A new approach to identify electrical PD signal patterns using frequency spectral analysis

Partial Discharge (PD) tests on HV equipment have been identified as a popular technique to monitor the deterioration process in insulation. Therefore it is very important to recognize and locate such partial discharges from an unknown origin in the insulation. The research on "PD pattern recognition" is mostly performed in time-domain. In this paper a new criteria has been proposed to analyse "Electrical PD Signals" in the frequency domain for "PD pattern recognition". An experiment is carried out on an 11 KV, single-core, 240 mm², XLPE cable under three different soil thermal resistivities and cable loading to acquire "Electrical PD Signals" in the shape of ellipse on a 50 Hz time base. The time-domain electrical PD signals are then transformed into frequency spectrum by short-time-Fourier transform (STFT) technique for analysis using eight related frequency-domain descriptors. The proposed spectral analysis results of the "Electrical PD Signals" acquired from the cable experiment have proved that the procedure should be made both for positive and negative polarization of the applied voltage. The obtained results also confirm the ability of the proposed "frequency-domain analysis" to identify a specific PD type from the associated frequency spectra and the related descriptors.