Instance Segmentation of Damaged Insulators for Power System Inspection and Maintenance

This paper introduces a novel approach for power system inspection using Mask-RCNN, an advanced model for instance segmentation. The accuracy of our model in identifying and diagnosing damaged low- to medium-voltage insulators is exceptional, as evidenced by its Mean Average Precision (mAP) score of 98.7<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>• Throughout the training process, the model exhibited minimal loss metrics, indicating its proficiency in predicting bounding boxes precisely, segmenting, classifying, and aligning masks. The model exhibited consistent reliability in distinguishing between undamaged and damaged components during the inference process, as seen by its consistently high detection rates for both “normal intact” and “damaged broken” insulators. The results demonstrate the practical usefulness of the model and offer promising prospects for enhancing power system maintenance and inspection protocols. The model's consistent and accurate identification of damaged insulators has the potential to significantly enhance the safety and efficiency of power systems, thereby revolutionizing the industry.