Data-driven state-space modeling of indoor thermal sensation using occupant feedback

Current thermal comfort models (Fanger's model or adaptive thermal comfort model) predict thermal sensation in a steady state environment. There has been increasing interests in developing models of dynamic thermal sensation (DTS) due to transient environment conditions, e.g., sudden ambient temperature changes. In this paper, we develop a data-driven Hammerstein-Wiener (HW) model to characterize the dynamic relation between ambient temperature changes and the resulting occupant thermal sensation. In the proposed HW state-space model, thermal sensation is defined as the state variable, and the output measurement corresponds to occupant actual mean votes (AMV), which could be corrupted by sensor noise including psychological habituation or expectation and other non-thermal factors. We have conducted a chamber experiment and the collected thermal data and occupants' thermal sensation votes are used to estimate the model coefficients of the Hammerstein-Wiener model. We evaluate the performance of the proposed HW model and also compared it to other thermal sensation models in the literature.