GIS Defect Monitoring and Identification Based on Multi-sensor Association

Abstract

During the operation of gas insulated switchgear (GIS), internal defects will occur due to seal failure, poor contact, metal particles, switching overvoltage and other reasons. Aiming at the problem of low accuracy of traditional GIS condition monitoring methods, which is easy to cause false detection and leakage judgment, this paper proposes a GIS insulation condition monitoring method based on multi-sensor data association analysis. The monitoring data of infrared and ultraviolet sensors are quantitatively calibrated and correlated. The temperature rise is calibrated by infrared photoelectric signal, and the partial discharge is calibrated by ultraviolet photoelectric pulse. By optimizing the weight of serious faults in the loss function, the BP neural network (BPNN) is improved, and the data association analysis of GIS insulation state detection is realized. The data processing and associating method is verified by the simulated discharge experiment of GIS epoxy insulator surface defects. The experimental results show that the accuracy of GIS insulation defect classification is significantly improved by this method, which is helpful to GIS insulation state assessment and fault prediction, and provides a useful reference for GIS equipment operation and maintenance.