international survey on failures in large powertransformers in service, Electra, no. 88, May 1983.

45] M. Arshad, S. Islam, A novel fuzzy logic technique for power transformer assetmanagement, in: Proc. 41st IAS Annual Meet. Ind. Appl. Conf., October 2006.

46] J.P. Van Bolhuis, E. Gulski, J.J. Smit, Monitoring and diagnostic of transformersolid insulation, IEEE Trans. Power Deliv. 17 (2002) 528–536.

47] H. Herman, M.J. Shenton, G.C. Stevens, R.J. Heywood, A new approach to con-dition assessment and lifetime prediction of paper and oil used as transformerinsulation, in: Proc. 7th Int. Conf. on Solid Dielectr., Eindhoven, Neth., June,

2001.

48] M. Arshad, S.M. Islam, A. Khaliq, Power transformer insulation response andrisk assessment, in: Proc. 8th Int. Conf. on Probab. Methods Applied to PowerSyst., Iowa State University, Ames, Iowa, September, 2004.

49] A.J. Kachler, I. Höhlein, Aging of cellulose at transformer service temperatures.Part 1: Influence of type of oil and air on the degree of polymerization of press-

[

[

[

er Systems Research 80 (2010) 456–464

board, dissolved gases, and furanic compounds in oil, IEEE Electr. Insul. Mag.21 (2005) 15–21.

50] P. Pahlavanpourl, Eklund, M.A. Martins, Insulating paper ageing and furfuralformation, in: Proc. Electr. Insul. Conf. and Electr. Manuf. & Coil Winding Tech-nol., September 2003.

51] M. Ali, C. Eley, A.M. Emsleyr, X. Xaio, Measuring and understanding the aging ofkraft insulating paper in power transformers, IEEE Electr. Insul. Mag. 12 (1996)28–34.

52] M. Mirzaiel, A. Gholamil, H. Tayyebi, Thermal degradation of cellulose paperinsulation in power transformers, in: Proc. Int. Conf. on Solid Dielectr., Winch-ester, UK, July, 2007.

53] R.D. Stebbins, D.S. Myers, A.B. Shkolnik, Furanic compounds in dielectric liq-uid samples: review and update of diagnostic interpretation and estimationof insulation ageing, in: Proc. the 7th Int. Conf. on Prop. and Appl. of Dielectr.Mater., Nagoya, June, 2003.

54] ANSI/IEEE C57.92, IEEE guide for loading mineral oil immersed power trans-formers, 1981.

55] T.A. Prevost, H.P. Gasser, R. Wicks, B. Glenn, R. Marek, Estimation of insulationlife based on dual temperature aging model, in: Proc. Weidmann-ACTI Inc. FifthAnnual Tech. Conf., Albuquerque, November, 2006.

56] T.K. Saha, Review of modern diagnostic techniques for assessing insulationcondition in aged transformers, IEEE Trans. Dielectr. Electr. Insul. 10 (2003)903–917.

57] A.M. Emsley, X. Xiao, R.J. Heywood, M. Ali, Degradation of cellulosic insulationin power transformers. Part 2: Formation of furan products in insulating oil,IEE Proc. Sci. Measur. Technol. 147 (2000) 110–114.

58] I. Chendong, Monitoring paper insulation aging by measuring furfural contentsin oil, in: Proc. 7th Int. Symp. on High Volt. Eng., 1991.

59] A.M. Emsley, G.C. Stevens, Review of chemical indicators of degradation ofcellulosic electrical paper insulation in oil-filled transformers, IEE Proc. Sci.Measur. Technol. 4 (1994) 324–334.

60] S. Yakov, M. Nigris, P.G. Antonioli, Study of the life consumption of distributiontransformers and assessment of end-of-life criteria, in: Proc. 12th Int. Conf. onElectr. Distrib., CIRED, Birmingham, UK, May, 1993.

61] ANSI/IEEE C57.91, IEEE guide for loading oil immersed overhead and padmounted distribution transformers, 1981.

62] K.T. Muthanna, A. Sarkar, K. Das, K. Waldner, Transformer insulation life assess-ment, IEEE Trans. Power Deliv. 21 (2006) 150–156.

63] W.J. McNutt, Insulation thermal life considerations for transformer loadingguides, IEEE Trans. Power Deliv. 7 (1992) 392–401.

64] P.K. Sen, S. Pansuwan, Overloading and loss-of-life assessment guidelines of

65] M.S. Bowman, Applied Economic Analysis For Technologists, Engineers, andManagers, Prentice Hall, 1999.

66] E.P. Degarmo, W.G. Sullivan, J.A. Bontadelli, Engineering Economy, Macmillanpublishers, 1993.

67] P.E.M. Kurtz, Handbook of Engineering Economics, McGraw-Hill, 1984.