Building energy and comfort management through occupant behaviour pattern detection based on a large-scale environmental sensor network

Detection of occupant presence has been used extensively in built environments for applications such as demand-controlled ventilation and security. However, the ability to discern the actual number of people in a room is beyond the scope of most current sensing techniques. To address this issue, a complex environmental sensor network is deployed in the Robert L. Preger Intelligent Workplace (IW) at Carnegie Mellon University. The results indicate that there are significant correlations between measured environmental conditions and occupancy status. It is shown that an average of 83% accuracy on the occupancy number detection was achieved by Gaussian Mixture Model based Hidden Markov Models during testing periods. To illustrate the consequent energy impact based on the occupant behaviour detection (i.e. number and duration of occupancy) in the space, an EnergyPlus model of the IW with an assumed standard variable air volume (VAV) system is created. Simulations are conducted to compare the energy consumption consequences between a prescribed occupancy schedule according to the ASHRAE 90.1 base case with the predicted occupancy behaviour. The results show that energy saving of 18.5% can be achieved in the IW while maintaining indoor thermal comfort.