Deep Learning Model Development for Detecting 22 kV Line-Post Insulator Faults

Abstract

In electrical power distribution systems, the most common insulators are used. Power outages can be caused by insulator damage in a variety of ways. The purpose of this study is to develop a deep learning model for detecting 22 kV line-post insulator faults. To investigate image sizing, the quantity of datasets, viewing points, and the number of appropriately trained individuals, the training datasets were created on IDE Roboflow. Moreover, the model was created by YOLOv5 on Google Colab. The images of insulators were divided into two groups including normal and faulty insulator. These insulator images were collected from the different point of view including on the floor and on the distribution line. The image size employed was 416x416 pixels, which corresponded to 148 real images and 251 augmentation images. The experiment defined different batch sizes with Epoch counts ranging from 50 to 300. The results demonstrated that the developed model could detect cracks in insulators mounted on the distribution lines in various spots.