Identification of Abnormal Electricity Consumption Behavior Based on Bi-LSTM Recurrent Neural Network

Abstract

Abnormal electricity consumption (AEC) results in a huge safety hazard for the power grid. In particular, it is significant for the power grid marketing department to identify users’ AEC behavior. In view of the AEC of power grid users, a prediction model based on bidirectional long short-term memory network (Bi-LSTM) feature extraction network is proposed in this paper, which identifies the AEC behavior based on the historical electricity consumption data of users. The framework of TensorFlow was used to construct a model for feature extraction and result prediction. Some power consumption units in a prefecture-level city were selected for analysis. Experimental analysis shows that, compared with support vector machine (SVM), BP neural network and long short-term memory (LSTM), the proposed model is more accurate and robust in detecting AECs.