Towards multi-domain models of thermal comfort : predicting occupants’ thermal satisfaction as a function of CO₂ concentrations Crosby, Sarah; Rysanek, Adam
In recent years, emerging studies have examined the multi-domain nature of thermal comfort and IEQ with the goal of reducing the gap between predictions of thermal comfort models and real perceptions of thermal comfort. In a recent work, we applied Bayesian inference methods to correlate and quantify the relationship between perceived thermal comfort, thermal indoor conditions, and non-thermal metrics of IEQ such as CO₂ concentrations and indoor noise levels. In the first phase, we made use of the COPE dataset, a field study of objective and subjective IEQ measurements from 800 occupants of open-plan offices conducted by the National Research Council of Canada in large Canadian and US cities. Bayesian regression of the COPE dataset yielded a significant correlation between perceived thermal comfort and measured values of indoor CO₂ correlations. In the second phase, the prior findings were updated by adding 150 new samples of IEQ measurements collected from occupants of office spaces at the University of British Columbia in 2019. Bayesian logistic regression of the expanded dataset revealed stronger evidence to suggest that perceived thermal comfort is independently correlated with measured indoor CO₂ concentrations. The statistical significance of these results is validated using several Bayesian model validation techniques which, in turn, validates the robustness and significance of our prior findings. This paper formulates and applies a new predictive model of thermal comfort, derived from the Bayesian logistic regression of the COPE and UBC datasets. The model is implemented in a building energy model which can be used by building modellers and performance simulation experts to predict thermal comfort in office settings based on thermal conditions and ventilation rates, which may result in energy savings while not sacrificing indoor air quality and well-being, an important challenge to building modellers, particularly now in a post-COVID-19 world.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International