Abstract—In this paper a novel method for finding the fault zone on a Thyristor Controlled Series Capacitor (TCSC) incorporated transmission line is presented. The method makes use of the Support Vector Machine (SVM), used in the classification mode to distinguish between the zones, before or after the TCSC. The use of Discrete Wavelet Transform is made to prepare the features which would be given as the input to the SVM. This method was tested on a 400 kV, 50 Hz, 300 Km transmission line and the results were highly accurate. Keywords—Flexible ac transmission system (FACTS), thyristor controlled series-capacitor (TCSC), discrete wavelet transforms (DWT), support vector machine (SVM). I. INTRODUCTION HE use of Flexible AC Transmission System (FACTS) devices has become quite popular in transmission systems. The Thyristor Controlled Series-Capacitor (TCSC) is one such device that has found usage in the transmission lines as it has enabled optimum use of the transmission line by allowing features like power control, dynamic compensation etc. The TCSC consists of a series capacitor in parallel with a series combination of a reactor and an anti-parallel connection of thyristors along with a protection feature in the form of a Metal Oxide Varistor (MOV). The MOV provides over- voltage protection to the series capacitor. However, the introduction of the TCSC has caused the protection schemes of transmission lines using traditional methods, to be difficult because they introduce sudden changes in the apparent impedance of the transmission line [1],inversion of voltage and current signals etc. [2,3]. Since, protection schemes, existing now, determine the state of the system through the voltage and current waveforms, the fault analysis has now essentially become a pattern recognition problem. Different kinds of approaches have been proposed to find out the faulty zone. [4] uses the travelling wave approach for the protection of the series compensated line, [5] and [6] make use of the SVM and a combination of the SVM and DWT respectively and [7] make use of a scheme called the Decision Tree (DT) to find out the faulty zone. Renju Gangadharan (renjumace@gmail.com), G.N. Pillai (gnathfee@iitr.ernet.in) and Indra Gupta (indrafee@iitr.ernet.in) are all with the Dept. of Electrical Engg. at the Indian Institute of Technology Roorkee, Roorkee-247667, India. The above mentioned works, however, do not produce satisfactory results for section identification with a maximum overall accuracy of 93.917% has been reported in [6]. The method proposed here also uses both the SVM and the DWT, but a thorough analysis was done to find out the best features that can aid the process. The pre-requisite for the proposed method is that the type of fault viz. a-g, b-g, c-g, a- b-g, a-c-g, b-c-g, a-b-c-g, a-b, a-c and b-c must be found. A data window of 2 cycles corresponding to 200 samples was taken for the analysis. . The idea of using 2 cycles’ information is to ensure that all the possible test cases can be incorporated. The data obtained after processing was given to the SVM for classification. The SVM algorithm classifies the data into two classes i.e. SVMs solve for two class problems. The section identification is for a particular type of fault, hence, a total of 10 SVMs are needed to determine the faulty section. In this paper, the second section gives a brief insight into the wavelet transforms. The third section gives an account of the classification using SVM. The fourth section describes the system studied and the simulation process. The fifth section presents the feature extraction process. The sixth section describes the results and the last section concludes the work II. DISCRETE WAVELET TRANSFORM Before discussing about DWT, a brief introduction of continuous wavelet transform or simply the wavelet transforms should be made. The wavelet transforms are realised by two sets of functions called the scaling function and the wavelet function. The expressions for these are, 2 00 2 (2 ) j j jk t k ϕ ϕ = − (1) 2 00 2 (2 ) j j jk t k ψ ψ = − (2) where φ represents the scaling function and ψ represents the wavelet function. φ 00 represents the basis function for the scaling function and similarly, ψ 00 represents the basis function for the wavelet function. It is also called the mother wavelet. The parameter j represents the resolution level and the parameter k is called the shifting parameter. The parameter k ranges from 0 to 2 j-1 while the parameter j starts at 0 and ideally goes to infinity but, in practice it is limited by the sampling rate. Hence, the scaling and the wavelet functions are the scaled and shifted versions of their corresponding basis functions. However, the two scale relation in the wavelet theory helps in defining the wavelet and the scaling function at a particular resolution level purely in terms of the scaling Renju Gangadharan, Graduate Student Member, IEEE , G. N. Pillai, and Indra Gupta Fault Zone Detection on Advanced Series Compensated Transmission Line using Discrete Wavelet Transform and SVM T World Academy of Science, Engineering and Technology International Journal of Electrical and Computer Engineering Vol:4, No:9, 2010 1331 International Scholarly and Scientific Research & Innovation 4(9) 2010 scholar.waset.org/1307-6892/10590 International Science Index, Electrical and Computer Engineering Vol:4, No:9, 2010 waset.org/Publication/10590