New Approach for Fault Location on Transmission Lines Not Requiring Line Parameters

Transcription

1 New Aroach for Fault Locatio o Trasmissio Lies Not equirig Lie Parameters Z. M. adojević, C. H. Kim, M. Poov, G. Presto, V. Terzija Abstract This aer resets a ew umerical algorithm for fault locatio o trasmissio lies. t does ot require lie arameters, which is a radical ste forward comared to the existig aroaches, which require this iformatio, so the algorithm ca be cosidered as a settigs-free algorithm. Lie arameters are oly aroximately costat; they differ with the loadig ad weather coditios. Thus, a aroach which does ot require them would be more robust, accurate ad flexible, tha those aroaches that do require lie arameter iformatio to determie the locatio to the fault. This is essetial for the fast ad secure elimiatio of faults o trasmissio lies, ad, cosequetly, the solutio leadig to a sigificat imrovemet of the quality of the eergy suly. The ew algorithm uses sychroised data samlig at both lie termials, which is a rerequisite for the successful algorithm alicatio. The aer resets the results of the iitial algorithm testig through the use of ATP-EMTP simulatios. Keywords: Fault locatio, trasmissio lies, GP sychroizatio, settigs-free algorithm.. NTODUCTON eliable algorithms for the aalysis of faults o overhead trasmissio lies have become a essetial art of moder trasmissio lie rotectio schemes. A itegral art of such algorithms is the fault locator, which determies the distace to the fault from the local lie termial(s). Various methods of fault locatio have bee develoed i the ast, some of which use data from oe lie termial ad some of which use data from two or more lie termials [1]-[1]. Further ublicatios o this imortat subject ca be foud i [11]. Oe commo factor betwee all of the various methods is the requiremet of the lie arameters ad the lie legth to determie the fault distace. However, the arameters of lies are ot always available ad they ca vary with differig loadig ad weather coditios. [12], Liao ad Elagova roosed a algorithm for locatig faults o trasmissio lies which does ot require the arameters of the lie i questio. Because the Z. M. adojević is with The Uiversity of Belgrade, Faculty of Electrical Egieerig, erbia, ( radojevic@ieee.org). C. H. Kim is Director of Ceter for Power T, ugkyukwa Uiversity, uwo, Korea ( chkimskku@yahoo.com). M. Poov is with the Delft Uiversity of Techology, The Netherlads, e- mail: (m.oov@tudelft.l). G. Presto ad V. Terzija are with The Uiversity of Machester, UK, ( g.resto@studet.machester.ac.uk, terzija@ieee.org). Paer submitted to the teratioal Coferece o Power ystems Trasiets (PT29) i Kyoto, Jaa Jue 3-6, 29 algorithm assumes asychroous data samlig, it requires the use of iterative techiques to locate lie-lie faults ad it caot be used to locate balaced three-hase faults. Also, it does ot cosider zero-sequece coulig. cotrast, the algorithm roosed i this aer ca be used to locate all fault tyes icludig balaced three-hase faults without recourse to sulemetary techiques, ad, because it uses ositive ad egative-sequece comoets, it is ot suscetible to zero-sequece coulig. The aim of this aer is to develo a ew umerical algorithm that ca determie the distace to the fault o a trasmissio lie without ay kowledge of the lie arameters. t was assumed that the ukow fault locatio will be determied from voltage ad curret hasors, sychroously measured at both lie termials. a umber of aroaches, the fault locatio is calculated by usig iformatio about the lie arameters (e.g. resistace ad iductace er uit legth, which are kow i advace) ad the measured voltage ad curret hasors. Here the fault locatio calculatio is determiistic ad based o a suitable formula for calculatig the fault locatio. reality, a umber of stochastic factors geerate a high level of ucertaity i determiig the fault locatio. For examle, the fault resistace, or fault arc, other o-liear effects o istrumet trasformers, as well as the radom errors itroduced durig A/D coversio, ca adversely affect the accuracy of fault locatio. This makes the roblem eve more challegig ad requires very careful algorithm desig. The aer discusses sychroised data samlig at each lie termial ad gives a detailed algorithm derivatio. Furthermore, it icludes a ectio i which the ew algorithm was tested ad validated through comuter simulated tests. ice the ew method is based o the use of the emergig ychroised Measuremet Techology (MT), a brief descritio of this is give i the ext ectio. The ew algorithm is develoed for the most frequet fault tye: the sigle lie to groud fault. From the methodology reseted i this aer, a more geeral solutio could be also derived. The authors itetio was to roose the ew aroach for fault locatio, to test it ad to assess the limitatio of the ew aroach.. YNCHONZED MEAUEMENT TECHNOLOGY ychroized Measuremet Techology (MT) is a ew cocet, based o the use of the Global Positioig ystem (GP), sychroisig differet itelliget Electroic Devices

2 (EDs) i a large etwork. The ew cocet icludes the followig mai buildig blocks: a) sychroized measuremet uits (MUs, also commoly kow as hasor measuremet uits - PMUs), b) data cocetrators, c) commuicatio etwork ad d) alicatios. CT1 CB1 satelite CB 2 VT1 VT 2 CT 2 The ew algorithm for the settigs-free fault locatio, which will be reseted i this aer, resets a ew MT alicatio, where it is assumed that the other required buildig blocks are available ad techically ready to fulfil the algorithm s requiremets. uch ractical imlemetatio issues will ot be discussed i this aer. t is assumed that the MT ifrastructure is caable of fulfillig the algorithm s requiremets (e.g. reliable ad fast commuicatio liks betwee remote lie termials). The Global Positioig ystem (GP) is a sace-based ositioig, avigatio, ad timig system, cosistig of 24 satellites that costatly revolve aroud the earth. ice a GP receiver rovides time sychroizatio with a accuracy of ±1 μs, PMUs usig GP-sychroizatio i ower systems ca be successfully alied to fault locatio/aalysis ad the rotectio of overhead trasmissio lies. The GP is used to sychroize the clocks of the measurig devices which samle voltages ad currets at the lie termials. For the urose of the ew settigs-free fault locatio algorithm it is assumed that the sychroisatio is achieved at the samle/samlig level. Fig. 1 two ychroized Measurig Uits (MUs) at each lie termial are reseted. This would be the architecture ecessary for the successful imlemetatio of the ew algorithm. Fig. 1 two MUs costatly acquire voltage ad curret samles from voltage ad curret istrumet trasformers istalled at each ed of the lie. The voltage ad curret samles collected from the MUs are forwarded to a cetral Data Cocetrator (DC), i which the rocessig of simultaeously samled voltages ad currets is carried out. This rocessig is actually the ew umerical algorithm for fault locatio. f faulty coditios are detected, the umerical algorithm for fault locatio is istatly started ad withi 2-4 ms the fault locatio is determied. How fast the ukow fault locatio will be determied deeds o the method for extractig the voltage ad curret hasors. this aer this issue will ot be searately cosidered. imly, the commoly used Fast Fourier Trasform will be imlemeted for this task. the ext aer ectio, the ew settigs-free umerical algorithm for fault locatio, based o sychroised data acquisitio from both lie termials, will be reseted. MU1 DC MU 2 Fig. 1. ychroized samlig arragemet (CT curret trasformer, VT voltage trasformer, CB - circuit breaker, MU sychroised measuremet uit, DC data cocetrator).. DEVATON OF THE NEW FAULT LOCATON ALGOTHM this ectio, a ew, settigs-free umerical algorithm for fault locatio will be reseted. t will give solutios for sigle lie to groud faults, which are the most commo tye of fault o overhead trasmissio lies. Let us assume a a-hase sigle lie to groud fault o a trasmissio lie at distace l from the left lie termial, as reseted i Fig. 2. Assumig that the lie is less tha 1 km log (such lies ca be cosidered as short lies), the shut caacitace ad the shut coductace of the trasmissio lie ca be eglected. Fig. 2 the fault locatio is deoted by F, the fault distace by l, D is the lie legth, ad subscrits ad deote the sedig- ad receivig ed of the lie, resectively. Voltages ad currets are sychroously samled at both lie termials. Usig stadard sigal rocessig techiques, the corresodig hasors ca be calculated from the voltage ad curret samles. this aer the Fast Fourier Trasform (FFT) was used to filter the voltage ad curret hasors from their samles. A traditioal aroach, i which voltages ad currets were samled with the samlig frequecy f s =3.2kHz (64 samles er 2ms) was used, settig the data widow size, T dw, to corresod to the fudametal eriod, T =2ms. As with ay method, the efficiecy of the ew method deeds uo the fast ad reliable determiatio of the voltage ad curret hasors. This aer will ot cosider the sesitivity of the ew algorithm to the quality of the hasors determiatio. t is kow, that the FFT aroach is sesitive to the existece of the decayig DC comoet i the fault curret. For this challege, a searate solutio ca be derived. Other solutios such as otimal estimatio theory [13] ca also be used for hasor extractio.

3 D V V z = V z( D ) (1) z = V z( D ) (2) c b a a b c V,, a, b, c F a, b, c a b c V,, where:,, V, V are the ositive- ad egative sequece hase voltages at both lie termials;,,, are the ositive- ad egative sequece hase currets at both lie termials; z are ositive- or egative sequece lie imedace, which are equal. Fig. 2. Three-hase reresetatio of the faulted lie. Based o the voltage ad curret hasors, by usig the symmetrical comoets method, the ositive, egative ad zero sequece symmetrical comoets of the voltages ad currets ca be determied. The ew method requires just the ositive ad egative sequece comoets, i.e. their equivalet circuits. o, the usymmetrical three hase circuit from Fig. 2 (it is usymmetrical, because the fault at the oit F is a lie to groud fault) ca be rereseted by three siglehase equivalet circuits ositive (), egative (), ad zero () sequece circuits, resectively. As stated, oly the ositive ad egative sequece comoets/circuits will be used. These are reseted i Figs. 3a ad 3b. V z z( D ) F V t is obvious, that i equatios (1) ad (2) just two terms ( z ad z ( D ) ) are ukow. They ca easily be determied by solvig equatios (1) ad (2). The two solutios are give i followig two formulas: ) ( V ) ( V z = (3) ) ( V ) ( V z( D ) = (4) Covetioal fault locatio methods based o oe or two termial data, would be solved usig the above equatios ad iformatio about the lie arameters (Ω/km) ad X (Ω/km). As discussed above, a more flexible solutio ca be achieved if the fault locatio algorithm is ideedet of lie arameters. This was achieved ad demostrated i the followig text. Let us exress the distace to the fault, l, as a ercetage of the lie legth, D, through the followig formula: Fig. 3a. Equivalet ositive sequece circuit of the faulted lie from Fig. 2. % = 1 (5) D z z( D ) F The above formula ca be reseted i the followig form: V V % = z 1 z + z( D ) (6) Fig. 3b. Equivalet egative sequece circuit of the faulted lie from Fig. 2. both of the equivalet circuits i Figs. 3a ad 3b, the lie ositive ad egative imedaces are equal. For the circuits deicted i Figs. 3a ad 3b the followig equatios hold: After icludig the exressios for z ad z ( D ) i equatio (6), the followig formula for the distace to the fault ca be obtaied: ( V ) ( V ) % = 1 (7) ( V )( + ) ( V )( + )

4 where ad deote egative ad ositive sequece variables, ad ad deote the sedig- ad receivig ed variables. Obviously, the algorithm does ot require the lie arameters or iformatio about the lie legth. t is just based o the rocessig of sychroously recorded voltage ad curret samles, determiig their 5 Hz hasors, ad calculatio of the ositive ad egative sequece voltage ad curret comoets. Equatio (7) rovides the exlicit exressio for fault locatio i the case of sigle lie to groud faults o overhead trasmissio lies. The ew settigs-free umerical algorithm was thoroughly tested usig a umber of comuter simulated cases, as reseted i the ext aer ectio. V. COMPUTE MULATED TET igle lie-to-groud bolted (metallic) faults were simulated at differet locatios alog the lie. each case the fault icetio time was set at t = 23 ms. t was assumed that the lie was loaded before the fault icetio. The samlig frequecy was f s = 3.2 khz. The data widow size was T dw = 2 ms (N = 64 samles er data widow). t is assumed that all hasors are erfectly sychroized (the sychroizatio error is equal to degrees). Figs. 5-8 show the simulated faulted hase voltages ad currets at both lie termials, which were used as a iut to the ew algorithm. t is obvious that all curves corresod to the sigle lie to groud fault. The fault was simulated at the 35th km, observed from the left lie termial (termial ). The algorithm testig was carried out through dyamic simulatio aalysis usig the ATP-EMTP software [14]. Fig. 4, the schematic diagram of a 4 kv, 1 km log, overhead trasmissio lie is reseted. Tables ad, the arameters of active etworks A ad B ad the lie arameters are listed, resectively. Network A Faulted lie Network B hase voltages (kv) Fig. 5. Faulted hase voltages at the lie sedig ed (). Fig. 4. igle lie diagram of the simulated faulted lie. Parameters TABLE PAAMETE OF NETWOK A AND B. Networks A U LL,M [kv] ϕ 1 [ ] -2 [Ω] L [H] , [Ω] L, [H] B hase voltages (kv) TABLE LNE PAAMETE. Parameter - ad -sequece -sequece resistace, Ω/km Fig. 6. Faulted hase voltages at the lie receivig ed (). iductace mh/km

5 hase currets (ka) Fig. 7. Faulted hase currets at the lie sedig ed (). From Fig. 9 it is obvious that accurate values were obtaied after a short covergece rocess. The algorithm covergece roerties are determied by the quality of the method for hasors extractio. As metioed above, the FFT algorithm was used for this urose, ad it is kow to be sesitive to the decayig DC comoet existig i the fault curret(s). this aer, methods of imrovig the quality of extractio of voltage ad curret hasors are ot cosidered. [13] some alterative solutios for this roblem, which are ot sesitive eve to chages of the system frequecy, are roosed. t is obvious that accurate values are obtaied. order to further demostrate the algorithm efficiecy, faults were simulated over a rage of distaces. The results (i %) for the differet fault locatios are reseted i Fig. 1. ice the lie legth was 1km, based o the results obtaied, it ca be cocluded, that the accurate results were obtaied for the full rage of fault locatios. hase currets (ka) Fig. 8. Faulted hase currets at the lie receivig ed (). fault distace (%) km 6km 35km 15km 1km Fig. 1. Calculated fault locatios (i %) for differet fault locatios. Based o the voltages ad currets show i Figs. 5-8, the ukow fault locatio was calculated usig equatio (7), as reseted i Fig. 9. fault distace (%) km fault the ext test examle, the algorithm sesitivity to the fault resistace ad fault arc was ivestigated. A arcig fault over fault resistace ( F =1Ω) at 6km was simulated. The legth of the arc was aroud 3 m. The fault arc was simulated as a curret ad arc legth deedet voltage source [15]. Fig. 11 the sedig lie termial voltages for this case are reseted. From the waveform of the faulted hase voltage, it ca be oted that it is slightly distorted, a tyical cosequece of the fault arcs. The equivalet distortios are obvious i the faulted hase voltage sigal from the receivig lie ed, as well. site of such fault coditios, the algorithm delivered the correct result (see Fig. 12, i which the calculated fault distace, 6km, is reseted), rovig that it is ot sesitive to fault ad arc resistaces. Fig. 9. Estimated fault locatio (the exact value used by EMTP was 35 km).

6 hase voltages (kv) Fig. 11. Faulted hase voltages at the lie sedig ed () As stated above, the algorithm reseted requires ideal sychroizatio, which might be a challege i real-time alicatios. Based o the comuter simulatio testig, it was cocluded that a iaccurate sychroizatio itroduces errors i the alicatio. For examle, for a sychroizatio error of t = 1ms, the error i fault distace is roud 3%. The authors are curretly derivig ew solutios, which will ot be sesitive to sychroizatio errors. fault distace (%) km fault Fig. 12. Estimated fault locatio (6 km arcig fault over fault resistace). V. CONCLUON this aer a ew efficiet settigs-free umerical algorithm for fault locatio o overhead trasmissio lies is reseted. The algorithm is based o sychroized measured voltages ad currets, samled at both lie termials, ad the use of fast commuicatio chaels betwee two sychroized measuremet uits (MUs) istalled at the lie termials. t does ot require lie arameters to determie the fault locatio. the aer, the solutio for the most commo fault tye, the sigle lie to groud fault, is reseted ad thoroughly tested. The algorithm is derived i the sectral domai ad based o the rocessig of voltage ad curret hasors. For this urose it alies the Fast Fourier Trasform aroach. Through algorithm testig it was roved that the algorithm efficietly ad accurately determies fault locatios ad that is uaffected by fault ad arc resistaces. The authors are curretly derivig a solutio that does ot require sychroized samlig. V. EFEENCE [1] K. Takagi, Y. Yomakoshi, M. Yamaura,. Kodow, ad T. Matsushima, Develomet of a ew tye fault locator usig the oe termial voltage ad curret data, EEE Tras. Power A. yst., vol. PA-11, , Aug [2] A. Girgis, D. Hart, adw. Peterso, A ew fault locatio techique for two- ad three-termial lies, EEE Tras. Power Delivery, vol. 7,.98 17, Ja [3] M. Abe, T. Emura, N. Otsuzuki, ad M. Takeuchi, Develomet of a ew fault locatio system for multi-termial sigle trasmissio lies, EEE Tras. Power Delivery, vol. 1, , Ja [4] J.-A. Jiag, J.-Z. Yag, Y.-H. Li, C.-W. Liu, ad J.-C. Ma, A adative PMU based fault detectio/locatio techique for trasmissio lies art : Theory ad algorithms, EEE Tras. Power Delivery, vol. 15, , Ar. 2. [5] A. Goalakrisha, D. Hamai, M. Kezuovic, ad. McKea, Fault locatio usig the distributed arameter trasmissio lie model, EEE Tras. Power Delivery, vol. 15, , Oct. 2. [6] M. Djurić, Z. adojević, ad V. Terzija, Distace rotectio ad fault locatio utilizig oly hase curret hasors, EEE Tras. Power Delivery, vol. 13, , Oct [7] M. Tawfik ad M. Morcos, ANN-based techiques for estimatig fault locatio o trasmissio lies usig Proy method, EEE Tras. Power Delivery, vol. 16, , Ar. 21. [8]. Tamroglak,. Horowitz, A. Phadke, ad J. Thor, Aatomy of ower system blackouts: Prevetive relayig strategies, EEE Tras. Power Delivery, vol. 11, , Ar [9] Chi-ha Yu, Chih-We Liu, u-li Yu, ad Joe-Air Jiag, "A New PMU-Based Fault Locatio Algorithm for eries Comesated Lies", EEE Tras. o Power Delivery, Vol. 17, No. 1, , Ja. 22. [1] D. Novosel, D.G. Hart, E. Udre, J. Garitty, Usychroized Two- Termial Fault Locatio Estimatio, EEE Tras. o Power Delivery, Vol. 11, No. 1, Jauary [11] T..idhu, at all, "Bibliograhy of elay Literature, 23 EEE Committee eort" EEE Tras. Power Delivery, vol. 21, , Ja. 26. [12] Y. Liao,. Elagova. Usychroized two-termial trasmissio-lie fault-locatio without usig lie arameters. EE Proc.-Geeral Trasmissio Distributio, Vol.153, No.6, November 26. [13] V.Terzija, "mroved ecursive Newto Tye Algorithm for Frequecy ad ectra Estimatio i Power ystems", EEE Tras. o strumetatio ad Measuremet, Vol. 52, No. 5, October 23, [14] htt:// [15] V.Terzija, H.-J.Kogli, O The Modelig of Log Arc i till Air Ad Arc esistace Calculatio, EEE Trasactios o Power Delivery, Vol. 19, No. 3, July 24, Page(s):