Model-Free HVAC Control Using Occupant Feedback

Optimal control of Heating, Ventilation, and Air Conditioning (HVAC) is an important step towards reducing the carbon footprint of buildings and requires balancing energy reductions and occupant comfort. Conventional thermostats for temperature set points provide a centralized single point of user input, often leading to significant thermal discomfort for occupants. We propose to instead include users in the HVAC control loop through distributed smart-phone based votes about their thermal comfort for aggregated control of HVAC. Unlike existing approaches that require in-situ sensors or build complex comfort models of individual users, we propose a model- and sensor-free HVAC control algorithm that uses simple user input (hot/cold) and adapts to changing office occupancy or ambient temperature in real time. We develop an iterative data fusion algorithm that finds optimal temperature in offices with multiple users and propose techniques that can aggressively save energy by drifting indoor temperatures towards the outdoor temperature. Our evaluation is based on empirical data collected in 12 offices over a 3-week period and shows that adaptive HVAC control can save up to 60% of energy at a relatively small increase of 0.3 ◦C in average occupant discomfort.