Time–frequency analysis of PD‐induced UHF signal in GIS and feature extraction using invariant moments

The ultra‐high‐frequency (UHF) method is efficient in partial discharges (PDs) detection in gas‐insulated switchgear (GIS). The features extraction of UHF signals is significant for propagation characteristics analysis and PD pattern classification. The PD‐induced UHF signals are acquired by the internal UHF sensors in an actual 252 kV L‐shaped GIS. The short‐time Fourier transform method is applied to process UHF signals and describe the propagation characteristics in L‐shaped GIS. Hu's invariant moments of energy density distribution are extracted as features in time–frequency plane. The features are utilised to discriminate different PD defect patterns in actual GIS model by the support vector machine classifier and achieve good results. Finally, a novel system of features extraction and classification of UHF signals is summarised.