Partial Discharge Fault Detection of Substation GIS Based on CEEMDAN Fusion Processing Algorithm of Multi-Frequency Signals

Abstract

Partial discharge is a common fault mode of GIS equipment, and timely and accurate detection of its status is of great significance for ensuring the safe operation of the power system. Therefore, a partial discharge fault detection method for substation GIS based on multi-frequency signal CEEMDAN fusion processing algorithm is proposed. By analyzing the typical GIS partial discharge fault state structure, segmented collection of substation GIS partial discharge data is carried out; Based on the window function method and nonlinear gain adjustment method, a limited impulse response filter with precise linear phase characteristics is selected for multi-frequency signal enhancement processing; Simultaneously combining wavelet reconstruction technology and Fisher criterion to improve the CEEMDAN algorithm, obtaining a fused signal containing frequency feature information; Using CNN network model to fuse feature signals as input, achieve accurate detection of partial discharge faults in substation GIS. The experimental results show that the detection accuracy of typical substation GIS partial discharge faults such as suspended discharge, hole discharge, metal particle discharge, and corona discharge obtained by the design method is higher than 95%. It can capture the partial discharge characteristics of GIS equipment, accurately judge its partial discharge state, accurately detect fault types, better generalize GIS equipment of different types and states, and have good robustness and practical detection effect.