Model predictive control is an important controlmethod to reduce building heating and cooling energy consumptions. However, the mechanism by which the energy savings are achieved is not well understood. This paper investigates such mechanism using building energy simulation. The simulation results show that the better constrained indoor temperature leads to lower heating and cooling loads, which results in reduced energy consumptions. Comparing with a conventional 2 Â°C dead-band control, a model predictive control, which restricts the indoor temperature within 0.2 Â°C range of the set point, obtains an 8.5% and a 13.6% annual energy savings in heating and cooling respectively.
Keywords indoor temperature variation, application of machine learning in buildings, heat pump variable speed control, building model predictive control, advanced HVAC control