Hybrid models for simulating indoor temperature distribution in air-conditioned spaces

Abstract

Air conditioning systems play a vital role in enhancing thermal comfort for building occupants in hot climates. However, their high energy consumption and impact on outdoor air temperatures highlight the need for intelligent, fast, and real-time information systems to ensure sustainable operation. In this study, we evaluate the performance of various models in simulating the thermal and airflow dynamics of a room regulated by an Air Conditioning (A/C) system. Three modelling approaches are examined: a state-of-the-art Computational Fluid Dynamics (CFD) model using OpenFOAM, a physics-based surrogate model utilizing Model Order Reduction techniques, and a hybrid model that combines the surrogate approach with a correction term based on real-time, on-site measurements. The results indicate that while both the CFD and surrogate models perform well overall, they fail to capture localized airflow features that in the measuring points are accurately predicted by the hybrid model. These findings highlight the potential of advanced data-driven models, particularly hybrid approaches, for the intelligent and sustainable management of building environments.