Development and Comparative Analysis of BMI-Based Thermal Comfort Correction Coefficients for Occupant-Centric Thermal Environment Control

Purpose: Thermal comfort is a critical determinant of occupant health and well-being, and the predicted mean vote (PMV) model has been widely adopted as a standard tool for its assessment. However, PMV consistently shows systematic discrepancies with actual thermal sensation votes (TSV) due to its limited capacity to reflect individual physiological characteristics such as body mass index (BMI). This study aims to address this limitation by developing BMI-based PMV correction coefficients. Method: Using the ASHRAE Global Thermal Comfort Database II, PMV prediction errors were analyzed across BMI categories. The corrected index PMV* is defined as PMV*=PMV+t, where t is a data-driven correction coefficient derived for each combination of BMI category and PMV range. Two models were developed in parallel: a Ridge PMV* model prioritizing interpretability and an XGBoost PMV* model focusing on prediction accuracy, both providing correction coefficient lookup tables. Result: The Ridge PMV* model reduced MAE and RMSE by approximately 8.7% and 8.3% compared to the baseline PMV, while the XGBoost PMV* model achieved superior performance with reductions of approximately 24.0% and 20.2%, respectively. The proposed approach is expected to serve as a practical foundation for occupant-centric thermal environment control, particularly when integrated with non-intrusive image-based BMI estimation technologies.