Research on Multivariate Image Processing and Pattern Recognition of Power Equipment Failure

Partial discharge is one of the important characteristics of insulation deterioration of power equipment. The detection of partial discharge is of great significance to judge the insulation fault of power equipment. In recent years, pattern recognition technology has been widely used in the field of partial discharge detection. In order to make the pattern recognition of partial discharge more accurate, an image fusion algorithm based on non-down sampling contour wave transform (NSCT) is used to fuse the optical partial discharge spectrum and UHF partial discharge spectrum, so as to realize the complementarity of partial discharge information under the two methods. Build GIS partial discharge detection platform, measure PRPD maps under three typical defect types, and convert PRPD 3D maps into φ- u two-dimensional map, and then use NSCT algorithm to fuse optical and UHF φ- u two-dimensional spectrum, and the photoelectric fusion partial discharge spectrum is obtained. The shape, texture and entropy features of the fusion map are extracted, and then the dimension of the feature space composed of these features is reduced by PCA. Finally, BP neural network classifier is used to verify the pattern recognition effect of optical, UHF and photoelectric fusion partial discharge spectra. The results show that the pattern recognition effect of photoelectric fusion partial discharge map is better than that of other two kinds of partial discharge maps under single detection method.