Arc feature extraction and flashover warning of direct current composite insulator based on image processing using visible discharge images

Abstract

Insulator flashover seriously affects the safe operation of power grid. To improve the arc development mechanism and achieve flashover warning, this paper researches the arc evolution characteristics during direct current (DC) flashover process of composite insulator using visible discharge images. Firstly, an arc segmentation algorithm based on edge brightness and local threshold is proposed. It aims to solve the problems of uneven arc brightness, insulator reflection and arc halos in visible discharge images. Then, to quantify segmented arcs, an arc feature extraction method with four feature quantities of arc width, projected length, offset distance and centroid ordinate is proposed. Finally, this paper also extracts and studies the variation characteristics of the four feature quantities in two types of DC flashover processes respectively. The result shows that the proposed feature quantities effectively describe the arc state. Additionally, the arc width change rate k exceeding 1.3kmax, the insulation failure percentage reaching 75%, and the arc offset distance reaching the set threshold can serve as the critical criteria for arc development stages, dividing the two types of DC flashover processes into three stages. Building upon the critical criteria, it is expected to achieve flashover warning.