A New Inference System for the Robust Identification of Defects Supporting PD in MV Cables

A method for the identification of a large number of defects supporting Partial Discharges (PD) in MV cables is presented in the paper. A specific set of parameters that summarize features related to the shape of the Phase-Resolved PD (PRPD) patterns, is organized in a tree structure on the assumption of a correlation between the PRPD shapes and the defect typologies. Tree inspection resorts to the Fuzzy-Logic to take into account the inherent uncertainty of the identification results. Predicates that address PRPD to the defect typologies and the relevant membership values are provided as the output taking into account even more than one option. Both training and working examples obtained testing artificial defects and real cables, are presented and discussed.