The current cable fault detection is usually based on a single sensor and signal source, coupled with the constraints of the field environment, which greatly enhances the difficulty of personnel inspection. In this paper, by integrating five kinds of multi-dimensional cable fault signal feature information. The extracted multi-dimensional information features are downscaled using the KPCA downscaling algorithm, SVM is used as the multi-dimensional information fusion cable fault identification algorithm, and the KPCA-SVM multi-dimensional information fusion identification model is optimized with kernel parameters through the APSO algorithm, and finally the APSO-KPCA-SVM multi-dimensional information fusion identification cable fault model based on APSO-KPCA-SVM is proposed. The experiments are validated using simulated datasets, and the fusion and dimensionality reduction of five fault signal features is used as the input feature vector of the cable fault identification model, which reduces the accuracy of the model by only 0.22% compared with the model before dimensionality reduction, while the training time is reduced by 413.14 s. The tests are conducted under the condition of noise interference as well as data loss, and compared with the SVM and the DBN, the APSO-KPCA-SVM has stronger anti-interference performance, when the signal-to-noise ratio is reduced to 20dB, only the model cable fault identification accuracy of this paper is still maintained at more than 90%.
