Electrical Transmission Line Fault Detection and Classification using Convolution Neural Networks and Support Vector Machine

Abstract

The detection and classification of mistaken conditions in electrical power transmission and distribution systems is a task of importance for trustworthy activity. In recent years, methods based on performance in relay security and performance of electrical components has become familiar. Moreover, they have fault in dealing with power system insufficiency. This paper proposes a new technique for the prediction of data-based transmission line faults in power systems using Convolution Neural Networks (CNN) networks and support vector machine (SVM). The chronological characteristics of multi-sourced data were taken into custody with CNN networks, which achieve well in extorting the features of time series for a long-time period. The powerful learning and mining capability of CNN networks is appropriate in power transmission and distribution for a heavy quantity of time series. SVM is introduced for classification in order to obtain the final predictive results, with a strapping generalization capability and sturdiness. The method tested on the VSB dataset and practical detection performance is attained.