Partial Discharge Pattern Recognition of High Voltage GIS Defects by Using GWO-SVM Method

This paper focuses on partial discharges (PD) pattern recognition of typical GIS defects by UHF sensor detection and intelligence fault algorithm. 4 typical PD defects, such as needle tip, air gap, particle, and suspension were simulated in the SF6 filled GIS chamber. The PD results indicated an obvious difference among the 4 PD defects. The corona discharge presented sharp discharge peaks during the 200–300°. According to the PRPD analysis, the eigenvalues of PD signals were calculated, including the skewness, the steepness, the local discharge factor, the cross-correlation coefficient, and the corrected cross-correlation coefficient etc. We employ a Grey Wolf Optimization algorithm (GWO) to optimize the parameter of kernel function in SVM algorithm. The proposed GWO-SVM method presents a better PD pattern recognition result. The predicted accuracy rate can be reached to 98.8%. This work can be used to guide GIS fault diagnosis.