인공 신경망을 이용한 가스절연 변압기의 부분방전 패턴인식 연구

Abstract

Partial Discharge(PD) phenomena occurred by different nature of insulating defects regarded as a random process by which Phase Resolved Partial Discharge Analysis(PRPDA) has been proposed and then commercially accepted for the diagnosis of the power apparatus since more than three decades. In this work, for the diagnosis of Gas Insulated Transformer(GITr), four different types of specimen were fabricated as a model of the possible defects that might possibly cause its sudden failures such as turn to turn insulation, inter coil insulation, free moving ball particle and positive protrusion. For this purpose, UHF coupler has been designed and fabricated to detect the partial discharges produced from the above defects introduced into the Gas Insulated Transformer(GITr) mock-up and experimental investigations have been carried out in order to analyze the related Partial Discharge(PD) patterns by means of both Phase Resolved Partial Discharge Analysis(PRPDA) and Artificial Neural Network respectively and then their comparisons are made systematically. For the pattern learning, database was established by use of self-designed defects which are reported can be occurred in GITr as for the input of neural network. Database was analyzed to distinguish patterns by means of PRPDA (Phase Resolved Partial Discharge Analysis) and stored to the form with to unite the average amplitude of PD pulse and the number of PD pulse.