Recognition of Partial Discharge in oil-paper insulation based on Analytic Hierarchy Process

Abstract

Partial Discharge (PD) is the main cause of insulation deterioration in oil-paper insulation which is the common structure in power transformers. PD signals contain sufficient information of insulation status and thus PD measurement becomes very important to early detection or even prevention of insulation failure. The type of PD can provide even more information about the PD defects and the analysis of them can provide effective guidance for taking further measurements. PD in different kinds of defects involves different discharge mechanism, which can be reflected by the difference of Phase Resolved Partial Discharges (PRPD) pattern and parameters. The differences provide the basis for the research of pattern recognition. In order to recognize PD, five kinds of typical defect models were prepared to simulate the typical actual defects of oil immersed transformer in the thesis. Through K-W test, 11 features with the strongest ability of classification are chosen from statistical operators of PRPD pattern. Based on the chosen features, in the case of being trained by small sample, Analytic Hierarchy Process (AHP) is applied to recognize these typical PD, and the results recognized by AHP are compared to those which are recognized by Artificial Neural Network (ANN) in the same condition. The results shows that the recognition accuracy rates gotten by AHP are all above 85%, are better than which is gotten by ANN, this lays a foundation for the typical Partial Discharges recognition under the premise of small sample.