Cable Insulation Fault Identification Using Partial Discharge Patterns Analysis

This study deals with internal defects existing or occurs in cable insulation due to stress over its operation. The most popular tool for identifying and assessing insulation-based flaws is partial discharge (PD) analysis for every power cable and solid insulant. Characterization of defects is of utmost significance for overall degradation intensity and possible deterioration evaluation. In this article, a machine learning-based diagnostics scheme is proposed to identify and characterize PD signals formed by different internal sources in solid insulation. The internal discharge sources are differently shaped and sized voids created in polymeric insulation. A dissimilar shaped cavity produces distinct PD patterns. The PD signal is recorded and denoised using the wavelet analysis to remove unwanted consistent noise interference efficiently. The feature matrixes are formed by implementing features obtained from statistical operators and phase-resolved PD (PRPD) signal characteristics based on different sizes. The proposed scheme is efficient with forms of support vector machines (SVMs) and ensemble algorithm tools to achieve the high accuracy of defect identification and classification. The accuracy band of the proposed machine learning-based diagnosis to identify and characterize the defect scale is from 96% to 92%.