Partial Discharge defects recognition using different Neural Network Model in XLPE cable under the DC stress

Abstract

There are many advantages and applications of DC-XLPE cables under DC stress, it is of paramount importance that it is required to have some reliable diagnosis system to check the insulation state of cable insulation in order to avoid the unpredictable service failures of the cable system. Thus, it has been conceived that partial discharge diagnosis is one of the major tools for this purpose. However, very few research works have been reported regarding the PD diagnosis on XLPE cable under the DC stress.
By keeping this fact in mind in the current work, PD detection has been carried out using artificial defects introducible into DC-XLPE cable system and then PD signals have been analyzed by use of Chaotic Analysis of Partial discharge (CAPD). Afterwards, the application of the artificial neural network has been done in order to improve the recognition rate of the PD defects by adding power spectrum data for the first time in this concerned area. In this method, the power spectrum data of the PD signal is combined with CAPD data as the training data for artificial neural networks models. And then different NN techniques have been applied for the recognition of PD defects by using CAPD data and CAPD data combined with power spectrum data. As a result, better recognition rate as well as low mean square loss by using CAPD data combined with Power spectral data, also the MLP techniques has shown best results among all other NN networks.