A neural network-based estimation of the level of contamination on high-voltage porcelain and glass insulators

In harsh environments, such as those found in Saudi Arabia, the periodic washing of insulators which is a necessity to ensure continuity in a power supply is very expensive. The accurate estimation of the severity of contamination can help electric utilities to properly schedule such washing and thus reduce the expense of washing and prevent insulator flashover. This paper presents a neural network algorithm for detecting the contamination level of high-voltage porcelain and glass insulators. The algorithm is based on images captured using a digital camera. Two types of features are extracted from each image: histogram-based statistical features and linear algebraic features based on singular value decomposition. Using the extracted statistical features, linear algebraic features, or a combination of both, three neural networks scenarios were successfully designed to associate the insulator images with the appropriate contamination levels and thus the possibility of flashover. Tests of the proposed estimation algorithm demonstrated a high success rate in estimating the contamination level of insulators. Finally, the developed algorithm has been deployed at the high-voltage station at King Fahd University (KFUPM). This deployment will eliminate the need for human intervention in determining the timing and location of required insulator cleaning.