Abstract

The maintenance of electrical insulators through washing represents a high economic cost for companies aiming to avoid flashover faults. An alternative is to use glass insulators coated with silicone rubber, which provides hydrophobicity and consequently reduces surface contamination. Over prolonged use, the surface can degrade, losing some of its hydrophobic characteristics. The challenge here is to determine whether the insulators in use are aged or contaminated. This paper presents a study of different indices obtained from the characteristics of the leakage current at different humidity levels to determine the surface condition of the insulator. Through a support vector machine (SVM) model, a classification of the state of the silicone layer was carried out, discriminating between aging and contamination levels. The indices used to train the SVM allow for the classification of contamination with 99% accuracy; however, a satisfactory classification of the aging level of the silicone layer is not achieved.