Recognition of Partial Discharge in Microphone Array Data through Fusion of Spatial and Temporal Correlation Features

Microphone arrays provide a non-intrusive and adaptable approach to detect partial discharges (PD) in power equipment. Nonetheless, current methods that rely on microphone arrays frequently overlook the distinct data characteristics inherent in microphone array recordings, especially in the context of PD classification. This paper endeavors to address these gaps by analyzing the temporal and spatial correlation features within microphone array data. Three distinct methods for PD pattern recognition in acoustic array data are proposed, utilizing one-dimensional convolutional neural network (1D-CNN) and the squeeze-and-excitation (SE) correlation extraction method. Based on the PD dataset collected from laboratory simulations, experiments are presented to validate the effectiveness of the proposed methods. Furthermore, by comparing different scenarios involving correlation analysis, the research highlights the significance of considering both spatial and temporal correlations to enhance the effectiveness of PD classification.