Quantifying structural uncertainties in Reynolds-averaged Navier–Stokes simulations of wind turbine wakes

Abstract

Reynolds-averaged Navier Stokes (RANS) based modeling is considered the mainstream computational fluid dynamics (CFD) approach for wind energy applications. Considering the inherent shortcomings associated with RANS models, quantification of uncertainties is of obvious importance if these models are to be used for design and optimization of wind farms. In the present work, structural uncertainties of RANS closure are quantified for simulations of wake flow behind a stand-alone wind turbine. The uncertainty is modeled by introducing perturbations to the Reynolds stress tensor. We specifically focus on perturbations in the eigenvalues of the tensor. The k-ω SST model and large-eddy simulation (LES) data are used as the baseline RANS model and reference data, respectively. Initially we compare the unperturbed RANS simulation against LES data, and show that the k-ω SST model generally tends to predict higher levels of isotropy in the turbulent wake. A comparison between LES data and differently perturbed k-ω SST simulations is made for the evolution of velocity deficit and turbulence intensity behind the turbine. A satisfactory coverage of LES profiles is observed when the amount of introduced perturbation is adjusted based on a priori comparison of the baseline RANS and LES results with the exception of the turbulent intensity immediately behind the turbine.