Long Short Term Memory Based Deep Learning Method for Fault Power Line Detection in a MV Overhead Lines with Covered Conductors

Abstract

The progress in computation power has motivated neural network based solutions in a diverse area of study. Partial discharges (PD) detection is a very old challenge for power operators. An efficient in time detection of PD helps in reliable power supply to the cities and secure power infrastructure. This paper proposes a PD pattern analysis for fault power line detection using Long Short Term based deep learning method in a Medium Voltage(MV) Overhead Covered Conductor(OCC). The method is tested on the real dataset provided by ENET centre VSB, which is the largest known available dataset in open source. For the training and validation, k-fold stratified cross validation method is used. The result of the proposed method is compared with the Seasonal and Trend decomposition using Loess (STL) and Support Vector Machine (SVM) based method. The LSTM method outperformed the traditional classifier SVM in the detection of a faulty power line.