Electrical behavior recognition based on bidirectional LSTM

To solve the problem of abnormal user power consumption caused by non-technical losses, a method for detecting abnormal power consumption behavior based on a bidirectional long-term short-term memory neural network (Bi-LSTM) is proposed. In this method, firstly, the missing data is processed by interpolation method, and the asymmetric distribution of abnormal electricity consumption behavior value is smoothed by quantile normalization. Then, the Bi-LSTM model is constructed by combining the LSTM neural network unit and bidirectional network, which is used to obtain the implicit relationship of abnormal electricity consumption behavior of users. Finally, the cross entropy is used to determine the optimal parameters, thereby detecting the abnormal power consumer. The experiments are conducted based on real energy consumption data, and the results show that the detection performance of this proposed method outperforms other methods in terms of accuracy and stability.