Field study of data-driven model predictive control for GEOTABS in an ultra-low energy office building

Abstract

The combination of a thermally activated building system (TABS) and ground source heat pump (GSHP), which is known as GEOTABS, has been demonstrated to be a promising technology to improve building energy efficiency and thermal comfort. However, the control of such systems is challenging due to the high thermal inertia and slow thermal dynamics. This paper proposes a novel data-driven model predictive control method and demonstrates it in an ultra-low energy office building called HouseZeroTM. First, the thermal dynamics of the target zone were modeled with a data-driven model based on historical operational data. Second, a multi-objective model predictive control optimization problem was formatted to maximize the energy efficiency subject to maintaining thermal comfort. Then, the implemented control algorithm was deployed to the building systems through the Internet of Things (IoT) infrastructure of the building. The experiment result concluded that the data-driven model predictive control saved 12.7% of energy compared to the rule-based control that was tuned over the last three years of building operation. The number of unmet hours in terms of thermal comfort during the occupied hours decreased by 74%.