A lightweight DRR-YOLOv11s model for power equipment failure and personal protective equipment detection in hydropower stations.

The increasing demand for intelligent operation and maintenance in hydropower plants highlights the limitations of conventional inspection approaches in real-time performance and reliability, especially under edge-computing constraints. This paper presents DRR-YOLOv11s, a lightweight object detector for electrical equipment fault detection and worker safety monitoring. By optimizing the model structure for efficient inference, DRR-YOLOv11s reduces parameters by 37.58% and computational cost by 28.64% compared with the baseline, resulting in 5.88 M parameters and 15.2 GFLOPs. Experiments on the Electrical Equipment Failure dataset show a precision of 92.50%, recall of 89.42%, mAP@0.5 of 93.46%, and mAP@0.5:0.95 of 73.45%, while achieving 160.22 FPS. Cross-dataset evaluation on a Personal Protective Equipment dataset further indicates good generalization. Overall, DRR-YOLOv11s achieves a favorable trade-off between accuracy and efficiency, supporting practical deployment for real-time monitoring in hydropower plants.