Recognition of Meteorological Conditions in Snowy Days for Transmission Based on Deep Learning

The stable operation of transmission lines under severe weather conditions is critical to the reliability of power systems. Accurately identifying the state of transmission lines during snowy and other extreme weather conditions can enhance the efficiency of grid maintenance and dispatch. This thesis proposes a computer vision-based method for recognizing the snow weather state of transmission lines. By constructing a dataset of images taken during clear and snowy days, this study compares and analyzes the performance of several deep learning models, including ResNet, RepLKNet, and TinyViT, in snow weather recognition. Experimental results demonstrate that the RepLKNet model exhibits the best accuracy and robustness, effectively distinguishing between clear and snowy weather images with an accuracy of over 99%. Compared to traditional methods, the proposed deep learning-based approach significantly improves the accuracy and efficiency of snow weather state recognition for transmission lines. The findings provide technical support for intelligent maintenance of power systems under extreme weather conditions.