Building structures and furniture have the capability of storing thermal energy and therefore could be called building thermal mass. In commercial buildings, room conditions remain in thermal comfort zone for a while after the end of office hours when the air-conditioning systems are tuned off. It is possible to recover the stored cooling energy from building thermal mass in commercial buildings. This study proposes a model-based control strategy for such purpose. A simplified building RC model and a black-box model are combined by a simple data fusion algorithm for easier implementation and more accuracy prediction. The feasibility of such energy recovery was validated on-site in a super high-rise commercial building in Hong Kong, and the proposed method was validated on a dynamic simulation platform built based on the same building. In the two on-site validations, the energy savings during the energy recovering period were 85.8% and 80.1%, respectively. In the simulation tests, by allowing indoor air temperature increase by 1 K, the proposed control strategy could save 83.8% of the cooling energy during the recovering period, which accounts for 7.23% of the total cooling energy consumption in the entire tested days. (C) 2019 Elsevier Ltd. All rights reserved.