Thermal mass of buildings and domestic hot water tanks represent interesting sources ofthermal energy storage readily available in the existing building stock. To exploit them to their full potential,advanced control strategies and a coupling to the power grid with heat pump systems represent the mostpromising combination. In this paper, model predictive control (MPC) strategies are developed and tested ina semi-virtual environment laboratory setup: a real heat pump is operated from within a controlled climatechamber and coupled with loads of a virtual building, i.e., a detailed dynamic building simulation tool.Different MPC strategies are tested in this laboratory setup, with the goals to minimize either the deliveredthermal energy to the building, the operational costs of the heat pump, or the CO2emissions related to theheat pump use. The results highlight the ability of the MPC controller to perform load-shifting by chargingthe thermal energy storages at favorable times, and the satisfactory performance of the control strategies isanalyzed in terms of different indicators, such as costs, comfort, carbon footprint, and energy flexibility. Thepractical challenges encountered during the implementation with a real heat pump are also discussed andprovide additional valuable insights

Peer Reviewed