Prediction of the Levels of Contamination of HV Insulators Using Image Linear Algebraic Features and Neural Networks

Contaminated high-voltage (HV) insulators in polluted areas may lead to flashovers if they are not cleaned periodically. Flashover often leads to lengthy service outages and thus has a considerable impact on power system reliability. Therefore, an accurate prediction of the contamination level of HV insulators is vital. In this study, a MATLAB-based algorithm for predicting the contamination level is proposed. The algorithm uses the extracted features (in this work, linear algebraic features) from images captured by digital cameras as an input to a neural network. When compared to existing methods reported in the literature, the designed neural network correlates successfully the captured insulator images and the contamination level when tested on unseen insulators.