Improving the thermal comfort model for students in naturally ventilated schools: Insights from a holistic study in the Mediterranean climate

Abstract

The thermal comfort models defined in existing standards are usually ineffective in predicting students' mean thermal sensation. Accordingly, overheating and overcooling are frequently reported, which affect students’ well-being and cause unnecessary energy waste. Existing studies have not further explored the model improvement strategies for educational buildings. Hence, this research aims to improve the mean thermal sensation prediction model for students in naturally ventilated schools by incorporating correlated factors, based on a holistic field survey in the Mediterranean climate. The results indicate that the modified extended predicted mean vote of thermal adaptations reinforced around thermal neutrality has the highest accuracy among all validated adaptive predicted mean vote models. However, the performance of the adaptive predicted mean vote models is limited by the deficiencies of the original predicted mean vote model. The overall performance of the improved models was doubled after introducing the identified correlated factors, but the intrinsic error of the predicted mean vote index can slightly affect the models. As a result, the improved model built directly with the identified correlated factors achieved the best performance. Moreover, the potential of the machine learning models is limited by the volume of the mean thermal sensation vote data that can often be obtained, therefore conventional models such as multiple linear regression can be considered an effective and practical tool for developing the mean thermal sensation vote models. The findings of this study benefit the well-being of students and the energy conservation of educational buildings.