An Approach for <scp>HVAC</scp> Operational Improvement Using Machine Learning Models to Achieve a Comfortable Indoor Thermal Environment and Air Quality

ABSTRACT Optimizing HVAC operation in buildings while maintaining thermal comfort and indoor air quality (IAQ) is a major challenge. This study presents a framework that combines Gaussian Process Regression (GPR) models with an exhaustive search method to improve HVAC operation. The models were trained on Building Energy Management System (BEMS) data, with key features selected using the RReliefF algorithm. The GPR models achieved high prediction accuracy, with an RMSE of 0.22°C and R 2 of 0.813 for indoor temperature, and an RMSE of 34.45 ppm and R 2 of 0.919 for CO 2 concentration. For operational improvement, the supply air volume of the Dedicated Outdoor Air Handling Unit (DAHU) was adjusted to maintain CO 2 concentration near 700 ppm, while the supply air temperature of the Air Conditioning Unit (ACU) was controlled to keep indoor temperature between 26°C–27°C. These adjustments reduced fluctuations and lowered mean values, confirming improvements in both thermal comfort and IAQ. The proposed method requires no additional sensors or complex physical models, and its low computational cost ensures feasibility on standard PC environments. Overall, the results demonstrate that combining GPR‐based prediction with exhaustive search enables simultaneous enhancement of thermal comfort and IAQ, with practical applicability verified using real BEMS data.