Model-based multivariable regression model for thermal comfort in naturally ventilated spaces with personalized ventilation

Abstract

This work proposes a method for developing an accurate correlation to predict thermal comfort (TC) as function of occupant physiological and environmental parameters. This method is implemented for a space that relies on hybrid natural ventilation (NV) and personalized ventilation (PV) cooling. Multivariable linear regression was adopted to develop the TC correlation while retaining variables based on the significance and interdependency. The correlation was found to be dependent on indoor temperature (T indoor), relative humidity (RH), facial temperature (T facial) and its rate of change (dT facial/dt). Sample data from the observations used in developing the correlation and outside-data were utilized to compare simulated and predicted TC over a scale from −4 (very uncomfortable) to +4 (very comfortable). The standard error in estimating TC was 0.4 with a maximum deviation of 1.0. The developed method can be used to derive TC correlations pertaining to other complex dynamic thermal environments with different applications.