Equivalent Salt Deposit Density Prediction of Silicone Rubber Insulators Under Simulated Pollution Conditions

Abstract

Insulator flashover happens when soluble and/or non-soluble contaminants cover the insulator surface which results in a reduction of the surface resistance. Significant research has been conducted to utilize leakage current (LC) to predict the contamination level on the outdoor ceramic insulators surface. This can help as a mean to warn overhead lines operators about the advent of insulator flashover. However, there have been few attempts to predict the contamination levels on the surface of non-ceramic insulators. This work aims to develop a non-intrusive technique to monitor and evaluate the surface condition of silicone rubber (SIR) insulators by predicting the equivalent salt deposit density (ESDD). Three different classifiers (K-Nearest Neighbor Classifier (KNN), Polynomial, and Neuro-fuzzy) have been utilized to predict the ESDD level of SIR samples after a salt fog test. Moreover, stepwise regression and principle component analysis (PCA) have been used as feature selection tools to optimize the classification process. The overall prediction accuracy improved from 68% to 95% when the number of classes reduced from four to two respectively.