INTEGRATED TRANSFORMER FLEET MANAGEMENT (ITFM SYSTEM Audrius ILGEVICIUS Maschiefabrik Reihause GbH, a.ilgevicius@reihause.co Alexei BABIZKI Maschiefabrik Reihause GbH a.babizki@reihause.co ABSTRACT The ipleetatio of decisio support for asset replaceet ad the aiteace of trasforer fleet aageet is based o a techical risk score (Health Idex ad criticality. This paper describes the applicatio of easureets obtaied fro olie oitorig ad regular diagostic ethods for idetificatio of suspect trasforers based o the Health Idex (HI. The coputer software syste supports risk-based aiteace ad replaceet prioritizatio copliat with PAS 55(Publicly Available Specificatio for the optial aageet of physical assets ad DIN VDE 0109. Iitial field test results deostrate the feasibility of this approach, ad its applicatio for asset aagers to prioritize replaceet decisios ad iprove aiteace progras. Data itegratio based o IEC 61970 (CIM eables the uber of APIs (applicatio prograig iterface betwee the software systes of differet vedors to be reduced to a iiu. The paraeters obtaied fro olie ad diagostic easureets are arraged ito physical subgroups before calculatig the overall HI of trasforers. INTRODUCTION The aageet of trasforer fleets with heterogeeous structures of differet age, physical coditio ad iportace requires a oder software syste to assess the ipact of the idividual trasforer o the perforace of the whole fleet. It is therefore iportat to detect irregularities ad aoalies of equipet, by updatig the HI with olie oitorig data before failures occur. The followig coditio-based approach for trasforer coditio assesset is proposed [1, 2]: - Level 1: olie oitorig ad olie diagostic (o outage required; - Level 2: routie off-lie diagostic easureets (short outage required; - Level 3: advaced off-lie easureets (a loger outage required. I order to ake assesset of the trasforers coditio ecooically viable data should be readily available via the software package I Level 1 the preprocessed data are based o both olie oitorig ad tie-based olie diagostic data, such as oil teperature, load, DGA, oil coditio, fura aalysis, visual ispectio or historical coditio data. I Level 2 the coditio based off-lie electrical ad dialectical easureets are applied. Fially, i Level 3 the advaced off-lie easureets, like FRA, SFRA, PDC/FDS or vibroacoustics, are applied. Depedig o the HI-score, recoedatios for the aiteace level are issued (see Table 1. CONCEPT OF ITFM SYSTEM The choice of the IT platfor for the fleet aageet syste is based o the followig requireets: - Ability to itegrate differet types of data: fro olie oitorig, offlie diagosis or historical data fro differet databases - Ability to ipleet algoriths ad odels for statistical ad reliability aalysis - Scalable database to hadle large ubers of assets - Capacity for display, visualisatio ad tredig aalysis of historical ad olie iforatio - Ability to eet IT cyber security requireets fro the utility side The ai cocept of the ITFM (Fig. 1 syste is based o the itegratio of olie data, cotiuous data fro differet sesors (e.g. DGA, OLTC, bushig, oisture via Itegrated Sart Module (ISM, a trasforer oitorig device, ad the correlatio of this iforatio with the historical ad offlie data. This requires the ipleetatio of soe statistical algoriths ad tredig aalysis to idetify the coditio of the physical paraeters. Next, the database ust allow the storage of data fro olie sesors, CSV or XML files with offlie easureets or the aual iput of iforatio fro visual ispectio data. The data odel for the database ust cotai IEC 61970/61968 CIM-based structure [3]. The itegratio of olie oitorig data cosiders IEC 61850 Ed.2 data objects for oitorig ad supervisio of equipet. Fially, the results ust be coverted ito eaigful iforatio, (for the asset aager or aiteace egieer, such as the HI ad risk idex (RI based o the colour-codig ad uerical schee CIRED 2015 1/5
H2, pp/teperature, C 23 rd Iteratioal Coferece o Electricity Distributio Lyo, 15-18 Jue 2015 Offlie DGA Offlie Oil Furas Routie tests Offlie Data Electrical Tests Visual Ispectio CSV, XML, SQL SCADA (Optioal Load Oil teperature Abiet teperature EAM, ERP Failure probability Historical Data Figerprit Data IEC 61970 IEC 61970 IEC 61968 Itegrated Fleet Maageet Syste Health Idex Database Risk Idex Algoriths Dashboards Treds Warigs Loss of life 40,0 35,0 30,0 25,0 20,0 15,0 10,0 5,0 T1 T2 T3 Tx Teperature, C H2, pp Teperature DGA,H2 Olie Data (ISM Load DGA Moisture Hot spot teperature OLTC IEC 61850 IEC 60870-5-104 DNP 3.0 0,0 17.01.2011 22.01.2011 27.01.2011 01.02.2011 06.02.2011 Tie tie Fig. 1 Cocept of ITFM Syste DYNAMIC CONDITION ASSESSMENT OF TRANSFORMER FLEET The ai challege i assessig the trasforer s coditio is estiatig the HI based o the olie oitorig data ad the regular diagosis data fro oil aalysis ad electrical easureets. Furtherore, the HI ust be related to probability of equipet failure. Figure 2 shows the ai steps i the trasforer coditio assesset. I steps 4 ad 5 the required iterface for the iput of ISM, offlie diagosis ad further siulatios should be cosidered. Trasforer failure odes I the first step the statistical iforatio o historical failure odes is collected. The exaples of such failure odes are: - Mechaical - Theral - Dielectric - Tap chagers - Bushigs - Accessories The advatage of such a structured coditio assesset approach is that it allows for better tuig of idividual failure odes to the probability of failure usig weightig factors. The iforatio about the failure rates ad how they are distributed betwee the trasforer copoets ca be obtaied, for exaple, fro [4]. Failure idicators ad coditio paraeters (olie ad offlie paraeters The secod step is idetifyig which failure idicators ad coditio paraeters, idicate the failure odes. I the ITFM Tool there are ore tha 100 coditio paraeters together with the trasforer desig, aiteace ad operatio, locatio of istallatio ad accessories. The followig olie oitorig data ca be used i the algorith: 1. Active part: - DGA - Moisture i oil, - Moisture i paper, - Hot spot teperature, - Agig rate - Oil level 2. Bushig - Leakage curret, - Capacitace, 3. Tap Chager - Teperature, - Motor torque, - Cotact wear - Vibroacoustic Coditio paraeters (CP Coditio paraeters like DGA, Oil tests or historical data are coverted to a score of diesioless uit (Fig.3. By coparig the physical paraeters with the thresholds fro iteratioal ad atioal stadards ad CIRED 2015 2/5
1 Review of the trasforer failure odes ad their aiteace practices 2 Deteriatio of the failure idicators ad coditio paraeters 3 Developet of scorig, thresholds ad weightig for each coditio paraeter 1. Mechaical coditio icludig active part ad widigs 2. Theral or solid isulatio coditio (agig of isulatio 3. Dielectric coditio (Oil ad DGA results 4. OLTC coditio 5. Bushig coditio 6. Coolig Syste coditio 7. Tak ad accessories coditio 8. Historical Reliability Olie ad Offlie Data (over 100 paraeters 4 Coditio assesset of the trasforer ad calculatio of the Health Idex Algoriths applied to physical risk groups Thresholds ad figerprit data Iputs fro Olie Moitorig Syste ISM database Iputs fro Offlie diagosis ad figerprit database 1. Mechaical 2. Theral 3. Dielectric 4. OLTC 5. Bushigs 6. Coolig Syste 7. Tak ad accesories 8. Historical Reliability 5 Estiatio of Failure Probability ad relatio to the Health Idex 6 Priorisatio by Risk Idex ad aiteace recoedatios Criticality coefficiet fro etwork reliability siulatio Trasforator HI Risk Idex Failure Cosequece (Iportace Criticallity (Ecooical Aspects Fig. 4 Structure of trasforer failure risk assesset Fig. 2 Mai steps of trasforer coditio assesset field experiece, the scorig values are obtaied. The score of idividual results is discreet ad coded usig classificatio syste fro 0 to 3 (where 0 is oral coditio ad 3 is a faulty coditio: see Fig. 3. Olie ad Off-lie testig results Coditio Fuctio Evaluatio of the test results ad applyig the weights to sigle easureet Coditio Idicator Idividual results are coverted ito uerical score Fig. 3 Coversio of test results ito uerical score Calculatio of the Health idex for a Trasforer After the first evaluatio of the idividual coditio paraeters, they are classified ito the followig risk groups (Fig 4: As etioed above, the risk groups are based o the trasforer failure odes. I each of the risk groups, weighted idividual coditio paraeter scores are averaged to the risk group paraeter score (RPS. The RPS is calculated accordig to the followig equatio [5]: 1 ( RPS (1 WCPCP 1 ( WCP CP ax Fially, the HI is the calculated accordig the followig equatio: 1 ( WRPS HI 100% (2 ( WRPS RPS 1 RPS,ax Here the WCP ad WRPS i the Eq 1 ad 2 deotes weightig coefficiets of CP ad RPS respectively. The Health idex is oralized betwee a score of 0 (ew ad 100 % (failed or fro 0 to 1 o a liear scale. The coditio value ca be ay uber i betwee. I case of the HI score >30%, the liear Pearso s correlatio betwee the sigle coditio paraeters ad CIRED 2015 3/5
risk groups is ade: r i 1 ( x x( y y i x i (3 ( 1 y Based o the correlatio coefficiet r the aiteace recoedatios are ade. Whe aggregatig the idividual paraeters to the risk group score (RPS ad further to trasforer HI, it is iportat that the failure of oe critical paraeter should lead to the total HI approachig 100% (failed. This is achieved usig the o-liear approach, where the weightig of the coditio paraeters of a specific risk group depeds o the score of a particular coditio paraeter. However, ot all the coditio paraeters lead to trasforer failure. I order to odel the criticality or ipact of the coditio paraeter to the overall trasforer coditio a probability ethod ust be applied. Oe of the possible approaches is the applicatio of the Bayesia coditioal probability rule [5]. Based o the trasforer fault three, a coplex Bayesia Network represetig all iput variables ca be built ad the aggregatio of the coditio paraeters to HI accordig to Bayesia forula is ade. The iterpretatio of the HI is preseted i the Fig. 5 ad Table 1, recoeded by Cigre Workig Group A2.44 (D10 [6], ca be referred to the coditio (probability of failure or to the aiteace actio required. However, the lik betwee aiteace actio ad the coditio is very specific ad depeds o respective copay policy. Estiatio of failure probability ad relatio to the Health idex Oce the HI of the trasforer is deteried, its effective age ca be deteried fro the relatioship betwee the probability of failure ad its HI. The relatioship betwee the health idex ad the fault rate is show by the probability of failure curve i Fig. 6. The coefficiets A, B ad C are calculated i the equatio (4 fro the failure rates of the equipet copoets [7]. ( HI Aexp( B HI C (4 Fig. 6 Health idex relative to probability of failure Prioritizatio by risk idex ad aiteace recoedatios Fig. 5 Coditio developet i tie (top ad coditio codig (botto Table 1 Health idex Colour Coditio Recoedatio 85 < HI < Failed Eergecy replaceet 100 70 < HI < 85 Faulty Replaceet o short ter 50 < HI < 70 Defective Icreased diagostic testig / repair actio required / Level 2,3 30 < HI < 50 As good as Noral aiteace / ew Level 2 0 < HI < 30 New Noral aiteace / Level 1 The risk score or idex ca be forulated as follows: Risk Idex CC (5 where: CC criticality coefficiet I the forulatio (5 above the ter probability of failure deotes the techical coditio of a asset, also idetified as HI fro 0 (oral to 100 % (failed, while the criticality coefficiet CC fro 0 (uiportat to 100 % (high iportace icludes ecooic factors related to the loss of power delivery, systes stability, repair or pealties. Fro the risk score, a risk atrix with the aiteace strategy ca be created icludig the prioritizatio of aiteace activities. CIRED 2015 4/5
Fig.7 Trasforer fleet aiteace prioritizatio The aiteace prioritizatio order i Fig. 7 ca be calculated fro the distaces d1-d3, cosiderig 45 C shift of the d axis with respect to the ordiate. The followig order of trasforer aiteace is calculated: T4, T2, T3 ad T1. RESULTS FROM THE FIELD The exaple for the evaluatio of trasforer fleet of 4 trasforers (220/66 kv, 100 MVA fro differet aufacturers ad of differet age, fro 8 to 40 years ad its copoets is show i the Fig. 8. The algorith was applied to calculate the coditio of each trasforer ode (risk of failure ad the total HI of each trasforer. It ca be see that trasforer 2 has the worst HI value due to load history (overload, which possibly lead to the overheatig of the solid isulatio (ratio of CO2/CO > 11. The sae applies to the trasforer 3, where the ratio of CO2/CO > 18. Here, the correlatio betwee the loadig, top oil teperature ad DGA results was applied. The trasforer 4 was becharked with the HI of 31% due to old age (over 40 years while trasforer T1 is the ewest with the age of 8 years ad was becharked with the best coditio. aageet cosiderig olie oitorig data ad offlie data. The advatage of this approach is that the oitorig syste based o ISM techology eables ay aoalies i the trasforer fleet to be detected before the copoets becoe defective. The HI is related to the probability of failure where differet properties of the trasforer (echaical, dielectric, theral, OLTC, bushig, coolig syste ad accessories are cosidered separately as risk groups. Liear correlatio betwee idividual coditio paraeters ad betwee risk groups is applied i case the HI or failure risk idicates defective coditio (>30%. The ost challegig issue is the Health idex aggregatio, where o-liear or Bayes probability ethods ust be applied i order to accurately calculate the effect of idividual coditio paraeter o the HI. Fially, the IT ad data security issues ust be cosidered whe itegratig the tool i the utility IT or asset aageet eviroet. Much of the iput data of the tool coes fro olie sesors ad is of cofidetial ature. Therefore, itegratio of oitorig data ito IT eviroet ust coply with utilities IT ad data security policies. REFERENCES [1] W. Fleischa, T. Krüger, A. Ilgevicius, Maschiefabrik Reihause GbH, 3 rd Iteratioal Colloquiu Trasforer Research ad Asset Maageet, Split, Croatia, October 15-17, 2014 [2] CIGRE Workig Group 2.34 Guide for Trasforer Maiteace [3] CIGRE Techical Brochure 227: Life aageet techiques for power trasforer, WG A2.18, Paris, 2003 [4] S. Tebohle, et.al, XVII Iteratioal Syposiu o High Voltage Egieerig, Haover, Geray, August 22-26, 2011 [5] CIGRE Techical Brochure 248: Guide o ecooics of trasforer aageet Workig Group A2.20, Paris, 2004 [6] R.E. Brow, Failure rate odellig usig equipet ispectio data, IEEE Tras.Power Systes, Vol.2, May 2000, pp.782-787 Fig. 8 Overview of trasforer fleet evaluatio CONCLUSIONS The paper presets a IT-based tool for trasforer fleet [7] CIGRE Techical Brochure: Guide o trasforer itelliget coditio oitorig syste Workig Group A2.44. Draft 10, 2014 CIRED 2015 5/5