Vision-based human activity recognition for reducing building energy demand

Occupancy behaviour in buildings can impact the energy performance and the operation of heating, ventilation and air-conditioning systems. To ensure building operations become optimised, it is vital to develop solutions that can monitor the utilisation of indoor spaces and provide occupants' actual thermal comfort requirements. This study presents the analysis of the application of a vision-based deep learning approach for human activity detection and recognition in buildings. A convolutional neural network was employed to enable the detection and classification of occupancy activities. The model was deployed to a camera that enabled real-time detections, giving an average detection accuracy of 98.65%. Data on the number of occupants performing each of the selected activities were collected, and deep learning-influenced profile was generated. Building energy simulation and various scenario-based cases were used to assess the impact of such an approach on the building energy demand and provide insights into how the proposed detection method can enable heating, ventilation and air-conditioning systems to respond to occupancy's dynamic changes. Results indicated that the deep learning approach could reduce the over- or under-estimation of occupancy heat gains. It is envisioned that the approach can be coupled with heating, ventilation and air-conditioning controls to adjust the setpoint based on the building space's actual requirements, which could provide more comfortable environments and minimise unnecessary building energy loads.