A Partial Discharge Pattern Recognition Method Based on Optimal Setting of characteristic Parameters and Classifiers

摘要

Gas Insulated Switchgear has been widely used in electrical power system because of its compact structure, small area, maintenance-free, reliable operation and other significant advantages. However, in the manufacture, transportation, installation, maintenance and long-term operation of GIS equipment, internal insulation defects and partial discharge will inevitably occur, and the hazards of different discharge types and corresponding treatment measures are different. In order to realize the effective and accurate diagnosis of the discharge type for UHF partial discharge ,the high-order moments and histogram entropy are proposed based on the traditional waveform features, moment features, typing features, statistical features and wavelet methods. And the support vector machine and BP neural network are used as classifiers for comparative study. The typical discharges such as tip discharge, surface discharge, internal air gap discharge and suspension discharge were simulated on the GIS in the laboratory, and 50 groups of data were collected for analysis. The results show that the recognition accuracy can be improved to 90% by adding higher-order moment and histogram entropy to the statistical operator. Compared with the traditional method, the recognition accuracy is improved by 1.25%. In addition, the combination of vector machine and BP neural network as classifier can effectively improve the accuracy of partial discharge defect recognition. Compared with using SVM and BP as classifiers alone, the recognition accuracy is improved by 8.75% and 10% respectively.