Data-driven optimization of building-integrated ducted openings for wind energy harvesting: Sensitivity analysis of metamodels

Metamodels are developed and used for aerodynamic optimization of a ducted opening integrated into high-rise building to maximize the amplification factor within duct. The duct consists nozzle, throat, diffuser. 211 high-resolution 3D RANS CFD simulations performed generate training testing datasets. space-filling design Genetic algorithm data sampling optimization, respectively. performance five commonly-used metamodels is systematically investigated: Response Surface Methodology (RSM), Kriging (KG), Neural Network (NN), Support Vector Regression (SVR), Aggregation (GARS). investigation based on (i) detailed in-sample out-of-sample evaluations metamodels, (ii) annual available power in wind (Pavailable), (iii) energy production (AEP) 3-bladed horizontal-axis turbine (HAWT) installed mid-throat optimum designs obtained by metamodels. results show that converging-diverging openings can magnify experienced speed enhance power. In addition, use different lead variation up 153% estimated Pavailable. For small dataset, crude yet still acceptable accuracy be achieved at very low computational time.