Fault Detection in Power Distribution Systems Using Sensor Data and Hybrid YOLO with Adaptive Context Refinement

Ensuring the reliability of power transmission systems hinges on the accurate detection of defects in insulators, which are vulnerable to environmental degradation and mechanical stress. Traditional inspection methods are time-consuming and often ineffective, particularly in complex aerial environments. To address these challenges, this paper presents a hybrid fault detection framework that integrates the YOLOv8 object detection model with a novel Adaptive Context Refinement (ACR) mechanism. While YOLOv8 enables real-time detection, the ACR module enhances precision by adaptively leveraging multi-scale contextual information surrounding detected objects. This joint architecture improves the detection of subtle and small-scale insulator defects, especially under challenging imaging conditions. The proposed system is evaluated across 25 YOLO model variants using high-resolution datasets, including real-world inspection images from power distribution networks. Results show that ACR significantly improves mean Average Precision, particularly in lightweight models like YOLOv10n, which achieved a 22.9% performance gain. The findings demonstrate the effectiveness of contextual refinement in reducing false positives and enhancing defect classification, supporting practical deployment in automated aerial inspection systems.