Multi-Source Partial Discharge Signal Separation and recognition Method Based on manifold Learning in Oil-pressboard Insulation System

Abstract

Oil pressboard insulation system is one of the most important systems in the power system., Various insulation structures such as oil and pressboard may produce partial discharge in the oil pressboard insulation system, releasing UHF and ultrasonic signals, and even multiple parts of the oil pressboard insulation system may have partial discharge faults at the same time, which will cause the partial discharge signals to cross and overlap affecting the identification and location of partial discharge faults. In this paper, a manifold learning based ultrasonic and UHF partial discharge signal separation and recognition strategy for oil pressboard insulation system is proposed. Firstly, the high-dimensional information of each PD signal sample is projected to the two-dimensional plane by the improved t-SNE algorithm, and then the equivalent Euler distance is calculated. Then, the partial discharge pulses from different PD sources are identified by improved possible CMeans method, and the effectiveness of the separation is verified by the experiments of partial discharge UHF and ultrasonic signals in the laboratory artificial defect model. The results show that the manifold learning oil pressboard insulation system can effectively separate the suspension, oil, and tip discharge sources.