Combined-LSTM based User Electricity Consumption Prediction in a Smart Grid System

The prediction of user electricity consumption is important for smart grid systems. Conventional methods usually relay on the prior data distribution and estimate parameters by time series models. It is still a challenge on how to directly learn electricity consumption patterns from users' history records, especially for the connection between peak and valley consumptions. Therefore, in this paper, we propose a novel deep neural network based model, named as Combined-LSTM, to predict the power consumption based on users' records. It could directly learn hidden relationships among peak and valley volume throng combined recurrent networks. Experiments in real-world scenario demonstrates the effectiveness of our proposed method. It outperforms several state-of-art methods in both prediction accuracy and computational complexity.