Prediction of CO₂ in Public Buildings

Abstract

Heritage from the COVID-19 period (in terms of massive utilization of mechanical ventilation systems), global warming, and increasing electricity prices are new challenging factors in building energy management, and are hindering the desired path towards improved energy efficiency and reduced building consumption. The solution to improve the smartness of today’s building and automation control systems is to equip them with increased intelligence to take prompt and appropriate actions to avoid unnecessary energy consumption, while maintaining a desired level of air quality. In this manuscript, we evaluate the ability of machine-learning-based algorithms to predict CO2 levels, which are classic indicators used to evaluate air quality. We show that these algorithms provide accurate forecasts (more accurate in particular than those provided by physics-based models). These forecasts could be conveniently embedded in control systems. Our findings are validated using real data measured in university classrooms during teaching activities.

Keywords

• air quality control
• machine learning algorithms
• forecasting methods