Search for: All records

Thermally dynamic building envelope is a promising technology to achieve building energy saving while improving thermal comfort. Their performance is highly dependent on the local climate conditions as well as on the way the dynamic properties are operated/controlled. The evaluation of whole building performance through building energy simulation can be useful to understand the potentials of different dynamic opaque envelope with active insulations in a specific context. This paper evaluates the potential to use model-free online reinforcement-learning (MFORL) control to regulate the behavior of dynamic building envelopes. Specifically, two control strategies were formulated and evaluated on dynamic opaque envelopes that consist of a concrete layer sandwiched between two active insulations on both sides of the thermal mass: (i) simple temperature-driven rule-based control, and (ii) MFORL control. The controllers were preliminarily tested in two scenarios with 10-day representative behavior under mild climate or during transitional seasons. The results show that MFORL control is promising in achieving adaptive thermal behavior for dynamic building envelopes and may have advantages over traditional rule-based controllers under complex environment.