Complex Network Aided Partial Discharge Signal Recognition Framework Employing Visibility Graph

Abstract

This letter presents a novel framework for automated recognition of partial discharge (PD) signals using complex network theory. PD signals measured using high frequency current transformer (HFCT) sensors were transformed into a complex network using visibility graph (VG). We propose a new edge weight of VG in this letter for the construction of weighted PD network. From the weighted adjacency matrices of different PD signals four network features were extracted. The statistical significance of the extracted features was further examined using one way analysis of variance (ANOVA) test. Finally, the classification of PD signals was done using three benchmark machine learning classifiers. Investigations revealed that the PD signals are accurately identified using the proposed method, which can be potentially implemented for condition monitoring of insulation in real time.