Localization of Phase-to-Phase Faults on a Medium Voltage Feeder with Distributed Generation

Transcription

1 Localization of Phase-to-Phase Faults on a Meium Voltage Feeer with Distribute Generation Jorun rene Marvik, Hans Kristian Høialen, Astri Petterteig Abstract This paper presents a localization proceure for phase-to-phase short-circuits on a meium voltage (MV) feeer with istribute generation (). The istance from the substation to the fault location is estimate from the funamental freuency voltage an current measure in the substation. Loas an -units connecte along the feeer are shown to have opposite impacts on the estimate istance. The current rawn by loas cause the estimate to become too small. s fee current to the fault, an cause the estimate to become too large. The net impact of a is larger than that of an eually size loa, since the current increases in a fault situation while the loa current ecreases. Two methos to compensate for these errors are presente. n compensation metho A, pre-fault measurements from the unit are utilize for estimating the -current uring fault. n compensation metho B, the magnitues of both -current an voltage uring fault are measure. The goal is to obtain sufficiently accurate fault localization, while taking into account that the use of measurements from other locations than the substation shoul be kept at a minimum. Keywors: Fault localization, Meium Voltage Network, Distribute Generation, Three-phase short-circuit, Two-phase short-circuit, Funamental freuency components. NTRODUCTON T has become more common to have generation units connecte at istribution level. n Norway, a lot of small hyro power plants are being built, an connecte to overhea istribution networks in rural areas. introuces some new challenges relate to voltage uality, stability an fault hanling in the istribution networks. With an increase focus on power uality, fast an efficient fault hanling becomes more important. Traitional manual sectioning of faults is very time-consuming. Toay, with remote control of breakers becoming more common, fast remote controlle sectioning is possible provie that accurate fault localization is available. n MV networks without, the challenge of fault localization is to minimize the impact of loas an fault resistance. Different methos [] an lots of literature on this are available. Fault localization methos using funamental freuency measurements in the substation, can utilize pre-fault measurements to minimize the impact from loa on the istance estimate [2]-[4]. This approach seems to work well for feeers without. Fault localization in networks with using funamental freuency measurements is treate in [5]. [6]-[8] eals with more avance methos base on fault transients, aaptive protection schemes an relay agents. This paper looks into possibilities for precise localization of three- an two-phase short-circuits, utilizing funamental freuency currents an voltages. t is shown how loa an introuce substantial errors in the istance estimate from traitional istance relays. Two methos to compensate for these errors are evelope an teste with simulate ata.. MV-FEEDER MODEL, SMULATED N PSCAD A simple 3 km long, 22 kv MV feeer with up to 3 units, shown in Fig., is moele in PSCAD. The -units are synchronous generators, operating at unity power-factor. High loa (HL) for the feeer is 6 MVA, an low loa (LL) is.5 MVA. The voltages an currents values uring fault are sample 4 ms (2 perios) after the fault inception. U S, S : substation voltage an current U, : voltage in connection point an current. f-main/f-sie: fault locate on the main branch/sie-branch. 2-ph./3-ph.: two-phase/three-phase short-circuit. 5 km 5 km 5 km This work is part of the project Distribution 22 at SNTEF Energy Research. The project is foune by the Norwegian Research Council, ABB, an several Norwegian utility companies an some other inustry partners. J.. Marvik an H. K. Høialen is with Norwegian University of Science an Technology (NTNU), Tronheim, Norway ( of corresponing author: jorun.marvik@elkraft.ntnu.no), (hans.hoialen@elkraft.ntnu.no). A. Petterteig is with SNTEF Energy Research, Tronheim, Norway: (astri.petterteig@sintef.no). Paper submitte to the nternational Conference on Power Systems Transients (PST29) in Kyoto, Japan June 3-6, 29 Fig.. Raial feeer with

2 . DSTANCE TO FAULT ESTMATON BASED ON MPEDANCE FROM SUBSTATON MEASUREMENTS The istance from the substation to the fault location, S, can be estimate from the voltage an current measure in the substation. n orer to minimize the influence from fault resistance, the imaginary part is use for istance estimation. Euation () give an estimate of the istance to a phase-tophase short-circuit when phases B an C are involve [9]. B C US U S S = m B C () S S X l Where X l is line reactance per unit of length, an superscripts B an C enote phase voltages an currents. The istance obtaine from () is use as a starting point for the calculations in both compensation methos escribe later in the paper. Loa an connecte along the feeer cause errors in the istance estimate. This is shown in Fig. 2-4., istance estimate error, is estimate minus real istance to the fault location. The error is positive when the estimate istance is to large, an negative when the estimate istance is too short. One specific case is marke by a small circle in the plots. This is the case with one -unit connecte 5 km from the substation, generating 3 MW at low loa. The fault is a 2- phase short-circuit 3 km from the substation. This mark is meant to make it simpler to compare the ifferent plot. A. One -unit Fig. 2 shows the istance estimate error when the faults are locate at the en of the feeer (3 km from the substation, on the main branch) in all cases, an the connection point is varie along the main branch. There is one -unit generating 3 MW (3M) or 6 MW (6M). All cases are at low loa ph. 3M 3-ph. 3M 2-ph. 6M 3-ph. 6M location, in km from substation Fig. 2. for faults at the en of the feeer, on the main branch. There is one -unit, an the connection point is varie. Low loa. The cause a positive error in the istance estimate, while the loa cause a negative error. The istance estimate error is largest when the -unit is connecte at the beginning of the feeer (5 km from the substation), an the fault is a two-phase short-circuit. The error increases substantially with the -size. While the maximum error is 8 km in the case with a 3 MW, it is 24 km with a 6 MW. Thus, a oubling of the -power results in a triplication of the istance estimate error in this case. f the is locate 25-3 km from the substation, the error is very small. Fig. 3 shows the istance estimate error when one 3 MW is connecte 5 km from the substation on the main branch, an the fault location is varie. All faults are 2-phase short-circuits Fig. 3. with the connecte 5 km from the substation on the main branch, an the fault location is varie. All faults are 2-phase short-circuits. The has negligible impact on the istance estimate for faults locate on the main feeer branch, before the connection point. The errors seen in these cases are mainly ue to loa. These are negative errors an largest for the high loa case. For faults on the sie-branch the has the same impact as for a fault locate after the -connection point, which is to increase the istance estimate error. A positive error means that the impact from the -unit ominates over the impact from the loa. Similar curves for 3-phase shortcircuits are shown in []. B. Three -units generating MW each Fig. 4 shows the istance estimate error with three units connecte to the feeer, as shown in Fig., when the fault location is varie. Each is generating MW, an all faults are 2-phase short-circuits Fig. 4. with 3 units, generating MW each, an the fault location is varie. All faults are 2-phase short-circuits.

3 The istance estimate error is positive for all fault locations, thus the -infee is always larger than the loa consumption. The error is increasing as the short-circuit is move farther away from the substation, confirming that a fault at the feeer-en is the worst-case. V. LOAD ESTMATON FOR FEEDER WTH ONE, UTLZNG PRE-FAULT -NODE VOLTAGE MEASUREMENT A simplifie feeer moel is use for analytical calculations. n this moel, loas are inclue as shunts in a pi-euivalent, as shown in Fig. 5. SL, SL, SL2, S L 2, Fig. 5. Simplifie representation of the feeer for analytical stuy. Loas are inclue as shunts in two pi-euivalents. f the pre-fault value of the voltage magnitue in the connection point, U,, is known, the loa between the substation an the, S L,, can be estimate from a secon orer euation []. To be able to solve this euation, the phase-angle of the loas, φ L,, must be known. The utility companies are assume to have knowlege about this angle. Since the -loa is known, the total loa can be estimate, an the remaining loa can be assigne to the loa after the connection point, S L2,. The estimate loa also inclues the loas on sie-branches. Knowing S L,, the -voltage phasor can be estimate: * (.5 ( ) U = U Z S U ) (2), S, l S, L, S, The angle of the estimate phasor is then combine with the measure -voltage magnitue. The loa current uring fault, L, is estimate using a static loa moel []: * NP NQ SL U U L = P * L, jql, U = (3) U U U U, U: voltage across the loa, pre-fault an uring fault P, Q : pre-fault active an reactive power of loa NP, NQ: voltage epenency factor for P an Q. The impact from an loa on the istance estimate can be compensate for when the loa an currents are known or estimate. The new istance estimate is obtaine from: B C US US Zcomp = (4) B B B C C C + + ( S L ) ( S L ) n this paper, the loa current is estimate in all cases. The current is either estimate (A) or measure (B). V. UTLZNG PRE-FAULT MEASUREMENTS FROM FOR FAULT LOCALZATON (COMPENSATON A) The -current uring fault neee in (4) has to be measure or estimate. n this chapter, pre-fault measurements from the -unit are utilize for estimating this current. Only magnitue values of voltage an current an the phase-angle are measure ( U,,,, φ, ). The avantage of not reuiring phasors, is that the time-synchronization between measurements from the an the substation o not nee to be as accurate as for instantaneous measurements. A. Assuming constant transient internal emf n the transient state, the generator can be represente by constant - an -axis transient emfs E ' an E ' behin the transient reactances x ' an x ', respectively. The rotor flux linkages in both axes can be assume to remain constant uring the transient state. The internal emfs corresponing to these linkages can also be assume to remain constant, an eual to the pre-fault values [2]: E = E E = E (5) This property, (5), is utilize for calculating the -current uring fault. The voltage uring fault can be estimate the same way as in (2), with the loa calculate using (3). To calculate the pre-fault transient internal emfs, the angle between a reference axis an the -axis, E, has to be foun. The angle is shown in Fig. 6. Phase A is use as the reference-axis that gives the reference angle for all measurements on the feeer. The -axis is leaing the -axis by 9º., E U, E, jx, U, Fig. 6. Phasor-iagram for the -unit, in the pre-fault state For a roun-rotor machine the angle of the -axis is foun by calculating steay-state -axis internal emf, E : E = U, + j x,, (6) By ecomposing the -current an -voltage to the - an -axes, the transient internal emfs can be calculate (7).

4 ( ) ( ) E = U, + x,, (7) E = U, x,, x,,x, inclue the transformer reactance in aition to the transient reactances of the generator. B. Estimation of 3-phase short-circuit current for Only positive seuence components are present uring a three-phase short-circuit, an the amplitue values are eual in all three phases. The can be represente by a positive seuence euivalent, as shown in Fig. 7. E + x + + U + E E x, x, +, +, Fig. 7. Positive seuence euivalent circuit for the transient state The - an -axis components of the currents are: = E U x ( ) ( ),,,, = E U, x, An the resultant current magnitue is: 2 2 (, ) (, ) U +, U +, = + (9) Table shows the calculate -current magnitues for ifferent -locations together with values from simulations, for comparison. The -locations are given in km from the substation, on the main feeer branch. TABLE ESTMATED AND MEASURED -CURRENT MAGNTUDES FOR 3-PHASE SHORT- CRCUTS AT THE FEEDER END location [pu] [pu] Low loa (LL) High loa (HL) estimate measure estimate measure The estimate values are larger than the measure values. The reason for this is probably that transient reactance values that are use in the calculation, correspons to a time in the short-circuit course earlier than 4 ms after fault inception, which is the time when the substation values are sample. C. Estimation of 2-phase short-circuit current for For a two-phase short-circuit the generator has to be represente by a negative seuence euivalent in aition to the positive seuence euivalent shown in Fig. 7. The (8) positive an negative seuence representations of the feeer are connecte in parallel at the fault location. The negative seuence representation of the is shown in Fig. 8. x U Fig. 8. Negative seuence euivalent circuit for the transient state The phase voltages in the -noe uring fault are estimate in a similar way as in (2). The positive an negative seuence components are given by (), where h = e j2π/3. + A B 2 C U = 3 ( U + hu + h U) () A 2 B C U = 3 ( U + h U + hu) Then the - an -axis components of the positive seuence -current are estimate from: + +, = ( E U, ) x, () + + = E U x ( ),,, The negative seuence -current is: ( ) ( ) = U j x = U j x x (2),, When the positive an negative seuence current components are known, the phase currents can be calculate: B 2 + = h + h + (3) C + 2 = h + h + The zero seuence component,, is eual to zero. Table 2 shows calculate -current magnitues together with the simulate values for ifferent -locations. TABLE 2 ESTMATED AND MEASURED -CURRENT MAGNTUDES FOR 2-PHASE SHORT- CRCUTS AT THE FEEDER END FOR LOW LOAD (LL) AND HGH LOAD (HL) location [pu] B C [pu] estimate measure estimate measure LL HL LL HL LL HL LL HL The estimate values are larger than the measure values, but the ifference is less than it was for three-phase shortcircuits. The negative seuence reactance of the generator, unlike the positive seuence reactance, is constant through the short-circuit course. Therefore, it is less critical that the measurements are sample at the exact same instant as the substation measurements for a two-phase short-circuit. D. Results with compensation A The istance estimate errors with compensation A for faults 3 km from the substation, on the main branch, are shown in

5 Fig. 9. There is one -unit, generating 6 MW or 3 MW, an the connection point is varie along the main branch. The results can be compare to Fig ph. 3M 3-ph. 3M 2-ph. 6M 3-ph. 6M location, in km from substation Fig. 9. with compensation A. All faults are at the en of the feeer, while the connection point of the is varie. Low loa. The compensation is best for two-phase faults. For threephase faults, all estimate -currents are higher than the real values. Thus there is an overcompensation of the impact from the. The result is a negative istance estimate error when the compensation is applie. Still, the result is much better than in Fig. 2, without any compensation. Fig. shows istance estimate errors with compensation A when the is locate 5 km from the substation, on the main feeer branch. The results can be compare to Fig angle is estimate in a similar way as in (2). Since the current magnitue is measure instea of estimate as for compensation A, the result is expecte to be better. A. One -unit The istance estimate errors with compensation B for faults 3 km from the substation are shown in Fig.. The figure can be compare to Fig. 2 an Fig ph. 3M 3-ph. 3M -2 2-ph. 6M 3-ph. 6M location, in km from substation Fig.. with compensation B. All faults are at the en of the feeer, while the connection point of the is varie. Low loa. As expecte, the result is much better than with compensation A (Fig. 9), especially for 3-phase short-circuits. Fig. 2 shows the istance estimate errors with compensation B when the is locate 5 km from the substation, an can be compare to Fig. 3 an Fig Fig.. with compensation A. The is connecte 5 km from the substation. The faults are 2-phase short-circuits at varying locations. The istance estimate error is significantly reuce when the compensation is applie, compare to Fig. 3. The errors were positive without compensation, but are generally negative with compensation. The ifference between the estimate -currents in phase B an C is larger than the simulate values, an the negative error is ue to overcompensation of the fault-currents. V. UTLZNG MEASUREMENTS FROM THE -NODE DURNG FAULT (COMPENSATON B) For compensation metho B the -current an -noe voltage magnitues uring fault (, U ) are measure, together with the phase angle (φ ). The -voltage phasor Fig. 2. with compensation B. The is connecte 5 km from the substation. The faults are 2-phase short-circuits at varying locations. Compare to the result with compensation A, shown in Fig., the errors are generally smaller. An exception is faults on the sie-branch at high loa, but the results are better than without compensation also in this case. The resulting errors are generally positive, so there is no overcompensation of the fault-currents. The overall errors are small, an the largest error for the cases shown is.25 km. Without compensation, the estimate error for the specific case marke by a circle in the plots is 4.5 km. With compensation it is reuce to.6 km (A) an. km (B). B. Three -units generating MW each Compensation metho B can be use for networks with

6 more than one -unit. nstea of estimating the pre-fault loa flows to the sections before an after each -unit, they are assume to be known. To be able to estimate the loa istribution, at least the currents flowing to the main- an the sie-branch nee to be measure. The assumption that the utilities know the loa in ifferent sections of the feeer is not unrealistic. Automatic meter reaing will provie ata of the loa consumption, which coul be utilize in the fault localization. f accurate loa ata is not available, the rating of istribution transformers in combination with typical loa profiles coul give an estimate of the loa in each section. Fig. 3 shows the istance estimate error with compensation B for a feeer with three -units, locate as shown in Fig.. Voltage an current magnitues, an phase angle of all three -units are measure uring fault Fig. 3. with compensation B for feeer with three -units. All faults are 2-phase short-circuits, an the fault location is varie. There is a large reuction of the istance estimate error compare to the curves shown in Fig. 4, without compensation. As before, the error is smaller with compensation for low loa than for high loa, while it initially was opposite. V. CONCLUSON The istance to the location of a phase-to-phase shortcircuit can be estimate from the substation voltage an current. Loas an -units connecte to the feeer cause errors in this istance estimate. The worst case for the example feeer is a two-phase short-circuit at the en of the main feeer branch, with the connecte close (5 km away) to the substation. Low loa is worse than high loa, since there is less loa to outweigh the impact of the infeecurrent. n a real case, measurement errors an inaccurate line ata may contribute to aitional errors in the istance estimate. This paper presents two methos for compensation of the errors ue to loa an. n compensation metho A, the pre-fault power-flow from the is utilize to estimate the -current magnitue uring fault. With this compensation, the istance estimate error is reuce in most cases compare to the situation with no compensation. The egree of improvement of the estimate, however, is epenent on the loa level an the location of the -unit. The estimate current tens to be too high, resulting in an over-compensation of the fault current. n compensation metho B, the -current magnitue uring fault is measure. This results in more accurate istance estimates than with compensation A, especially for three-phase short-circuits, an the result is much less epenent on the location of the -unit. t is also shown that if the loa in the ifferent feeer sections is known, this metho can be successfully use for fault localization with three -units. Both presente compensation methos result in a reuction of the istance estimate error. The results are best for the low loa case. For some of the high loa cases only a very small improvement is obtaine. The pi-euivalent representation of loas is a simplification, an the weakness of this moel becomes more significant as the loa is increase. Still, the simple pi-euivalent seems to be an acceptable loa moel. Since phasor angles can be estimate from the simplifie pi-euivalent moel, only magnitue values of -current an -voltage are reuire. This means that the timesynchronization between - an substation measurements can be less accurate than with instantaneous values. Finally, since the compensation methos are meant for localization, an not protection, communication spee is not critical. V. REFERENCES [] M. Lethonen, D. Cortinas et.al., Fault management in electrical istribution systems, Final report of CRED Working Group WG3 Fault Management, 5 th ntern. Conf. on Electricity Distribution, France, June 999, 4 pages. [2] T. Takagi, Y. Yamakoshi, M. Yamaura, R. Konow, T. Matsushima, Development of a new type fault locator using the one-terminal voltage an current ata, EEE Trans. on Power Apparatus an Systems, No. 8, pp , Aug [3] R. Das, Determining the Location of Faults in Distribution Systems, PhD-thesis, University of Saskatchewan, Canaa, 998, 26 pages [4] M. Lehtonen, M. Norman, J. Pitkänen, T. Antikainen, S. Hänninen, O. Vähämäki, S. Sauha-aho, New techniues for fault location in istribution networks. NORDAC, Sween, Aug. 26, 8 pages [5] D. Penkov, B. Raison, C. Anrieu, J.-P. Rognon, B. Enacheanu, impact on three phase fault location. use for fault location purposes?, ntern. Conf. on Future Power Systems, Nov. 25, 6 pages. [6] H. Y. Li, P.A. Crossley, N. Jenkins, Transient Directional Protection for Distribution Feeers with Embee Generations, 4 th Power Systems Computation Conference, June 22, 6 pages. [7] S.M. Brahma, A.A. Girgis, Development of aaptive protection scheme for istribution systems with high penetration of istribute generation, EEE Trans. on Power Delivery, Vol. 9, Jan. 24, pp [8] N. Perera, A.D. Rajapakse, A.M. Gole, Wavelet-base relay agent for isolating faulty sections in istribution gris with istribute generators, The 8 th EE ntern. Conf. on AC an DC Power Transmission, March 26, pp [9] S. H. Horowitz, A. G. Phake, Power system relaying, Secon Eition, John Wiley & Sons nc., 995, USA, pp 3-5 [] J. Marvik, H. K. Høialen, A. Petterteig, Evaluation of simple fault location on a feeer with, using funamental freuency components, NORDAC, Norway, Sept. 28, pages [] P. Kunur, Power System Stability an Control, McGraw-Hill, 994, USA [2] Machowski, J., Bialek, J. W., Bumby, J. R., Power system ynamics an stability, 997, John Wiley & Sons Lt, Englan.