Improving thermal model predictions for naturally ventilated buildings using large eddy simulations

Abstract

Natural ventilation involves complex flow and heat transfer processes that are challenging to model accurately and efficiently. Computational fluid dynamics (CFD) simulations can accurately represent the flow and heat transfer physics but are computationally expensive; while building thermal models are computationally efficient but they use reduced-order models to represent complex flow phenomena. In this study, we aim to use CFD to develop building-specific, accurate correlations for the flow and heat transfer rates in buoyancy-driven natural ventilation flow, such that these correlations can be used to improve the accuracy of a dynamic thermal model. First, a CFD model of an operational building is validated with full-scale experiments. Next, the validated CFD model is used to run nine large eddy simulations, from which correlations for the convective heat transfer and the natural ventilation flow rate and heat transfer are derived. The results suggest that standard correlations for these processes can differ significantly from the building-specific values. The CFD-based correlations are then used in a dynamic thermal model to improve its accuracy. The resulting CFD-informed dynamic thermal model accurately predicts indoor air temperature, surface temperature and heat transfer, while the model employing the standard correlations (without inputs from CFD) only predicts the indoor air temperature accurately, not the surface temperature and heat transfer. The fast yet accurate CFD-informed dynamic thermal model can be used to run many simulations in a short time, supporting uncertainty quantification to quantify the effect of variability in the operating conditions and real-time prediction of the natural ventilation process.