Comparing different stochastic models for the identification and separation of concurrent partial discharge phenomena

Results of investigations performed in order to evaluate the ability of three different probability functions in the identification and separation of different Partial Discharge (PD) phenomena are reported in this paper. PD-pulse height distributions coming from different sources are processed resorting to the two-parameter and the three parameter Weibull function, as well as the three-parameter Weibull-like function. Combined PD phenomena, caused by different PD mechanisms, are separated by means of a stochastic procedure based on the mixed Weibull model. The results are compared with those derived from PD measurements performed on the electrode configurations that produce a single PD mechanism at a time. Identification is pursued resorting to the shape parameter of the different distributions. It is shown that the Weibull-like function provides the best performances in terms of quality of fit, separation and identification.