Conference Paper

Proceedings of Building Simulation 2019: 16th Conference of IBPSA

Bayesian Inference For Predicting An Inter-Individual Variation Of Thermal Sensation Votes

Jongyeon Lim ¹, Yasunori Akashi ¹, Naoki Yoshimoto ^2
1 The University of Tokyo, Japan
² Hitachi, Ltd,.

DOI: https://doi.org/10.26868/25222708.2019.210235
Abstract: This paper proposes a probabilistic model that can provide the inter-individual variation of thermal sensation under a certain thermal condition. In contrast to the conventional model of predicting thermal sensation votes (TSVs) deterministically, our method in which an ordinal probit regression is adopted can capture stochastic characteristics for TSVs of a large group model. Model parameters were estimated using a hierarchical Bayesian inference technique. The results, based on a subset of the ASHRAE Global Thermal Comfort Database II, show that the proposed model provides accurate prediction of TSVs’ variation. The PMV–PPD relationship was compared among the proposed model, field data, and Fanger’s model. Unlike Fanger’s model that has the lowest PPD in the neutral prescribed by the PMV, the highest satisfaction of field data was shifted toward slightly cool side and our prediction model captures this skewed trend.
Keywords: Thermal sensation, Hierarchical Bayesian modeling, Parameter estimation, Uncertainty assessment, Nudge
Pages: 2136 - 2142
Paper: