Advancements in Power Grid Monitoring: UAV-Enhanced Transformer Neural Networks for Knife-Switch Status Detection

With the advancement of the smart grid, efficient and accurate monitoring of the status of electric power knife switches has become particularly important. Traditional monitoring methods for knife switch status are hampered by high costs, susceptibility to environmental influences, and complex data processing. This study introduces an intelligent monitoring method for electric power knife switch status based on unmanned aerial vehicle (UAV) imagery and a transformer neural network model. Utilizing UAVs for image acquisition, coupled with advanced image preprocessing techniques, the efficiency of data collection is enhanced while ensuring the quality of image data. The transformer neural network model introduced, particularly the ReViT model, employs residual attention connections and multi-head self-attention mechanisms, effectively improving the accuracy of knife switch status recognition under complex environmental conditions. To comprehensively evaluate the model's performance, this study developed an integrated assessment metric, SSICS, which considers the state matching rate, weighted consistency ratio, and harmonic stability coefficient, providing a thorough evaluation of the model's performance under various conditions. Experimental results demonstrate that the proposed method exhibits superior performance in safety, efficiency, cost, robustness, and accuracy, significantly outperforming existing technologies. Future research will focus on enhancing the system's generalization capability and real-time processing capacity, further broadening its application in intelligent grids and automated facility maintenance.