PD Pattern Recognition Method Using Shape Feature in Generator Stator Bar

Generator is the most important high voltage electrical equipment. Partial discharge (PD) detection can evaluate the operation state of the Generator and prevent large accidents. The partial discharge signals of five typical insulation defect models and one normal model in generator stator bar were collected in the laboratory, and the phase resolved partial discharge (PRPD) pattern in stator bar was constructed. In this paper, considering the asymmetry of positive and negative discharge pulse amplitude, a characteristic parameter is proposed, and combined with other shape statistical characteristic parameters to form a feature vector. The back propagation neural network optimized by particle swarm optimization algorithm (PSO-BP) is used as recognition classifier to identify six stator bar models. The results proof that the method is very effective on pattern recognition in multiple generator stator bar models. Particle swarm optimization algorithm can effectively reduce the prediction error, and improve the characteristics of BP neural network that is too sensitive to the initial weights and thresholds.