Dynamics of Carbon Dioxide Concentrations: Measurements and Simulation of a University Classroom

Abstract

Indoor CO₂ concentrations in classrooms in a school or a university tend to be in a high range which affects student’s cognitive performance. Several models for simulating the dynamics of CO₂ concentrations are developed to optimize the concentrations in the classroom. This study aimed to thoroughly evaluate two simplified modeling choices for numerical simulation of CO₂ concentrations in the classroom, including Teleszewski & Gładyszewska-Fiedoruk’s model and CIBSE AM 10 model. In this study, the real-time input method and fixed input method were both evaluated using the measured data at one-minute intervals from the selected classroom in the General Science Building, Faculty of Science, Chulalongkorn University, Thailand, during January to February 2020. The best performing model was then used to suggest suitable numbers of occupants for certain time spent in the classroom. The Teleszewski & Gładyszewska-Fiedoruk’s model and CIBSE AM 10 model perform well when employing the real-time input method (R² = 0.9965 ± 0.0028, 0.9912 ± 0.0072 respectively) while being unreliable with the fixed input method (R² = -2.7768 ± 5.2673, -20.9980 ± 22.4243 respectively). However, the real-time input method is not practical since the process to measure the CO₂ concentration every minute is tedious. Hence, the fixed input method is leveraged for simulation with the Teleszewski & Gładyszewska-Fiedoruk’s model which yields better performance than the CIBSE AM 10 model in both input methods. Moreover, the model could be further improved based upon residual modeling which resulted in the best predictive performance at 60 minutes (R² = 0.7547 ± 0.1631). The simulations suggest that the classroom should contain no more than to support a 60-minute study before taking a break when assumed no air ventilation before the class. A suitable number of occupants could be raised if the air ventilation is performed before the class begins.