A state-aware method for mobile distribution network emergency repair tools based on multisource sensor signal fusion

The application of artificial intelligence technologies in distribution network emergency repairs has increasingly become a key factor in enhancing efficiency and reliability. This paper proposes a state-awareness model based on multi-source sensor data fusion, combining multi-head attention mechanisms with LSTM for feature fusion and temporal modeling to accurately identify the status of distribution network repair tools. The model is optimized based on existing deep learning frameworks, with specific enhancements to LSTM and attention mechanisms. Experiments demonstrate that the optimized model achieves 94.8% accuracy across diverse environments while maintaining high recognition capability in low-temperature conditions. Compared to traditional models, it delivers approximately 30% faster inference speed and 20ms reduced latency, exhibiting strong robustness and real-time responsiveness.