Transformer Failure prevented by Gas-in-Oil On-Line Monitoring The impact of accuracy in evaluating the evolution of a fault

Oil filled power transformers are one of the most important components of electricity generation, transmission and distribution network. The analysis of gases from petroleum products has been performed for decades using gas chromatography. However, this technique was not applied specifically to transformer mineral oil until the late 1960s/early 1970s and is now commonly called dissolved gas-in-oil analysis (DGA). Some of the early developers of the technique were Dr. James Morgan of Morgan Schaffer Systems, Canada, and researchers J.E. Dind, R. Daust and J. Regis from the Canadian utility Hydro-Quebec. The analysis of dissolved gases, ASTM D3612 - Analysis of Gases Dissolved in Electrical Insulating Oil by Gas Chromatography, has been shown to be one of the most sensitive, as well as easiest to obtain, measures of potential trouble in electrical equipment. Certain gases may be formed due to natural aging but also as a result of faults which may damage the equipment. Some are of immediate concern (such as arcing) while others can have a long term effect (overheating). Periodic analyses of the oil provide a means monitoring the health of this equipment. The DGA technique involves extracting or stripping the gases from the oil and injecting them into a gas chromatograph (GC). Detection of gas concentrations usually involves the use of a flame ionization detector (FID) and a thermal conductivity detector (TCD). Removing the gas from the oil is one of the most difficult and critical portions of the procedure. The original method, now ASTM D3612A, required that the oil be subjected to a high vacuum in an elaborate glasssealed system to remove most of the gas from the oil. The detection of certain gases generated in an oil-filled transformer in service is frequently the first available indication of a possible malfunction that may eventually lead to failure if not corrected. Arcing, corona discharge, severe overloading, and over-heating in the insulation system are some of the mechanisms that can result in chemical decomposition of the insulating materials and the formation of various combustible and noncombustible gases. Normal operation may also result in the formation of some gases. In a transformer, generated gases will be found dissolved in the insulating oil. They may also be found in the gas blanket above the oil or in gascollecting devices. The detection of an incipient fault, if present, involves an evaluation of the amount of generated gas present and the continuing rate of generation. An indication of the source of the gases and the kind of insulation involved can sometimes be gained by determining the composition of the generated gases.

Page|1

Fault t gases are produced p by y: Thermal & electrical l stresses; Exposure e to air; Cellulosic c insulation n starts degrading g; nant induce ed chemical Contamin reactions. This will cause the t chain to break and trigge er molecula ar rearrangem ment.

Met thod A is s by far t the longes ststan nding techn nique and is s still widel ly used d today and d offers accu urate, reliabl le sam mple preparation if f followe ed rigo orously. As required by ASTM M (D3612 2), labo oratories mu ust use prep pared Gas i in Oil Standards o of known c concentratio on regu ularly to va alidate thei ir extractio on effic ciency and d establish a traceabl le perf formance of f their proce ess. The e use of cont trol charts a as part of a qual lity assuran nce program m gives any labo oratory the a ability to sh how that an anal lysis is in statistical con ntrol. Some of th he advantag ges that can come from brin nging a testin ng process i into control are: Ensurin ng the identification of ns; analytical variation Analysis s and exami ination of the proc cess statistics improves process control during normal operatio ons as well a as during process modificatio ons thus allowing g for improv ved method d perform mance; Using di s in oil issolved gas standard ds as part of f the producti ion flow ben nchmarks the anal lytical proce ess and allows fo for greater co onfidence in n the resul lt; A test th hat shows go ood statistica al control is s predictable e and can be relied on n; The cha art of a test i in statistical l control c can enables s us to determin ne the exper rimental error.

Hydr rocarbons & Hydrogen:

Carb bon oxides: Atmo ospheric gas ses (non fau ult gases)

Hydr rogen is the e key fault gas as it is s alway ys present in n every type e of fault. Three extract tion methods are e appro oved: Method M A (Vacuum m Extra action), Method M B (Stripping g meth hod) and Method M C (H Head Space e meth hod). Th hey each have their r advantages and d disadvan ntages but t

Page|2

DGA A Oil Stand dards are co ommercially y available (Morga an Schaffer True North h DGA A Oil Standar rds). asurement is made, , When a mea whet ther using g an ohm mmeter, a micro ometer, or r another laboratory y instru ument, the ere is a te endency to o accep pt readings without w thin nking about t their accuracy. Even with h the most t expen nsive equip pment and under u ideal l cond ditions, eve ery measu urement is s subje ect to errors and inaccuracies. l The precision of an analytical edure is a measure of its s proce repro oducibility, i.e., the dis stribution of f results about th he mean or r how close e sured value es are to each e other. . meas Mon nitoring the precision of analyses s on a day to day y basis can be difficult. . The precision of o a method d is often a funct tion of the e smallest measurable e incre ement, thus s the prec cision may y vary. . Also, anal lytical result ts can often n be at a or near the detec ction limits s maki ing it difficu ult or impos ssible to set t contr rol limits. Accuracy is a quali itative term m and refe fers to the e agree ement of a measured value v to the e true, accepted, or know wn (correct) ) value e. One appr roach used to estimate e the accuracy of o a meth hod is the e analy ysis of know wn materials s. Refer r to Figure 1 The e objective of a qualit ty assuranc ce prog gram for la aboratory t testing is t to redu uce errors in measu urements t to allow wable leve els and to o provide a mea ans of c confirming that th he mea asurements made ha ave a hig gh prob bability of being as ac ccurate as is nece essary. Q Quality as ssurance is com mprised of tw wo separate e but relate ed activ vities: Quality control, w which is th he applicat tion of proc cedures in a repetitiv ve and consisten nt fashion for controll ling the mea asuremen nt process a and its qua ality so it t satisfies th he recipients needs, a and Quality assessment t, the proces ss that con nfirms the te esting system m is opera ating within n acceptabl le limits an nd that the task of qua ality con ntrol is bein ng performe ed effective ely. This ne eeds to be a an on-going g operation. Not t only does quality assu urance cove er all laboratory operation ns but als so sam mple collect tion and id dentification n, train ning and oth her areas as s well. Rem member, D DGA is a three ste ep proc cess and each step has to b be perf formed in complian nce with a
Page|3

rigorous procedure to ensure reliable results. 1- Oil sampling 2- Gas extraction (ASTM D3612) 3- Gas Measurement (GC) Oil sample identification is essential in order to take full advantage of this process to update the database and enable further analysis. Required information: Date & Time Temperature Pressure Serial Number Equipment ID Apparatus type Location Description Manufacturer Rated kV MVA Year manufactured Fluid volume Fluid type Owner DGA is an integral part of Condition Based Maintenance programs. To be successful, it must be adapted to your needs and common mistakes avoided: Analysis intervals are too infrequent Poor sampling techniques Delay in getting samples to the laboratory Delays in getting the results back Poor information submitted to the lab (incomplete) Lack of correct tests Poor interpretation of the tests Failure to integrate with other condition based technologies.

Transformers are all different and every unit has a distinct signature. When a unit is first energized, its important to take an oil sample and perform a DGA to establish a baseline for this unit. Thereafter, DGAs are performed annually or more frequently depending on the criticality and/or the detection of an incipient fault. Historical DGA results should be accessible to calculate trends. Various DGA results interpretation methods are used for condition assessment. The most popular ones are: IEC 60599 Ratios IEEE C57.104, Limits, rates and TDCG Rogers Ratios Key Gas Method Duval Triangle Trend Analysis NEW GUIDELINES FOR INTERPRETATION OF DGA CIGRE Task force 15.01.01,Octr 1999 Companies guidelines More Diagnostic reliability is affected by the accuracy of the DGA measurement results as demonstrated by a study conducted by CIGRE and covered in an article published by Dr. Michel Duval and Dr. Jim Dukarm.. (IEEE August 2005, M. Duval, J. Dukarm, Improving the reliability of transformer in gas-in-oil diagnosis)

Page|4

lab results can n be comp pared to th he mon nitor readin ngs, at least a annually. Date & Time Monitor r model no. Serial nu umber Transfor rmer ID Location n Recorde ed readings (all) Date Downlo oad data abase fil le (recorde ed readings s and even nt log) Review alarm settin ngs Not te: When th he on-line m monitor has a sam mpling valve e, the oil sa ample shoul ld alwa ays be take en from th hat valve fo or com mparison. Con nclusion DG GA is the mo ost importan nt oil test fo or insu ulating fluid d in electric cal apparatu us as i it provides a wealth o of diagnosti ic info ormation to detect inci ipient faults. As s such, it has s become a standard i in the utility ind dustry thro oughout th he wor rld. Wh hether DGA A results a are obtaine ed m a labor ratory, por rtable DGA A from anal lyzer or o on-line mo onitor, ther re mus st be no room for misleadin ng conc clusions. Lab boratories must be e able t to dem monstrate at all-t time their effec ctiveness in n providing g precise an nd accu urate resu ults. Vari iances an nd inco onsistencies in the resu ults will lea ad to a wrong d diagnostic, corrupt th he DG GA database e and invalid date trendin ng calc culations. A As DGA c can also b be
Page|5

Rega ardless of the method used, u precise e and accurate DGA results are crucial l to yo our decision making pro ocess. Preci ision and accuracy y is also o impe erative to o enable trending g calcu ulation for early warn ning of all l types s of develo oping faults. Wrong g DGA A results will lead to wrong g diagn nostic and ultimately to wrong g decis sion. Coun nting only on periodi ic DGA is s often n not su ufficient to o prevent t catas strophic failures and the fault t cond dition can ch hange very quickly. q unately, the evol lution of f Fortu techn nology has s enabled utilities to o instal ll on-line monitors m (Ke ey fault gas s or multigas) m for f all crit tical transformers. Furth hermore, ut tilities now w have the possib bility to per rform a lab b preci ision DGA in the field d when the e situat tion is critic cal. When on-line monitoring m is i used, the e samp pling proced dure must be adapted d to en nable valid dation of th he readings s perio odically. In addition to o the sample e information sen nt to the laboratory, , take note of the e readings so that the e

perfo ormed in th he field usin ng portable e gas chromatogr c aphs, it is also highly y impo ortant to o calibra ate these e instru uments reg gularly to ensure e their r preci ision and accuracy, and thus s main ntain results trustworth hiness. In bo oth cases, th he regular use u of DGA A Oil Standards S is s imperativ ve as it will l valid date the in ntegrity of the DGA A proce ess and dete ect any dev viation from m the precision p and d accuracy objectives. o How wever, perio odic oil sam mpling and d DGA A performed d by a labo oratory or a porta able fault ga as analyzer r will rarely y enable early fau ult detection n. On-line e moni itoring is the t only method m that t will provide p con ntinuous info ormation to o detec ct and moni itor incipien nt faults. As with labor ratories an nd portable e analy yzers, data from on-lin ne monitors s must t be valida ated regula arly. It is s there efore highl ly recomm mended to o inclu ude the on n-line moni itor current t readi ing every time t an oil l sample is s collected from the transfor rmer. The e s sho ould always s be taken n oil sample from the instrum ment sampl ling port as s this is i the best point p for ha aving access s to oil l representa ative of the transformer r cond dition. When an alarm m is triggere ed by a key y gas on-line mo onitor, an oil sample e t be coll lected to enable a must diagn nosis, as th he values of o all fault t gases s are req quired to draw any y concl lusion and d make the t proper r decis sion.

Wit th on-line m multi-gas an nalyzers (On nLine e DGA M Monitors), the curren nt prac ctice is also to take an o oil sample t to valid date the ins strument re eadings usin ng a po ortable or la aboratory D DGA. This is som mewhat puzz zling as the purpose of a DG GA monitor r is to provide DGA A resu ults that ca an be used to diagnos se and monitor th he evolution of a fault. One e must th herefore co onclude tha at conf fidence in o on-line DGA A results wi ill only y be possi ible if strin ngent DGA A accu uracy verif fication pr rocesses ar re imp plemented. Only if laboratories, port table GCs and on-lin ne monitor rs cont tinuously m maintain their precisio on and accuracy, will utilitie es be able t to take e full advan ntage of D DGA results, inde ependently o of their sour rce.

Page|6

References
1Technical bu ulletin MS-25 Dr James E. Morgan. A guide to the interpretation of f transformer fa ault data; Improving the t assessment of transformer r condition by y enhancing labo oratory accuracy y using dissolved gas in oil standards. Peter r Lazarski, Nat tional Grid Marc Cyr, Morgan n Schaffer (200 08 Doble Enginee ering Company 75th Annua al International Doble Client t Conference ANSI/I EEE C57.104-1978 - American A National Stan ndard guide) for th he detection and determination n of generated gas ses in oilimmersed tran nsformers and the eir relation to the serviceability of the equipment t; Using Dissolv ved Gas Analysis to Detect Active Faults in Oil-I Insulated Electric cal Equipment Lance R. Lew wand Doble Eng gineering IEEE August 2005, M. Du uval, J. Dukarm, Improving the reliability of tra ansformer in gasin-oil diagnos sis.

2-

3-

45-

This T technical pap per was presented d during the NE ETA PowerTest Electrical E Mainten nance & Safety Conference e in Washington, DC on Feb bruary 21st 2011

_____ ___________ ____________ __________

Sale es Director Eastern E USA and a Mexico

Claud de Hermann n

Morgan n Schaffer In nc.

_____ ___________ ____________ __________

Page|7