Partial Discharge Characteristics and Development Process of GIS Insulator with Diverse Defects

Gas insulated switchgear (GIS) is a widely used apparatus in both middle voltage system and high voltage system. As for the defects diagnosis and condition monitoring of GIS, ultra-high frequency (UHF) detection is generally preferred. However, the conventional research on the diagnosis of partial discharge (PD) evolution is not precise enough. Thus the aim of this paper is to study the development process of the PD in various ways. An experimental platform is designed to complete both UHF and pulse current detection simultaneously, and typical artificial defects are prepared. Analysis of PD signals using the algorithm of Convolutional Neural Networks (CNN) has been carried out. An intensified Charge Coupled Device (ICCD) camera is also utilized in the whole system to capture the process of the discharge. The results show that it is effective in identifying the PDs of different types. Another thing observably is that whatever type of defects, the main frequency of the UHF signals is a constant which equals to 250 MHz, but the intrinsic mechanism of it needs to be further studied.