Demand-driven building control is an emerging approach to mitigate the increasing pressures on buildings and facilities for requirements of energy and comfort services. This study proposes a framework that integrates online learning capabilities to make building systems adapt to occupants’ actual energy and comfort demand. Based on the framework, two types of control strategies are developed: occupancy-based and thermalpreference-based demand-driven controls. Both of them have been implemented in an office building, keeping occupants in the loop of building operation under realistic conditions. This paper first introduces the proposed framework, and then presents two types of controls applied in for a case study. Lastly, lessons learnt from conducting them in the field tests are discussed.
Keywords Smart Buildings, Occupant Behavior, Thermal Comfort, Demand Response, Machine Learning