Multispectral Detection and Diagnosis of Multisource Discharges on Epoxy Fiberglass Board

To enable the diagnosis of multi-source partial discharge (PD) on the surface of epoxy fiberglass boards, this study investigates optical PD signals and diagnostic methods associated with typical defects and their combinations. The electric field distribution for typical insulation defects is analyzed, and both emission spectroscopy measurements and PD experiments are performed. The results demonstrate that different PD modes exhibit distinct spectral characteristics. For needle, crack, and metal contamination defects, the detection limits using a full-band silicon photomultiplier (SiPM) are determined to be 64.37 pC, 78.12 pC, and 159 pC, respectively, with corresponding response coefficients of 450 mV/pC, 300 mV/pC, and 160 mV/pC. The SiPM response coefficient is found to vary with detection wavebands and PD modes. Multispectral analysis shows that although the 300-400 nm band accounts for the highest proportion across all discharge modes, the distribution characteristics of each wavelength band differ. A decision tree model is developed for discharge pattern diagnosis, achieving an accuracy of 95.09%.