An Integrated Hardware-Software System to Identify the Underlying Distribution of PD Pulse Height Records

Abstract

In this paper, we have incorporated a software capability to extend the PDPAS developed in [11] where the incoming PD pulses are classified into PD pulse height distributions whilst computing the parameters. The Maximum Likelihood Estimation (MLE) technique is used in fitting the distributions to the PD data. The Quasi-Newtown optimization technique is used to numerically compute the optimal parameters of a distribution. Post fitting the distributions to the data Bayesian Information Criteria (BIC) is used to assess the Goodness of Fit. Pulse height data were artificially generated from different distributions and were tested with the developed system and this yielded consistent and reliable results. Identification of type of distribution of pulse height dataset will provide an insight into the degradation mechanism occurring in the insulation system and has immense practical utility. The details of mathematical formulations, software system design, results of testing and its implications are presented in the paper.