Subzone division optimization with probability analysis-based K-means clustering for coupled control of non-uniform thermal environments and individual thermal preferences

Collective air distribution is widely used for thermally comfortable indoor environments but is limited to unified thermal comfort requirements. Advanced collective air distribution has the potential to generate non-uniform thermal environments for individual thermal preferences. This study proposes a subzone division optimization method for nonuniform thermal environments to improve thermal preference satisfaction. The proposed method divided non-uniform thermal environments into subzones using probability analysis-based K-means clustering, minimizing and maximizing the thermal environment differences within and among the subzones, respectively. The number of subzones was determined to maximize thermal preference satisfaction with the coupled control between the nonuniform environments and individual thermal preferences while limiting the percentage of thermal discomfort. The results based on experiments of advanced collective air distribution, i.e., stratum ventilation, showed that thermal preference satisfaction and the percentage of thermal discomfort increased with an increasing number of subzones. Compared with the conventional method (i.e., uncoupled control without subzone division optimization) and recent method (i.e., coupled control without subzone division optimization), the proposed method improved the thermal preference satisfaction by 34.7 % and 11.7 %, respectively, for four subzones. Moreover, with the proposed method, increasing the system control flexibility increased thermal preference satisfaction. Compared with the variable air volume system, the constant air volume system and the system with variable supply airflow rate and temperature improved thermal preference satisfaction by 5.8 % and 22.5 %, respectively. The proposed method contributes to extending the capability of collective air distribution for individual thermal preferences. © 2023 Elsevier Ltd.