research article fuzzy algorithm for power transformer...

Hindawi Publishing CorporationAdvances in Fuzzy SystemsVolume 2013 Article ID 421621 7 pageshttpdxdoiorg1011552013421621

Research ArticleFuzzy Algorithm for Power Transformer Diagnostics

Nitin K Dhote1 and Jagdish B Helonde2

1 St Vincent Pallotti College of Engineering and Technology Wardha Road Nagpur 441108 India2 Principal ITM College of Engineering Nagpur India

Correspondence should be addressed to Nitin K Dhote nitindhoteyahoocom

Received 1 March 2013 Revised 13 May 2013 Accepted 28 May 2013

Academic Editor M Onder Efe

Copyright copy 2013 N K Dhote and J B Helonde This is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

Dissolved gas analysis (DGA) of transformer oil has been one of the most reliable techniques to detect the incipient faults Manyconventional DGA methods have been developed to interpret DGA results obtained from gas chromatography Although thesemethods are widely used in the world they sometimes fail to diagnose especially when DGA results fall outside conventionalmethods codes or when more than one fault exist in the transformer To overcome these limitations the fuzzy inference system(FIS) is proposed Twohundred different cases are used to test the accuracy of variousDGAmethods in interpreting the transformercondition

1 Introduction

The power transformer is a vital equipment of the electricalpower system A transformer may function well externallywith monitors while some incipient deteriorations mayoccur internally to cause fatal problems in later developmentNearly 80 of faults result from incipient deteriorationsTherefore faults should be identified and avoided at theearliest possible stage by some predictive maintenance tech-niqueDissolved gas analysis (DGA) is a reliable technique fordetection of incipient faults in oil-filled power transformerIt is similar to a blood test or a scanner examination ofthe human body it can warn about an impending problemgive an early diagnosis and increase the chances of findingthe appropriate cure The working principle [1ndash4] is basedon the dielectric breakdown of some of the oil moleculesor cellulose molecules of the insulation due to incipientfaults When there is any kind of fault such as overheatingor discharge inside the transformer it will produce corre-sponding characteristic amount of gases in the transformeroil These gases are detected at the per part million (ppm)level by gas chromatography [5ndash10] It is a technique ofseparation identification and quantification of mixturesof gases The commonly collected and analyzed gases arehydrogen (H

2) methane (CH

4) acetylene (C

2H2) ethylene

(C2H4) ethane (C

2H6) carbon monoxide (CO) and carbon

dioxide (CO2) Through the analysis of the concentrations of

dissolved gases their gassing rates and the ratios of certaingases the DGA methods can determine the fault type ofthe transformer Even under normal transformer operationalconditions some of these gases may be formed inside

Therefore it is necessary to build concentration normsfrom a sufficiently large sampling to assess the statistics

2 DGA Interpretation

If an incipient fault is present in the transformer concen-tration of gases dissolved in the oil significantly increases Agiven gas volume may be generated over a long time periodby a relatively insignificant fault or in a very short time periodby a more severe fault Once a suspicious gasrsquos presence isdetected it is important to be certain whether the fault thatgenerated the gas is active by calculating the total dissolvedcombustible gases (TDCG) and rate of TDCG (10) which isgiven by

119877 =(119878119879minus 1198780) sdot 119881 sdot 10

minus6

119879 (1)

2 Advances in Fuzzy Systems

Table 1 Coding rule for IEC method [11]

Codes Range of gas ratios1198771 1198772 1198773

0 lt01 01ndash1 lt11 01ndash3 lt01 1ndash32 gt3 gt1 gt3

Table 2 Classification of faults by IEC method [11]

Faulttype Characteristic fault 1198771 1198772 1198773

1 Normal ageing (N) 0 0 02 Partial discharge (PD) of low energy density 0 1 03 PD of high energy density 1 1 04 Discharges of low energy (D1) 1-2 0 1-25 Discharge of high energy (D2) 1 0 26 Thermal fault of low temperature (TL) lt 150∘C 0 0 17 TL between 150∘C and 300∘C 0 2 08 TL between 300∘C and 700∘C 0 2 19 Thermal fault of high temperature (TH) gt 700∘C 0 2 2

where119877 is the rate (litersday) So is the TDCG of first samplein ppm 119878

119879is the TDCG of second sample in ppm 119881 is tank

oil volume in liters and 119879 is the time (days)The rate of generation of TDCGgreater than 28 litersday

indicates that the transformer has an active internal fault andrequires additional inspection by DGA methods

Many interpretative methods employ an array of ratiosof certain key combustible gases as the fault type indicatorsThese five ratios are

1198771 = C2H2C2H4

1198772 = CH4H2

1198773 = C2H4C2H6

1198774 = C2H6C2H2

1198775 = C2H2CH4

Rogersrsquo method [13ndash15] utilizes three ratios 1198771 1198772 and 1198773The method gives fault for the specific combination of thesegas ratios Dornenburg [13ndash15] utilizes four ratios 1198771 1198772 1198774and 1198775 This procedure requires significant levels of gases tobe present for the diagnosis to be validThemethod gives faultafter comparing these ratios to the limiting values

Amongst ratio methods IEC Standard 60599 [11] is mostwidely used It also utilizes three ratios 1198771 1198772 and 1198773 Thecoding rule and classification of faults by the IECmethod aregiven in Tables 1 and 2

Incipient faults can be reliably identified by visual inspec-tion [16] of the equipment after the fault has occurred inservice as follows

(a) PDndashpossible X wax formation and sparking inducingsmall carbonized punctures in the paper

(b) D1ndashlarger punctures in the paper tracking or carbonparticles in oil

(c) D2ndashextensive carbonization metal fusion and possi-ble tripping of the equipment

(d) TLndashfor TL lt 300∘C the paper turns brown for TL gt300∘C the paper carbonizes

(e) THndashoil carbonization metal coloration or fusion

TheDuval Triangle [17ndash19] method utilizes three ratiosof certain gases for DGA interpretation of transformers filledwithmineral oilThe triangular coordinates corresponding toDGA results in ppm can be calculated by (2) as follows

C2H2=100 sdot 119909

(119909 + 119910 + 119911)

C2H4=100 sdot 119910

(119909 + 119910 + 119911)

CH4=100 sdot 119911

(119909 + 119910 + 119911)

(2)

where 119909 119910 and 119911 are concentrations of C2H2 C2H4 and

CH4in ppm respectively Along with three ratios given by

Duval Triangle method fourth ratio [12] is also used forfault diagnosis which is given by

H2=

100 sdotH2

(H2+ C2H6+ CO + CO

2) (3)

All these techniques are computationally straightforwardHowever these methods in some cases provide erroneousdiagnoses as well as no conclusion for the fault type

To overcome these limitations FIS is proposed

3 Diagnostic Procedure

Flow chart of proposed system diagnosis is shown in Figure 1The input data include concentration of dissolved gasesC2H2 C2H4 C2H6 CH4 H2 CO and CO

2of the sample

Information such as tank oil volume date of samplingand date of installation of transformer is asked for furtherinference

In the first step the system calculates TDCG and com-pares with the standard permissible limits (IEEE standard2008) For normal level of TDCG (lt720 ppm) permissiblelimits for individual gases are checked The normal levelof TDCG and individual gases indicates the satisfactoryoperation of a transformer Once an abnormal level of TDCGor individual gas has been detected the next step is todetermine the rate of generation of TDCG (1) by analysis ofthe successive sample For the normal rate of TDCG (less than28 litersday) further diagnosis is bypassed For an abnormalrate of TDCG the proposed FIS is adopted to diagnose theprobable faults In the last step severity degrees are assignedto the diagnosed faults On the basis of severity of faultsappropriate maintenance actions are suggested

4 Fuzzy Inference System (FIS)

Intelligent algorithms for example expert system [20] FIS[21 22] artificial neural networks [23ndash25] probabilistic

Advances in Fuzzy Systems 3

Normal

Smaller

DGAsample

Greater

Normal ageing(1)

Abnormal

Abnormal

Normal

1

Maintenanceactions

FIS

TDCG

Rate ofTDCG

Individualgas conc

Figure 1 Flow chart of proposed system diagnosis

neural networks [26] evolving neural networks [27] artificialneural FIS [28ndash31] wavelet networks [32] and combinedneural networks and expert system [33] have been usedto interpret DGA results These algorithms are not entirelysatisfactory These methods are mostly suitable for trans-formers with single fault In case of multiple faults onlydominant fault is indicated by these methodsThesemethodsare based on specific set of codes defined for certain gas ratiosFurther no quantitative indication for severity of fault andmaintenance suggestions is given by these methods

The proposed fuzzy diagnostic method is prepared usingtheMATLABFuzzy Logic Toolbox [34] Sugeno type FIS [35ndash37] is used as a fuzzy inference method

The rule for the zero-order Sugeno model is given belowIf input 1 = 119909 and input 2 = 119910 then output 119911 = constant

The output level 119911 of each rule is weighted by firingstrength 119908

119894of the rule For an AND rule firing strength is

given as

119908119894= AND method [1198651 (119909) 1198652 (119910)] (4)

where 1198651(119909) and 1198652(119910) are the membership functions forinput 1 and input 2The final output of the system is weightedaverage of all the rules output which is given by

119884 =sum119873

119894=1119908119894119911119894

sum119873

119894=1119908119894

(5)

where 119884 is final output and119873 is the number of rulesThe FIS consists of 3 ratios 1198771 1198772 and 1198773 as inputs

The coding rule for ratios is kept the same as IEC Method(Table 1) One of the major drawbacks of IEC method isthat when gas ratio changes across coding boundary thecode changes sharply between 0 1 and 2 In fact the gasratio boundary should be fuzzy Depending on the relativevalues of ratios IEC codes 0 1 and 2 are replaced by fuzzy

Input0 1 2 3 4

1Low Med High

09 11 27 33

Mem

bers

hip

degr

ee

variable R3

Figure 2 Membership function for ratio 1198773

codes Low Medium (Med) and High Due to uncertainty inmeasurements of gas concentrations by gas analyzers the gasratios would have a relative uncertainty of plus or minus 10[38] Membership function for code 0 of ratio 1198771 is given bythe linear declining function

120583Low (1198771) =

1 997888rarr 1198771 le 009

(011 minus 1198771)

(011 minus 009)009 le 1198771 le 011

0 1198771 ge 011

(6)

Membership function for code 1 of ratio 1198771 is given by trap-ezoidal function

120583Med (1198771) =

0 997888rarr 1198771 le 009

(1198771 minus 009)

(011 minus 009)997888rarr 009 le 1198771 le 011

1 011 le 1198771 le 27

(33 minus 1198771)

(33 minus 27)27 le 1198771 le 33

0 1198771 ge 33

(7)

Membership function for code 2 of ratio 1198771 is given by linearincreasing function

120583High (1198771) =

0 997888rarr 1198771 le 27

(1198771 minus 27)

(33 minus 27)997888rarr 27 le 1198771 le 33

1 1198771 ge 33

(8)

The codes of the ratios 1198772 and 1198773 are also fuzzified as LowMed and High variable depending on the range of ratios forthese codes Membership function for ratio 1198773 is shown inFigure 2

The FIS comprises of single output which has 5 fault typesas membership functions Weight in the range of 0 to 1 isassigned to each fault type on the basis of severity of the faultThe five types of faults used in FIS are TL ⟨02⟩ PD ⟨04⟩ D1⟨06⟩ TH ⟨08⟩ and D2 ⟨10⟩

The major drawback of the IEC method is that 16 IECcode combinations out the possible 27 do not indicate anyfault Only 11 inference rules are suggested by the IEC(Table 2) out of the 27 (3 times 3 times 3) possible rules To overcomethis limitation existing 11 rules are modified in terms offuzzy variables and additional 16 new rules are obtained as a


Table 3 Comparison of accuracy of different methods

119879119877

119879119875

119860119875

119860119862

Doernenburg 77 111 6937 3850Roger 89 145 6138 4450IEC 142 170 8353 7100Duval Triangle 172 200 8600 8600Li et al [12] 181 200 9050 9050Proposed method 187 200 9350 9350

result of extensive consultations with utility experts existingliterature and approximately 1500 DGA case histories Eachrule consists of two components which are the antecedent(IF part) and the consequent (THEN part) The rules havinga similar output are clubbed together and kept in order ofincreasing value of the severity of fault The fuzzy rules aregiven below

Rule 1 if 1198771 = Low 1198772 = Low 1198773 = Med then fault = TL

Rule 2 if 1198771 = Low 1198772 = Med 1198773 = Low then fault = TL

Rule 3 if 1198771 = Low 1198772 = Med 1198773 = Med then fault = TLFIS derives output from judging all the fuzzy rules by

finding the weighted average of all 27 fuzzy rules output

5 Case Studies Results and Discussions

FIS is developed based on the proposed interpretative rulesand diagnostic procedure of an overall system To demon-strate the feasibility of the system in diagnostic 200 DGA gasrecords supplied by themajor power companies in India havebeen tested

Accuracy is calculated in two different ways as follows

(a) When considering only the number of predictionspercentage accuracy is given by

119860119875=100 sdot 119879

119877

119879119875

(9)

where 119879119877is the number of correct predictions and 119879

119875

is the total number of the predictions(b) When considering the total number of cases percent-

age accuracy is given in by

119860119862=100 sdot 119879

119877

119879119862

(10)

where 119879119862is the total number of cases

Accuracy values of different methods for total 200 casesare compared and summarized in Table 3 Results from threecase studies are presented here

51 Case Study-I A 10MVA 132KV11 KV transformer is inservice for 11 years Tank oil volume is 12000 liters On loadtap changer inside main tank had intermittent sparking onsome of the contacts One nail was found on the shield of

bottom tank and few burnswere observed on the nail and thebolt DGA data obtained in ppm after the fault on 13022009is as follows C

2H2-4 C2H4-54 CH

4-09 H

2-78 C

2H6-67

CO-670 CO2-1243

Step 1 TDCG in ppm = 882 TDCG is above normal(gt720 ppm)

Step 2 The transformer is sampled again on 20022009 todetermine rate of TDCG Concentrations of dissolved gasesin ppm are as follows C

2H2-6 C2H4-73 CH

4-14 H

2-158

C2H6-75 CO-831 CO

2-1430

TDCG in ppm = 1157 rate of TDCG = 471 litday whichis greater than the normal level (28 litday)

Step 3 FIS is applied for fault diagnosis The output of FIS isgiven by rule viewer which is shown in Figure 3 Rule viewershows 1198771 = 0082 (Low) 1198772 = 0088 (Low) and 1198773 = 0973which lies on the boundary of the fuzzy ratios Low and MedDark dots in the first and eighth rows of the fault columnshow that rules 1 and 8 are satisfied which indicates possiblefaults TL and PD respectively This result matches the actualfault of the transformer Weighted average of both rules isgiven as 0327 Weight of both faults can be calculated asfollows

weight of TL = (04 minus 0327)(04 minus 02)

= 0365

weight of PD = 1 minus 0365 = 0635(11)

The weights point towards the strong possibility of fault PDand the relatively less possibility of fault TL The key featureof the proposedmethod is that it can diagnose multiple faultsunlike conventional DGA methods

Step 4 Severity of faults is Medium Maintenance actionssuggested are as follows

(1) Observe caution(2) Retest oil monthly(3) Determine load dependence

52 Case Study-II A 40MVA 220KV11 KV transformeris in service for 23 years Tank oil volume is 28000 litersThis transformer had overheated off circuit tapping switchingcontacts DGA data obtained in ppm after the fault on11062010 is as follows C

2H2-31 C2H4-53 CH

4-304 H

2-163

C2H6-15 CO-524 CO

2-786



2H2-34 C

2H4-69 CH

4-353 H

2-197

C2H6-22 CO-618 CO

2-931

TDCG in ppm= 1292 rate of TDCG= 1885 litday whichis greater than the normal level (28 litday)


123456789

101112131415161718192021222324252627

Fault0082 0089 0973 0327

TL

PD

D1

TH

D2

R1 R2 R3

Figure 3 Rule viewer for case-I

123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault0005 1657 3095 0595R1 R2 R3

Figure 4 Rule viewer for case-II

Step 3 FIS is applied for fault diagnosis The output of FISis given by rule viewer which is shown in Figure 4 Ruleviewer shows 1198771 = 0493 (Med) 1198772 = 179 (High) and 1198773= 314 which lies on the boundary of the fuzzy ratios Medand High Dark dots in the fault column show that rules 7and 22 are satisfied which indicates possible faults TL andTH respectively This result matches the actual fault of the

transformerWeighted average of both rules is given as 0636Weight of both faults can be calculated as follows


= 0273


The weights point towards the strong possibility of fault THand the relatively less possibility of fault TL

Step 4 Severity of faults is High Maintenance actions sug-gested are as follows

(1) Observe extreme caution(2) Retest oil weekly(3) Plan outage

53 Case Study-III A 25MVA 220KV132KV transformeris in service for 15 years Tank oil volume is 20000 litersThis transformer had an X - wax deposition Traces ofdischarges were found on paper of high voltage cable DGAdata obtained in ppm after the fault on 18032009 is asfollows C

2H2-15 C

2H4-19 CH

4-172 H

2-1903 C

2H6-14 CO-

180 CO2-635



2H2-26 C2H4-23 CH

4-2221 H

2-2257

C2H6-22 CO-220 CO

2-821


Step 3 FIS is applied for fault diagnosis The output of FISis given by rule viewer which is shown in Figure 5 Ruleviewer shows 1198771 = 113 (Med) 1198772 = 00979 which lies onthe boundary of the fuzzy ratios Low andMed and 1198773 = 105which lies on the boundary of the fuzzy ratios Low and MedDark dots in the fault column show that the rules 9 10 11and 13 are satisfied which indicates possible faults PD andD1This result matches with the actual fault of the transformerWeighted average of both rules is given as 0529 Weight ofboth faults can be calculated as follows

weight of PD = (06 minus 0529)(06 minus 04)

= 0355

weight of D1 = 1 minus 0355 = 0645(13)

The weights point towards the strong possibility of fault D1and the relatively less possibility of fault PD




123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault113 0098 105 0529R1 R2 R3

Figure 5 Rule viewer for case-III

6 Conclusion

The proposed FIS is developed using ldquoMATLABrdquo It candiagnose the incipient faults of the suspected transformersand suggest proper maintenance actions The fuzzy three-ratio method is proposed to diagnose multiple faults andfaults that cannot be diagnosed by the conventional DGAmethods Proposed FIS provides fault diagnosis for all thecases Accuracy of the proposed method is better than thatof other diagnostic methods

References

[1] IHohlein ldquoUnusual cases of gassing in transformers in servicerdquoIEEE Electrical Insulation Magazine vol 22 no 1 pp 24ndash272006

[2] C Mayoux ldquoDegradation of insulating materials under elec-trical stressrdquo IEEE Transactions on Dielectrics and ElectricalInsulation vol 7 no 5 pp 590ndash601 2000

[3] M K Pradhan ldquoAssessment of the status of insulation dur-ing thermal stress accelerated experiments on transformerprototypesrdquo IEEE Transactions on Dielectrics and ElectricalInsulation vol 13 no 1 pp 227ndash237 2006

[4] I Fofana J Sabau D Bussieres and E B Robertson ldquoThemechanism of gassing in power transformersrdquo in Proceedings ofthe IEEE International Conference on Dielectric Liquids (ICDLrsquo08) pp 1ndash4 July 2008

[5] V G Arakelian ldquoEffective diagnostics for oil-filled equipmentrdquoIEEE Electrical Insulation Magazine vol 18 no 6 pp 26ndash382002

[6] M Agah G R Lambertus R Sacks and K Wise ldquoHigh-speedMEMS-based gas chromatographyrdquo Journal of Microelectrome-chanical Systems vol 15 no 5 pp 1371ndash1378 2006

[7] M Agah G Lambertus R Sacks K D Wise and J APotkay ldquoHigh-performance temperature-programmed micro-fabricated gas chromatography columnsrdquo Journal of Microelec-tromechanical Systems vol 14 no 5 pp 1039ndash1050 2005

[8] M Duval ldquoNew techniques for dissolved gas-in-oil analysisrdquoIEEEElectrical InsulationMagazine vol 19 no 2 pp 6ndash15 2003

[9] V G Arakelian ldquoThe long way to the automatic chromato-graphic analysis of gases dissolved in insulating oilrdquo IEEEElectrical Insulation Magazine vol 20 no 6 pp 8ndash25 2004

[10] N Lelekakis D Martin W Guo and J Wijaya ldquoComparison ofdissolved gas-in-oil analysis methods using a dissolved gas-in-oil standardrdquo IEEE Electrical Insulation Magazine vol 27 no 5pp 29ndash35 2011

[11] ldquoMineral oil-impregnated electrical equipment in service guideto the interpretation of dissolved and free gases analysisrdquo IECPublication 60599 2007

[12] X Li H Wu and D Wu ldquoDGA interpretation scheme derivedfrom case studyrdquo IEEE Transactions on Power Delivery vol 26no 2 pp 1292ndash1293 2011

[13] ldquoIEEE guide for interpretation of gases generated in oilimmersed transformerrdquo ANSIIEEE Standard C5710420082009

[14] T K Saha ldquoReview of modern diagnostic techniques forassessing insulation condition in aged transformersrdquo IEEETransactions on Dielectrics and Electrical Insulation vol 10 no5 pp 903ndash917 2003

[15] N A Muhamad B T Phung T R Blackburn and K XLai ldquoComparative study and analysis of DGA methods fortransformer mineral oilrdquo in Proceedings of the IEEE LausannePower Tech pp 45ndash50 July 2007

[16] M Duval and A dePablo ldquoInterpretation of gas-in-oil analysisusing new IEC publication 60599 and IEC TC 10 databasesrdquoIEEE Electrical Insulation Magazine vol 17 no 2 pp 31ndash412001

[17] M Duval and J J Dukarm ldquoImproving the reliability oftransformer gas-in-oil diagnosisrdquo IEEE Electrical InsulationMagazine vol 21 no 4 pp 21ndash27 2005

[18] M Duval ldquoA review of faults detectable by gas-in-oil analysis intransformersrdquo IEEE Electrical Insulation Magazine vol 18 no3 pp 8ndash17 2002

[19] M Duval ldquoThe duval triangle for load tap changers non-mineral oils and low temperature faults in transformersrdquo IEEEElectrical Insulation Magazine vol 24 no 6 pp 22ndash29 2008

[20] C E Lin J M Ling and C L Huang ldquoExpert system fortransformer fault diagnosis using dissolved gas analysisrdquo IEEETransactions on Power Delivery vol 8 no 1 pp 231ndash238 1993

[21] Y-C Huang H-T Yang and C-L Huang ldquoDeveloping a newtransformer fault diagnosis system through evolutionary fuzzylogicrdquo IEEE Transactions on Power Delivery vol 12 no 2 pp761ndash767 1997

[22] H-T Yang and C-C Liao ldquoAdaptive fuzzy diagnosis system fordissolved gas analysis of power transformersrdquo IEEE Transac-tions on Power Delivery vol 14 no 4 pp 1342ndash1350 1999

[23] Y Zhang X Ding Y Liu and P J Griffin ldquoAn artificialneural network approach to transformer fault diagnosisrdquo IEEETransactions on Power Delivery vol 11 no 4 pp 1836ndash18411996

[24] J L Guardado J L Naredo P Moreno and C R FuerteldquoA comparative study of neural network efficiency in powertransformers diagnosis using dissolved gas analysisrdquo IEEETransactions on Power Delivery vol 16 no 4 pp 643ndash647 2001

[25] D V S S Siva Sarma and G N S Kalyani ldquoANN approachfor condition monitoring of power transformers using DGArdquoin Proceedings of the IEEE Region 10 Conference (TENCON rsquo04)vol 3 pp C444ndashC447 November 2004


[26] W M Lin C H Lin and M X Tasy ldquoTransformer-faultdiagnosis by integrating field data and standard codes withtraining enhancible adaptive probabilistic networkrdquo IEE Gen-eration Transmission Distribution vol 152 no 3 pp 335ndash3412005

[27] Y-C Huang ldquoEvolving neural nets for fault diagnosis of powertransformersrdquo IEEE Transactions on Power Delivery vol 18 no3 pp 843ndash848 2003

[28] R Naresh V Sharma and M Vashisth ldquoAn integrated neuralfuzzy approach for fault diagnosis of transformersrdquo IEEETransactions on Power Delivery vol 23 no 4 pp 2017ndash20242008

[29] D Bhalla R K Bansal and H O Gupta ldquoApplication ofartificial intelligence techniques for Dissolved Gas Analysis oftransformersmdasha reviewrdquoWorldAcademy of Science Engineeringand Technology vol 62 pp 221ndash229 2010

[30] D R Morais and J G Rolim ldquoA hybrid tool for detectionof incipient faults in transformers based on the dissolved gasanalysis of insulating oilrdquo IEEE Transactions on Power Deliveryvol 21 no 2 pp 673ndash680 2006

[31] R A Hooshmand M Parastegari and Z Forghani ldquoAdaptiveneuro-fuzzy inference system approach for simaltaneous diag-nosis of the type and location of faults in power transformersrdquoIEEE Electrical Insulation Magzine vol 28 no 5 pp 32ndash422012

[32] W Chen C Pan Y Yun and Y Liu ldquoWavelet networks inpower transformers diagnosis using dissolved gas analysisrdquoIEEE Transactions on Power Delivery vol 24 no 1 pp 187ndash1942009

[33] Z Wang Y Liu and P J Griffin ldquoA combined ANN and expertsystem tool for transformer fault diagnosisrdquo IEEE Transactionson Power Delivery vol 13 no 4 pp 1224ndash1229 1998

[34] Mathworks Fuzzy Logic Toolbox Userrsquos Guide[35] D Driankov H Hellendoorn and M Reinfrank An Introduc-

tion to Fuzzy Control Narosa Publishing House New DelhiIndia

[36] G J Klir and T A Folger Fuzzy Sets Uncertainity andInformation PHI New Delhi India

[37] M McNeill and E Thro Fuzzy Logic A Practical Approach[38] ldquoIEEE guide for dissolved gas analysis in transformer load tap

changersrdquo IEEE Standard C57139-2010

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



Table 1 Coding rule for IEC method [11]

Codes Range of gas ratios1198771 1198772 1198773

0 lt01 01ndash1 lt11 01ndash3 lt01 1ndash32 gt3 gt1 gt3

Table 2 Classification of faults by IEC method [11]

Faulttype Characteristic fault 1198771 1198772 1198773

1 Normal ageing (N) 0 0 02 Partial discharge (PD) of low energy density 0 1 03 PD of high energy density 1 1 04 Discharges of low energy (D1) 1-2 0 1-25 Discharge of high energy (D2) 1 0 26 Thermal fault of low temperature (TL) lt 150∘C 0 0 17 TL between 150∘C and 300∘C 0 2 08 TL between 300∘C and 700∘C 0 2 19 Thermal fault of high temperature (TH) gt 700∘C 0 2 2

where119877 is the rate (litersday) So is the TDCG of first samplein ppm 119878

119879is the TDCG of second sample in ppm 119881 is tank

oil volume in liters and 119879 is the time (days)The rate of generation of TDCGgreater than 28 litersday

indicates that the transformer has an active internal fault andrequires additional inspection by DGA methods

Many interpretative methods employ an array of ratiosof certain key combustible gases as the fault type indicatorsThese five ratios are

1198771 = C2H2C2H4

1198772 = CH4H2

1198773 = C2H4C2H6

1198774 = C2H6C2H2

1198775 = C2H2CH4

Rogersrsquo method [13ndash15] utilizes three ratios 1198771 1198772 and 1198773The method gives fault for the specific combination of thesegas ratios Dornenburg [13ndash15] utilizes four ratios 1198771 1198772 1198774and 1198775 This procedure requires significant levels of gases tobe present for the diagnosis to be validThemethod gives faultafter comparing these ratios to the limiting values

Amongst ratio methods IEC Standard 60599 [11] is mostwidely used It also utilizes three ratios 1198771 1198772 and 1198773 Thecoding rule and classification of faults by the IECmethod aregiven in Tables 1 and 2

Incipient faults can be reliably identified by visual inspec-tion [16] of the equipment after the fault has occurred inservice as follows

(a) PDndashpossible X wax formation and sparking inducingsmall carbonized punctures in the paper

(b) D1ndashlarger punctures in the paper tracking or carbonparticles in oil

(c) D2ndashextensive carbonization metal fusion and possi-ble tripping of the equipment

(d) TLndashfor TL lt 300∘C the paper turns brown for TL gt300∘C the paper carbonizes

(e) THndashoil carbonization metal coloration or fusion

TheDuval Triangle [17ndash19] method utilizes three ratiosof certain gases for DGA interpretation of transformers filledwithmineral oilThe triangular coordinates corresponding toDGA results in ppm can be calculated by (2) as follows

C2H2=100 sdot 119909

(119909 + 119910 + 119911)

C2H4=100 sdot 119910

(119909 + 119910 + 119911)

CH4=100 sdot 119911

(119909 + 119910 + 119911)

(2)

where 119909 119910 and 119911 are concentrations of C2H2 C2H4 and

CH4in ppm respectively Along with three ratios given by

Duval Triangle method fourth ratio [12] is also used forfault diagnosis which is given by

H2=

100 sdotH2

(H2+ C2H6+ CO + CO

2) (3)

All these techniques are computationally straightforwardHowever these methods in some cases provide erroneousdiagnoses as well as no conclusion for the fault type

To overcome these limitations FIS is proposed

3 Diagnostic Procedure

Flow chart of proposed system diagnosis is shown in Figure 1The input data include concentration of dissolved gasesC2H2 C2H4 C2H6 CH4 H2 CO and CO

2of the sample

Information such as tank oil volume date of samplingand date of installation of transformer is asked for furtherinference

In the first step the system calculates TDCG and com-pares with the standard permissible limits (IEEE standard2008) For normal level of TDCG (lt720 ppm) permissiblelimits for individual gases are checked The normal levelof TDCG and individual gases indicates the satisfactoryoperation of a transformer Once an abnormal level of TDCGor individual gas has been detected the next step is todetermine the rate of generation of TDCG (1) by analysis ofthe successive sample For the normal rate of TDCG (less than28 litersday) further diagnosis is bypassed For an abnormalrate of TDCG the proposed FIS is adopted to diagnose theprobable faults In the last step severity degrees are assignedto the diagnosed faults On the basis of severity of faultsappropriate maintenance actions are suggested

4 Fuzzy Inference System (FIS)

Intelligent algorithms for example expert system [20] FIS[21 22] artificial neural networks [23ndash25] probabilistic


Normal

Smaller

DGAsample

Greater

Normal ageing(1)

Abnormal

Abnormal

Normal

1

Maintenanceactions

FIS

TDCG

Rate ofTDCG

Individualgas conc







given as

119908119894= AND method [1198651 (119909) 1198652 (119910)] (4)


119884 =sum119873

119894=1119908119894119911119894

sum119873

119894=1119908119894

(5)



Input0 1 2 3 4

1Low Med High

09 11 27 33

Mem

bers

hip

degr

ee

variable R3



120583Low (1198771) =

1 997888rarr 1198771 le 009

(011 minus 1198771)

(011 minus 009)009 le 1198771 le 011

0 1198771 ge 011

(6)


120583Med (1198771) =

0 997888rarr 1198771 le 009

(1198771 minus 009)

(011 minus 009)997888rarr 009 le 1198771 le 011

1 011 le 1198771 le 27

(33 minus 1198771)

(33 minus 27)27 le 1198771 le 33

0 1198771 ge 33

(7)


120583High (1198771) =

0 997888rarr 1198771 le 27

(1198771 minus 27)

(33 minus 27)997888rarr 27 le 1198771 le 33

1 1198771 ge 33

(8)






119879119877

119879119875

119860119875

119860119862











119860119875=100 sdot 119879

119877

119879119875

(9)


119875



119860119862=100 sdot 119879

119877

119879119862

(10)





2H2-4 C2H4-54 CH

4-09 H

2-78 C

2H6-67

CO-670 CO2-1243



2H2-6 C2H4-73 CH

4-14 H

2-158

C2H6-75 CO-831 CO

2-1430




= 0365






2H2-31 C2H4-53 CH

4-304 H

2-163

C2H6-15 CO-524 CO

2-786



2H2-34 C

2H4-69 CH

4-353 H

2-197

C2H6-22 CO-618 CO

2-931



123456789

101112131415161718192021222324252627

Fault0082 0089 0973 0327

TL

PD

D1

TH

D2

R1 R2 R3


123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault0005 1657 3095 0595R1 R2 R3





= 0273






2H2-15 C

2H4-19 CH

4-172 H

2-1903 C

2H6-14 CO-

180 CO2-635



2H2-26 C2H4-23 CH

4-2221 H

2-2257

C2H6-22 CO-220 CO

2-821




= 0355

weight of D1 = 1 minus 0355 = 0645(13)





123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault113 0098 105 0529R1 R2 R3


6 Conclusion


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Normal

Smaller

DGAsample

Greater

Normal ageing(1)

Abnormal

Abnormal

Normal

1

Maintenanceactions

FIS

TDCG

Rate ofTDCG

Individualgas conc







given as

119908119894= AND method [1198651 (119909) 1198652 (119910)] (4)


119884 =sum119873

119894=1119908119894119911119894

sum119873

119894=1119908119894

(5)



Input0 1 2 3 4

1Low Med High

09 11 27 33

Mem

bers

hip

degr

ee

variable R3



120583Low (1198771) =

1 997888rarr 1198771 le 009

(011 minus 1198771)

(011 minus 009)009 le 1198771 le 011

0 1198771 ge 011

(6)


120583Med (1198771) =

0 997888rarr 1198771 le 009

(1198771 minus 009)

(011 minus 009)997888rarr 009 le 1198771 le 011

1 011 le 1198771 le 27

(33 minus 1198771)

(33 minus 27)27 le 1198771 le 33

0 1198771 ge 33

(7)


120583High (1198771) =

0 997888rarr 1198771 le 27

(1198771 minus 27)

(33 minus 27)997888rarr 27 le 1198771 le 33

1 1198771 ge 33

(8)






119879119877

119879119875

119860119875

119860119862











119860119875=100 sdot 119879

119877

119879119875

(9)


119875



119860119862=100 sdot 119879

119877

119879119862

(10)





2H2-4 C2H4-54 CH

4-09 H

2-78 C

2H6-67

CO-670 CO2-1243



2H2-6 C2H4-73 CH

4-14 H

2-158

C2H6-75 CO-831 CO

2-1430




= 0365






2H2-31 C2H4-53 CH

4-304 H

2-163

C2H6-15 CO-524 CO

2-786



2H2-34 C

2H4-69 CH

4-353 H

2-197

C2H6-22 CO-618 CO

2-931



123456789

101112131415161718192021222324252627

Fault0082 0089 0973 0327

TL

PD

D1

TH

D2

R1 R2 R3


123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault0005 1657 3095 0595R1 R2 R3





= 0273






2H2-15 C

2H4-19 CH

4-172 H

2-1903 C

2H6-14 CO-

180 CO2-635



2H2-26 C2H4-23 CH

4-2221 H

2-2257

C2H6-22 CO-220 CO

2-821




= 0355

weight of D1 = 1 minus 0355 = 0645(13)





123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault113 0098 105 0529R1 R2 R3


6 Conclusion


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





119879119877

119879119875

119860119875

119860119862











119860119875=100 sdot 119879

119877

119879119875

(9)


119875



119860119862=100 sdot 119879

119877

119879119862

(10)





2H2-4 C2H4-54 CH

4-09 H

2-78 C

2H6-67

CO-670 CO2-1243



2H2-6 C2H4-73 CH

4-14 H

2-158

C2H6-75 CO-831 CO

2-1430




= 0365






2H2-31 C2H4-53 CH

4-304 H

2-163

C2H6-15 CO-524 CO

2-786



2H2-34 C

2H4-69 CH

4-353 H

2-197

C2H6-22 CO-618 CO

2-931



123456789

101112131415161718192021222324252627

Fault0082 0089 0973 0327

TL

PD

D1

TH

D2

R1 R2 R3


123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault0005 1657 3095 0595R1 R2 R3





= 0273






2H2-15 C

2H4-19 CH

4-172 H

2-1903 C

2H6-14 CO-

180 CO2-635



2H2-26 C2H4-23 CH

4-2221 H

2-2257

C2H6-22 CO-220 CO

2-821




= 0355

weight of D1 = 1 minus 0355 = 0645(13)





123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault113 0098 105 0529R1 R2 R3


6 Conclusion


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




123456789

101112131415161718192021222324252627

Fault0082 0089 0973 0327

TL

PD

D1

TH

D2

R1 R2 R3


123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault0005 1657 3095 0595R1 R2 R3





= 0273






2H2-15 C

2H4-19 CH

4-172 H

2-1903 C

2H6-14 CO-

180 CO2-635



2H2-26 C2H4-23 CH

4-2221 H

2-2257

C2H6-22 CO-220 CO

2-821




= 0355

weight of D1 = 1 minus 0355 = 0645(13)





123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault113 0098 105 0529R1 R2 R3


6 Conclusion


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




123456789

101112131415161718192021222324252627

D2

TL

PD

TH

D1

Fault113 0098 105 0529R1 R2 R3


6 Conclusion


References















































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in
























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



research article fuzzy algorithm for power transformer...

Documents