An Optimised Fault Classification Technique Based on Support-Vector-Machines

As a new general machine-learning tool based on structural risk minimization principle, Support Vector Machines (SVMs) have the advantageous characteristic of good generalization. For this reason, the application of SVMs in fault classification and diagnosis field has becomes one growing reach focus. In this paper a new approach to real-time fault detection and classification is presented for high speed protective relaying in power transmission systems using SVMs. The integration with an online wavelet-based pre-processing stage [1,2,3] enhances the SVM learning capability and classification power. The classification criterion is based on using only the phase angles between the three line currents in the transmission line. The paper begins with the exploration of classifying different fault types (LG, LL, and LLG) using the SVMs. It proceeds with the classification concepts of the nine types of faults. Extensive theoretical studies and simulations using ATP and MATLAB-SVM Toolbox on an EHV transmission line model have proved that the veracity of the SVM classifier is very significant for fault classification.