This paper presents an enhanced version of YOLOv8 for detecting damage in overhead power lines. The proposed method integrates three key improvements: an adaptive threshold mechanism, GSConv, and Slim Neck. The adaptive threshold mechanism dynamically adjusts the detection threshold based on image brightness and contrast, improving model robustness. GSConv, a novel convolution method, balances model speed and accuracy by introducing grouping and random permutation. Slim Neck is a lightweight network structure that reduces model complexity and computational load while maintaining performance. These enhancements enable the YOLOv8 model to achieve high accuracy in detecting 'thunderbolt' and 'break' types of cable damage. Experimental results show that the improved YOLOv8 model achieves an average detection accuracy (mAP) of 90.2% for 'thunderbolt' and 86.5% for 'break', with high recall and precision rates. Compared to the original YOLOv8 model, these improvements significantly enhance the model's performance, demonstrating its practical value and strong generalization ability in detecting overhead power line damage. The method also shows high practical value in future research directions. The paper also evaluates the performance of the enhanced YOLOv8 model against other state-of-the-art models, showing its superior performance in terms of accuracy, speed, and efficiency. The model is further tested on video processing, achieving high frame rates and peak performance, indicating its effectiveness in real-time applications. Overall, the enhanced YOLOv8 model demonstrates significant advantages in detecting overhead power line damage, particularly in complex scenarios and class imbalance problems.This paper presents an enhanced version of YOLOv8 for detecting damage in overhead power lines. The proposed method integrates three key improvements: an adaptive threshold mechanism, GSConv, and Slim Neck. The adaptive threshold mechanism dynamically adjusts the detection threshold based on image brightness and contrast, improving model robustness. GSConv, a novel convolution method, balances model speed and accuracy by introducing grouping and random permutation. Slim Neck is a lightweight network structure that reduces model complexity and computational load while maintaining performance. These enhancements enable the YOLOv8 model to achieve high accuracy in detecting 'thunderbolt' and 'break' types of cable damage. Experimental results show that the improved YOLOv8 model achieves an average detection accuracy (mAP) of 90.2% for 'thunderbolt' and 86.5% for 'break', with high recall and precision rates. Compared to the original YOLOv8 model, these improvements significantly enhance the model's performance, demonstrating its practical value and strong generalization ability in detecting overhead power line damage. The method also shows high practical value in future research directions. The paper also evaluates the performance of the enhanced YOLOv8 model against other state-of-the-art models, showing its superior performance in terms of accuracy, speed, and efficiency. The model is further tested on video processing, achieving high frame rates and peak performance, indicating its effectiveness in real-time applications. Overall, the enhanced YOLOv8 model demonstrates significant advantages in detecting overhead power line damage, particularly in complex scenarios and class imbalance problems.