Wavelet Based Transmission Line Protection Scheme Using Centroid Difference and Support Vector Regression

This paper presents a Wavelet based transmission line protection algorithm which uses centroid difference for fault detection and support vector regression for the fault location. The sample of three phase currents signals of both the terminals of the line are synchronized and decomposed with Wavelet Transform to obtain the absolute values of approximate coefficients over moving window of a cycle. For the decomposition of current signal dbl mother wavelet is used. Two centroid at each cycle is computed using k-means clustering. The centroid difference is computed at both terminal added to obtain fault index. The fault index is compared with a threshold to detect the faulty phase and to classify the fault. The same approximate coefficients of post fault current transient obtain over half a cycle are used for the estimation of fault location with the help of support vector regression. A large number of case studies involving changes in fault impedance, fault incidence angle and fault location have been conducted to establish the performance of the proposed algorithm.