Application of fractal theory in detecting low current faults of power distribution system in coal m

Availableonlineatwww.sciencedirect.comMININGscⅰ enceDirectSCIENCE ANDTECHNOLOGYELSEVIERMining Science and Technology 19(2009)0321-0325www.elsevier.com/locate/jcumtApplication of fractal theory in detecting low currentfaults of power distribution system in coal minesLIU Jian-hua, LIANG Rui, WANG Chong-lin, FAN Di-pengSchool of information and Electrical Engineering, China University of mining Technology, Xuzhou, Jiangsu 221008, ChinaAbstract: Single-phase low current grounding faults areoften seen in power distribution system of coal mines. These faults aredifficult to reliably identify. We propose a new method of single-phase ground fault protection based upon a discernible matrix ofthe fractal dimension associated with line currents. The method builds on existing selective protection methods. Faulted feeders aretherefore, suitable for distribution networks having different neutral grounding modes '. he current signals were first processedthrough a fast Fourier transform and then the characteristics of a faulted line were identified using a discernible matrix. The methodof calculation is illustrated. The results show that the method involves simple calculations, is easy to do and is highly accurate. It is,Keywords: low current faults; zero sequence transient current; fractal dimension; discernible matrix; selective protection1 Introductionrent and are sensitive to sudden signal changes andweak signals. However, wavelet analysis is difficultThe neutral point ungrounded method and the to be applied to smooth and continuously changingc-suppression coil grounding method are widelysignals 2-4. An increased transition resistance at thesed in MV power distribution networks in coal grounding point will decrease the size of the suddenmines.These are called low current neutral grounding change in the zero sequence current. In addition, calsystems. In these systems the fault rate of single culations based on transient wavelet analysis of a sin-phase grounding is always the highest, accounting for gle phase grounding fault are complex5-9.Conse-more than 85% of the incidents. Until now, no quently, there would be many difficulties if this techmethod has been found to easily discriminate a nique were to be used in production scale projectsfaulted feeder from a normal one. Scholars at home With this in mind, we intend to develop an algorithmand abroad have researched how to identifyinvolving less calculation that is based on fractal the-faulted feeder. Two main methods have been develory, which may be used in single-phase groundped, one based on transient signals and the other ing-fault protectionostationary signals. For persistent grounding faults,The fractal theory used in wave analysis of a singleeither resistive or from an arc-suppression coil, pro- phase transient grounding fault is first describedtection may be done using analysis methods based on Then a method of faulted feeder selection when sin-stationary signals. These techniques include zero se- gle phase transient grounding faults exist will be deand current injection memo a rscribed. The method uses discernible matrix techsequence admittance, negatisequence admittancen1qTransient grounding faults are difficult to charac- 2 Variation algorithm of fractal theoryterize because the fault time is so short that they arehard to identify. Most conventional methods of tran-fractal theory is a novel way to describe complexsient grounding protection function mainly through system morphology of irregular structures. "Fractal"extracting high frequency components of from the is a term coined in 1975 by the mathematician Benoitfault information. These techniques include wavelMandelbrot, who was working for the IBM companyanalysis. They Judge the faulted feeder through at the time, in 1975(13-16)analysis of transient signals in the zero sequence cur-中国煤化工滩 n get the same deReceived 02 September 2008: accepted 15 November 2008CNMHGChinaCorrespondingauthorTel:+86-516-83885605;E-mailaddresscumtlr@126.comMining Science and TechnologyVoLI9 No. 3tails from a certain part of a signal as we obtain from feeder through a 35 kv/6 kv Y/d transformer. Thethe whole signal, which means that the signal is frac- transformer has a 10 MVA capacity and its powerL. The research object is a non-smooth and irregular factor is 0.8. Two cables and two overhead lines aregeometric aggregation. It is also a non-linear system included in the apparatusthat includes regular and irregular fractals. Regularfractals are formed from geometric curves and graphs20 kmollowing strict definitions. Examples include Koch○ (CAblecurves and Cantor graphs. Most fractal signals found30 kmin nature are irregular, for example coast lines and thevariation of a heart rateIn this paper, character dimensions of irregularwaves are studied as a measure for wave identification. The signals cannot be expressed by strict func-tional expressions, but we can prove that the signalsare fractal by measuring and studying them. Highfrequency transient waves in a power system are alsoFig. 1 A sketch of a single phase grounding fain the distribution networkirregular fractals. High frequency transient signalsmainly exist during the first period to the second pe-Assume that the asymmetry of the power networkriod of a ground fault in a single phase power systemto earth is 1 under normal conditions and that itThe signal satisfies the requirements to be treated capacitance current is 50 A. The circuit parametersusing fractal principles. By analyzing fault waves and are, where the feeder distributed capacitance isand taken as a selective protection criterion for the earth of cable 1 is 20 uF; the capacitance to earth ofsingle phase grounding fault. There are many meth- cable 2 is 16 pF; the capacitance to earth of line 3 is 6ods for computing the irregular fractal dimensionF; and, the capacitance to earth of line 4 is 4 uF. Anincluding the box account, the isle, the sandbox, the over-expiation arc-suppression coil is used and itsarea slew radius and the variation transform methodsparameters are: a resistance of 6. 777 Q2 and an induc-We chose the variation transform for our computatiotance of 0.20816Hof the dimensionWe simulate a metallic, single phase groundingWe use a curve to describe the transient develop- fault, under different circumstances of groundingment of zero sequence voltage and zero sequence phase-voltage and phase-angle of the neutral pointcurrent when a fault occurs. Beginning by putting a ground, by an arc-suppression coil with resistance. Arectangle of width R over the fractal curve with the fault of the neutral point to ground is simulated by aheight of the rectangle fixed by the difference be- 1000 ohm resistance and a fault of bus grounding. Atween the highest and lowest points of the fractal half cycle of the zero sequence current waveform in aurve falling within the frame, the total area S(R)is single phase during transient grounding is shown ascomputed by summing the areas of all the rectangles Fig. 2of width R scanned over the curve. A sequence of S(R)The transient component from the grounding curvalues are found by repeating the aforementioned rent is much bigger than the stationary current com-process and changing the width(R). It should be con- ponent during the transient single-phase fault.Thefirmed that the rectangle has moved farther than its transient in the fault voltage and current is very short,width when performing the operation described abovebut still contains abundant information The transientThe formula is obtained by dividing S(r) by rcapacitance current is the sum of the capacitance cur-rent between the non-faulted phases and the faultedN(R)=S(R)/Rphase. The frequency of the discharge capacitanceAfter plotting a curve of In N(R)-In(I/R)we get the current in the faulted phase can reach thousands offractal dimension D as the slope of the lineHertz. This current flows through the fault point andthe bus and it is rapidly attenuated. The charge3 Simulation experimentspacitance current flowing in the non-faulted phasesforms a loop through the power supply and is attenu-Simulation experiments were performed in a 6 kv ated slowly; it has a low surge frequency. When ahigh voltage laboratory to verify the reliability of the single phase grounding fault occurs and the faulttechnique. The test equipment was a 6 kv neutral phase-voltage exceeds a certain value the slope of thepoint grounded by an arc-suppression coil systemzerotransient current of the faulted andwith four output-lines. The arc-suppression coil is nor中国煤化工 when this value apcontrollable and its out-of-resonance condition is ad- proonditionjustable(Fig. 1). The parameters are: the power net- appC NMH Gansient current curvework voltage is 35 kV and it is supplied to each seems almost the same when the bus is groundedLIU Jian-hua et alApplication of fractal theory in detecting low current faults of power125020.20202040.206020.2020.2040.206(a)At0°b)At30°1200.20.2020.2040.2060.20.2020.2040.206(d) At maximumFig 2 Waves of a faulted feeder and a normal feeder when a single phanding fault occurs under different conditions4 Faulted feeder selection for single phase and the fractal dimension of each feeder is obtainedtransient grounding based on fractalby using the variation fractal algorithm. Fig. 1 showstheoryzero sequence current waves after a single phasegrounding fault. From Fig. 1, we know that the zeroSuppose that the transformer substation bus con- sequence current consists of high frequency signalssists of n output-lines labeled (1, 2, ., n), zero seduring the first half wave. The frequency is around 1quence currents(io, iI,..., in) can be obtained from to 2 kHz and these high frequency signals have de-zero sequence current transformers at each feed back cayed to almost zero by the last half of the periodConsidering the characteristics of a single phase4.1 Data pre-processinggrounding fault, we call the first period of the faultthe transient process. A synchronized current meas-The distribution network in a coal mine is signifiurement is collected from each feeder at a samplingcantly different from a general 10 kV distribution speed of 30 kHz. This makes the length of the datanetwork. In a coal mine the network is supplied from600 points per feeder. The calculation of fractal dia single terminal with many short HV transmission mension is detailed belowlines radiating from the terminus. In addition, the load1)The rectangle width is set to two data-pointsnd power supply structure of the network change The graph is divided into 300 rectangle windows andfrequently. At the same time working conditions in the height of each window is the difference betweenthe coal mine are specialthe two data S(R) and N(r) are calculatedThe main load is centralized unederground maki2) The width is set to four data-points. The graph isit is necessary to pre-process the data sampled during divided into 125 rectangle windows and the length isa grounding fault. In this paper, an improved fast computed as in(1), except that the height of the rec-from interfering signals. The data are not contami- within the box. S(R)and N(R)are calculate om valueFourier transform is used to avoid contaminationnated by decaying DC components when using this 3)The width is set to 6 data-points. The graph isdivided into 100 rectangles. And, again, S(R)and N(R)4. 2 A practical method for calculating fractal calculateddimensionThis process is repeated until the width of the rec-tangle is 100 data-Doints. Curves of In(1/R)and InN(R)Identifying the faulted feeder during single phase four中国煤化工 The curves are fit to atransient grounding requires sampling every feeder at linethe fractal dimensionigh speed for several periods after the moment the DCNMHOand plotted as Fig 3fault occurred. A rise in zero sequence voltage is used to illustrate the method. Other plots are not includedas a start-up threshold signal. After the sampling, lists here because of length limitationMining Science and TechnologyVol 19 No. 38Ez0n(1/R)In(I/R)Fig 3 Determining the fractal dimension of different feedersThe fractal dimension of the faulted feeder is 1.876 definedas computed from the transient data. Fractal dimensions of the non-faulted feeders are 1.217. 1.235 andMD(,j=D-D(,j=1,2,……,n)(2)1. 229. The results in the other situations are collectedThe matrix is a square matrix of order n withagonal elements that are zero. The elements locatedsymmetrically in the upper- and lower-triangular maTable 1 Fractal dimension of a single phase groundedto earth and through 1000 Q2trix are equal in magnitude but reversed in sign. Sthe upper triangular matrix is ignored, preserving theGround fault initial phase 8 ()lower oneTo eartThrough 1000 Q2The discernible matrix defined this way is used fo306090selecting the faulted feeder. The row and columnDI.217 1. 189 1.202 1 178 1. 121 1.145 1.210 1.20 number of the maximum element in the lower trian-D212351.1671.1921.1981.1222131,16gular matrix are recorded. The column number corD1.8761.7841.8061.82314221.726 1.740 responds to the faulted feeder when the maximumD4 1 229 1.201 1.297 1.188 1.186 1.183 1.235 1.188 element is positive. The row number corresponds tothe faulted feeder when the maximum element isnegative. The elements of the discernible matrix4.3 Feeder selection for a discernible matrix offound by subtracting of every pair of two outputthe fractal dimensionfeeder fractal dimensions. Compared to an absolutTo enlarge the differences between the faulted comparison of the fractal dimension for each feederfeeder and the non-fault feeder found from the fractal this method magnifies differences between the faulteddimension, the signal is magnified by a discernible feeder and the non-faulted feeders. Table 1 shows thematrix.MD(i,D)is the element of row i and columni fractal dimension of various test feeders. The dis-in the discernible matrix. The discernible matrix is cernible matrices from these data are0.0180.022(0)-0.6590.6410MoI(30-0.5950.6170-0.0120.0060.6470-0.0120.0340.58300.0140.020M。1(60)=MD1(90)=-0.6040.6140-0.6450.6250-0.095-0.1050.5090-0.010010.63500-0.0360M2(0)=M。2(30°)0.3770.3410-0.0650.0640.2360-0.0380.0020.33900Ma2(60)=/00030中国煤化工MD2(90(4)-0.516-0.5130CNMHG0.02500220.4910L0.0180.0230.5520LIU Jian-hua et alApplication of fractal theory in detecting low current faults ofAs shown in Eq (3), MDI(3, 1)has the maximumfault in coal mine distribution network. Journal ofabsolute value for 0=0 and is a negative numberUniversity of mining Technology, 2008, 18(3):Hence the output- line three (line3)is the faulted[3] Lazkano A, Ruiz J, Aramendi E, Leturiondofeeder. In the same way, the faulted feeder is lineEvaluation of a new proposal for an arcing faultthree when=30°,θ=60°orO=909.Thetion method based on wavelet packet analysis. Eurfaulted feeder is line three according to Eq (4)Elect Power,2004,3(14):161-174.Therefore, the proper faulted feeder is obtained using[4] Zhang H F, Pan Z C, Tian Z G Discerniblethe method even when the grounding phase-angle ofmethod based method for detecting singlethe fault is different. The results are not affected bygrounding feeder. Electric Power, 2006, 39(11): 33-3the grounding resistance value. To check the co[5] Ruz F, Quijano A, Gomez E. A new algorithm for highness of the algorithm we also examined neutralmpedance fault detection in compensated neutralgrounding by an arc-suppression coil and an isolatedgrounded M.v. power systems.rtranneutral. the results in these two conditions also show2003,13(1):23-28.that the faulted and non-faulted feeders have gre[6] Mazon A J, Zamora J J, Zamora l, Albizu L, Diez M.Electronic circuit tests in the laboratory for superimposdifferences and that the method can differentiate be-ing voltages techniques. In: Proc 12th IEEE Mediterratween them. The only defect of this approach is itsnean Electrotechnical Conf Dubrovnik, 2004: 12-15low sensitivity when bus grounding occurs.[7] Zhao H M, Zhang B H, Duan J D, Zhou D C, Zhang wH. A new scheme of faulty line selection with adaptively5 Conclusionscapturing the feature band for power distributionnetworks. Proceedings of the CSEE, 2006, 26(2): 41-46(In Chinese)The fractal dimension of feedersomp[8] Jota PR S, Jota F G Fuzzy detection of high impedanceand used to identify a fault in the transmission line.faults in radial distribution feeders. Elect Power System,The feeder selection method for small grounding99,4903):169-174currents is based on the discernible matrix fractal di- [9 Welfonder T, Leitloff V, feuillet R, Vitet S. Locationmension. The correctness and availability of thestrategies and evaluation of detection algorithms forearth faults in compensated MV distribution systemsmethodology are verified by a series of experiments,IEEE Trans Power Del, 2000, 15(4): 1121-1128as shown above. The method does not have any spe- [10] Segui T, Bertrand P, Guillot M, Hanchin P,Bastardcial requirements and has less computational com-damental analysis for distance relaying with paplexity. It is a greatly improved method compared totrical estimation. IEEE Trans Power Del, 2001, 16(1)feeder selection methods based on transient99104It has great feasibility when using high san[Il] Zhang Q C, Yao B, Yang J F, Wang Z H. Fault lindentification and fault location of non-direct groundpling speed analog to digital converters and requiresneutral system based on R/s dimension Transactions olittle computing power. The technique also has dis-China electrotechnical Society, 2007, 22(12): 164-169advantages, such as having a low sensitivity to busIn Chinesefaults and misjudging a fault when there are high [121 Zamora l, Mazon A J, Sagastabeitia k, Pico O, Saenz Jcurrent disturbances. So there will be much work toR. Verifying resonant grounding in distribution systems.do before using it in productionIEEE Comput Appl Power, 2002, 15(4): 45-50[13] Hao J L, Yuan Z H. Study on line selection of groundedfault in non-effectively earthed system based on waveletAcknowledgementsneural network. Electrical Measurement Instrumentation, 2007, 44(11): 1-5.(In Chinese)Financial support for this work, provided by the [4] LaiTM, Snider L A, Lo E, Sutanto D High-impedanceNational Natural Science Foundation of China(50504015)and the Youth Science and Technologyfrequency range and RMs conversion. IEEEPower Del,2005,20(1):397-407Research Program of China University of Mining and [15] Yang X H, Cai X. Application of fractal theory in theTechnology (0C060996), are gratefully acknowlfault classification for distribution systems. Automationof Electric Power Systems, 2006, 30(4): 59-64.(InReferences[16] Wu Z S, Yang X C, GuiJ F Distinguishing groundn different feeders of power distribution networks1 novel principle of arc suppression coil based on L.AWang C L, Liang R, Liu J H, Li X B, Zhang Dfractal information. Journal of Tsinghua University2007,47(4):470473.( In chinese[17] Liang R,China University of Mining Technology, 2005613-616. (In Chinese)on. High Voltage Engineering.[2] Li X B, Wang C L Comparison between open phasefault of arc suppression coil and single phase to earth中国煤化工CNMHG