Conventional heating, ventilation and airconditioning (HVAC) features such as use of static occupancy schedule profile to control HVAC operation or traditional controls may not be enough to cope with requirements of the next generation-built environment. This work introduces a demand-driven deep learningbased framework which can be integrated with building energy management systems (BEMS) to accurately predict occupancy’s activity for HVAC systems which can minimize unnecessary loads and produce satisfactory thermal comfort conditions for occupants. The developed framework utilises a trained deep learning algorithm and an artificial intelligence (AI)-powered camera. Tests are performed with new data fed into the framework which enables predictions of typical activities in buildings such as walking, standing, sitting and napping. To initially test and validate the framework, building energy simulation was used with various occupancy profile schedules under a modelled UK office building with 4 occupants. Initial results present occupancy heat gains were 23.5% lower when Deep Learning Influenced Profile (DLIP) was used as compared to static office occupancy profile. Further developments include; framework enhancement to increase detection accuracy and to provide automated set point adjustment for HVAC system. Initial data indicates the method could resolve occupancy related problems within buildings and enhance building performances through accurate occupancy activity prediction.
Keywords Artificial intelligence, deep learning, building energy management, occupancy detection, activity detection, HVAC systems