Pattern identification method of partial discharge based on the features of UHF envelope signals

To determine the relationship between partial discharge type and envelope signal of partial discharge is important to evaluate the insulation state of gas-insulated switchgear (GIS). In this paper, discretization and differential matrix reduction is first conducted over the feature matrix composed of feature vectors characterizing UHF PD envelope signals using rough set theory for dimensionality reduction. Then the reduced feature vectors are used for pattern identification of four different types of UHF PD envelope signals in combination with BP neural network classifier. The results show that this method has a high identification rate.