Intelligent Detection and Processing Technology of Abnormal Data in Low-Voltage Distribution Networks

Anomaly detection in distribution networks is crucial for ensuring the stable operation of power systems. To improve the accuracy and efficiency of anomaly identification, we designed a distribution network anomaly detection system based on deep learning. This system employs an improved LSTM model, incorporating attention mechanisms and residual connections, significantly enhancing model performance. The application of a microservice architecture achieved high scalability and real-time processing capabilities. Experimental results show that the improved model reduces the root mean square error by 23.5% on actual datasets, with an average detection time of only 50 milliseconds per instance. During 30 days of actual operation, the system achieved an accuracy of 98.7% and an F1 score of 98.3%. Performance evaluations demonstrate the system's excellent real-time capabilities, scalability, and reliability. These achievements provide strong support for the intelligent management of distribution networks, significantly improving grid operation efficiency and reliability.