Accurate Hot Spot Detection on High-Voltage Power Transmission Lines Using YOLOv8-Based Deep Learning

Early detection of thermal anomalies—commonly referred to as hot spots—in high-voltage power transmission systems is essential to avoid catastrophic equipment failures, reduce energy losses, and ensure the safety and reliability of the electrical grid. Manual inspection methods using thermal cameras or visual checks are still widely practiced; however, they are labor-intensive, timeconsuming, and prone to human error. In this context, deep learning-based object detection algorithms have emerged as a promising solution to automate the inspection process with high accuracy and speed. This study proposes a robust and scalable approach utilizing the YOLOv8 architecture for accurate hot spot detection in thermal infrared imagery. A custom dataset comprising 2,560 real-world thermal images was created, encompassing diverse environmental conditions and electrical components such as current transformers, circuit breakers, and lightning arresters. Unlike conventional multi-class detection setups, this work focuses exclusively on the “hot spot” class to maximize detection performance and deployment efficiency The dataset was divided into training (70%), validation (15%), and test (15%) sets, and enhanced with data augmentation techniques, including random rotation, horizontal flipping, and brightness normalization. The model was trained using the Ultralytics YOLOv8 implementation, and achieved a precision of 95.4%, recall of 93.4%, and a mean average precision ([email protected]) of 97.4% on the test set. These results confirm the high detection capability and generalization performance of the system across varying scenarios. The lightweight architecture of YOLOv8 also enables real-time inference, making it well-suited for deployment on UAVs and portable thermal inspection devices. Overall, the proposed system offers a reliable, efficient, and fully automated alternative to traditional inspection methods, contributing to the advancement of intelligent grid maintenance strategies.