Predictive Control of Natural Ventilation In Existing Housing: Optimizing CO₂ Concentration And Energy Use Intensity

Naturally ventilated residential buildings often rely on occupant-controlled window operation to regulate indoor environmental quality (IEQ), resulting in inconsistent ventilation patterns and uncertain trade-offs among CO₂ exposure, thermal comfort, and energy consumption. These challenges are particularly significant in existing residential buildings, where limited mechanical ventilation and variable occupant behavior can contribute to inadequate ventilation and elevated indoor pollutant concentrations. This paper presents a predictive control framework for natural ventilation that integrates environmental monitoring, physics-based airflow estimation, building energy simulation, machine learning, and multi-objective optimization to support performance-informed window operation strategies. The framework was tested in a naturally ventilated dormitory room. High-resolution data, including indoor CO₂ concentration, temperature, occupancy patterns, and window states, were collected. Physics-based CO₂ mass-balance calculations were used to estimate ventilation requirements, while a calibrated energy model quantified the heating-energy implications of multiple ventilation scenarios, including closed-window, continuously open-window, temperature-based, and occupant-driven behaviors. A multi-objective optimization framework was developed to balance three competing criteria: maximum CO₂ above ambient differentials per ASHRAE Standard 62.1, thermal comfort per ASHRAE Standard 55, and incremental heating energy demand. Results demonstrate that occupant-driven ventilation patterns frequently failed to maintain acceptable IEQ, with monitored CO₂ concentrations regularly exceeding 2000 ppm during occupied periods. Unrestricted window opening reduced indoor CO₂ concentrations but increased heating demand and annual EUI, highlighting trade-offs between ventilation performance and energy consumption. Compared to temperature-based ventilation operation, the machine-learning framework achieved lower heating demand while providing a more adaptive response to changing occupancy patterns and environmental conditions. Although findings are specific to the monitored dormitory environment, the methodological framework for environmental monitoring, predictive modeling, and control optimization is adaptable across residential contexts. The research demonstrates how data-driven natural ventilation control can improve IEQ while minimizing energy penalties in existing residential buildings.