Sensitivity of night cooling performance to room/system design: Surrogate models based on CFD

Night cooling, especially in offices, attracts growing interest. Unfortunately, building designers face considerable problems with the case-specific convective heat transfer by night. The multizone building energy simulation programs they use actually need extra input, from either costly experiments or computational fluid dynamics (CFD) simulations. Alternatively, up-front research on how to engineer best night cooled spaces can thrust the application of night cooling. The authors of underlying paper set up a global surrogate-based optimization procedure to find room/system design solutions which induce a high convective heat flux during night cooling in a generic open plan office. This fully-automated configuration of data sampling, geometry/grid generation, CFD solving and surrogate modelling generated several surrogate models. These surrogate models indicated how the convective heat flux in the night cooled open plan office related to the ventilation concept, the thermal mass distribution, the geometry and the driving force for convective heat transfer. Actually these surrogate models merely guided the data sampling towards the global optimum. However, they also provided additional rough-hewn insights into the global behaviour. The results indicated that cases with thermal mass at the floor produce convective heat fluxes which are significantly higher than the ones with thermal mass at the ceiling. Among these cases, the performance of cross ventilation surpasses the ones of both single sided ventilation and under floor ventilation. Among the cases with a thermally massive ceiling, single sided ventilation seems superior. However, while cross ventilation is generally a robust ventilation concept, single sided ventilation is particularly sensitive to the geometry.