A New Box-Counting-Based-Image Fractal Dimension Estimation Method for Discharges Recognition on Polluted Insulator Model

Abstract

This study presents an innovative approach to identify electrical discharges by proposing an algorithm incorporating fractal geometry concepts. Based on the box-counting method, our algorithm is developed to detect and track the progression of electrical discharges leading to flashover. This is achieved by calculating the fractal dimension of discharge images which are visual representations of electrical activity recorded during experiments on a planar glass insulator model subjected to different levels of contamination. First, the RGB image is transformed into a binary matrix using the NIBLAK binarization algorithm. Subsequently, the acquired matrix is converted into a square matrix, and its fractal dimension is computed for various resolutions. The final fractal dimension of the image is calculated using the least squares method. This latter is applied to the fractal dimensions (FDs) across all resolutions. According to our algorithm, discharge images have FD values ranging from 1.15 to 1.25. FD increases are observed with applied voltage and non-soluble deposit density (NSDD). The density and activity of discharges also increase with FD. Specifically, a discharge is considered “no-arc” if FD is less than 1.2 and “arc” otherwise.