Residential Appliance-Level Load Forecasting with Deep Learning.

Short-term forecasting of the electric load in a household received significant research interest, with applications that include smart grid systems and the possibility to reduce the energy cost to the homeowner. Most previous research focused on forecasting the load at the level of the whole household. In this paper, we propose a novel approach for forecasting the load of individual electronic devices. Our approach uses a recurrent deep neural network with Long Short Term Memory (LSTM) cells. We train and validate the system using real-world datasets, and show that the approach outperforms the baseline forecasting approaches.