Pulse waveform based identification and classification technique of PD for the high voltage power apparatuses

For insulation diagnosis and condition assessment in partial discharge (PD) measurement of the high voltage power apparatuses, one of the major tasks may be the PD signal (PDs) identification, the PDs origin separation and the estimation of condition of the insulation. The conventional phase-based statistic analysis of PDs, which emphasizes the detection of PDs amplitudes, is often not enough to carry out effective diagnosis, risk assessment and condition-based maintenance. So, firstly the equivalent bandwidth-the equivalent time length (W-T) separation technique, which is used to classify and identify the PD pulses in virtue of the pulse clusters in the W-T two-dimensional Cartesian coordinate system, is studied theoretically in this paper. Moreover, the similar parameters, which are necessary for the PD pulses to make a compact cluster in W-T plane, are calculated. Secondly, the pulse waveforms in the same cluster are pre-selected and normalized, and then a Sampling Counting Ratio (SCR) classification technique is investigated to reveal the multi-peak and oscillation information about the PD current pulse waveforms. Thirdly, three types of discharge model such as air gas discharge, corona discharge and surface discharge, are constructed and PDs are extracted from the experiment presented in this paper. Then the extracted PD pulses are analyzed by W-T and SCR classification technique. Results show that W-T and SCR classification technique is efficient for analyzing the defect types, PD sources and degree of the degradation of the insulation in high voltage apparatuses.