A. H. Osman, Student Member, IEEE O. P. Malik, Fellow, IEEE Dept. of Electrical and Computer Engmeermg University of Calgary, T2N 1N4, Canada Abstract: An apphcatlon of wavelet transform to digital distance protection for transmission hnes is presented in this paper. Fault simulation is earned out using the Power System Computer Aided Design program (PSCAD). The simulation results are used as an input to the proposed wavelet transform protection-relaying technique. The technique Nbased on decomposing the voltage and current signats at the relay location using Wavelet Filter Banks (WFB). From the decomposed signals, faults can be detected and classified. Also the fundamental voltage and current phasors, which are needed to calculate the impedance to the fault point can be estimated. Results demonstrate that wavelets have high potentiat m distance relaying. Keywords: Distance Protection Relaying, Wavelets. 1. INTRODUCTION Power Systems, Great attention has been paid to distance relaying techniques for the protection of transmission lines. The main target of these techniques is to calculate the impedance at the fundamental frequency between the relay and the fault point. According to the calculated impedance, the fault is identified as internal or external to the protection zone. This impedance is calculated from the measured voltage and current signals at the relay location. In addition to the fundamental frequency, the signals are usually contain some harmonics and DC component, which affect the accuracy of the phasors estimation [1]. Recently, distance relays have experienced much improvement due to the adoption of digital relaying, Signal processing is one of the most important parts of the operation of the digital distance protection. Until recently, Fourier analysis and Kalman filtering methods were the mam tools in signal processing for distance relaying [2]. The trip/no trip decision has been improved compared to the electromechanical or solid-state relays. However, the reach accuracy of these methods 1s affected by the different fault conditions particularly in the presence of high frequency and DC offset in the signals. Wavelet transform as an alternative solution to the metmoned approaches for distance relaying of transmission lines is introduced in this paper. Wavelets are a recently developed mathematical tool for signal processing. Compared to Fourier analysis, which relies on a single basis function, wavelet analysis uses basis functions of a rather wide functional form. The basic concept in wavelet transform is to select an appropriate wavelet function mother wavelet and then perform analysis using shifted and dilated versions of this wavelet. Wavelet can be chosen with very desirable frequency and time characteristics as compared to Fourier techniques. The basic difference is that, in contrast to the short time Fourier transform which uses a single analysis window, the wavelet transform uses short windows at high frequencies and long windows at low frequencies. The basic functions in wavelet transform employ time compression or dilation rather than a variation in time frequency of the modulated signal [3], This work describes tbe application of wavelet transform in detecting and classifying faults as well as extracting the voltage and current fundamental phasors needed to calculate the impedance to the fault point. Il. WAVELET TRANSFORM Wavelet transform was introduced at the beginning of the 1980s and has attracted much interest in the fields of speech and image processing since then, Its potential applications tOPOWer industry have been discussed recently [4 7]. A brief introduction to the wavelet transform is given here, more details can be found in [8, 9]. There are two fundamental equations upon which wavelet calculations are based; the scaling function, tit) and the wavelet function Y(t): q(t) =N5 ;hk 47(2t -k) (1) !4) =fi pkd2t - k) (2) The functions are two-scale difference equations based on a chosen scaling function tp, with properties that satisfy certain admission criteria and discrete sequence: h, and gk = (-I)* h,.Lrepresenting discrete filters that solve each equation. The scaling and wavelet functions are the prototype of a class of orthonormal basis functions of the form: 0-7803-7031-7/01/$10.00 (C) 2001 IEEE 0-7803-7173-9/01/$10.00 2001 IEEE 115 ( ~J,k(f) =2j/2 ) p2Jtk, j,kcz yj,k(t) = 2J2 ( ) y2~t k, j,kez (3) (4) where the parameter j controls the dilation or compression of the function in time scale and amplitude. The parameter k controls the translation of the function in time. z is the set of integers. Wavelet system may or may not have compact support. Compact support means that the wavelet system has finite number of nonzero coefficients. A wavelet system with compact support allows wavelets to localize in both time and frequency, so only is dealt with here. Several techniques have been used in the literature to create wavelet systems, These include cubic splines, complex exponentials, and parameter-space constructions. Once a wavelet system is created, it can be used to expand a function f(t)in terms of the basis functions: where the coefficients c(l), and d(j, k) are calculated by inner product as: (6) (7) If the wavelet system has compact support and an upper limit is placed upon the degree of dilation j, then the expansion equation becomes: (8) The expansion coefficients c(l) represent the approximation of the original signal ~(t) with a resolution of one point per every 2J points of the original signal. The expansion coefficients d(j, k) represent details of the original signal at different levels of resolution. These coefficients completely and uniquely describe the original signal and can be used in a way similar to the Fourier transform. The wavelet transform then is the process of determining the values of c(1) and d(j,k) for a given f(t)and wavelet system. The expansion equation naturally leads to a recursive algorithm for the wavelet transform, given certain assumptions. First, the function f(t) is taken as a sequence of discrete points sampled at 2 points per unit interval. These points can be viewed as the inner product of @and ~(t). That is, the sample points are approximation, or c(l) coefficients, of the continuous function ~(t). This allows c(l) and d(j, k) terms to be calculated by direct convolution of ~(t) samples with the coefficients h~and g~. It was discovered in [10] the wavelet transform can be implemented with a specially designed pair of Finite Impulse Response (FIR) filters called a Quadrature Mirror Filters (QMF) pair. QMFs are distinctive because the frequency responses of the two FIR filters separate the high frequency and low frequency components of the input signal. The dividing point is usually halfway between OHz and half the data sampling rate (the Nyquist frequency). The outputs of the QMF filter pair are decimated (or desampled) by a factor of two; that is, every other output sample of the filter is kept, and the others are discarded. The low-frequency (low-pass) filter output is fed into another identical QMF filter pair. This operation can be repeated recursively as a tree or pyramid algorithm, yielding a group of signals that divides the spectrum of the original signal into octave bands with successively coarser measurements in time as the width of each spectral band narrows and decreases in frequency. The tree or pyramid algorithm can be applied to the wavelet transform by using the wavelet coefficients as the filter coefficients of the QMF filter pairs as shown in [11]. The same wavelet coefficients are used in both low-pass and high-pass (actually, band-pass) filters. The low-pass filter coefficients are associated with the h~ of the scaling function #t). The output of each low-pass filter is the c(l), or approximation components, of the original signal for that level of the tree. The high-pass filter is associated with the g~ of the wavelet function Y(t).The output of each high- pass filter is the d(j,k), or detail components, of the original signal at resolution 2J. The c(l) of the previous level are used to generate the new c(l) and d(j, k) for the next level of the tree. Decimation by two corresponds to the multiresolutional nature of the scaling and wavelet functions. The reverse wavelet transform essentially performs the operations associated with the forward wavelet transform in the opposite direction [12]. In wavelets applications, different basis functions have been proposed and selected. Each basis function has its feasibility depending on the application requirements. In the proposed scheme, the Daubechies-4 (db4) wavelet [10] served as the wavelet basis function for the detection and classification of faults as well as for the estimation of the voltage and current phasors. The db4 FIR filter coefficients are given in Table 1. Table 1. db4 FIR filter coefficients k h, ~ o -0.0106 -0.2k304 1 0.0329 0.7148 2 0.0308 -0.6309 3 -0.1870 -0.0280 4 -0.0280 0.1870 5 0.6309 0.0308 6 0.7148 -0.0329 7 0.2304 -0.0106 0-7803-7031-7/01/$10.00 (C) 2001 IEEE 0-7803-7173-9/01/$10.00 2001 IEEE 116 Ill. PROPOSED WAVELETS DISTANCE PROTECTION ALGORITHM In order to investigate the applicability of the proposed wavelet transform distance protection algorithm, a simulation of transmission line model for different fault locations and different loading conditions is done. Fault simulations were carried out using PSCAD. The simulated power network is shown in Fig.1. The network consists of two areas connected by a transmission line. The transmission line was modeled as a distributed parameters line, representing a 225km long 240kV ideally transposed transmission line with impedance, ZL=(4. 15 + J84.66)f2. The three phase voltage and current signals at the relay location, which is at bus A, are sampled at 960Hz (16 samples/fundamental power cycle). These six signals are then filtered using a pre-band-pass filter with cut-off frequency 80Hz to attenuate the dc component. The output- filtered signals are the input to the proposed wavelet distance protection algorithm. G1 A T. Line B G2 Zh 240kV, 225km Sources X/R=60 Fig. 1 Transmission line model system The proposed technique is divided into two sections: 1) The first section is the detection of the fault by observing the output of the high-pass filter (details) of the first decomposition level. This decomposition level has the ability to detect any disturbances in the original signal waveform. Fig. 2 shows an example for an original current waveform during a fault and the output of the high-pass filter (D 1) of the first decomposition level. It is clear that the fault moment has been detected easily by this decomposition. By passing all three phase currents and voltages at the relay location through the wavelet first decomposition level, it is possible to detect and classify the faulty phase(s). The data window length till the moment of faul~ detection is half a cycle (8 samples). Oriaina!current smal 01 Detail (Dl) * Eo g -01 0 20 40 60 80 100 120 140 160 180 Fig.2 Original current waveform and D1 during a fault 2) The second section of the algorithm is the estimation of the fundamental frequency voltage and current phasors. It can be done by observing the output of the low-pass filter at the second decomposition level, Theoretically, The frequency band of the fundamental frequency should be obtained from the fourth level of decomposition, and this requires a very long data window, which is not possible to be applied on line. The second level of decomposition gives a very good approximation for the phasors. At this level the high frequencies in the signal are eliminated by the high- pass filters of the first and second decomposition levels and the DC component has already been eliminated by pre- band-pass filtering the signal. The data window length for estimating the phasors before the fault is half a cycle (8 samples). However, at the moment of fault detection, the last half cycle data window is kept and it starts to increase by one new sample until it reaches a full cycle (16 samples) and then it continues with this window length. This is equivalent to an adaptive data window, and gives the advantage of fast clearance to the near faults, while the faults at the end of the line need about% to one full cycle to be classified. The estimation of the phasors is based on capturing the peak of each signal (magnitude), and locating its position from the beginning of the data window (phase). For each new sample, the impedances for ground and phase faults that are seen by the relay at A are calculated using the well-known equations: i- For phase to ground faults v z phase = ~ phase (9) Pha.w(l)+ Iphaw(x + K(o) Zpha,, (o) , K(o)=ZL(o)/zL(l) where Vp,,,,,,,is the estimated phase voltage phasor, 1,,.,,(1), 1,,,,,,,,(2), 1,,.,.(0), are the positive, negative, and zero sequence estimated phase currents respectively, and K(O) is the ratio between the zero sequence to the positive sequence impedance of the protected transmission line. ii- For phase faults (example: phase a to phase bfault) where V,,, and V~are the estimated voltage phasors, l., and Ib are the estimated current phasors. IV. SIMULATION RESULTS The described wavelet transform approach to distance protection for transmission lines is applied for the model of Fig, 1, The performed tests include different fault types, different fault locations, and different loading conditions, 0-7803-7031-7/01/$10.00 (C) 2001 IEEE 0-7803-7173-9/01/$10.00 2001 IEEE 117 1) The first set of tests is carried out with a power angle of 1(Ydegrees between the two sources, and fault resistance Rf=O, for different fault locations. Fig.3 shows the impedance calculated at the relay location for a single line to ground fault at 50km from the relay location. Fig.4 shows the impedance fault trajectory for this fault, from which one can recognize how fast the fault trajectory entered the tripping area, It took about 3ms to classify the fault as an internal fault. Fig. 5 shows the impedance calculated at the relay location for a double line to ground fault at 100km, and Fig.6 shows its impedance fault trajectory. It took about 7ms to classify the fault as an internal fault. Fig.7 shows the impedance calculated at the relay location for a single line to ground fault at 150km from the relay location, and Fig.8 shows its impedance fault trajectory. It took about 12ms to classify the fault as an internal fault. This also clarifies the idea of using longer data window in order to classify the far fault locations. 2) The second set of tests is carried out with a power angle of 20 between sources, and fault resistance Rf=O, for different fault locations. Fig. 9 shows the calculated impedance at the relay location for a single line to ground fault at the beginning of the line near bus A, and Fig. 10 shows its impedance fault trajectory. The trajectory entered the tripping zone immediately after the fault. Fig. 11 shows the calculated fault impedance at the relay location for a double line to ground fwlt at 50km from the beginning of the line, and Fig. 12 shows its fault impedance trajectory. The trajectory entered the tripping zone after 3ms. Fig. 13 shows the calculated impedance at the relay location for a phase-to- phase fault at 100km, and Fig. 14 shows its fault impedance trajectory. The trajectory entered the tripling zone after 8ms. The results show that wavelet transform enabled the detection of faults, and estimation of the phasors at the relay location. The wavelet distance relay successfully classified every type of fault at any distance. o 0 10 20 30 40 50 60 Tme Samples Fig,3 Single line to ground fault (A-G) at 50km Fault trajectory for SLG (A-G) at 50km R (ohm) Fig.4 Fault trajectory for single line to ground fault (A-G) at 50km Double line to around fault 1A-C-G) at 100km. Rf=O Tme Samples Flg5 Double Line to ground fault (A-C-G) at 10Okm R (ohm) F]g.6 Fault trajectory for double line to ground fault (A-C-G) at 100km 0-7803-7031-7/01/$10.00 (C) 2001 IEEE 0-7803-7173-9/01/$10.00 2001 IEEE 118 Tme Samples -:20 0 20 40 60 80 100 120 140 R (ohm) Fig.7 Single line to ground fault (A-G) at 150km Fig. 10 Fault trajectory for single line to ground fault (A-G) at the beginning of the line R (ohm) Tme Samples Fig.8 Fauk trajectory for single line to ground fault (A-G) at 150km Fig. 11 Double Line to ground fault (A-B-G) at 50km SLG fault (A-G) at the beginning of the line, Fif=O Tme Samples Fault trajectow for DLG (A-B-G) at 50km, Rf=O \ j . . . .. II -20 I : ,.: ! I -20 0 20 40 60 80 100 120 140 R (ohm) Fig.9 Single line to ground fault (A-G) at the beginning of the line Fig. 12 Fault trajectory for double line to ground fault (A-B-G) at 50km 0-7803-7031-7/01/$10.00 (C) 2001 IEEE 0-7803-7173-9/01/$10.00 2001 IEEE 119 Double line fault (B-C) at 10Okm, Rf=O 2500 ! ! ! ! ! ~ -I o 0 10 20 30 40 50 60 Tme Samples Fig. 13 Double Ltne fault (B-C) at 100km Fault trajectory for DL,!sak@-C) at 100km, Rf=O 140, , , ,., . .... , I ! 1 :20 0 20 40 60 80 100 120 140 R (ohm) Fig. 14 Fault trajectory for double hne fault (B-C) atl 00km V. CONCLUSIONS Use of the wavelet transform in distance protection relaying of transmission lines is introduced in this paper. The ability of wavelets to decompose the signal into frequency bands (multi-resolution) in both time and frequency allows accurate fault detection as well as estimation of the signal phasor at the fundamental frequency. The results show that the proposed technique used to introduce wavelets in distance protection is applicable and encouraging. Various types of faults on the transmission line can be classified accurately within one cycle according to the fault location. Further studies are in progress in testing the relay for high resistive ground faults and line energization. [1] [2] [3] [4] [5] [6] [7] [8] [9] V1. REFERENCES D. V. Coury, and D. C. Jorge Artificial neural network approach to distance protection of transmission lines, fEEE Trans. Power Delivery, Vol. 13, No. 1, January 1998, pp. 102-108. A, A. Girgis, and R. G. Brown, Application of Kalman filtering in computer relaying, fEEE Trans. on Power Apparatus and Systems, PAS-1 OO(7), July 1981, pp.3387-95, L L hi , E Styvaktakis, and A, G, Sichanie Wavelet transform for high impedance fault identification, Proceeding of the 4 tntemational conference on advances in power system control, operation and management Hong Kong, November 1997, pp. 188- 191. D. Wai, and X. Yibin, A novel technique for high impedance fault identification, tEEE Trans. on Power Delivery, Vol. 13, No, 3, July 1998, pp. 738-744. S. Santoso, E. J. Powers, W. M. Grady, and P. Hofmann, Power quality assessment via wavelet transform analysis, fEEE Trans. on Power Delivery, Vol. 11, No. 2, April 1996, pp. 924-930, D, C, Robertson, 0, I. Camps, J. S. Mayer, and W. B, Gish, WaveIet and electromagnetic powers ystem transients, IEEE Trans, on Power Delivery, Vol. 11, No. 2, April 1996, pp. 1050-1056, 0. Chaari, M. Meunier, and F. Brouaye, Wavelets: A new tool for the resonant grounded power distribution systems relaying, tEEE Trans. on Power Delivery, Vol. 11, No. 3, July 1996, pp. 1301-1308, G. Strang, and T. Nguyen, Wavelet and filter banks, Wellesleg- Cambridge Press, 1997. C. Sidney Burros, and R. A. Gopinath, Introduction to wavelets and wavelet transforms a primer , Prentice Hall, 1988, [10] I. Daubechies, Ten lectures on wavelets, Capital City Press, 1992. [11 ] S. G. Mallat, A theory for multiresolution signal decomposition: the wavelet representation, tEEE Trans. on Pattern Analysis and Machine intelligence Vol. 11, No. 7, July 1989, pp. 674-693, [12] Mac. A. Cody, The fast wavelet transform, Dr. Dobbs Journal, April 1992. VI1.BIOGRAPHY A. H. Osman(SM01) received his B. SC. and M. SC. in electrical engineering from Helwan University, Cairo, Egypt in 1991 and 1996 respectively. Currently, he is working towards a Ph.D. degree at the University of Calgary, Canada, His areas of interest include power system engineering, digital protection relaying, and power electronics, O. P. Malik (M66-SM69-F87) graduated in electrical engineering from Delhi Polytechnic, In&la, in 1952, and obtained the M.E. degree in electrical machine design from the University of Roorkee, India, in 1962. Ju 1965 hereceived the Ph,D. degree from the University of London, England, and D.I.C. from the Jmperial College of Science smd Technology, London, or at the University of Calgacy in 1974 and is a faculty professor emeritus. He is a Fellow of the Institution of Electrical Engineers (London), and a registered professor Engineer in the provinces of Alberta and Ontario, Canada. 0-7803-7031-7/01/$10.00 (C) 2001 IEEE 0-7803-7173-9/01/$10.00 2001 IEEE 120