Location of Single Line-to-Ground Faults on Distribution Feeders Using Voltage Measurements

Transcription

1 1 Locaton of Sngle e-to-ground Faults on Dstrbuton Feeders Usng Voltage Measurements R. A. F. Perera, Student Member, EEE, L. G. W. da Slva, M. Kezunovc, Fellow, EEE, and J. R. S. Mantovan, Member, EEE Abstract-- Ths paper proposes a dedcated algorthm for locaton of sngle lne-to-ground faults n dstrbuton systems. The proposed algorthm uses voltage and current phasors measured at the substaton level, voltage magntudes measured at some buses of the feeder, a database contanng electrcal, operatonal and topologcal parameters of the dstrbuton networs, and fault smulaton. Voltage measurements can be obtaned usng power qualty devces already nstalled on the feeders or usng voltage measurement devces dedcated for fault locaton. Usng the proposed algorthm, lely faulted ponts that are located on feeder laterals geographcally far from the actual faulted pont are excluded from the results. Assessment of the algorthm effcency was carred out usng a 38 buses real-lfe dstrbuton feeder. The results show that the proposed algorthm s robust for performng fast and effcent fault locaton for sustaned sngle lne-to-ground faults requrng less than 5% of the feeder buses to be covered by voltage measurement devces. ndex Terms-- Fault Locaton, Power Dstrbuton, Voltage Measurement.. NTRODUCTON N the recent past, electrc power regulatory agences have ncreased the requrements for electrc utltes to mantan electrc power supply ndces wthn qualty standards. One way of attendng to these requrements s carryng out fault locaton n an effcent way decreasng the customer outage tme. As a consequence, technques, algorthms and methodologes for fault locaton have receved sgnfcant attenton of researchers. The man dfferences found among varous proposals for fault locaton on dstrbuton feeders are related to the number of used measurement varables and requred nstrumentaton, as well as the approach for fndng fault locaton. Algorthms, methodologes and technques for fault Ths wor was supported n part by the CAPES (grant BEX769/5-3), FEPSA (grant 7/5) and CNPq. R. A. F. Perera s wth Electrcal Engneerng Department FES/UNESP, lha Soltera, Sao Paulo, Brazl (e-mal: ddgo@yahoo.com). L. G. W. da Slva s wth Electrcal Engneerng Department FES/UNESP, lha Soltera, Sao Paulo, Brazl (e-mal: lgwesz@gmal.com). M. Kezunovc s wth Electrcal and Computer Engneerng Department Texas A&M Unversty, College Staton, Texas, USA (e-mal: ezunov@ece.tamu.edu). J. R. S. Mantovan s wth Electrcal Engneerng Department FES/UNESP, lha Soltera, Sao Paulo, Brazl (e-mal: mant@dee.fes.unesp.br). locaton utlzng measurements of voltage and current phasors of fundamental frequences at the sendng node of the feeder or at the substaton, as well as the derved equatons usng symmetrcal components, sutable load modelng, and even fault ndcators nstalled on the feeders were proposed n [1]- [5]. Reference [6] proposes an assocaton of fault ndcators nstalled along the feeder, fault dstance computaton, statstcal ndcaton of frequency of fault occurrences on a lne secton, and probablstc analyss to carry out fault locaton. Reference [7] presents an approach based on supermposed components of voltage and current phasors measured at the sendng node of the feeder, derved phase-doman equatons and, as well as the sutable load model. Fault locaton algorthms that use voltage and current phasors measured at the sendng node of the feeder at the substaton were presented n [8]-[1]. Loads are modeled consderng ther voltage dependence and phase-doman equatons were derved, whch enables the algorthms to be applcable for feeders composed of double and sngle-phase laterals. Addtonal analyss for selectng the most lely fault locaton among the fault locatons canddates was also proposed. Reference [11] presents a method for fault locaton that uses fault ndcators nstalled along the feeder and fuzzy logc to model the nformaton from these devces. n [1] a technque usng voltage and current measurements at the sendng and recevng nodes of each lne secton of the feeder was presented. n ths way, t s possble to locate the faulted secton very accurately. Reference [13] presents a fault locaton methodology that combnes a power systems analyss program along wth database search method and pattern recognton technque. Ths methodology locates the faulted pont by means of matchng the measured voltage at the sendng node of a substaton wth voltages for the same pont calculated usng fault smulatons for each feeder node. Results are presented for locaton of faults nvolvng zero resstance. Ths paper presents a technque for fault locaton on overhead dstrbuton feeders usng measurements of pre- and durng-fault voltage and current phasors at the sendng node of a substaton along wth durng-fault voltage measurements at the nodes along the feeder. The proposed technque was tested on an overhead, 13.8 V, 38 nodes real-lfe feeder and the results show that ths technque s robust and effcent for carryng out fault locaton n a fast and accurate way.

2 . FAULT LOCATON TECHNQUE USNG VOLTAGE MEASUREMENTS Fault locaton on transmsson lnes based on comparson of measurements and smulatons of both voltages and currents was proposed n [14]. Smlar approach for fault locaton on dstrbuton feeders was presented n [15]. Ths approach uses pre- and durng-fault voltage and current phasors measured at the sendng node of the feeder; durng-fault voltage magntudes measured at some feeder nodes, and derved equatons usng symmetrcal components. n order for the proposed technque to be applcable for the three-phase untransposed feeders contanng double and sngle-phase lateral branches, all equatons that were derved usng symmetrcal components were replaced by phase-doman equatons. Fg. 1 depcts the data flow n the technque proposed n ths paper. Pre-Fault Voltage and Current Phasors Measured at the Sendng Node of the Feeder Substaton Pre- Fault Power Ratng Estmaton Feeder Database (Topology, Transformer nal Powers, e mpedances, Transformer Average Loadngs, etc.) Durng-Fault Voltage and Current Phasors Measured at the Sendng Node of the Feeder Dstrbuton Transformer Power Ratng Estmaton Three-Phase Load Flow Fg. 1 Data flow n the proposed technque. Voltage Magntudes Measured at some Nodes on the Feeder Fault Resstance Computaton Fault Locaton Ranng Durng-Fault Voltage Msmatch Computaton Fault Smulaton and Durng-Fault Voltage Magntudes Computaton for all Measurement Nodes Followng are dervatons of the algorthm proposed for locaton of sngle lne-to-ground faults. A. Measurements at the Substaton Pre- and durng-fault voltage and current phasors measurements at the substaton are obtaned from ntellgent Electronc Devces (EDs) placed on each feeder. EDs can measure and store the necessary quanttes for carryng out fault locaton. n order to extract voltage and current phasors of fundamental frequency from the sgnals sampled by EDs, Dscrete Fourer Transform for one-cycle data wndow s used. B. Measurements along the feeder Durng-fault magntudes are measured at some feeder nodes usng voltage measurement devces dedcated only for fault locaton purposes or usng power qualty devces already nstalled on the feeder. Fault locaton accuracy depends on where the voltage measurement devces are placed on the feeder. An optmzaton strategy for placement of voltage measurement devces usng genetc algorthm s presented n [16]. Communcaton channels should be avalable for sendng the measured data from remote feeder nodes to the computer responsble for processng the fault locaton algorthm. Voltage magntudes are the rms values calculated by means of equaton (1) for one-cycle data wndow. 1 1 n RMS = v n = V (1) v : voltage nstantaneous value; n: number of samples n one-cycle data wndow. C. Substaton Pre-Fault Power Ratng Estmaton Pre-fault voltage and current phasors measured at the substaton level are used for estmatng pre-fault power ratng at the substaton. Ths s an estmaton of the pre-fault power ratng at the substaton because the measured power at ths pont contans the total losses along the feeder. The larger the losses of the feeder, the larger s the effect on the accuracy of the fault locaton process. D. Dstrbuton Transformer Power Ratng Estmaton One of the sources of the fault locaton algorthm errors s the dffculty n estmatng an exact loadng of each transformer durng fault events. n ths paper, loadng for each dstrbuton transformer of the feeder s based on ts nomnal power ratng stored n a dstrbuton system database, substaton pre-fault power ratng, and average loadng (β) estmated tang nto account electrc blls of each customer suppled by the transformer. Thus, the complex power ratng for each transformer can be calculated usng the followng equaton: SS S Meas S = β S [ cos( ϕss ) + sn( ϕss )] nl () β S = 1 S : Transformer nomnal apparent power ratng; S SS Meas : Apparent power ratng measured at the substaton; cos( ϕ SS ) : Power factor for the measured power ratng at the substaton; nl: Total number of transformers supplyng the feeder; β : Average loadng of the transformer. E. Load Flow n ths wor, load flow s calculated usng ATP program, and the load model s one wth a constant mpedance. F. Fault Resstance Computaton Dfferently from the methodologes proposed n [8]-[1] for each nvestgated node, a fault resstance value should be calculated.

3 3 V V V V V V [ L ] [ L ] [ L ] V V V V l l V l V F R F Load l Load Fg. Three-phase model of the dstrbuton networ. Fg. depcts a generc three-phase dstrbuton feeder. Consderng node s the substaton sendng node, fault resstance value can be computed accordng to the followng steps: ) Startng from the sendng node at the substaton, durngfault voltage for node s calculated usng the followng equaton; V V = V V aa ab ac ab bb bc ac bc cc V : Durng-fault voltage on phase α at node ; (3) αα : Self-mpedance of phase α of the lne secton; : Mutual-mpedance between phase α and ε; αε : Durng-fault current on the lne secton between nodes and. ) Usng computed voltages for node and consderng L constant mpedance load model, load current abc s updated for fault condton by means of the followng equaton; L a L V a L V L = b (4) L Vb V L c Vc V : Pre-fault voltage on phase α at the node ; α Lγ : Pre-fault current on phase α of the load L γ ; α Lγ : Durng-fault current on phase α of the load L γ. ) Tang nto account fault occurrence at the node, durng-fault current through branch s computed usng the followng equaton; α V V = V V V a b c a b c : Pre-fault current on phase α of the lne secton ; (5) : Durng-fault current on phase α of the lne secton. v) After load currents connected at the node and downstream branch currents of the node have been updated sngle lne-to-ground fault current to the node s computed usng the followng equaton; m = (6) m Ψ α: Faulted phase; m: Downstream branch of the node ; n: Load connected at the node ; Φ: Set of all loads connected at node ; Ψ: Set of all branches connected downstream of node. v) Fault resstance for node s computed by means of the followng equaton; V R f = Re (7) v) Now, consderng that the fault dd not occur at the node, and tang nto account that the fault occurred at the node, durng-fault current through branch should be updated accordng to the followng equaton; = V V V V V m a a m Ω m b b m Ω m c c m Ω Ω: Set of branches connected downstream at the node (8)

4 4, excludng the branch. v) Return to the step ) and the same procedure s executed untl one fault resstance value s computed for each feeder node. h) Fault Smulaton and Voltage Computaton for measurement nodes: Analogously to the load flow, fault smulaton s carred out usng ATP program. Durng-fault voltages for the measurement nodes are computed and stored durng fault smulaton process. ) Computaton of durng-fault voltage msmatches between measured and calculated voltages for each measurement node: Usng measured and calculated voltages for each measurement node the durng-fault voltage msmatches can be computed usng the followng equaton: δ = V V (9) meas, calc V : Voltage magntude measured at node ; meas V, calc : Voltage magntude calculated for node consderng fault at node. ) Selecton of the lely fault locaton: Lely fault locaton s selected based on the analyss of the values of δ for all measurement nodes. The values of δ for the faulted node should have the same magntude. Then, selecton of the lely fault locaton node s done usng the followng equaton: fb = mn {max{ δ } mn{ δ }} = 1,..., nm = 1,..., nb nm: Total number of voltage measurement devces; nb: Total number of feeder nodes.. TESTS RESULTS (1) The proposed technque for locaton of sngle lne-toground faults was tested on an overhead, 13.8 V, 38 nodes, real-lfe dstrbuton feeder shown n fg. 3. Fault smulatons were carred out usng ATP program. Loads were modeled as constant mpedances and average loadng of each dstrbuton transformer (β), varyng between.9 and 1.1, was estmated based on electrc blls of each customer connected to the transformer. Pre-fault load flow and faults to be located were carred out by multplyng average loadng of each transformer by a random number (λ), varyng between.95 and 1.5, accordng to the followng equaton: S = β λ S (11) S : Complex power assumed for the transformer ; S : nal complex power for the transformer ; β : Average loadng of the transformer ; λ : Random number. Fault locaton calculatons were carred out tang nto account that the total power ratng measured at the sendng node at the substaton was allocated for each feeder transformer accordng to (), and fault resstance equal 1 ohms Fg. 3 Real-lfe feeder used for testng the proposed fault locaton technque. Tang nto account that fve voltage measurement devces were placed at nodes 11, 19, 166, 176 and 5, fault smulatons were performed for each feeder node and the proposed technque for fault locaton was appled to locate the faults. Two hundred thrty-sx faults were smulated on the feeder (from node 3 to node 38) and the results are shown n table. TABLE FAULT LOCATON RESULTS Obtaned results for 36 fault smulatons Total Accurate fault locaton 8 Actual faulted node raned as second faulted node 19 Actual faulted node raned as thrd faulted node 6 Actual faulted node raned as fourth faulted node 3 Table shows that among the 36 smulated faults on the feeder, 8 of them were accurately pnponted whle 19 of them were placed n the second, sx of them n the thrd and three of them n the fourth poston of the ran. Tang nto account the load devaton between fault smulaton and fault locaton process along wth the feeder sze and the number of laterals present on the feeder, these twenty-eght wrong ndcatons do not fect the relablty and robustness of the

5 5 proposed technque because all the nodes raned before the actual faulted node are wthn the same geographc area,.e., nodes on laterals geographcally far from the faulted node were not ponted as the lely fault locaton. Fg. 3 shows the values of fb, (1), for the lely faulted nodes raned for a fault located accurately. Fg. 4, 5 and 6 show the values of fb for the lely faulted nodes raned for three dfferent faults, whch were not located accurately fb (V) 6 frst canddate node for the fault locaton, the node 117 s raned as the second one and the node 13 s raned as the thrd one. Although node 13 s ndcated n the thrd poston of the faulted node ranng, the dstance between nodes 131 and 13 s 96.6 meters, and the dstance between nodes 117 and 13 s 59.4 meters. 3 5 fb (V) Faulted nodes ranng Faulted nodes ranng Fg. 6 Faulted nodes ranng for a fault occurrng at the node. Fg. 3 Faulted nodes ranng for a fault occurrng at the node fb (V) Faulted nodes ranng Fg. 4 Faulted nodes ranng for a fault occurrng at the node 9. Fg. 4 shows the faulted node ranng for fault occurrence at node 9. t can be noted that the node 1 s raned as the frst canddate node for fault locaton. Although node 1 s ndcated as the lely faulted node the dstance between nodes 9 and 1 s only 4 meters fb (V) Faulted nodes ranng Fg. 5 Faulted nodes ranng for a fault occurrng at the node 13. Fg. 5 shows the faulted node ranng for fault occurrence at node 13. t can be noted that the node 131 s raned as the Fg. 6 shows the faulted node ranng for fault occurrence at node. t can be noted that the node 3 s raned as the frst canddate node for fault locaton, the node 1 s raned as the second one, the node 19 s raned as the thrd one and the node s raned as the fourth one. n spte of node beng ndcated n the fourth poston of the faulted node ranng, the dstance between nodes and 3 s 95.1 meters, the dstance between nodes 1 and s 7.8 meters, and the dstance between nodes 19 and s 49.1 meters. Analyzng the twenty-eght cases where the actual faulted nodes were not located accurately, t can be verfed that the results stll present good estmates. Although frst, second or thrd poston of the faulted nodes ranng are flled wth nonfaulted nodes, these nodes are always n close proxmty of the faulted node. The usage of voltage measurements for carryng out fault locaton always provdes as results the nodes that are n the neghborhood of the faulted node. Possble nodes located far from the faulted node are excluded from the results wthout the use of any auxlary methods of dagnoss for selectng the lely faulted pont. By usng the results from the fault locaton technque, the mantenance crews are able to locate ether the actual faulted node or they can have a relable ndcaton of the faulted area n a fast and effcent way. Wth ths approach the tme spent on performng swtchng for restoraton and the outage tme for the customers are consderably decreased. V. CONCLUSONS A technque for locatng sngle lne-to-ground faults on dstrbuton feeders usng voltage measurements along the feeder was presented n ths paper. At least two measurement devces should be nstalled for performng fault locaton and the maxmum number s defned by tang nto account techncal and economc constrants. Results show that the technque s robust and effcent n

6 6 fndng the faulted node and/or the area near to the actual faulted node. The usage of voltage measurements along the feeder enables the proposed technque to locate faults excludng the nodes located far from the actual faulted node wthout usng an auxlary dagnoss method, le matchng waveform scenaros and analyses of the pre- and post-fault power measured at the substaton. These characterstcs of the presented technque are very mportant for electrc utltes for meetng standards mposed by the regulatory agences. As t was observed n the presented results, n some cases the actual faulted node was not ponted accurately. Ths has occurred due to the man error source that fects the accuracy of fault locaton technques on dstrbuton feeders. Ths error source s related to the transformers loadng estmaton. t can lead the fault locaton technque to ndcate the lely fault locaton as beng upstream or downstream from the actual faulted pont. n some cases, t appears that the frst node n the node ranng s further away from the actual faulted pont than the second, thrd or fourth one, but the dstance s so small that t does not have sgnfcant effect on the qualty of the obtaned results. One mportant characterstc of the proposed technque s that t can be easly mplemented usng power qualty measurement devces already nstalled on the feeder snce they can provde data requred for carryng out fault locaton. Ths way, the power qualty devces play two functons, namely, power qualty analyss and fault locaton. Ths multple use of exstng devces s an mportant aspect that electrc utltes are consderng whle mang future nvestments n the technques and devces for mprovng power supply qualty and networ performance. V. ACKNOWLEDGMENT The frst author gratefully acnowledges the Vstng Researcher poston, as well as the facltes support provded by the Electrcal and Computer Engneerng Department at Texas A&M Unversty. V. REFERENCES [1] R. Das, M. S. Sachdev and T. S. Sdhu, A fault locator for radal subtransmsson and dstrbuton lnes n Proc. EEE Power Engneerng Socety Summer Meetng, Seatle, Washngton, pp [] M. M. Saha, F. Provoost and E. Rosolows, Fault locaton method for MV cable networ n Proc. 1 EE Seventh nternatonal Conference on Developments n Power System Protecton, Amsterdam, The Netherlands, pp [3] M. M. Saha, E. Rosolows and J. zyows, A new fault locaton algorthm for dstrbuton networs n Proc 4 EE Eghth nternatonal Conference on Developments n Power System Protecton, Amsterdam, The Netherlands, pp [4] A. A. Grgs, and C. M. Fallon, Fault locaton technques for radal and loop transmsson systems usng dgtal fault recorded data, EEE Trans. Power Delvery, vol. 7, pp , Oct [5] A. A. Grgs, C. M. Fallon, and D. L. Lubeman, A fault locaton Technque for rural dstrbuton feeders, EEE Trans. ndustry Applcatons, vol. 9, pp , Nov [6] M. Lehtonen, A. Matsnen, E. Antla, J. Kuru, P. Vuorenpaa, E. Matnlass and S. Pettssalo, Automatc fault management n dstrbuton networs n Proc. 1 CRED - 16 th nternatonal Conference and Exhbton on Electrcty Dstrbuton. [7] R. K. Aggarwal, Y. Aslan, and A. T. Johns, New concept n fault locaton for overhead dstrbuton systems usng supermposed components, EE Proc. Generaton, Transmsson and Dstrbuton, vol. 44, pp , May [8] J. hu, D. L. Lubeman, and A. A. Grgs, Automated fault locaton and dagnoss on electrc power dstrbuton feeders, EEE Trans. Power Delvery, vol. 1, pp , Apr [9] S. J Lee, M. S. Cho, S. H. Kang, B. G. Jn, D. S. Lee, B. S. Ahn, N. S. Yoon, H. Y. Km, and S. B. Wee, An ntellgent and effcent fault locaton and dagnoss scheme for radal dstrbuton systems, EEE Trans. Power Delvery, vol. 19, pp , Apr. 4. [1] E. C. Senger, G. Manassero, C. Goldemberg, and E. L. Pelln, Automated fault locaton system for prmary dstrbuton networs, EEE Trans. Power Delvery, vol., pp , Apr. 5. [11] P. Jarventausta, P. Verho, and P. Partanen, Usng fuzzy sets to model the uncertanty n the fault locaton process of dstrbuton networs, EEE Trans. Power Delvery, vol. 9, pp , Apr [1] V. N. Gohoar and V. V. Gohoar, Fault locaton n automated dstrbuton networ n Proc. 5 EEE nternatonal Symposum on Crcuts and Systems, pp [13] H. L, A. S. Mohar and N. Jenns, Automatc fault locaton on dstrbuton networ usng voltage sags measurements, n Proc. 5 CRED 18 th nternatonal Conference and Exhbton on Electrcty Dstrbuton. [14] M. Kezunovc, S. Luo and D. R. Sevc, A novel method for transmsson networ fault locaton usng genetc algorthms and sparse feld recordngs n Proc. EEE Power engneerng Socety Summer Meetng, Chcago, llnos, pp [15] R. A. F. Perera, L. G. W. da Slva and J. R. S. Mantovan, Localzacao de faltas fase-terra em almentadores radas de dstrbucao utlzando aquscao de dados esparsos sncronzados, n Proc. 4 XV Congresso Braslero de Automatca, Gramado, Brazl. [16] R. A. F. Perera, L. G. W. da Slva, M. Kezunovc and J. R. S. Mantovan, Optmzed placement of voltage measurement devces for determnng locaton of sngle lne-to-ground faults on overhead electrc power dstrbuton feeders, submtted to the 6 XV Congresso Braslero de Automatca, Salvador, Brazl. V. BOGRAPHES Rodrgo Aparecdo Fernandes Perera (S 6) receved hs B.Sc. and the M.S. degrees from UNESP/lha Soltera-SP (Brazl), n 1 and 3, respectvely. He s worng towards a Ph.D. degree n Electrcal Engneerng at UNESP/lha Soltera-SP (Brazl). Currently, he s wth the Electrcal and Computer Engneerng Department at Texas A&M as a Vstng Researcher. Hs general research nterest s n the electrc power systems relablty and control areas, manly expert systems for fault locaton on dstrbuton feeders. Lus Gustavo Wesz da Slva receved the B.Sc. degree n 1999 from UNDERP-MS and the M.S. and Ph.D degrees n Electrcal Engneerng from UNESP/lha Soltera-SP (Brazl), n and 5, respectvely. Hs general research nterest s n the areas of relablty and control of electrcal power systems. Mladen Kezunovc (S 77-M 8-SM 85-F 99) receved the Dpl. ng. Degree n electrcal engneerng from the Unversty of Saraevo, Bosna- Herzegovna, n 1974, and the M.S. and Ph.D. degrees from the Unversty of Kansas, Lawrence, n 1977 and 198, respectvely. Currently, he s the Eugene E. Webb Professor and Drector of Electrc Power and Power Electroncs nsttute at Texas A&M Unversty. Hs man nterests are dgtal smulators and smulaton methods for relay testng as well as applcaton of ntellgent methods to power system montorng, control, and protecton. Dr. Kezunovc s also a Fellow of EEE and a member of the CGRE-Pars. José Roberto Sanches Mantovan (M 6) receved the B.Sc. degree n 1981 from UNESP/lha Soltera-SP, and hs Ph.D degree n Electrcal Engneerng from UNCAMP/Campnas-SP, n Currently he s an Assocate Professor n the Electrcal Engneerng Department of UNESP/lha Soltera- SP. Hs general research nterests are n the area of plannng and control of electrcal power systems.