Application of S-Transform for Detection of High Impedance Faults in Power System Network

Application of S-Transform for Detection of High Impedance Faults in Power System Network

Nabamita Banerjee Roy(1st Author) Associate Professor, Electrical Engineering, Narula Institute of Technology

Kolkata, India

Kushal Kumar (2nd Author)

Student, M. Tech, 2nd Year, Electrical Engineering, Narula Institute of Technology

Kolkata, India

Abstract Detection of High Impedance Faults mainly in distribution systems is a troublesome and challenging task that has caught researchers attraction over past years. The random behavior of HIF current, as well as its low magnitude, causes difficulties in the detection of faults by conventional fault detection methods. Conventional distance relays, overcurrent relays, and ground fault relays cannot be used for HIF detection due to its sensitivity and selectivity issue owing to the low value of fault current. There are several Signal processing techniques available for HIF detection such as Fourier transforms, wavelet transforms. In this paper, S-Transform has been used for feature extraction from the voltage signals simulated under different fault conditions in a power system network. S -Transform has the combined features of FFT and Wavelet Transform and is very popular in analysis of transient signals. High Impedance Faults for different values of fault resistances and fault locations have been simulated. The effect of noise on the features extracted has been also studied. All the simulations have been done in MATLAB.

Keywords Feature extraction, High Impedance Fault (HIF), S-Transform.

1. INTRODUCTION

   Fast and accurate detection of faults allows relays to isolate the faulty part from the system which is essential for the protection of equipment and maintain continuity of power supply to healthy part of system. Additionally, accurate fault classification also provides significant information regarding the location of fault that accelerates up the repair works. High Impedance faults mainly occur in transmission and distribution lines and represent a persistent issue in power system. HIF occurs when an overhead line makes contact with a tree branch or when a broken conductor makes contact with ground or a crane [1]. This fault is characterized by having an impedance that is sufficiently high and remains undetected by common protection methods such as overcurrent relays fuses etc. The arcing behaviour of High Impedance fault causes a random change in effective impedance and the magnitude of fault current becomes uncontrolled in nature. Variations in the 50Hz and harmonic components occur in conjunction with the HIF. Due to the dynamics of fault arc, these variations are not stationary, but rather time varying [2]. The fault current in HIF might range from zero to overcurrent relays pickup setting. As a result, HIF will not be detected by overcurrent relays. Energized conductors may fall within the reach of employees if this type of fault occurs. Furthermore, because arcing frequently occurs with these faults, it provides a fire threat. As a result, detecting HIF is vital from standpoint of both public safety and operational reliability. The methods of detecting, classifying and locating faults in power transmission networks

   have been thoroughly studied. Efforts are underway to develop an intelligent protection system capable of properly detecting, classifying and localizing defects. Researchers have been able to adopt a more complete and dedicated approach in studies related with conventional fault protection solutions thanks to advancements in signal processing techniques, artificial intelligence and machine learning. Furthermore, two well- known shortcomings of online mechanism of fault detection are addressed. The difficulty in acquiring the necessary data is the first limitation. To overcome this, Intelligent Electronic Devices (IED) are used to gather information at different buses in grids [3]-[4]. These technological improvements ensure that online monitoring methods based on sensor networks respond quickly to problematic events and that they function properly. Fault detection based on feature extraction using signal processing techniques has been established to be fast, accurate and convenient. There are several signal processing tools like Fast Fourier Transform (FFT), Wavelet Transform (WT) and S-Transform (ST). WT has been implemented in [5]-[8] for fault detection in transmission in which different kinds of signal features have been extracted for fault classification. Different versions of FFT have been employed in [9]-[12] for fault classification in transmission lines. In the present paper, ST has been used for feature extraction from the voltage signals under different fault conditions. An overview of ST has been given in section II. The simulation of the power system network and the different fault conditions have been explained in section III. The method of feature extraction using ST and the effect of noise on the magnitude of the features has been discussed in section IV. Results and discussion have been given in section V. Conclusion and future scope of work have been provided in section VI.

2. OVERVIEW OF S-TRANSFORM

   The S-transform (ST) can be obtained as a short-time Fourier transform (STFT) with a frequency-dependent window width, or as a phase correction to the wavelet transform. When compared to the STFT and the WT, ST has certain distinct advantages. The S-transform can be used to get local spectral properties, which is important for accelerating transients [13]- [14]. The ST is utilized to alleviate the drawbacks of the DWT, such as its sensitivity to noise and inability to precisely show harmonic features [15]-[16]. Defective lines and sections were chosen using S-transform contour energy and standard deviation, respectively [17]. ST was utilized to determine the amplitude, impedance, and fault sites by S. Samantaray et al. [18]. Discrete ST of a time series generates a complex matrix(S-matrix). The rows of S-matrix are the frequencies and the columns are the corresponding times and each column

   represents the local spectrum at one point in time. S-Transform of a discrete time series h (k, T) is given by equation (1).

   (1)

   N=Signal Length T=Sampling Time interval

   where j, m=0,1,2, . N-1 and n=1, 2, . N-1

3. SIMULATION OF POWER SYSTEM NETWORK AND HIGH IMPEDANCE FAULTS

   1. Simulation of power system network for study

      A 400 kV. 50Hz three phase system is shown in Fig.1. In this system, B1 and B2 are sending end and receiving end buses connected by a three-phase transmission line of length 300Km.

      Fig. 1. Single Line diagram of 400 kV, 50 Hz, 3-phase power system

      network

      TABLE I. SYSTEM PARAMETERS

      System Components	Specifications
      Generator	Impedance = (0.2+j4.49) , X/R ratio = 22.45.
      Transmission Line	Length: 300 Km, R1 = 0.02336/km, R2 = 0.02336/km, R0 = 0.38848/km, L1 = 0.95106mH/km, L2 = 0.95106mH/km, L0 = 3.25083mH/km, C1 = 12.37nF/km, C2 = 12.37nF/km, C0 = 8.45 nF/km
      Balanced Load	Load Impedance = (720+j11) , p.f.= 0.9, MVA rating = 200

      The parameters of the system are given in TABLE I. The total time period of simulation in MATLAB is 0.04 sec.

   2. Simulation of power system faults

      The following ten types of faults have been simulated in steps of 10 Km from sending end (B1) of power system network as shown in Fig 1.

      1. L-G Faults: AG, BG, CG.

         /li>

      2. L-L Faults: AB, BC, CA.

      3. L-L-G Faults: ABG, BCG, CAG.

      4. L-L-L Faults: ABC.

   As it is required to simulate the High Impedance Fault, the fault resistance has been considered to be 200 ohms, 500 ohms, and 1000 ohms. The voltage waveforms for two types of fault conditions have been shown in Fig. 2.

   (a)

   (b)

   Fig. 2. Plots of Three phase voltage waveform during (a) AG type (b) AB type of fault occurring at 100 Km from B1 in the power system network with Fault resistance=500ohms

4. APPLICATION OF S-TRANSFORM IN ANALYSIS OF HIGH IMPEDANCE FAULTS

   1. Selection of features

      In the present study, S-Transform is applied on voltage signal for each phase. After a signal is processed by S- Transform, suitable features are needed to be extracted. A successful outcome depends on the nature of features that are extracted. The choice of features should be done in such a way that the desired output is obtained with acceptable accuracy.

      S-Transform is implemented on voltage signal of each phase. The output of S-Transform of each voltage signal of each phase is a complex matrix, S-matrix. The size of S-matrix is 257×512 where as the number of samples for each voltage signal is 512. The absolute value of S-matrix of each phase is obtained. The maximum value of each row of absolute matrix is obtained and the size of matrix is 1×257(row matrix).

      The magnitudes of the absolute value of S-matrix for double line fault condition (AB fault) has been plotted and shown in Fig. 3.

      Fig. 3. Magnitudes of the absolute value of S-matrix with respect to sampled frequencies for phase A in case of AB type of fault occurring at 100 km. from B1 in the power system network with Fault resistance = 500 ohms.

      Finally, the maximum value of row matrix is found which is a single numeric value, and this process is repeated for voltage signal of each phase during no fault conditions as well as all types of fault conditions simulated in this work. Hence the table of features is generated for all the fault conditions and the plots of the features of four types of fault conditions are shown in Fig.4. to Fig.5.

      It is to be noted that the average value of this feature during no fault signal is 0.45.

      The features that have been obtained finally for different fault conditions and fault locations are shown below in Fig. 4. to Fig. 5.

      (a)

      (b)

      Fig. 4. Plot of magnitude of features with respect to fault location in case of

      (a) AG fault (b) AB Fault occurring at 100 km from B1 in the power system network with fault resistance = 500 ohms

      (a)

      (b)

      Fig. 5. Plot of magnitude of features with respect to fault location in case of

      (a) ABG fault (b) ABC Fault occurring at 100 km from B1 in the power system network with fault resistance = 500 ohms

   2. Effect of noise on voltage signals

   All the voltage signals under different fault conditions have been implemented with 20 dB White noise and once again features are calculated from those signals using S-Transform. The magnitudes of the features in presence of noise have been plotted for different fault locations and fault conditions. They have been compared with those obtained without noise. All the plots are shown in Fig.6. to Fig.7.

   (a)

   (b)

   Fig. 6. Plot of magnitudes of features with respect to fault location in case of

   (a) AG type (b) AB type of fault occurring at 100 km from B1 in the power system network with fault resistance = 500 ohms in presence of noise

   (a)

   (b)

   Fig. 7. Plot of magnitudes of features with respect to fault location in case of

   (a) ABG type (b) ABC type of fault occurring at 100 km from B1 in the power system network with fault resistance = 500 ohms in presence of noise

5. RESULTS AND DISCUSSIONS

   The following observations have been made from the profile of features shown in Fig.4. and Fig.5.

   TABLE II. COMPARISON OF THE MAGNITUDES OF FEATURES OF THREE PHASES DURING DIFFERENT TYPES OF FAULT CONDITIONS WITH THOSE OF PHASES DURING NORMAL CONDITIONS WITHOUT NOISE

   Nature of fault	Magnitude of feature in comparison to the feature of the voltagesignal during no fault condition
   	Phase A	Phase B	Phase C
   AG	Lowest	Higher than normal value	Higher than normal value
   BG	Higher than normal value	Lowest	Higher than normal value
   CG	Higher than normal value	Higher than normal value	Lowest
   AB	Lowest	Lowest	Higher than normal value
   BC	Higher than normal value	Lowest	Lowest
   CA	Lowest	Higher than normal value	Lowest
   ABG	Lower than normal value	Lower than normal value	Higher than normal value
   BCG	Higher than normal value	Lower than normal value	Lower than normal value
   CAG	Lower than normal value	Higher than normal value	Lower than normal value
   ABC	Lower than normal value	Lower than normal value	Lower than normal value

   It is observed from the Fig.6. to Fig.7. that in presence of noise the magnitudes of the features do not change under the same fault conditions and hence all the fault conditions can be accurately identified.

6. CONCLUSION AND FUTURE SCOPE OF WORK In the present work, different types of high impedance

faults have been simulated with fault resistance 500 ohms. Suitable features have been selected from the voltage signals by using S-Transform under no fault and different fault conditions. The features have been thoroughly studied and it has been observed that the fault conditions can be accurately identified. Also, the fault conditions have been simulated by

incorporating 20 dB white noise and the features have been obtained from those signals using ST. It has been observed that the magnitudes of the features remain almost unaffected in presence of noise. Hence, it can be concluded that by using these features obtained from S-Transform all the types of faults can be accurately identified even in presence of noise.

In the future scope of work, the signal features that have been obtained would be used to train a suitable classifier involving neural network or support vector machine so that fault can be notified automatically under different conditions. The study can be further extended by considering other values of fault resistances like 200-ohms, 1000 ohms etc. Also, the study can be implemented in other different kinds of power system networks involving a greater number of buses and transmission lines.

REFERENCES

1. A. V. Mamishev, B. D. Russell, and C. L. Benner, Analysis of high impedance faults using fractal techniques. IEEE Transactions on Power Systems, vol. 11, no. 1, pp. 435-440, 1996.

2. High impedance fault detection technology, Rep. Power Syst. Relaying Committee (PSRC) Working Group D15, Mar. 1996.

3. Chen, K.; Huang, C.; He, J. Fault detection, classification and location for transmission lines and distribution systems: A review on the methods. IET High Volt. 2016.

4. Kezunovic, M. Smart fault location for smart grids. IEEE Trans. Smart Grid 2011.

5. Youssef, Omar A.S., 2001. Fault classification based on wavelet transforms. Transmission and Distribution Conference and Exposition IEEE/PES 1, 53536.

6. El Safty, S., El-Zonkoly, A., 2009. Applying wavelet entropy principle in fault classification. Electr. Power Energy Syst., 604607, Elsevier.

7. Ashok, V., Bangarraju, K.G.V.S., Murthy, V.V.N., 2013. Identification and classification of transmission line faults using wavelet analysis. ITSI Trans. Electr. Electron. Eng. 1 (1), 117122.

8. Dileep Kumar, A., Raghunath Sagar, S., 2014. Discrimination of faults and their location identification on a high voltage transmission lines using the discrete wavelet transform. Int. J. Educ. Appl. Res. 4 (January June (1)), 107111.

9. Jose, Prince, Bindu, V.R., 2014. Wavelet-based transmission line fault analysis. Int. J. Eng. Innov. Technol. 3 (February (8)), 5560.

10. Yu, S.; Gu, J.C. Removal of decaying DC in current and voltage signals using a modified Fourier filter algorithm. IEEE Trans. Power Deliv. 2001, 16, 372379.

11. Jamehbozorg, A.; Shahrtash, S.M. A decision-tree-based method for fault classification in single-circuit transmission lines. IEEE Trans. Power Deliv. 2010, 25, 21902196. [CrossRef].

12. Hagh, M.T.; Razi, K.; Taghizadeh, H. Fault classification and location of power transmission lines using artificial neural network. In Proceedings of the 2007 International Power Engineering Conference (IPEC), Singapore, 36 December 2007; pp. 11091114.

13. S. Mallat, A Wavelet Tour of Signal Processing. London, Ed. New York: Academic, 1998.

14. Stockwell, R.G.; Mansinha, L.; Lowe, R.P. Localization of the complex spectrum: The S transform. IEEE Trans. Signal Process. 1996, 44, 998 1001.

15. Dash, P.K.; Das, S.; Moirangthem, J. Distance protection of shunt compensated transmission line using a sparse S-transform. IET Gener. Transm. Distrib. 2015, 9, 12641274.

16. Moravej, Z.; Pazoki, M.; Khederzadeh, M. New pattern-recognition method for fault analysis in transmission line with UPFC. IEEE Trans. Power Deliv. 2015, 30, 12311242.

17. Nabamita, R.; Kesab, B. Detection, Classification, and Estimation of Fault Location on an Overhead Transmission Line Using S-transform and Neural Network. Electr. Power Components Syst. 2015.

18. Samantaray, S.R.; Dash, P.K. Transmission line distance relaying using a variable window short-time Fourier transform. Electron. Power Syst. Res. 2008, 78, 595604.

AUTHOR

Dr. Nabamita Banerjee Roy is presently Associate Professor in the Electrical Engineering Department of Narula Institute of Technology, Kolkata, India. She has obtained her graduation in Electrical Engineering from B.E. College, Shibpur, Howrah, West Bengal (presently IIEST Shibpur) in 2002. She has obtained both M.E.E and PhD from Jadavpur university, Kolkata in 2004 and 2017 respectively. She has a rich and diverse academic career as a faculty in Electrical Engineering Degree level and as an administrator (acting principal) at the diploma engineering level, since 2004. She has also served as HOD, Electrical Engineering Department at Narula Institute of Technology. She has supervised many B.Tech and M.Tech projects. She has published papers in National/International conferences and journals along with a Book Chapter in Springer. Her areas of research include signal processing, power system faults, neural network, soft computation, and high voltage engineering.

Kushal Kumar is presently final year student of M. Tech in Power Systems at Narula Institute of Technology. He has completed his M. Tech project under the supervision of Dr. Nabamita Banerjee Roy. He is enthusiastic to pursue further research work in the domain of signal processing, power system faults, neural network and soft computation