Partial Discharges analysis and parameters identification by continuous Ant Colony Optimization

Abstract

The technique of ant colony optimization has been employed in this paper to efficiently deal with the problem of parameters identification in partial discharge, PD, analysis. The latter is a continuous optimization problem. From the technical point of view the identification of these parameters allows the modeling of the phenomenon of Partial Discharges in dielectrics. In this way it is possible the early diagnosis of defects in Medium Voltage cable lines and components and thus it is possible to prevent possible outages and service interruptions. Analytically, the problem consists of finding the Weibull parameters of the Pulse Amplitude Distribution (PAD) distributions allowing the identification and classification of the defects in dielectrics. The accuracy in this identification is crucial for correct classification of defects. The proposed algorithm, called DACS, Dynamic Ant Colony Search, allows the easy investigation of complex problems both in discrete and continuous search spaces. It dynamically redefines the search tree through which the ants (agents) move using an adaptive parameter in order to increase exploration or exploitation. In order to check the efficiency of the proposed algorithm in solving continuous optimization problems, many Partial Discharges, PD, experimental tests at various temperatures have been performed on some lumped capacity specimens. In this way, the experimental cumulative probability of amplitude histograms has been compared with those attained using the Weibull analysis. All the applications show that the error is quite limited and that the calculation times are considerably low compared to other techniques employed for the same purpose.