To address challenges of limited computational resources on drones and infrared imagery interference in power equipment detection, this study proposes HRT-YOLO, a lightweight adaptive algorithm. We enhance YOLOv11 via three innovations: (1) Inspired by RepViT[1], the Re-parameterized Depthwise Separable Convolution block (LRepViTBlock ) was proposed, enabling the decoupling of the token-channel mixer, improving computational efficiency, and enhancing the model's learning ability through the reparameterization technique; (2) A hybrid ratio-tunable convolution (HRTConv) was proposed, enabling computation-accuracy trade-offs through channel allocation factor ; (3)A C3K2 module incorporating HRTConv was proposed, effectively reducing the computational complexity while maintaining negligible accuracy loss.When the is 0.75,evaluations on thermal imagery show HRT-YOLO achieves 95.2% mAP@0.5 and 77.6% mAP@0.5:0.9 at 4.7 GFLOPs, reducing computations by 25.4% versus baseline while outperforming YOLOv5 by 5.8% in mAP@0.5:0.9. The solution balances real-time performance and accuracy for drone inspections.