Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines

G Model

ARTICLE IN PRESS

ASOC 3114 1–8

Applied Soft Computing xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Applied Soft Computing journal homepage: www.elsevier.com/locate/asoc

Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines

1

2

3

Q1

4

Hassan Fathabadi ∗ Engineering Department, Kharazmi University, Tehran, Iran

5 6

a r t i c l e

7 21

i n f o

a b s t r a c t

8

Article history: Received 6 February 2015 Received in revised form 20 June 2015 Accepted 29 July 2015 Available online xxx

9 10 11 12 13 14

20

Keywords: Discrete wavelet transform (DWT) Soft computing method Filter based technique Power transmission lines Short-circuit fault detection

22

1. Introduction

15 16 17 18 19

In this study, two approaches are presented to detect short-circuit faults in power transmission lines. The two proposed methods are completely novel from both theoretical and technical aspects. The first approach is a soft computing method that uses discrete wavelet transform with Daubechies mother wavelets db1, db2, db3, and db4. The second approach is a hardware based method that utilizes a novel proposed two-stage finite impulse response filter with a sampling frequency of 32 kHz, and a very short process time about three samples time. The two approaches are analyzed by presenting theoretical results. Simulated results obtained by simulating a three-phase 230 kV, 50 Hz power transmission line are given that validate the theoretical results, and explicitly verify that the filter based approach has an accuracy of 100% in presence of 10% disturbance while the accuracy of the wavelet transform based approach is maximally 97%, but it has less complication and implementation cost. Another comparative study between this work and other works shows that the two proposed methods have higher accuracy and very shorter process time compared to the other methods, especially in presence of 10% disturbance that actually occurs in power transmission lines. © 2015 Published by Elsevier B.V.

Q3 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38

A transmission line is one of the important elements of an electric power system. Since 1945, electric power systems have been extremely developed because the steep increase in population of most countries has caused a huge increase in electric energy demand [1,2]. As a result, power transmission lines have been rapidly developed both in number and length. One important factor of an electric power transmission system is to continuously deliver the electric power to consumers. A problem related to electric power systems is faults occurrence in power transmission lines which is an unfavorable and inevitable issue. Short-circuit faults are the worst types of the faults occurring in power transmission lines. Short circuit faults have many harmful effects on the electrical distribution systems and devices. Some harmful effects are: shortening the life of the electrical devices, increase in power losses, and additional heat produced by cables, wires, insulators, transformers, etc. [3]. When a short-circuit fault occurs in a transmission

∗ Current address: National Technical University of Athens, Q2 Tel.: +98 9714321; fax: +98 9714321. E-mail addresses: [email protected], [email protected]

Greece.

line, power outage is the first result which is carried out by protection relays, and consequently, there is an interrupt in delivering the electric power to consumers. Thus, utilizing some methods or using some devices having capability to determine faults location quickly and accurately are necessary. In recent years, applications of artificial neural network (ANN) in power systems such as stability and transient response analysis have attracted many attentions [4,5]. As an application, fault analysis in power transmission systems using ANN has been subject of some researches [6–9]. Based on a radial basis function (RBF) neural network with orthogonalleast-square (OLS) learning procedure, a simple fault classification method was reported in [10]. The method identifies various patterns of associated voltages and currents, but it does not have the capability of fault location, and furthermore, some faults cannot be classified because the method can not identify all possible patterns related to different faults. The RBF neural network was also compared with the back-propagation (BP) neural network, and it was shown that the RBF neural network classifies faults better than BP neural network [10]. Discrete wavelet transform (DWT) as a soft computing tool has been used in different applications [11–13]. In some researches, wavelet transform has been combined with other topics such as ANN and fuzzy systems to form hybrid frameworks [14–19]. A fault detection and classification method which uses

http://dx.doi.org/10.1016/j.asoc.2015.07.039 1568-4946/© 2015 Published by Elsevier B.V.

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61

G Model ASOC 3114 1–8 2 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127

ARTICLE IN PRESS H. Fathabadi / Applied Soft Computing xxx (2015) xxx–xxx

discrete wavelet transform in combination with ANN was reported in [20]. The technique is absolutely theoretical and fault location is impossible. A method based on S-transform (ST) and support vector machines (SVMs) for classification and identification of a faulty section in a transmission line with a fixed series capacitor placed at the middle of the line was presented in [21]. To detect faults, distinctive features of line voltages, currents, and zero sequence current are used. The relevant features of these signals are obtained using ST, and then, the obtained features are used as input for multiple SVM classifiers, and their outputs are combined to classify the fault type [21]. Using frame-based sequence classification (FBSC), the Alternative Transient Program (ATP), and a public dataset, a framework was proposed for event classification in [22]. It was shown that the method can be used for classifying short-circuit faults in transmission lines. A fault location model of a transmission line that uses Elman recurrent network (ERN) was presented in [23]. Wavelet transform was again used for selecting distinctive features of the faulty signals, and then, ERN is utilized to determine the fault location using the features obtained by wavelet transform. The model can be only used for locating balanced shortcircuit faults [23]. A technique for fault detection which uses finite impulse response artificial neural network (FIRANN) was proposed in [24]. For training the FIRANN, the training patterns obtained from more than one relaying position were used in order to obtain better result. The proposed FIRANN-based method uses voltage and current samples at 2 kHz to detect faults [24]. Faults classification in transmission lines using data mining was reported in [25]. The ATP simulator was utilized to produce a comprehensive labeled dataset which is necessary to classify faults. A comparison between different processing methods and algorithms such as wavelets, decision trees, and neural networks showing better performance of neural networks was addressed in [25]. A survey in the literature shows that there are other similar researches as follows that the accuracy of each proposed algorithm in absence of any disturbance has been also reported. Particle swarm optimization (PSO) together with ANN was used to classify faults in [26]. Fuzzy logic was applied for fault classification in [27]. A method for protection of power systems using SVMs was addressed in [28]. Detection and classification of faults in power transmission lines using functional analysis and computational intelligence was reported in [29]. It is worthwhile to note that in a practical power transmission line, the disturbance is always available in the transmission line, but as mentioned, the above works have been reported the accuracy of their algorithms in absence of any disturbance. In this study, two novel approaches are presented for shortcircuit faults detection in power transmission lines. The first approach is a soft computing method that uses DWT with Daubechies mother wavelets. The novelty of this application of DWT is that the detail coefficients down sampled and obtained by utilizing wavelet transform are only used to detect a short-circuit fault, and furthermore, a difference function between the detail coefficients in different times is performed to completely eliminate the effect of the usual disturbance that actually occurs in a transmission line. In fact, using detail coefficients and completely eliminating the effect of the disturbance are the two distinct differences between the DWT based method proposed in this work and the similar applications of DWT such as that reported in [30]. For example, the method presented in [30] uses approximate coefficients, and there is not any consideration for robustness of the method against the actual disturbance of the transmission line. It is worthwhile to note that using approximate coefficients for fault detection in a power transmission line cannot be practically used because approximate coefficients are actually the low-frequency components of the power signal transmitted on the line. On the other hand, the power signal itself is a low-frequency signal, so the

magnitude of the power signal effectively impacts on the obtained approximate coefficients. The second approach is an entire hardware based method that uses a proposed two-stage finite impulse response (FIR) filter. The first stage is a FIR comb filter, and the second stage consists of four FIR band-pass filters. For the first time, the center frequencies of the four FIR band-pass filters are obtained in this research. The structure of the proposed two-stage FIR filter is completely novel, and there is not any other entire hardware based method or filter with similar structure or capability reported in the literature. For example, the research addressed in [31] deals with estimating the time-instants of abrupt changes in the signal recorded in digital fault recorders (DFRs), the method does not detect any type of faults such as short-circuit faults. It only analyzes the recorded signal to determine the time-instants before fault occurrence, initiation of fault, circuit-breaker opening, and auto-reclosure of the circuit-breakers. At first, a simple first-order adaptive whitening filter with sampling frequency of 2.5 kHz is used to reduce the noise of the signal recorded in DFRs, and then, the detailed and smoothed components of the signal are extracted using DFT. Finally, a timeinstant of abrupt changes is detected when the wavelet coefficients exceed the universal threshold of Donoho and Johnstone. Thus, the simple filter used in [31] only reduce the nose, but the two-stage FIR filter presented in this work itself alone detects a short-circuit fault in a power transmission line without any assistance, even using DFRs. Similarly, the first approach (DWT based method) proposed in this work is a single-handed method. The rest of this paper is organized as follows. The wavelet transform based approach is presented in Section 2. Section 3 deals with the second approach. Simulated and comparative results are presented in Section 4 to validate theoretical results, and to verify the superior performance of the two proposed approaches. Section 5 concludes the paper.

2. Proposed wavelet transform based approach for short-circuit faults detection

1



am 0

 x(k)

n − kb0 am 0 am 0

129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160

161 162

Discrete wavelet transform (DWT) of a signal such as x(t)is defined as [30,31]: DWT (m, n) =

128

163 164

 (1)

It can be summarized that DWT is an extended form of the discrete Fourier transform (DFT), so that, it can be applied to the signals such as non-stationary signals with non-iterative features that DFT can not be used to analyze them. When a short-circuit fault occurs in a power transmission line, some transient features appear in the magnitudes-time and phase-time characteristics of the threephase current flowing through the power transmission line. In this section, DWT is used to extract the mentioned transient features of the three-phase current, and then, the obtained features are used to detect short circuit faults. In fact, the three phase currents are decomposed using DWT to extract the produced high frequency transient features caused by the short-circuit fault(s), so choosing mother wavelet type is very important issue. In this research, Daubechies (db) wavelet is considered as mother wavelet because it has better features for protection applications [30]. There are four different types of Daubechies wavelet which are called db1, db2, db3, and db4 [31]. All the four types with a0 = 2, b0 = 1, and onelevel operation are used to detect short-circuit faults, but as will be shown the mother wavelet db4 has better performance and accuracy. To clarify the subject, DWT with the one-level mother wavelet db4 has been applied to the signal x(t) shown in Fig. 1(a), the approximation coefficient (Ax ) and detail coefficient (HFx ) resulted

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

165

166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187

G Model

ARTICLE IN PRESS

ASOC 3114 1–8

H. Fathabadi / Applied Soft Computing xxx (2015) xxx–xxx

3

Fig. 1. DWT with the one-level mother wavelet db4 and down sampling factor of two: (a) signal x(t); (b) approximation coefficient; (c) detail coefficient.

188 189 190 191 192 193 194 195 196 197 198 199 200

201

202 203

204

205

206

207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223

from down sampling by the factor of two are shown in Fig. 1(b) and (c). In practice, disturbance is always available in power generation and distribution systems such as power transmission lines, so to detect a short-circuit fault, it is important to eliminate or reduce the effect of the usual disturbance. In this research, at any time, a difference function between the two sequential values of the detail coefficients at the two last sampling times is performed to eliminate the effect of the usual disturbances available in the transmission line. Thus, for each phase current a logic function that subtracts the two sequential values of its related detail coefficients is defined to detect the related short-circuit fault. For the phase current R, the logic function is defined as:



L.FW −R =

1 0

when |HFR (n) − HFR (n − 1)| > Mth when |HFR (n) − HFR (n − 1)| ≤ Mth

(2)

where Mth is the threshold value. For the phases S and T, the related logic functions are similarly defined as:



L.FW −S = and L.FW −T =

1 0

when |HFS (n) − HFS (n − 1)| > Mth when |HFS (n) − HFS (n − 1)| ≤ Mth

(3)

Fig. 2. Frequency response of the comb filter.

3. Proposed filter based technique for short-circuit faults detection

224 225

The transfer function of a FIR comb filter with the length of N is [32]: H(z) =

N−1 

z −n

(6)

226 227

228

n=0

Eq. (6) can be rewritten as: H(z) =

1 − z −N = 1 − z −1



1 1 − z −1



229

(1 − z −N ) = GI (z) · GD (z)

(7)

where GI (z) and GD (z) are the transfer functions of the integration and differentiation parts of the comb filter, respectively. In fact, the integration and differentiation parts act as accumulator and differentiator, respectively. In this research, a FIR comb filter with the length of 8 and a sampling frequency of fs = 32 kHz is considered as following: 1  −n z , 8

230

231 232 233 234 235 236

7



1

when |HFT (n) − HFT (n − 1)| > Mth

0

when |HFT (n) − HFT (n − 1)| ≤ Mth

H(z) = (4)

and thus

It is clear that a short-circuit fault occurs in the transmission line when at least one of the three mentioned logic functions becomes “1”, so the final logic function that becomes “1” when a short-circuit fault occurs is defined as: L.FW = L.FW −R + L.FW −S + L.FW −T

(5)

where “+” is “OR” logic function. The final logic function needs a process time to detect a shortcircuit fault in a power transmission line. It is clear that the process time depends on the down sampling factor because the duration between the two components occurring in the detail coefficients, which are used by the proposed logic functions, strictly depends on the down sampling factor. The process time of the above DWT based method proposed in this work with the down sampling factor of two is maximally 1/10 TC , where TC is the period of the electric current transmitted through the power transmission line. The disturbance of the transmission line does not any effect on the process time.

(8)

237

n=0

H(z) =

1 − z −8

 1

1 · = · 8 1 − z −1 8

1 1 − z −1



238

(1 − z −8 )

(9)

By comparing Eq. (9) with Eq. (7), the transfer function of the accumulator is found as: 1 GI (z) = · 8



1 1 − z −1



(10)

and similarly, the transfer function of the differentiator is obtained as: GD (z) = 1 − z

−8

(11)

The frequency response of the proposed comb filter implemented in MATLAB software is shown in Fig. 2. It is clear that the proposed comb filter has following properties: • It acts as a FIR low pass filter. • It acts as three FIR band pass filters with the central frequencies of fo1 = 5.80 kHz, fo2 = 10.00 kHz and fo3 = 13.93 kHz.

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

239

240 241

242

243 244 245 246 247 248

249 250 251

G Model ASOC 3114 1–8 4

ARTICLE IN PRESS H. Fathabadi / Applied Soft Computing xxx (2015) xxx–xxx

Fig. 3. Proposed two-stage FIR filter implemented in MATLAB software.

252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289

The proposed two-stage FIR filter designed and implemented in MATLAB software is shown in Fig. 3. As will be shown, the proposed two-stage FIR filter itself alone detects short-circuit faults in a power transmission line. The first stage is the same FIR comb filter with the length of 8. The frequency responses of the FIR low pass filter (LBP) and the other three FIR band pass filters (BPFs) are shown in Fig. 4. In fact, the output of the low pass filter (AP[n]) gives the main body of the input signal x(t), so it is often called “approximation coefficient”. On the other hand, the outputs of the three band pass filters (HFD1 [n], HFD2 [n] and HFD3 [n]) give the high frequency details of the input signal, so they are called “detail coefficients”. To check the performance of the proposed two-stage FIR filter, a sinusoidal waveform with the frequency of 10 Hz distorted by a disturbance the magnitude of which is 10% of the signal magnitude is used as the input signal x(t). The waveforms of the filter outputs are shown in Fig. 5. AP[n] that is shown in Fig. 5(b) explicitly gives an approximation of x(t) while HFD1 [n], HFD2 [n] and HFD3 [n] shown in Fig. 5(c)–(e) give the high frequency details of x(t). More than 300 distinct experiments carried out in this study by simulating different power transmission lines verified that all types of the short-circuit faults produce a transient state in the short-circuited current, and then, the three important frequency components pulsated at the frequencies of fo1 = 5.80 kHz, fo2 = 10.00 kHz and fo3 = 13.93 kHz appear in the spectral characteristic of the short-circuited current. The three mentioned frequencies are fixed, and do not depend on the disturbance of the transmission line. In fact, when a short-circuit fault occurs in a transmission line, the short-circuited current suddenly increases, and therefore a transient state appears in the short-circuited current that produces three frequency components pulsated at the constant frequencies of fo1 = 5.80 kHz, fo2 = 10.00 kHz and fo3 = 13.93 kHz. These three frequencies only depend on the ratio of x/r, where x and r are, respectively, the reactance and resistance of the power transmission line. It is reminded that the ratio x/r is constant for a standard power transmission line, and does not change by variation in the length of the transmission line because the reactance and resistance both proportionally grow by increasing the length of the transmission line. Thus, the three mentioned

frequency components can be used to detect a short-circuit fault as follows. The three-phase current of the power transmission system are measured in Proteus 6 environment, and then, they are transmitted from Proteus 6 software to the two-stage FIR filter implemented in MATLAB software. The input data of the proposed two-stage FIR filter are sampled with the sampling frequency of 32 kHz, and then, three output discrete signals are produced for each current (HFD1C [n], HFD2C [n] and HFD3C [n]) by the FIR filter that are actually the detail coefficients. For detecting a fault, two sequential sampled values of the phase current are considered for assessment. If the three differences between the three detail coefficients of the two sequential sampled current obtained by the proposed two-stage FIR filter all are absolutely more than the threshold values, then a fault will be detected. For example, for phase R, the three pre-logic functions are defined as:

 L.F1R [n] =

1 when |HFD1CR [n]| − |HFD1CR [n − 1]| > C1st 0

 L.F2R [n] =

when |HFD1CR [n]| − |HFD1CR [n − 1]| ≤ C1st

1 when |HFD2CR [n]| − |HFD2CR [n − 1]| > C2nd 0

when |HFD2CR [n]| − |HFD2CR [n − 1]| ≤ C2nd

291 292 293 294 295 296 297 298 299 300 301 302 303 304

(12)

305

(13)

306

and

307

 L.F3R [n] =

290

1 when |HFD3CR [n]| − |HFD3CR [n − 1]| > C3rd 0

when |HFD3CR [n]| − |HFD3CR [n − 1]| ≤ C3rd

(14)

where HFD1CR , HFD2CR and HFD3CR are the three detail coefficients of the phase current R, and C1st , C2nd and C3rd are the threshold values of the three detail coefficients. The logic function for phase R is defined as: L.FR [n] = L.F1R [n] · L.F2R [n] · L.F3R [n]

(15)

where “·” is “AND” logic function. In other words, if all the pre-logic functions of phase R become “1”, then a short-circuit fault will be

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

308

309 310 311 312

313

314 315

G Model ASOC 3114 1–8

ARTICLE IN PRESS H. Fathabadi / Applied Soft Computing xxx (2015) xxx–xxx

5

Fig. 4. (a) Frequency response of FIR low pass filter (LPF). (b) Frequency response of FIR band pass filter 1 (BPF 1). (c) Frequency response of FIR band pass filter 2 (BPF 2). (d) Frequency response of FIR band pass filter 3 (BPF 3).

316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334

detected in phase R. For the two other phases (S and T), the logic functions are similarly defined as: L.FS [n] = L.F1S [n] · L.F2S [n] · L.F3S [n]

(16)

and L.FT [n] = L.F1T [n] · L.F2T [n] · L.F3T [n]

(17)

out to determine the best threshold values (C1st , C2nd and C3rd ) by simulating different power transmission lines. In each experiment, the threshold values were continuously changed to find the best values that give the minimum number of errors in fault detection. Finally, after passing 312 distinct experiments, the best threshold values that provide the highest sensitivity for the proposed fault detector were obtained as:

Now, the final logic function used in fault detection module to detect a short-circuit fault can be defined as:

⎧ C = 43 mA ⎪ ⎨ 1st

L.F[n] = L.FR [n] + L.FS [n] + L.FT [n]

⎪ ⎩

(18)

where “+” is “OR” logic function. It is clear that the final logic function needs two samples time to produce its binary output, the proposed filter itself needs less than one sample time to sample the three phase currents, and to produce the three detail coefficients of each phase current, and thus, the proposed two-stage FIR filter needs a very short process time about three samples time. Since the sampling frequency is 32 kHz, so the process time regardless of the disturbance is about 94 ␮s (tprocess = 3 × (1/32 kHz) ≈ 94 ␮s). To achieve the highest sensitivity of the proposed filter based fault detector, more than 300 distinct experiments were carried

C2nd = 68 mA

(19)

335 336 337 338 339 340 341

342

C3rd = 94 mA

4. Simulated results A power transmission system has been considered to simulate in Proteus 6/MATLAB environments. The simulated power system is a three-phase 230 kV, 50 Hz power system supplied from both sides by sources A and B. The power system includes a three-phase transmission line with the length of 50 km. The simulated power transmission system used in this study in which a short-circuit fault has happened between phase T and ground (T–G fault) is shown in Fig. 6. The sources A and B have same nominal power

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

343

344 345 346 347 348 349 350 351

G Model ASOC 3114 1–8

ARTICLE IN PRESS H. Fathabadi / Applied Soft Computing xxx (2015) xxx–xxx

6

Fig. 7. Three-phase current of the simulated power transmission system when a short-circuit fault occurs between phase R and ground (R–G fault).

Fig. 5. (a) Input signal of the filter (x(t)). (b) Filter output (AP[n]). (c) Filter output (HFD1 [n]). (d) Filter output (HFD2 [n]). (e) Filter output (HFD3 [n]).

Table 1 Parameters of the simulated power system. Transmission line positive and negative sequence impedance (˝/km) Transmission line zero sequence impedance (˝/km) Source positive sequence impedance () Source zero sequence impedance () Apparent power (MVA) Rated voltage (kV) Phase of source A (degree) Phase of source B (degree) Frequency (Hz) Transmission line length (km)

352 353 354 355 356 357 358 359 360 361

2.92 + j31.26 27.92 + j102.46 14.023 ∠ 85◦ 24.107 ∠ 85◦ 320 230 30 0 50 50

and the phase of the source A is 30 degree more than the source B. The parameters of the simulated power system are summarized in Table 1. To provide simulation verifications, the three-phase current waveform of the simulated power transmission system when a short-circuit fault occurs between phase R and ground (R–G fault) measured in Proteus 6 environment is shown in Fig. 7. The length of the short-circuit fault location is 30 km and it happens at 60 ms. The measured three-phase current is transmitted from Proteus 6 environment to MATLAB environment, and then, it is used as the input for the proposed two-stage FIR filter in which the sampling

frequency is 32 kHz. The three detail coefficients of each phase current produced as the filter outputs (HFD1C [n], HFD2C [n] and HFD3C [n]) are shown in Fig. 8. Since L . FR = 1, so L . F = 1, and thus, the short-circuit fault has been successfully detected using the proposed filter based technique. For the same short-circuit fault, the three logic functions L . FW−R , L . FW−S and L . FW−T obtained using DWT with the one-level mother wavelet db4 and Mth = 10 mA are shown in Fig. 9 that explicitly shows that the short-circuit fault has been successfully detected again because L . FW−R = 1, and so L . FW = 1. In a practical power transmission line, the disturbance always occurs in the transmission cables, and the usual disturbance is generally about 5–7% of the phase current magnitude. But in this research, to intensify the effect of the disturbance, and to evaluate the performance of the two proposed techniques, a white noise with a magnitude of 10% of the phase current magnitude was added to the three phase currents. Thus, in practical case, when one of the proposed approaches is implemented, the accuracy of fault detection is certainly equal or more than those obtained for the simulated cases. It is reminded that the white noise is a stochastic signal that its statistical mean and standard deviation are zero and one, respectively. 100 short-circuit faults were produced in all the phases (R, S and T) and different locations of the simulated transmission line by connecting the related phase cable to ground or other phases at different points of the transmission line. The obtained results of the short-circuit fault detection using the two proposed techniques in presence of 10% disturbance and in absence of disturbance are summarized in Table 2. The results reported in Table 2 show that the filter based approach which is a hardware technique has an accuracy of 100% which is more than the accuracy of the wavelet transform approach (maximally 97%) that is a soft computing

Fig. 6. Simulated power transmission system.

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392

G Model

ARTICLE IN PRESS

ASOC 3114 1–8

H. Fathabadi / Applied Soft Computing xxx (2015) xxx–xxx

7

Table 2 Detection accuracy of the two proposed approaches in presence of 10% disturbance and without disturbance. Approach

Filter based DWT with db4 DWT with db3 DWT with db2 DWT with db1

Faults number

100 100 100 100 100

True fault detection number

Detection accuracy

With 10% disturbance

Without disturbance

With 10% disturbance

Without disturbance

100 97 94 96 91

100 100 99 98 96

100% 97% 94% 96% 91%

100% 100% 99% 98% 96%

Table 3 Comparison between different methods for faults detection in a three-phase 230 kV, 50 Hz power transmission line. Reported works

[26] [27] [28] [29] [30] This work

393 394 395 396 397 398 399

Technique used in the work

PSO and ANN Fuzzy logic SVMs Elliptical 2-D structure Wavelet transform and linear discriminant analysis FIR filter DWT with db4

Process time

Detection accuracy

With 10% disturbance

Without disturbance

With 10% disturbance

Without disturbance

No reported No reported No reported No reported No reported

5 ms 5 ms 20 ms 28 ms 5 ms

No reported No reported No reported No reported No reported

99.91% 98% 98.703% 98.44% 100%

94 ␮s 2 ms

94 ␮s 2 ms

100% 97%

100% 100%

technique. On the other hand, DWT with the mother wavelet db4 itself has higher accuracy compared to the other mother wavelets (db1, db2 and db3). It is worthwhile to note that the results reported in Table 2 have been obtained with the threshold values of 43 mA, 68 mA and 94 mA for the filter based approach, and Mth = 10 mA for the wavelet transform approach, so the threshold value for the wavelet transform approach is even less than the threshold values

Fig. 8. Detail coefficients of the three-phase current produced by the proposed filter based technique.

for the filter based approach. On the other hand, although the filter based approach has higher accuracy but the wavelet transform approach has less complication and implementation cost. The simulated results reported in Table 3 compare the two proposed approaches with the five recent methods reported in [26–30]. In fact, the comparative results provide a comparison between the accuracy and process time of the different methods both in presence of 10% disturbance and in absence of disturbance. As mentioned, in a practical power transmission line, the disturbance is always available in the transmission cables, and the usual disturbance is generally about 5–7% of the phase current magnitude. It is worthwhile to note that the accuracy and process time of the five mentioned methods have been reported in absence of disturbance that is an ideal and impractical case, but the accuracy and process time of the two methods proposed in this study are presented both in presence of 10% disturbance and in absence of disturbance. The two methods were presented in this work

Fig. 9. Three logic functions L . FW−R , L . FW−S and L . FW−T obtained using DWT with the one-level mother wavelet db4 and Mth = 10 [mA].

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416

G Model ASOC 3114 1–8

H. Fathabadi / Applied Soft Computing xxx (2015) xxx–xxx

8 417 418 419 420 421 422 423 424 425 426 427

428

429 430 431 432 433 434 435 436 437 438 439 440 441 442 443

444

445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467

ARTICLE IN PRESS

(filter based and DWT based method) that each is a single-handed method. In other words, each method itself alone detects a shortcircuit fault in a power transmission line without any assistance such as using ANN, DFRs, etc., so it is logically expected that they have shorter process time compared to other combinational methods which use hybrid frameworks formed by combining different topics such as wavelet transform, ANN, fuzzy systems. The comparative results summarized in Table 3 explicitly verify that the two methods presented in this work have higher accuracy and shorter process time compared to the other methods, especially in presence of 10% disturbance that is the practical case. 5. Conclusion In this paper, two novel approaches that are based on discrete wavelet transform and two-stage FIR filter were presented to detect short-circuit faults in power transmission lines. The theoretical results were presented, and the accuracy of the two approaches were evaluated by simulating a three-phase 230 kV, 50 Hz power transmission line with the length of 50 km. The simulated results validated the theoretical results, and verified that the filter based approach has an accuracy of 100% with a process time about 94 ␮s, while the accuracy of the wavelet transform based approach is less than 98% in presence of 10% disturbance, and the process time is 2 ms. On the other hand, the wavelet transform based approach has less complication and implementation cost. Another comparative study showed that the two methods presented in this work have higher accuracy and shorter process time compared to the other methods, especially in presence of disturbance. References [1] W.R.M. Zaki, A.H. Nawawi, S.S. Ahmad, Economic assessment of Operational Energy reduction options in a house using Marginal Benefit and Marginal Cost: a case in Bangi, Malaysia, Energy Convers. Manag. 51 (2010) 538–545. [2] H. Fathabadi, Ultra high benefits system for electric energy saving and management of lighting energy in buildings, Energy Convers. Manag. 80 (2014) 543–549. [3] B.K. Bose, Modern Power Electronics and Drives, Prentice-Hall, 2007. [4] S. Messalti, A. Gherbi, S. Belkhiat, Artificial neural networks for assessment power system transient stability with TCVR, J. Electr. Eng. 13 (2013) 76–82. [5] H. Zhengyou, G. Shibin, C. Xiaoqin, B. Zhiqian, Q. Qingquan, Study of a new method for power system transients classification based on wavelet entropy and neural network, Int. J. Electr. Power Energy Syst. 33 (2011) 402–410. [6] Z. Qingchao, Z. Yao, S. Wennan, Transmission line fault location for singlephase-to-earth fault on nondirect-ground neutral system, IEEE Power Eng. Rev. 17 (1997) 40–41. [7] J.A. Morales, E. Orduna, C. Rehtanz, Classification of lightning stroke on transmission line using multi-resolution analysis and machine learning, Int. J. Electr. Power Energy Syst. 58 (2014) 19–31. [8] W.P. Ferreira, M.D.C.G. Silveira, A.P. Lotufo, C.R. Minussi, Transient stability analysis of electric energy systems via a fuzzy ART-ARTMAP neural network, Electr. Power Syst. Res. 76 (2006) 466–475. [9] C. Buccella, A. Orlandi, Diagnosing transmission line termination faults by means of wavelet based crosstalk signature recognition, IEEE Trans. Compon. Packag. Technol. 23 (2000) 165–170.

[10] W.M. Lin, C.D. Yang, J.H. Lin, M.T. Tsay, A fault classification method by RBF neural network with OLS learning procedure, IEEE Trans. Power Deliv. 16 (2001) 473–477. [11] P.P. Preˇseren, B. Stopar, Wavelet Neural Network employment for continuous GNSS orbit function construction: application for the assisted-GNSS principle, Appl. Soft Comput. 13 (5) (2013) 2526–2536. [12] N.A. Al-geelani, M.A.M. Piah, Z. Adzis, M.A. Algeelani, Hybrid regrouping PSO based wavelet neural networks for characterization of acoustic signals due to surface discharges on H.V. glass insulators, Appl. Soft Comput. 13 (12) (2013) 4622–4632. [13] N.A. Al-geelani, M.A.M. Piah, R.Q. Shaddad, Characterization of acoustic signals due to surface discharges on H.V. glass insulators using wavelet radial basis function neural networks, Appl. Soft Comput. 12 (4) (2012) 1239–1246. [14] F. Palmieri, U. Fiore, A. Castiglione, A.D. Santis, On the detection of card-sharing traffic through wavelet analysis and Support Vector Machines, Appl. Soft Comput. 13 (1) (2013) 615–627. [15] J. Sanz, R. Perera, C. Huerta, Gear dynamics monitoring using discrete wavelet transformation and multi-layer perceptron neural networks, Appl. Soft Comput. 12 (9) (2012) 2867–2878. [16] S. Srivastava, M. Singh, M. Hanmandlu, A.N. Jha, New fuzzy wavelet neural networks for system identification and control, Appl. Soft Comput. 6 (1) (2005) 1–17. [17] M. Singh, S. Srivastava, M. Hanmandlu, J.R.P. Gupta, Type-2 fuzzy wavelet networks (T2FWN) for system identification using fuzzy differential and Lyapunov stability algorithm, Appl. Soft Comput. 9 (3) (2009) 977–989. [18] B.U. Shankar, S.K. Meher, A. Ghosh, Wavelet-fuzzy hybridization: featureextraction and land-cover classification of remote sensing images, Appl. Soft Comput. 11 (3) (2011) 2999–3011. [19] O. Akyilmaz, H. Kutterer, C.K. Shum, T. Ayan, Fuzzy-wavelet based prediction of Earth rotation parameters, Appl. Soft Comput. 11 (1) (2011) 837–841. [20] F. Martín, J.A. Aguado, Wavelet-based ANN approach for transmission line protection, IEEE Trans. Power Deliv. 18 (2003) 1572–1574. [21] R. Cöteli, A combined protective scheme for fault classification and identification of faulty section in series compensated transmission lines, Turk. J. Electr. Eng. Comput. Sci. 21 (2013) 1842–1856. [22] J. Morais, Y. Pires, C. Cardoso Jr., A. Klautau, A framework for evaluating automatic classification of underlying causes of disturbances and its application to short-circuit faults, IEEE Trans. Power Deliv. 25 (2010) 2083–2094. [23] S. Ekici, S. Yildirim, M. Poyraz, A transmission line fault locator based on Elman recurrent networks, Appl. Soft Comput. 9 (2009) 341–347. [24] A.L. Orille Fernandez, N.K.I. Ghonaim, A novel approach using a FIRANN for fault detection and direction estimation for high-voltage transmission lines, IEEE Trans. Power Deliv. 17 (2002) 894–900. [25] Y. Pires, J. Morais, C. Cardoso, A. Klautau, Data mining applied to the electric power industry: classification of short-circuit faults in transmission lines, Stud. Comput. Intell. 169 (2009) 107–122. [26] J. Upendar, C.P. Gupta, G.K. Singh, G. Ramakrishna, PSO and ANN based fault classification for protective relaying, IET Gener. Transm. Distrib. 4 (10) (2010) 1197–1212. [27] S.R. Samantaray, A systematic fuzzy rule based approach for fault classification in transmission lines, Appl. Soft Comput. 13 (2013) 928–938. [28] U.B. Parikh, B. Das, R. Maheshwari, Fault classification technique for series compensated transmission line using support vector machine, Int. J. Electr. Power Energy Syst. 32 (6) (2010) 629–636. [29] A.S. Gomes, M.A. Costa, T.G.A. Faria, W.M. Caminhas, Detection and classification of faults in power transmission lines using functional analysis and computational intelligence, IEEE Trans. Power Deliv. 28 (3) (2013) 1402–1413. [30] A. Yadav, A. Swetapadma, A novel transmission line relaying scheme for fault detection and classification using wavelet transform and linear discriminant analysis, Ain Shams Eng. J. 6 (1) (2015) 199–209. [31] A. Ukila, R. Zivanovicb, Abrupt change detection in power system fault analysis using adaptive whitening filter and wavelet transform, Electr. Power Syst. Res. 76 (9) (2006) 815–823. [32] X. Liang, S. Liu, Y. Li, Z. Liu, S. Jian, Characteristics of a high extinction ratio comb-filter based on LP01–LP11 even mode elliptical multilayer-core fibers, Opt. Fiber Technol. 21 (2015) 103–109.

Please cite this article in press as: H. Fathabadi, Two novel proposed discrete wavelet transform and filter based approaches for short-circuit faults detection in power transmission lines, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.07.039

468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533