Thermal-comfort optimization design method for semi-outdoor stadium using machine learning

Improving the thermal comfort of spectators is an important aspect of the semi-outdoor stadium design procedure. Although previous studies provided suggestions, a valid optimization method for improving thermal comfort has not been proposed. The main difficulty is the lack of a specific evaluation index for the stadium's thermal comfort and a comprehensive simulation with all the weather conditions in a stadium's use-cycle. This study defined that the PC_ave is averaged percentage of comfortable seats (UTCI temperature between 9 °C and 26 °C) within one year of use. In this paper, the Tianjin Tuanbo tennis stadium was taken as the research object, and the PC_ave was selected as an evaluation index. The objectives were to reveal the relationship between the stadium's shape and thermal performance with an accurate calculation model and to propose a valid morphological optimization method for improving the thermal-comfort performance. Energy simulations and computational fluid dynamics simulations were performed. The appropriate simulation range was identified, and the test mesh was adjusted. The simulation results were close to real measurements. Artificial neural networks and a genetic algorithm were used for optimization, and the PC_ave of the optimized stadium was improved by 8.96%.