Fault Locator Using Travelling Waves: Experience in the Belgian Transmission Network

Transcription

1 Faul ocaor Using Traelling Waes: Experience in he Belgian Transmission Nework X. Busamane-Mparsakis, J.C. Maun Absrac Traelling Waes (TWs) are elecro-magneic ransiens ha are generaed when here is a sudden change of olage in he nework, such as when a faul occurs. Faul locaors based on TWs hae recenly emerged as an alernaie o ypical impedance faul locaors hanks o heir higher precision in mos cases. The deelopmen of new algorihms based on TWs is improed by he undersanding of heir behaiors, in paricular wih he analysis of high frequency records of fauls. Such records are no readily aailable. We performed a measuremen campaign in he Belgian ransmission nework o acquire records of TWs generaed when fauls occur. This paper repors he experience acquired during he campaign, and discusses faul locaion algorihm improemens o accoun for he unique opology of he moniored line. The algorihm improemen reduces he error caused for non-homogeneous lines, which can be significan for big lines. The records analysis showed hree imporan facors ha influence he shape of he recorded TWs: he disconinuiy in he line caused by a T- juncion, he bandwidh of he measuremen ransformers, and he secondary cables conneced o he curren ransformers. Therefore, hose effecs should be carefully considered when deeloping new algorihms based on TWs. Keywords: Faul locaion, Traelling waes, Field experience, Faul records. A I. INTRODUCTION fer a faul occurs on he power nework, he precise knowledge of he faul locaion is of significan imporance, especially for long lines. I allows a fas dispaching on locaion o sole he problem. When a faul occurs, he sudden change of olage generaes elecro-magneic ransiens: he Traelling Waes (TWs) [1], []. Those TWs propagae in boh direcions from he faul poin, wih a elociy close o he speed of ligh in oerhead lines [1], [3]. The exac elociy depends on he geomery of he line. The increasing performance of measuremen deices made i possible o record TWs in subsaions. Traelling Wae Faul locaors (TWF) hae hus recenly emerged as an alernaie o ypical faul locaors based on impedance measuremens. They hae a beer precision, which is no affeced by he faul resisance or load flow. The mos commonly used TWF algorihm, he ype-d [3], is based on measuring he arrial ime of he firs TW a boh ends of he fauly line. For a good precision, he signals mus X. Busamane-Mparsakis and J.C. Maun are wih he Deparmen of bio, elecro and mechanical sysems, Free Uniersiy of Brussel, Belgium. ( s: xbusama@ulb.ac.be, jcmaun@ulb.ac.be) Paper submied o he Inernaional Conference on Power Sysems Transiens (IPST017) in Seoul, Republic of Korea June 6-9, 017 be recorded wih a high sampling frequency (ypically higher han 1MHz), and a good ime synchronizaion is needed beween he records a boh ends of he line. While he classic ype-d algorihm is simple, he aenuaion and disorion of TWs make i difficul o precisely deermine heir arrial ime. Due o heir high elociy, an error of 1µs in he arrial ime leads o an error of ~300m for he faul locaion. Many recen sudies aim a deeloping algorihms ha minimize his error [4]-[7]. I is dangerous o deelop such algorihms based on simulaions only, which represen a simplified ersion of realiy due o he assumpions made. Many phenomena occurring when recording TWs are unexpeced or oherwise difficul o model correcly. For example, he measuremen ransformers affec he high frequency signals. They are designed o perform a power frequencies, and few works hae been done o sudy heir bandwidh a high frequencies. Those sudies agree ha curren ransformers hae a good bandwidh and proide a correc reproducion of curren TWs, while olage ransformers hae a poorer bandwidh and will significanly affec he olage TWs recorded [3], [8]. The exac bandwidh of he measuremen ransformers is deicespecific, and an accurae high frequency model of hose ransformers is a difficul ask. Anoher effec seldom alked abou is he secondary cable ringing, which significanly affecs he signals recorded: fas reflecions of TWs occur in he conrol cables conneced o he secondary side of curren ransformers [4], [9]. The analysis of faul-induced ransiens recorded in acual subsaions is imporan o beer he undersanding of TWs, and o he deelopmen of faul locaion and proecion algorihm based on TWs, and o he improemen of simulaion ools. Such high-frequency faul records are no readily aailable and are difficul o acquire. We performed a measuremen campaign in he Belgian ransmission nework o acquire records of TWs generaed when fauls occur. We obained ime-synchronized records of one lighning-induced faul and of wo line energizing. This paper repors he experience acquired during he campaign, and discusses improemens o he classic ype-d algorihm o accoun for he unique opology of he moniored line. II. MEASUREMENT CAMPAIGN We performed a measuremen campaign o acquire records of TWs generaed inside he power nework when fauls occur. The measuremen deices were insalled in subsaions on he Belgian ransmission nework o monior a 70kV line.

2 A. Nework Topology The line ha was moniored during he measuremen campaign presens a unique opology (Fig. 1). A T-juncion is presen inside he line, bu he addiional branch is shor (165m) and ends in a power ransformer ha was no conneced o he nework during he es period. Addiionally, one secion is a parallel line while he oher secion is a single line. Classical iron-cored curren ransformers and inducie olage ransformers were used o reduce he signals o be recorded. The currens are measured enering he line. Fig. 1. Topology of he moniored 70 kv line. Measuremens are performed in saions A and B. Due o pracical consrains on he rigger, differen measuremen equipmen was used for recording fauls and line energizing, and sauraion is presen in he energizing records. All he measuremens are synchronized using GPS anenna wih a precision of 100ns. B. Records Oeriew Six useful records were acquired: four faul records and wo line-energizing records. Due o echnical difficulies on he rigger, we only acquired ime-synchronized records of one faul (Fig. ). A succeeding similar faul record was acquired in subsaion A only. In addiion, wo line energizing records were acquired (Fig. 3). Those synchronized records will be deeloped in his paper. The power frequency was filered ou of all he records presened. Fig.. Synchronized faul record during a lighning srike. III. PROPAGATION SPEED During he energizing of a line, he hree phases are no closed simulaneously. Each phase closure generaes new TWs. One energizing record can proide us wih hree propagaion eens o measure he propagaion ime. For our second record, only one phase closure is usable due o he olage sauraion of he records. A. Propagaion Time Energizing Records The propagaion ime was measured for each een (Table I). The oal propagaion ime inside he moniored line is aeraged as 7.15 µs, and is subjec o errors caused by he ime synchronizaion, he sag of he line and he error made when measuring i. This gies an aerage propagaion speed inside he line of m/µs. The oal propagaion ime is he sum of he propagaion ime inside he parallel line secion and inside he single line secion. The TW elociy inside each secion is differen due o he change in geomery. If we assume a consan speed in he line, an error is added o he faul locaion. Fig. 3. Synchronized record during he re-energizing of he line. TABE I PROPAGATION TIME MEASURED WITH THE ENERGIZING RECORDS Phase closed A (een 1) 71.8 B (een 1) 7.3 C (een 1) 71.9 C (een ) 7.5 prop (µs)

3 B. Propagaion Speed - Simulaions A ypical oerhead line was modelled in EMTP wih a J- Mari line model, using he geomery of a specific line proided in [10]. The CC rouine of EMTP compues he line parameers (he disribued resisance, inducance, capaciance and conducance) based on is geomery. Single and parallel lines were modelled, and heir geomery was aried. The line parameers are affeced by his change in geomery, which modifies he propagaion speed inside he line [11]. The propagaion speed inside he lines for each modelled geomery is displayed in Table II. This analysis is jus an example o illusrae he ariabiliy of he propagaion speed, and real resuls will differ. TABE II PROPAGATION SPEED INSIDE INES MODEED WITH EMTP Parameer Value Propagaion speed Single line Conducor diameer 10 mm 98.5 m/µs 0 mm 97.8 m/µs 30 mm 97.4 m/µs Phase disance 1 m 96.7 m/µs m 97.8 m/µs 3.5 m 98 m/µs Parallel line ines disance 5 m 91.7 m/µs 9 m 86 m/µs 16 m 80.3 m/µs The propagaion speed in single lines has lile ariabiliy wih he changes in geomery. In parallel lines i is slower, and aries significanly wih he disance beween lines. The aerage propagaion speed measured in he moniored line (93.46 m/µs) is in agreemen wih hose resuls, since he line is composed of boh single and parallel line secions. C. Non-homogeneous ine Analysis If we assume a consan speed in he line, an error is added o he faul locaion. I is impossible, in pracice, o measure he propagaion ime in each secion independenly. Two opions are aailable o accoun for he non-homogeneousness of he line: Model he lines in order o deermine heir parameers; Assume he elociy in one secion of he line; The firs mehod consiss in modelling he line o deermine he propagaion speed in each secion. I requires he precise knowledge of he geomery of he line o deermine he propagaion speed based on he parameers. We propose a soluion where we assume he elociy inside he single line secion. We make his assumpion hanks o he small ariabiliy of propagaion speed in single lines. The elociy inside he parallel secion is hen compued as in (). 1 prop sec1 sec 1 (1) 1 prop 1 Where prop is he oal propagaion ime measured inside he line (s), sec1 and sec are he propagaion ime respeciely in he parallel line secion and in he single line secion (s), 1 and are he propagaion speed in each secion (m/s), 1 and are he secion lenghs (m). In his paper, we chose =97.8 m/µs (based on [10]), which gies 1 =9 m/µs. IV. FAUT OCATION AGORITHM A. Classic Type-D Algorihm The mos frequenly used TWF algorihm is he so-called ype-d [3]. The TWs generaed a faul poin propagae in boh direcions and reach saions A and B (Fig. 4). Such algorihms require measuremens a boh ends of a line wih a good ime synchronizaion, bu hae he adanage of requiring only he arrial ime of he firs TW. The faul locaion is compued as in (3). Fig. 4. Type-D faul locaor [1]. ( ) D A B (3) where D is he faul locaion from saion A (m), is he wae elociy (m/s), is he line lengh (m), and A and B are he arrial imes of he firs inciden wae a each saion (s). B. Type-D Algorihm for Non-Homogeneous ines Equaion (3) is correc when he propagaion speed is consan along he whole line. For non-homogeneous lines (such as he moniored line), he propagaion speed differs depending on he secion. The ypical ype-d algorihm has o be modified o accoun for his speed difference. For a faul ha occurs on secion 1 of a line ha consiss of wo disinc secions, he propagaion imes before reaching each saion are expressed in (4) and (5). This leads o a faul locaion as expressed in (7). A D 1 () (4) D 1 B (5) 1 D D 1 B A (6) 1 1

4 D (7) Where A and B are he propagaion ime from faul poin o saions A and B (s), D is he faul locaion (m), 1 and are he lenghs of each secion (m), and 1 and are he propagaion speed in each secion (m/s). If he faul occurs on secion, he faul locaion is found wih (8). The use of his algorihm herefore requires a selecion of which secion is a faul. This can be done wih he measuremen of Δ and he knowledge of he opology. D 1 (8) Where is he oal lengh of he line (m). C. Type-A Algorihm Type-A faul locaion algorihms use he measuremens on one side of he line only. The faul locaion is compued based on he measuremen of he arrial ime of he firs TW, and he arrial ime of he firs reflecion of ha TW on he faul poin, as depiced on Fig. 5. Fig. 5. Type-A faul locaor [1]. D (9) Where 1 is he arrial ime of he firs inciden wae (s) and 3 is he arrial ime of he firs reflecion on faul poin (s). The main challenge in using ype-a algorihms is o recognize he TW reflecion on faul poin from all he oher reflecions occurring in he nework. In he example depiced in Fig. 5, he TW reflecion on saion B reaches saion A (a ) before he desirable reflecion on faul poin (a 3 ). To aoid he need o idenify each reflecions, he arrial ime of he firs reflecion can be approximaed hanks o he faul locaion found wih he ype-d algorihm. The ype-a algorihm can hen be used as alidaion and correcion. I remoes a source of error coming from he ime synchronizaion. D. Applicaion o Simulaions The ypical and updaed ype-d algorihms were applied o a simulaion model. The simulaion models a nonhomogeneous 50 km line wih EMTP using he J-Mari line models described in secion III. B. The firs par of he line is a parallel line where he propagaion speed is lower, and he second par is a single line (Fig. 6). The exac line models are no imporan o he discussion. The poin of his secion is o illusrae he imporance o use an updaed ype-d algorihm for non-homogeneous lines where he elociy is no consan in he whole line. The faul occurs a 0km from saion A. Fig. 6. ine modelled in EMTP To illusrae he error inroduced by each algorihm, he propagaion speed inside each secion is assumed unknown. The only aailable informaion is he measured oal propagaion ime inside he line during energizing, and he measuremen of Δ during he faul. Boh algorihms are applied o find he faul locaion (Table III). TABE III CASSIC AND UPDATED TYPE-D AGORITHMS APPIED TO SIMPE SIMUATIONS ine daa Secion 1 elociy 80.3 m/µs Secion elociy 98.5 m/µs Measuremens Toal propagaion ime µs Δ= B- A 30.4 µs Classic ype-d compuaion Aerage speed 89.1 m/µs Faul locaion 0.63 km Updaed ype-d compuaion Secion speed (assumpion) 97.8 m/µs Secion speed (esimaion) m/µs Faul locaion 0.04 km When applying he classic ype-d algorihm, he elociy inside he line is assumed homogeneous. This assumpion inroduces an error of 630m for his paricular example. The updaed ype-d algorihm assumes only he elociy inside he single line secion. This assumpion inroduces a smaller error of 40m for he faul locaion. E. Applicaion o he Faul Record The updaed ype-d algorihm was applied o our synchronized faul record, using he elociies preiously compued. The faul locaion compued is inside he range proided by he TSO, and is alidaed by a ype-a faul locaion algorihm (Table IV). A correcion of 105m is applied wih he ype-a algorihm. TABE IV TWF AGORITHMS APPIED TO THE FAUT RECORD TSO faul locaion 8.18 ± 1.63 km Updaed ype-d 7.83 km Expeced 1 s reflecion µs Measured 1 s reflecion µs Type-A algorihm 7.75 km

5 V. FAUT RECORDS ANAYSIS We analyzed he faul records o undersand and illusrae he differen effecs of he power nework and subsaions on he TWs recorded. A. T-Juncion Fas reflecions are caused by he T-Juncion. The T- Juncion is a poin of disconinuiy in he sysem. A any poin of disconinuiy, reflecions occur [1]. When he TW reach he T-Juncion, par of he wae will coninue hrough he line and par of he wae will be moing oward he power ransformer in C (Fig. 1). Fas reflecions ensue beween he juncion and he power ransformer. This is only obserable in he remoeend which has he T-Juncion in-beween i and he faul poin (saion B in our case). The reflecions should occur eery 1.1 µs (eq. (10)) and are obsered in Fig. 7. (10) 3 T 1.1µs Where T is he ime beween wo reflecions caused by he T-Juncion (s), 3 is he lengh of he shor line conneced o he juncion (m), and is he propagaion speed inside he single line secion (m/s). reflecions. The differen cable lenghs in each subsaion generaes ripples of differen period (Fig. 9). Fig. 8. The olage waes in subsaion A are differen from he olage waes in subsaion B. Fig. 9. The effec of curren ripples caused by he CT secondary cabling is obsered in boh saions. D. Firs Reflecion on Faul Poin The expeced arrial ime of he 1 s refleced TW in saion A (=548.3 µs) was compued on secion IV E. This reflecion is no obserable in saion B due o he T-Juncion, creaing oo many pre-reflecions before he reflecion on faul poin. Fig. 7. The T-juncion creaes fas curren reflecions in subsaion B. B. Volage Transformers I is agreed in he lieraure ha olage ransformers hae a wors bandwidh han curren ransformers [3], [8]. In our records, we found ha in subsaion A, he olage waes were clearly deecable, een if less seep han he curren waes. In subsaion B howeer, he olage waes are an order of magniude lower wih bigger reflecions proporionally (Fig. 8). This difference in olages acquired and he frequencies ha appear are no explained wih he nework and are belieed o be heaily affeced by he olage ransformers. The undersanding of he olage waes recorded will require addiional work o be performed on he subjec. C. Curren Transformers Curren ransformers hae a beer bandwidh for TWs, bu heir secondary cables creae ripples on he curren waes [4], [9]. A one end of hose secondary cables, he TW sees a high impedance (from he curren ransformer), and a he oher end i sees a low impedance (he relay), which creaes he fas In saion A, his reflecion can be deeced beer wih he currens han wih he olages (Fig. 10). I is challenging o recognize his firs reflecion from all he oher eens occurring wih he TWs, bu he arrial ime of he firs reflecion was successfully used as a alidaion of he ype-d algorihm. Fig. 10. The firs reflecion should occur a =548.3µs based on he compued faul locaion.

6 E. Successie eens Successie records for disinc lighning srokes proide similar shapes (Fig. 11), which show ha he resuls are reproducible for differen bu close lighning srokes (1.4 km apar in his case) and mus be explained by he nework and he subsaions. Daa from he phase C faul hae been inered in order o beer compare he resuls. In hose records, we obsere ha he curren and olage waes display similar shapes. This is expeced, since he TWs are reduced wih he same measuremen ransformers. The curren ripples caused by he CT secondary cables hae he same reflecion period, bu wih differen ampliude. The period depends on he lengh of he secondary cables and is consan for a gien subsaion. The ampliude depends on he faul locaion (a furher faul will delay he arrial of ground mode waes) and on he inciden wae ampliude. recorded TWs: he disconinuiies inside he line, he bandwidh of he measuremen ransformers, and he secondary cables conneced o he curren ransformers. Those effecs are difficul o model, and should be carefully considered when deeloping new algorihms based on TWs. The records presened in his paper can be used o improe and alidae simulaion models for high frequencies suiable for TWs sudies. VII. ACKNOWEDGMENT The auhors graefully acknowledge Elia, he Belgian ransmission sysem operaor, for he opporuniy o perform a long-erm measuremen campaign in heir subsaions. VIII. REFERENCES Fig. 11. The signals in subsaion A are similar for similar eens (lighning srokes ~1.4 km apar). F. Waes polariy The polariy of he waes are in coherence wih he analysis found in he lieraure. In [13], we see ha for fauls occurring on he line, he TW from one phase has a differen polariy from he oher wo. In [14], we see ha he curren and olage waes measured in one subsaion hae a differen polariy for forward fauls. VI. CONCUSIONS This paper presened an updae of he classic ype-d faul locaion algorihm. This updaed algorihm akes ino accoun non-homogeneous lines. I was applied o EMTP simulaions and field records, and was alidaed wih a ype-a algorihm. I decreases he error caused by non-homogeneous line when using a ype-d TWF. This paper also presened faul records of TWs generaed during a faul, and during he energizing of a line in he Belgian ransmission nework. The analysis of he records showed ha hree facors significanly affec he shapes of [1]. Van der Sluis, Transiens in Power Sysems. Chicheser: John Wiley & Sons d, 001. [] A. M. Elhaffar, Power ransmission line faul locaion based on curren raeling waes, Ph.D. disseraion, Dep. of Elec. Eng., Uni. of echnology, Helsinki, 008. [3] G. Krzyszof, R. Kowalik, D. Rasolomampionona, and S. Anwar, Traeling wae faul locaion in power ransmission sysems: an oeriew, J. Elecr. Sys., ol. 3, no. 7, pp , 011. [4] S. Marx, B. K. Johnson, A. Guzmán, V. Skendzic, and M. V Mynam, Traeling Wae Faul ocaion in Proecie Relays : Design, Tesing, and Resuls, in 16h Annual Georgia Tech Faul and Disurbance Analysis Conference, 013, pp [5] G. Krzyszof and D. D. Rasolomampionona, Traelling wae faul locaion algorihm in HV lines - Simulaion es resuls for arc and high impedance fauls, in IEEE EuroCon 013, 013, no. July, pp [6] G. Zhang, H. Shu, and Y. iao, Auomaed double-ended raeling wae record correlaion for ransmission line disurbance analysis, Elecr. Power Sys. Res., ol. 136, pp. 4 50, 016. [7] F. V opes, S. Member, and W.. A. Nees, Faul ocaion on Transmission ines Based on Traelling Waes, in Inernaional Conference on Power sysems Transiens, 011. [8] M. A. Redfern, S. C. Terry, F. V. P. Robinson, and Z. Q. Bo, A aboraory Inesigaion ino he use of MV Curren Transformers for Transien Based Proecion, in Inernaional Conference on Power sysems Transiens, 003. [9] D. J. Spoor, J. Zhu, and P. Nichols, Filering effecs of subsaion secondary circuis on power sysem raeling wae ransiens, in 005 Inernaional Conference on Elecrical Machines and Sysems, 005, p Vol. 3. [10] W. Dommel, Oerhead ransmission lines," in EMTP heory book. Vancouer, Briish Columbia. 1981, pp , [11] Working group D6, AC Transmission ine Model Parameer Validaion, for he IEEE Power & Energy Sociey, Sep [1] G. Krzyszof, R. Kowalik, and D. Rasolomampionona, Traelling wae faul locaion in h lines. [13] Y. iu, G. Sheng, Y. Hu, X. Jiang, Y. Sun, and S. Wang, Idenificaion of back flash and shielding failure on ransmission line based on ime domain characerisics of raeling wae, in IEEE Power and Energy Sociey General Meeing, 014. [14] E. O. Schweizer, B. Kaszenny, A. Guzmán, V. Skendzic, M. V Mynam, and S. E. aboraories, Speed of ine Proecion Can We Break Free of Phasor imiaions?, in 68h Annual Conference for Proecie Relay Engineers, 015, pp