A data-driven state-space model of indoor thermal sensation using occupant feedback for low-energy buildings

Abstract

A data-driven state-space Wiener model was developed to characterize the dynamic relation between ambient temperature changes and the resulting occupant thermal sensation. In the proposed state-space model, the mean thermal sensation state variable is governed by a linear dynamic equation driven by changes of ambient temperature and process noise. The output variable, corresponding to occupant actual mean vote, is modeled to be a static nonlinearity of the thermal sensation state corrupted by sensor noise. A chamber experiment was conducted and the collected thermal data and occupants’ thermal sensation votes were used to estimate model coefficients. Then the performance of the proposed Wiener model was evaluated and compared to existing thermal sensation models. In addition, an Extended Kalman Filter (EKF) was applied to use the real-time feedback from occupants to estimate a Wiener model with a time-varying offset parameter, which can be used to adapt the model prediction to environmental and/or occupant variability. Future studies can use this model to dynamically control the Heating Ventilating and Air Conditioning (HVAC) systems to achieve a desired level of thermal comfort for low-energy buildings with actual occupant feedback.