Accurate Single-End Fault Location and Line-Length Estimation Using Traveling Waves

Transcription

1 Accurate Sngle-End Fault Locaton and Lne-Length Estmaton Usng Travelng Waves E. O. Schwetzer, III, A. Guzmán, M. Mynam, V. Skendzc, and B. Kasztenny Schwetzer Engneerng Laboratores, Inc. C. Gallacher Avsta Utltes S. Marx Bonnevlle Power Admnstraton Presented at the 13th Internatonal Conference on Developments n Power System Protecton Ednburgh, Unted Kngdom March 7 1, 216

2 Accurate sngle-end fault locaton and lne-length estmaton usng travelng waves E. O. Schwetzer, III*, A. Guzmán*, M. Mynam*, V. Skendzc*, B. Kasztenny*, C. Gallacher, S. Marx** *Schwetzer Engneerng Laboratores, Inc., 235 NE Hopkns Court, Pullman, WA USA, Avsta Utltes, 1411 E Msson Ave., Spokane, WA 9922 USA, **Bonnevlle Power Admnstraton, 135 Lndsay Boulevard, Idaho Falls, ID 8341 USA Keywords: fault locatng, sngle-end method, travelng waves, reflectons, tme reference. Abstract Sngle-end fault locatng usng travelng waves can potentally be more accurate than double-end fault-locatng methods. Travelng-wave reflectons from taps or neghborng termnals can present a challenge to ths approach, but characterzng these reflectons ahead of tme, usng events such as lne energzaton, smplfes the analyss. In addton to fault locatng, travelng-wave recordng can also be used for estmatng lne length. The analyses of feld events, presented n ths paper, show how we dentfed reflectons for accurate fault locaton and lne-length estmaton. Lne patrol staff have valdated the results obtaned usng our methodology. 1 Introducton Accurate fault locatng s crtcal for the relable operaton of power systems. Identfyng permanent faults s mportant for power system restoraton; however, locatng temporary faults wth successful autoreclosng may not have a smlar level of urgency. Travelng-wave-based fault locaton (TWFL) methods offer accuracy n the order of one to two tower spans, ndependent of the lne length. Ths accuracy makes fndng the damaged nsulators comparatvely less challengng and less tme-consumng than usng alternatve methods such as mpedance-based methods. When lne patrols fnd and replace damaged nsulators, reoccurrences of transent faults at the same locatons are elmnated, whch results n mproved power system relablty. The sngle-end TWFL method uses the tme dfferences between the frst arrved wave and the successve reflectons from the fault and/or remote termnal to compute the fault locaton. Ths method s appealng because t depends only on local nformaton. It does not requre a communcatons channel and a precse, common tme reference for the local and remote TW data. Although t may be more challengng to mplement, the sngle-end TWFL method has less error than the double-end TWFL method because t estmates the frst and reflected wave arrval tmes usng only one hardware devce and ts nternal clock. Ths provdes a chance for some errors to self cancel. Furthermore, the sngle-end method s the only opton avalable when energzng a faulted lne. The challenge comes wth dentfyng reflectons, especally f wdebandwdth voltage measurements are not avalable to solate the ncdent TW from the measured voltage and current TWs. Ths paper dscusses a novel sngle-end TWFL method that uses current TWs to dentfy reflectons from the fault, the remote end, and any dscontnuty on the transmsson lne. The proposed algorthm excludes the reflectons from neghborng statons, taps, and other dscontnutes by usng a reference tran of TWs captured durng one or more of these events: lne energzaton, external faults, automatc reclosng (wth and wthout faults), and external swtchng of reactor or capactor banks. The algorthm uses ths nformaton along wth the TW data captured durng the fault to estmate the fault locaton. The TW event data also nclude nformaton of the wave propagaton tme along the lne. We use ths nformaton for determnng the lne length and dstances between taps. The proposed approach s applcable to overhead and tapped transmsson lnes. It s also sutable for underground and underground-overhead power lne applcatons. We present data from a km ( m), 115 kv tapped transmsson lne and a km (72.77 m), 161 kv two-end transmsson lne to demonstrate the valdty of ths method. These data, along wth data captured durng lne energzaton and other reclosng events, provde enough nformaton for accurate fault locaton and lne-length estmaton. We valdate the TWFL estmaton results wth actual lne patrol fndngs. 2 eflected and transmtted TWs When passng through a juncton pont (e.g., a tap on the lne), ncdent current TWs are reflected and transmtted accordng to the characterstc mpedances of the connected lne segments, see Fg. 1. For the ncdent TW, 1I, the reflecton Γ and transmsson T coeffcents are calculated by usng (1) and (2) [1], [2]. 1 I Z 1 Z3 Z 2 Fg. 1. Current TW passng through a juncton pont. 1 I Z 1 1 Z 3 Z 2 2 T 3 T 1

3 Z Z 1 P Γ (1) Z1 + ZP 2 Z1 T Z + Z 1 where Z 1 s the characterstc mpedance of Segment 1, and Z P s the equvalent mpedance of the parallel of the characterstc mpedances of Segments 2 and 3, Z 2 and Z 3, as shown n (3). Z P P Z2 Z3 Z + Z 2 3 The reflected current, 1, and transmtted currents, 2 T and 3T, are calculated accordng to (4), (5), and (6), respectvely. I 1 1 (2) (3) Γ (4) Z T (5) T P I 2 1 Z2 L m l m L t L2 t L3 t 1 t 2 Fg. 2. Lattce dagram showng ncdent, reflected, and transmtted TWs and arrval tmes at the local termnal to estmate lne length and fault locaton. l s the lne length and m s the dstance to the fault from L. Z T (6) T P I 3 1 Z3 B L m l m... In Secton 3, we dscuss the challenges that the sngle-end TW fault-locatng method faces because of transmsson and reflecton of TWs. B L 3 Sngle-end fault locatng and lne-length estmaton t L2 t Fault locaton estmaton The sngle-end TW fault-locatng method uses local TW arrval nformaton from the frst and the reflected waves [3], [4], and [5]. Fg. 2 shows that the TW reachng L, at, s both transmtted and reflected. The reflected wave reflects off the fault and travels back to L, arrvng at tme t L2. The tme t L2 s the travel tme from L to the fault and back. Notce that the reflected wave from the remote termnal arrves at t L3. To estmate fault locaton, ths sngle-end fault-locatng method uses the tme dfference between the frst arrved TW and the successve reflecton from the fault, as shown n (7). m t t 2 L2 L1 where:, t L2 are the TW arrval tmes at L, v s the TW propagaton velocty. We can, therefore, locate the fault usng nformaton from one end, as long as we are not confused by other reflected waves. For example, suppose that there s a short lne rght behnd L, so short that ts far end, B, s closer to L than the dstance from L to the fault (see Fg. 3). In ths case, the TW from B could reach L before t L2. These TWs need to be sorted out dependng on ther drecton n order to accurately locate the fault. v (7) t L3 t 2 Fg. 3. eflecton from the external network element (B) reaches L before reflecton from the fault. These extra TW reflectons are network dependent. Schwetzer successfully developed a sngle-end fault locator for hghvoltage dc (HVDC) lnes [3], [4]. The devce uses voltage and current to separate ncdent and reflected TWs. Usng only currents for TW fault-locatng lmts our ablty to separate the ncdent and reflected TWs. Under these lmtatons, we use an alternatve approach n whch we dscard reflectons from known juncton ponts and dentfy reflectons from the fault for smplfyng the sngle-end fault-locatng analyss. For dentfyng reflectons from known termnatons (network topology), we use TW data captured durng the followng power system events: Lne energzaton External faults Automatc reclosng External reactor/capactor swtchng Prevous lne faults The feld cases n Sectons 5 and 6 show oscllograms wth current TWs recorded from lne energzaton and faults. 2

4 3.2 Lne-length estmaton In addton to fault-locatng nformaton, sngle-end TW event reports nclude nformaton requred for estmatng the lne length. Ths estmated length provdes feedback to the lne patrol staff for verfyng the actual lne length used for fndng faults along the lne. Fg. 2 llustrates TW arrval tmes at the local termnal L from the fault ( and t L2) and from the remote termnal (t L3). The tme t L3 s two tmes the travel tme along the lne mnus two tmes the tme from the fault to L. The tme t L3 corresponds to the dstance 2l 2m. Wth these TW arrval tmes, we use (8) to estmate the lne length. l v ( tl2 tl1) + ( tl3 tl1) 2 (8) If we dvde (7) by (8), we can express the per unt length fault locaton n terms of the t L2 and t L3 delays, as shown n (9). m ( tl2 tl1) l ( t t ) + ( t t ) L2 L1 L3 L1 4 Determnng TW arrval tmes usng a dfferentator smoother [6] Ths approach orgnated n leadng-edge trackng technques used n radar [7]. It overcomes most effects of sgnal dstorton and allows for nterpolaton between samples. Ths approach was frst used n fault locatng n the dc fault locator descrbed n [3] and [4]. Subsequent mplementaton of the dfferentator smoother n the ac fault locator descrbed n [6] further proved ths approach n the feld. Fg. 4a shows a block dagram sutable for demonstratng the method. The current s frst low-pass fltered or smoothed; then ts output s dfferentated. Smoothng reduces the effects of waveform dstortons and causes the current rsng edge to smooth out and become less steep. Softenng the rsng edge at frst may seem contrary to the objectve of determnng the tme of arrval; however, t spreads the edge over several samples, makng the tme-nterpolaton process possble. The smoothed waveform s then dfferentated, turnng the step-lke current waveform nto a soft pulse-lke shape. That pulse-lke dervatve has ts peak at the nstant of the steepest slope of the current waveform. The peak of the dervatve s relatvely nsenstve to ampltude changes, beng about halfway along the edge, no matter how tall the current step s. Fg. 4c shows the dervatve output n detal and adds the ponts n tme where the samples were taken. It also shows a par of lnes; ther ntersecton s an excellent measure of t A, the TW arrval tme. When usng flters to smooth the dervatve of the current, the output resembles a parabola, as shown n Fg. 4b and Fg. 4c. Therefore, n our mplementaton, we use a parabola-based nterpolaton method for calculatng the arrval tme. The algorthm selects a few samples pror to the peak sample and a few samples followng the peak. It further uses the leastsquares estmaton (LSE) method to ft a parabola to the (9) selected ponts, ncludng the maxmum sample, and calculates the arrval tme (t A) usng the best-fttng parabola (see Fg. 5). (a) (b) (c) Current (t) d/dt Smoother (LPF) S(t) (t) Dfferentator d/dt d/dt t A S(t) d/dt -of- Peak Estmator Fg. 4. Dfferentator smoother: (a) block dagram, (b) typcal waveforms, and (c) tme-of-peak estmaton. Output From the Dfferentator Smoother (A) of Arrval Best Ft Parabola t t Arrval t A (μs) Fg. 5. Accurate tme-stampng of the TW usng the best-ft parabola. 5 Sngle-end TW fault locatng usng feld events from a lne wth taps 5.1 Brasada-Harney Lne Fg. 6 shows the 115 kv, km ( m) Brasada- Harney transmsson lne along wth ts neghborng system. t A 3

5 The tap locatons are termnated wth transformers servng dstrbuton feeders. The lne s a secton of a longer transmsson path (about 422 km or 262 m), wth several tapped loads, n Central and Eastern Oregon. Half of the path s 115 kv and the rest of the path s 138 kv. The terran conssts of rural hgh-alttude desert and rrgated farm land. Hstorcally, the lne experences ten or more faults per year. In many cases, faults on the lne have been attrbuted to nsulator flashover caused by lghtnng, contamnaton from wldlfe, or dust accumulaton mxed wth ran or dew. The relays successfully reclosed for all these faults. edmond Yew Ave. Tap Brasada Brothers Tap Hampton Tap ley Tap Harney Fg kv transmsson network ncludes the Brasada- Harney lne wth adjacent lnes and taps Usng TWs captured durng lne energzaton for dentfyng TW reflecton from a fault To address the challenge of dentfyng and excludng the reflectons from the neghborng statons, tap locatons, and other dscontnutes, we used a reference tran of waves captured durng lne energzaton wth the remote termnal open. We also could have captured the reference tran of waves durng external faults or automatc reclosng (wthout a fault). The reference waves provde nformaton on the lne topology and adjacent taps. By usng these reference waves along wth the waves captured durng an nternal fault, and algnng the frst wave captured from both events, we were able to dscard the common reflectons from both events. Followng ths approach, we dentfed the reflecton from the fault. The reflectons could be drectly from the fault or from the wave reflected from the remote termnal. Fg. 7 shows the phase current TWs captured durng lne energzaton that we used as the reference sgnal. Note that the closure of the B-phase and C-phase poles provded a tran of waves showng the locaton of taps and other termnatons. In ths lne energzaton event, the A-phase pole closed at the zero voltage pont on the wave. Fg. 8 shows the tran of current TWs recorded at Brasada durng a lne energzaton algned wth the A-phase current TWs recorded durng an A-phase-toground fault. Secondary Amperes C-Pole Closure B-Pole Closure Yew Ave. Tap Brothers Tap Hampton Tap Harney ley Tap Fg. 7. Lne energzaton event dentfyng TW reflectons from taps resultng from the B-phase pole closure. Secondary Amperes eflecton From Fault, t L Fg. 8. A-phase current TWs for an A-phase-to-ground fault algned wth the current TWs recorded durng lne energzaton. Once we dentfed the reflecton from the fault usng the above concept, we used the wave arrval tmes assocated wth the frst arrved wave ( s) and the reflected wave (t L s) from the fault to compute the fault locaton usng (7). Wth v tmes the speed of lght, we estmated the fault locaton at 96.4 km (59.68 m). Addtonally, snce we knew the locaton of the lne termnatons and taps, we provded the lne patrol staff wth the fault locaton from the Hampton tap Usng mpedance-based fault locatng for dentfyng TW reflecton from a fault We used the event reports from a C-phase-to-ground fault close to the Brasada termnal for analyzng operatng frequency sgnals and TWs. For ths event, the fault contrbuton from Harney was so low that the lne relay at Harney operated after the Brasada breaker opened (sequental trppng occurred). Fg. 9 shows the recorded phase currents and voltages and Fg. 1 shows the TW currents recorded at the Brasada termnal. 4

6 IAW IBW ICW VAY VBY VCY Fg. 9. Currents (prmary A) and voltages (prmary kv) for a C-phase-to-ground fault for performng the frst fault locaton estmaton. Prmary Amperes Fg. 1. Current TWs for a C-phase-to-ground fault used for fault locatng. For ths event, we used a sngle-end mpedance-based method [8] to dentfy the reflecton from the fault. As mentoned earler, ths event resulted n sequental trppng, whch translates to havng the Harney termnal feedng the fault wth the Brasada termnal open (radal system). Ths system condton s well suted for the sngle-end mpedance-based method. We used the phase voltages and currents captured at the Harney termnal n the radal system condton along wth the postve- and zero-sequence lne mpedances, and computed the fault locaton at km (15.9 m) from Harney and km (6.95 m) from Brasada. Wth ths mpedance-based sngle-end fault-locaton estmate, we predcted the reflecton assocated wth the fault to be at about t s. We found the reflecton at t L s. Fg. 11 shows the frst wave and the reflectons from the fault and from the Yew Avenue tap (located rght behnd the Brasada termnal). After dentfyng the reflecton from the fault, we calculated the fault locaton by usng the arrval tme of the frst TW and the tme assocated wth the frst reflecton from the fault: s and t L s. The estmated fault locaton was km (6.92 m) from Brasada. Ths fault locaton translates to tower 8/6 (eghth tower after the sxth mle), where lne patrol staff found the flashed over nsulator shown n Fg. 12. Secondary Amperes eflecton From Fault, t L2 eflecton From Yew Ave. Tap Fg. 11. C-phase alpha current durng the C-phase-toground fault recorded at the Brasada termnal along wth the dentfed reflectons from the fault. Fg. 12. Flashed-over nsulator on the C-phase, tower 8/6 (eghth tower after the sxth mle) of the Brasada-Harney lne. 6 Lne-length estmaton usng a feld event We analyzed a feld event of a B-phase-to-ground fault on the 161 kv, km (72.77 m) Goshen-Drummond lne (see Fg. 13) where we knew the fault locaton for estmatng the lne length; the fault occurred because of gunshots to nsulators. The lne patrol reported the fault locaton at km (38.16 m) from Goshen and 55.7 km (34.61 m) from Drummond. Goshen 161 kv elay G Goshen 115 kv Swan Valley 161 kv elay D Drummond 115 kv Fg. 13. Transmsson network ncludes the Goshen- Drummond lne and CT and relay connectons. Notce that the lne termnaton at Drummond s an autotransformer [9]. 5

7 Fg. 14 shows the lattce dagram for ths fault and correspondng alpha-mode B-phase TW currents as seen by the Goshen and Drummond relays. We used the algorthm to estmate the wave arrval tmes descrbed n Secton 4. The dagram shows the fltered currents accordng to ths algorthm. In these plots, you can clearly see the local and remote reflectons at Goshen: the frst reflected peak comes from Drummond and the second reflected peak comes from the fault. Usng the followng data from Goshen, we estmated a lne length of km (72.97 m) by usng (8) and a fault locaton of km (38.4 m) from Goshen by usng (7) s t L s t L s v tmes the speed of lght The estmated lne-length error s.28 percent or 328 m; ths error could be attrbuted to conductor saggng. mcroseconds Goshen αb (A) km Drummond αb (A) 2 2 Fg. 14. TWs for a B-phase-to-ground fault on the Goshen- Drummond lne used for estmatng lne length. 7 Concluson The analyses of the feld events presented n ths paper demonstrate that the fault locatng algorthm that uses the dfferentator-smoother and wave arrval tme estmator, together wth the presented method for dentfyng reflectons from the fault, smplfy fault-locatng analyss and provde accurate sngle-end TW fault locaton estmatons (errors of less than 2 m). Lne patrol fndngs valdated these results. Feld event analyss also demonstrates that TW event reports provde nformaton for estmatng lne length wthn percent. Conductor saggng s the man contrbutor of the dfference between the geographcal and actual lne lengths. eferences [1] L. V. Bewley, Travelng Waves on Transmsson Systems. Dover Publcatons, Mneola, NY, [2] A. Greenwood, Electrcal Transents n Power Systems, 2nd Edton, John Wley & Sons, Inc., ISBN , [3] M. Ando, E. O. Schwetzer, III, and. A. Baker, Development and Feld-Data Evaluaton of Sngle-End Fault Locator for Two-Termnal HVDC Transmsson Lnes, Part I: Data Collecton System and Feld Data, IEEE Transactons on Power Apparatus and Systems, Vol. PAS 14, Issue 12, December 1985, pp [4] M. Ando, E. O. Schwetzer, III, and. A. Baker, Development and Feld-Data Evaluaton of Sngle-End Fault Locator for Two-Termnal HVDC Transmsson Lnes, Part II: Algorthm and Evaluaton, IEEE Transactons on Power Apparatus and Systems, Vol. PAS 14, Issue 12, December 1985, pp [5] M. Aurangzeb, P. A. Crossley, and P. Gale, Fault Locaton on a Transmsson Lne Usng Hgh Frequency Travellng Waves Measured at a Sngle Lne End, proceedngs of the 2 IEEE Power Engneerng Socety Wnter Meetng, Vol. 4, Sngapore, January 2, pp [6] E. O. Schwetzer, III, A. Guzmán, M. V. Mynam, V. Skendzc, B. Kasztenny, and S. Marx, Locatng Faults by the Travelng Waves They Launch, proceedngs of the 4th Annual Western Protectve elay Conference, Spokane, WA, October 213. [7] F. E. Nathanson, adar Desgn Prncples: Sgnal Processng and the Envronment, McGraw-Hll Book Co., [8] E. O. Schwetzer, III, A evew of Impedance-Based Fault Locatng Experence, proceedngs of the 15th Annual Western Protectve elay Conference, Spokane, WA, October [9] S. Marx, B. K. Johnson, A. Guzmán, V. Skendzc, and M. V. Mynam, Travelng Wave Fault Locaton n Protectve elays: Desgn, Testng, and esults, proceedngs of the 16th Annual Georga Tech Fault and Dsturbance Analyss Conference, Atlanta, GA, May by Avsta Utltes, Bonnevlle Power Admnstraton, and Schwetzer Engneerng Laboratores, Inc. All rghts reserved TP6732 6