Research on Diagnosis and Positioning Method of Typical Partial Discharge Defects in Gas-insulated Switchgear

Abstract

Gas-insulated switchgear (GIS) is one of the important equipment in power system. Using UHF partial discharge detection method to find and diagnose internal insulation defects in time is an important means during the operation and maintenance of GIS equipment. Different discharge types in GIS show different partial discharge characteristics. This paper proposes a diagnosis and positioning method of discharge faults in GIS equipment through experiments. Firstly, UHF partial discharge signals of different typical discharge models in GIS are obtained by UHF sensor, and then 3D GIS PRPS patterns are drawn. After that, PRPS patterns are transformed into 3 decomposition patterns through Gabor transform, and the texture and shape features of these decomposition patterns are extracted. Based on extracted characteristics, different kinds of machine learning algorithms are applied to identify the discharge type. The results show that the recognition accuracy of three typical discharge types by different machine learning algorithms is high. The characteristics extracted from decomposition patterns obtained by Gabor transform well reflect the discharge type and has high discrimination. Combined with relevant positioning algorithms of GIS, the fault location can be better realized when the rough location of defects is known when accurate identification of partial discharge defects is obtained. This diagnosis method can provide a reliable reference basis for the early warning and accurate location of GIS faults.