A deep reinforcement learning approach to using whole building energy model for HVAC optimal control

Whole building energy model (BEM) is difficult to be used in the classical model-based optimal control (MOC) because of its high-dimension nature and intensive computational speed. This study proposes a novel deep reinforcement learning framework to use BEM for MOC of HVAC systems. A case study based on a real office building in Pennsylvania is presented in this paper to demonstrate the workflow, including building modeling, model calibration and deep reinforcement learning training. The learned optimal control policy can potentially achieve 15% of heating energy saving by simply controlling the heating system supply water temperature.