The Research Of Fault Diagnosis In Partial Discharge For GIS Based On Wavelet Transform And Neural Network

Higher and higher quality electrical power supply is needed with the development of the society. So gas insulated switchgear (GIS) has been widely used in the power system due to its characteristics advantages, such as small volume, high quality, etc. However, in case of an insulation breakdown, it causes many serious damages to power system. So the research of PD detection and its pattern recognition for GIS is very important to guarantee its safe operation, to find out its insulation condition and defects type in time and to guide its maintenance.The main work in the thesis is as follows:A wavelet and wavelet packet threshold de-noising are researched and realized, and two application cases are presented. According to the characteristics of partial discharge signals, different statistical methods are used in feature extraction for different feature vectors based on wavelet packet transform.For the shortcomings of slow convergence rate and falling into local minimum easily of BP neural network, genetic algorithms and neural network are combined to form GA-BP network. Using the global search ability of genetic algorithm, weights and thresholds of neural network are optimized and the two inherent shortcomings of network are solved. It shows that the train epochs of GA optimized BP NN decreased a lot by comparing simulation results with MATLAB, and many disadvantages of traditional BP NN are overcome by using GA-BP. The GA-BP method improves BP NN a lot, enhances the accuracy rate of fault diagnosis to GIS.At present, the results of locating the PD in GIS are not satisfying due to the processing methods and the strong noise at the sites. To solve this problem, the wavelet transform is used to filter the PD signal to get the signal with fewer disturbances. The thesis deeply studies on time difference algorithm based on the relative increase in energy and applies it to locate the PD in GIS creatively, which can further improve the location accuracy.The computer simulation certifies the correctness and validity of the analyses and syntheses methods in this thesis and indicates that the research has its practical application.