Power Cable Partial Discharge Pattern Recognition Based on Topological Data Analysis for Time Series

In the partial discharge (PD) patterns recognition of power cables, phase resolved partial discharge and statistical features often affect recognition accuracy due to insufficient discrimination. Therefore, a method based on time series topology data analysis (TDA) for feature extraction and recognition of PD is proposed. Firstly, a method combining symbolic entropy and Particle Swarm Optimization (PSO) for the selection of reconstruction parameters is proposed. The pre-processed PD signal in time-domain is reconstructed in phase space to generate a three-dimensional PD data point cloud. Secondly, based on the TDA method, persistent homology features are extracted to generate persistence diagram and persistence barcodes, which are calculated and visually expressed as Betty curve. Finally, the Betty curve is employed as the input of 1D-CNN model for recognition. The experimental results show that the proposed method is more accurate in the selection of the time-delay parameter for phase space reconstruction, and the TDA features achieve good discriminability. Compared with other models that use phase spectrograms and statistical features as inputs, the overall recognition accuracy can be improved by up to 15.34%, reaching 98.55%.