Do More Sophisticated Models Improve the Accuracy of Wind Resource Estimates?
In the wind energy industry, a popular approach is to rely on linear wind flow models (e.g. WAsP, MsMicro) to estimate the averaged wind speed, wind rose and other boundary layer characteristics near a site. Linear wind flow models are relative simple and computationally efficient but still perform reasonably well where the wind is not significantly affected by steep slopes, thermally driven flows, low-level jets, etc.
More than ten years ago, Risoe DTU National Laboratory and AWS Truepower independently pioneered a method to couple a prognostic numerical weather prediction (NWP) model with a diagnostic microscale model (linear or a mass-conserving) for wind resource characterization at horizontal grid spacings on the order of 50 to 100 m. In principle, fully compressible, non-hydrostatic NWP models can simulate a broad range of meteorological phenomena from synoptic to micro scales but the required computing power is substantial as the grid spacing is refined. To circumvent this issue, the NWP model is run using nested grids with the smallest grid spacing on the order of 1 km. Then, the microscale model is initialized with averaged conditions from the NWP model to downscale the wind flow from 1-km to 50-m grid spacing based on high resolution terrain and land use data sets. Coupled NWP and microscale models have proven to be more accurate than the industry standard WAsP model in complex terrain.
Computational Fluid Dynamics (CFD) and Large-Eddy Simulation (LES) models are considered the next generation of wind flow models. CFD models are designed to solve both the mass and momentum conservation equations. For idealized cases, they perform well and give a high level of detail on the turbulence characteristics of the flow. However, most CFD models are run in a steady-state mode and are thereby inherently incapable of capturing unsteady effects, like intermittent separation and eddy generation / transport. In addition, CFD models do not solve the energy conservation equation although in many instances it is important to include temperature gradient effects that model atmospheric boundary layer flows. So far, there is no clear evidence that CFD models perform better than the industry standard WAsP model.
On the other hand, LES models have their origin in meteorology and weather prediction so they include the (unsteady) Navier-Stokes equations with physical parameterizations (e.g. radiation, cloud microphysics, convection, etc.). LES models are run at very high resolution and are therefore able to explicitly resolve the energetically important eddies of the flow while parameterizing the small ones. However, LES models are mainly used as a research tool since the required computing power represents a major hurdle.
The present study aims at characterizing the mean and turbulent wind flow over complex terrain using: i) WAsP (linear model), ii) SiteWind (a coupled mesoscale/microscale system); iii) Meteodyn WT (CFD model), and iv) Advanced Regional Prediction System (ARPS; configured as a Very Large Eddy Simulation or VLES model). The sites include a steep hill/ridge near Lake Superior and a mountainous area in Wyoming. In theory, the VLES is the most sophisticated approach since it relies on a single state-of-the-science numerical model with appropriate physics and dynamics to handle most scales of motion relevant for wind mapping while overcoming the limitations of diagnostic microscale models. The output from each model is compared to onsite meteorological observations. The presentation will focus on an analysis of wind speed and wind rose variability across the sites, as well as the vertical profiles and diurnal cycles. The results will be used to address the question posed in the title of the paper.