Research On Method Of PD Pattern Recognition Based On Wavelet Neural Network

With the development of electric power industry, the voltage level continues to improve, whether from the aspects of safety or economic terms, the monitoring of high voltage equipment insulation parameters is important. Therefore, to monitor the state of the high voltage equipment, it needs to acquiring signals of the partial discharge, both analysis and processing the collected data from operating high voltage apparatus, to conclude the degree of deteriorationand the remaining life of high voltage equipment insulation. Hence research on the method of partial discharge pattern recognition technology has a valuable academic value and important significance.On the basis of wavelet theory and neural network, constructed the Morlet wavelet neural network, written Matlab code to complete the partial discharge pattern recognition by using wavelet neural network.Described the structure of the wavelet neural network, the training of learning and the principle of the algorithm, discussed the effects of the input vector, the excitation function and the learning rate.From the study of the characteristics of partial discharge, constructed five different electrode models to simulate the partial discharge characteristics in different electric field distribution.Collected the partial discharge information and extracted the fingerprint by DDX-7000 partial discharge detector.On the basis of feature separability criterion, through the analysis of the fingerprint data, select and determine its mode characteristics. Through the analysis of the wavelet neural network training and recognition process, and contrast between the BP neural network and the RBF neural network, proved that in the wavelet neural network pattern recognition, the feature selection, hidden layer neuron number, the number of iteration parameter directly affects the network recognition rate.Through the research of this thesis, and contrast with BP and RBF neural network, it is proved that the wavelet neural network in partial discharge pattern recognition technology has significant advantages, provided theoretical basis and technical supportfor the application of wavelet neural network in the field of pattern recognition of partial discharge.