Determination of Type of Partial Discharge in Cubicle-Type Gas Insulated Switchgear (C-GIS) using Artificial Neural Network

Abstract

Partial discharge (PD) measurement is performed as one of diagnostic tool for condition monitoring in cubicle-type gas insulated switchgear (C-GIS). This paper presents artificial neural network (ANN) to identify the type of PD occurring in C-GIS. PD measurement results obtained from C-GIS in a field were fed to the ANN as input parameters to determine the type of PD or external noises. The C-GIS consists of two compartments of power cables and two compartments of transformers. PD measurements were conducted in the C-GIS using a commercial PD measurement system consisting of an oscilloscope and three different kinds of sensors (transient earth voltage sensor, surface current sensor and high frequency current transformer) to generate input signal waveform parameters for ANN (field data). An attempt was made to locate the PD source using the output of the sensors set at different sites. PD measurements using four kinds of artificial PD sources were also conducted in laboratory using the same PD measurement system to obtain another signal waveform parameters (laboratory data) which were used for training the ANN. Thereafter, an attempt was also made to identify PD source occurring in C-GIS using field data as input for the trained ANN. As a result, the developed ANN was found to predict the kind of PD source as void-discharge with 99% probability.