Cohort-based personal comfort models for HVAC occupant-centric control

One important goal of the built environment is to provide thermal comfort, since it affects occupant satisfaction, health, and performance. Existing approaches make use of pre-defined rules with pre-defined set-points, i.e., the temperature set-points in an indoor environment, which in turn contributes to one third of the total energy consumption of buildings. Research has explored advanced modelling and sensing techniques for prediction models in order to define better temperature set-points but such prediction remains a fundamental challenge due to the individual preferences of occupants. In this current work I systematically explored data requirements for thermal preference modeling and prediction by generating synthethic data and by leveraging occupants similarities. The results show that augmenting existing datasets provides around 4% improvements in prediction models and grouping users based on pre-calculated similarities allows an average increase of 5% for half the occupants in prediction with less historical data (or sometimes none). As next steps, my research aims at combining these new modeling approaches with occupant-centric control for a complete human-in-the-loop control framework.