Characteristic Parameters Extraction and Pattern Recognition of Partial Discharges based on Envelope of Ultra-high Frequency Signal

With the help of the Hilbert transform, the low-frequency signals modulated by high-frequency signals can be demodulated. This article proposed a method to extract the time-domain features of ultra-high frequency (UHF) signals regarding the partial discharges (PD) in a gas insulated switchgear (GIS). Utilizing the Hilbert transform, the envelope of a PD signal can be solved, and further the corresponding characteristic parameters of some key time domains can be acquired. An experimental platform for the detection of artificial insulation faults in a GIS was established. Thus, the experimental data of 3 types of typical PD signals were gained, and then used for the analysis and verification of the presented algorithm. The result shows that different types of PDs were accurately recognized. It can be concluded that the method presented is practical and effective.