Journal of Korean Institute of Architectural Sustainable Environment and Building Systems 한국건축친환경설비학회논문집 ISSN:1976-6483(Print) 2586-0666(Online)

This paper focuses on developing and validating a CO₂ concentration prediction model to optimally control and operate mechanical ventilation systems in residential buildings. Various datasets and models were utilized throughout the research process. Initially, CO₂ concentration was predicted based on a basic dataset comprising the day of the week, time, temperature, and indoor CO₂ levels. Subsequently, a more comprehensive dataset, incorporating ventilation system operation counts and natural ventilation presence, was employed for predictions. Machine learning and deep learning models such as ANN, DNN, LSTM, KNN, and LGBM were trained using these datasets, and their predictive performance was compared across different areas of the residence, namely bedrooms, living rooms, and corridors. The findings demonstrate that even using the basic dataset alone, it is possible to predict CO₂ levels in residential buildings with a high degree of confidence. These results are expected to contribute to the development of future algorithms for operating next-generation mechanical ventilation systems that aim to minimize energy consumption while maintaining a clean indoor environment. This paper is anticipated to increase interest and understanding in environmentally conscious indoor environmental control systems and offer essential research insights for the advancement of building systems and environmental technology in the future.