LAUSR

Abstract Predicting the heating and cooling (HC) energy performance of a building is essential in the understanding and energy-efficient control of HC systems. The aims of this study were to develop and propose advanced and accurate energy prediction models using deep learning techniques. Also, to assess the importance and the significance of the relevant variables used in the models. The models were developed based on measured data collected in an educational building and were classified into different prediction time groups at 3-min, 15-min, 30-min, and 1-h time intervals. The inputs used in the models for the HC system and the EHP were selected through a variable selection process based on domain knowledge and correlation analysis. The results also indicate that the operational factors of the HC system had greater influence on the energy consumption than the indoor and outdoor temperatures. The performances of developed models indicate that a deep learning approach can be effectively applied to predict and understand the electric energy consumption of a HC system. Furthermore, the variable selection process and the important variables identified through it can be applied to energy prediction of HC systems in other buildings.