Intelligent electricity consumption forecasting and electricity-theft analysis method based on deep learning

Residential electricity consumption data have great mining value for electricity-theft analysis and electricity consumption forecasting. This paper designs an intelligent electricity consumption forecasting and electricity-theft analysis method based on deep learning. Firstly, the power consumption data of a single user is converted into multi-period data through preprocessing, and the data forecasting is realized by designing a deep neural network model. The electricity-theft data are first amplified through user’s power data and electricity-theft labels. And then the time series are converted into two-dimensional tensors for classification and recognition by deep CNN. The classification accuracy of electricity-theft reaches 92.2% compared with RNN method.