A Study on the rECOGNITION of Anomalous Data in Distribution Network Balance Area Based on LSTM Combined with Variational Self-Coder

Aiming at the problem of low accuracy of load data anomaly detection methods in different balancing zones of distribution networks, we propose a load anomaly detection method for distribution networks by combining the long and short-term memory network and the variational self-encoder. The optimized short-term and long-term memory network is optimized by the residual structure to improve the feature learning ability, and the optimized short-term and long-term memory network replaces the neural network layer of the variational self-encoder, which can better obtain the temporal correlation of the load data. The superiority of the proposed detection method is verified by comparing with the conventional detection method. The results show that compared with the conventional load data anomaly detection method, the proposed method has better detection accuracy and can provide a certain reference for the development of distribution network balance area.