GIS partial discharge detection and recognition based on the kernel combination and multiple feature fusion method

GIS partial discharge pattern recognition is an important part of its insulation state evaluation, authors set up 252kV GIS partial discharge detection simulation experiment platform, and detected the partial discharge of four kinds of typical insulation defect models based on the ultra high frequency method and ultrasonic method, and then obtained the corresponding statistical map based on the characteristics of the signals, also extracted the corresponding feature parameters; then blended UHF and ultrasonic characteristic parameters based on the kernel combination parameters method which is optimized by K-fold cross-validation and particle swarm optimization method, then input parameters after fusion and single UHF and ultrasonic parameters to the pattern recognition classifier respectively. The results show that the recognition rates is higher than single feature after multiple characteristic parameters fusion, and the recognition rate is more than 92%.