Microbiome-Aware HVAC Control in Green Buildings Using Reinforcement Learning

Occupants spend most of their time indoors, and indoor air quality (IAQ) frameworks and green building certifications such as LEED account for chemical pollutants and energy savings when evaluating the indoor environment and neglect the indoor microbiome. This paper introduces an open-source, reinforcement learning-based HVAC simulation environment designed to treat microbial stability as a controllable target. This includes four components, a multi-zone building airflow model, a quantitative Microbiome Stability Index (MSI), pathogen dominance, and temporal variability. A Deep Q-Network (DQN) agent adjusts zone-level ventilation while an energy-accounting module tracks consumption. The AI-based strategy achieves a 1.2% higher MSI than the fixed-setpoint baseline (2.0 ACH), while lowering average ventilation to just 1.30 ACH, striking a balance between bio-air quality and energy efficiency. MSI peaks at 80% RH, exceeding the standard 40–60% relative humidity range, which suggests existing standards may prioritize mold suppression at the expense of microbial diversity. Across simulated climates, coastal and humid-tropical conditions most readily sustain stable microbiomes, whereas arid and cold-dry climates impose steep energy penalties. The zone level analysis demonstrates that the different spaces can have a different energy input to reach the same MSI level, indicating the optimization availability. These findings outline a path toward microbiome-informed IAQ control for healthy buildings.