Pattern Recognition Method of UHF PD Envelope Signal in GIS Based on Rough Set Reduction

Qing LI,Da-peng DUAN,Wu-bin QIU,Kai-sui CAI,Ge-hao SHENG,Xing-quan HUANG,Xiu-chen JIANG

DOI: https://doi.org/10.13296/j.1001-1609.hva.2012.03.003

2012-01-01

Abstract:Determining the relationship between partial discharge (PD) type and PD envelope signal is an important premise to assess the insulation state of gas-insulated metal-enclosed switchgear (GIS) and to make rational maintenance strategy. The matrix of 37 feature vectors, with which a UHF PD envelope signal is characterized, is discretized, and its difference matrix is reduced with the dimension reduction rough set theory. Then the reduced feature vectors combining with the BP neural network classifier are applied to pattern recognition of four different types of UHF PD envelope signals. The results show a high recognition rate of the proposed method.