A study on the influence of the numerical scheme on the accuracy of blade-resolved simulations employed to evaluate the performance of the NREL 5 MW wind turbine rotor in full scale

Abstract

We present a numerical investigation on the influence of the numerical arrangement in the accuracy of the NREL 5 MW wind turbine rotor performance evaluation in full scale, using a Computational Fluid Dynamics (CFD) methodology which employs the Finite Volume Method (FVM) implemented in the OpenFOAM software. The nonlinearity of the Navier–Stokes equation, which requires special treatment and still represents a subject under intense debate in the scientific community, is tackled by using two different discretization schemes for computing the convective term. In addition, since the advection scheme numerically interacts with the turbulence closure model employed in the numerical simulations, we considered the URANS k-ω SST and DES k-ω SST models to perform the blade-resolved simulations. A hybrid central/upwind scheme, namely the LUST scheme, was considered for discretizing the convective term for both URANS and DES simulations, while the second order accurate upwind given by the LUD scheme was considered only in the URANS simulation. Considering the LUST scheme, an investigation on the temporal and spatial discretization was performed for both turbulence closure models. The performance of the NREL 5 MW rotor for offshore application in full scale was assessed in terms of power production, generated thrust, and forces distribution along the blade span, for the operation condition of optimal wind-power conversion efficiency. We provide detailed information regarding the flow features and computational cost, and verified the results from each case by comparing the CFD results against values obtained using the blade element momentum theory implemented in OpenFAST. For the spatial discretization considered in one of the meshes, the LUD scheme showed low accuracy in the results, being more susceptible to the grid influence for the URANS simulations compared to LUST, while the URANS-LUST and DES-LUST approaches were both suitable options for all the meshes investigated. The DES-LUST approach was less affected by the variation in the size of the time step employed. Additionally, finer flow structures were captured with the DES-LUST simulations at an affordable computational cost.