Optimisation of a numerical model for analysis of partial discharge phenomena in a flat cavity

The failure probability of the insulation in HV components can be studied through partial discharge (PD) measurements to detect defects. A numerical model, previously developed by the authors for the simulation of PD activity in a spherical void embedded in epoxy resin, is implemented for a flat cavity under a divergent electric field. This simulates very well the real working conditions encountered in the insulation of electrical machines. The model is developed taking into account all the geometric and electrical parameters, and a fine examination on physical parameters playing a fundamental role on the PD phenomenon is carried out. In particular, a more accurate formulation of the work function and a finite elements analysis (FEA) of the electric field inside the cavity has been applied. A new approach is investigated to reproduce the uncertainty of the discharge phenomenon. A probability function distribution, obtained by Weibull analysis, gains the best results. In order to obtain the best fitting of the model output to the experimental PD data, a strong procedure based on an heuristic algorithm is performed. This algorithm detects the minimum of a quality function that evaluates the distance between the PD patterns of the numerical model and of the experimental data. After a description of the numerical model adopted, a comparison between the experimental and the simulated data are presented together with comments and remarks