Skip to Main Content
Frequency-domain analysis of ultra-high frequency (UHF) signals for source identification of partial discharge (PD) occurring in SF6 inside gas-insulated substation (GIS) has been widely covered in literature. In this, Fast Fourier Transform and Discrete Wavelet Transform based techniques have been extensively applied to derive classifying features from transformed patterns. On the other hand, it appears feasible to develop a time-domain classifier, which derives features directly from the original waveshape. The time-domain classifier is conceptually simple, and requires potentially less computing resources and simpler algorithmic interface with other intelligent techniques due to elimination of frequency-domain transformation. A novel classifier to extract features directly from time-domain waveforms is proposed for classifying SF6 PD from air corona and among the three types of SF6 PD, regardless of changes in PD locations and measurement conditions. Three sets of classifying features are proposed. Encouraging results have been achieved with comprehensive experimental data, which verifies and proves the usefulness and feasibility of the time-domain classifier.