Application of the Chaotic System–Based Error Trace Diagrams for Partial Discharge Feature Extraction

This study established four models representing common defect types in 25‐kV cross‐linked polyethylene power cable joints and analysed partial discharge (PD) signals associated with such defects. An acoustic emission sensor was used to measure acoustic signals induced by the PD phenomenon in the power cable joints. A chaotic synchronisation system was applied to analyse dynamic errors associated with the signals. Motion trajectories were calculated using the master and slave systems in the Chen‐Lee chaotic system and then plotted as three‐dimensional error trace diagrams for each of the four models. Two chaotic gravity distances were obtained as features and a backpropagation neural network algorithm was used for cable joint pattern recognition. Furthermore, acoustic signals were measured from 160 sets of cable joints and random white noise was added to test the robustness of the feature extraction algorithm against noise. This study also evaluated the use of fractal dimensions for extracting features from 3D PD patterns. The results confirmed that the proposed method achieved high accuracy, was straightforward to implement and effectively distinguished between different cable joint models.