Research On Pattern Recognition Of PD In Oil-Paper Insulation Based On Statistical And Fractal Feature

Partial discharge (PD) inside insulation is considered as a major cause of insulation deterioration in electrical equipment and of importance to the safety and reliability of running electrical equipment. Oil-paper insulation is the most important part in transformer, and therefore, the analysis of PD characteristics, mechanism and influence on the oil-paper insulation constitutes the foundation of PD diagnosis technique, and is significant to improve the accuracy of oil-paper insulation aging diagnosis. Based on the PD test for defect models of oil-paper insulation, spectrum features of PD gray image and texture extraction method are studied, 13 characteristics are extracted whose internal relations and physical essence are analyzed. Incorporating with BPNN, this paper fulfills the discharge sample recognition. The main content are as follows:(1) 5 typical PD defect models are designed. According to the analysis of testing method, it acquires data, constructs gray images and analyzes the discrepancy between different discharge images.(2) Entropy, standard deviation, mean absolute deviation of the PD gray images are extracted and the acquired characteristics are analyzed for five types of discharge pattern recognition.(3) The original gray images and their differential gray images of five types discharge are acquired whose fractal as well as multi-fractal dimension are calculated respectively. Moreover, relationship of sample cycles and fractal dimension is also revealed.(4) A BPNN used for PD pattern recognition is designed. Both the fractal and statistical features are used as input vector of BPNN to fulfill the recognition of 5 defect models.According to the above theoretical and experimental research, the correctness of proposed method is explained and illustrated systemically.