Pattern Recognition of Partial Discharge Type of Typical Defects in 10 kV Cable Joints of High-Speed Railway

In response to the problem of low accuracy in identifying common discharge types of typical defects in 10 kV cable joints of high-speed railways, four discharge models were established based on common discharge types of cable joints, a partial discharge test platform was built, and the discharge spectra of the model were obtained. Extract the distribution and moment features of the spectrum, and compare the recognition accuracy of the discharge model feature sequence between the back propagation (BP) neural network algorithm and the support vector machine (SVM) algorithm for optimizing grid search parameters. Produce physical models of four typical defects in cable joints and obtain their discharge spectra, extracting spectral distribution features and moment features. Compare the recognition accuracy of BP neural network and SVM based on grid search parameter optimization for cable joint defect discharge types. The research results show that BP neural network can accurately identify the combined features of the cable joint defect discharge spectrum, and the recognition rate is higher than other traditional algorithms. It has certain engineering application value in the field of cable partial discharge pattern recognition.