Modeling the 3D microscale wind structures at a large wind energy farm with complex terrain using the NCAR WRF-RTFDDA-LES model

Compared to other typical weather applications, wind energy prediction requires higher accuracy of wind-speed forecasts in the lowest 300m of the atmosphere since wind power is proportional to the cube of the wind speed. Meanwhile, turbulences and wind shear across the wind turbine blade disks can significantly reduce the power production. Furthermore, it is observed that winds frequently vary significantly across a wind farm of a few kilometers and such variation can affect the total power production of the wind plant. Nevertheless, both our knowledge and modeling capabilities about such microscale wind characteristics are very limited. In fact, mesoscale Numerical Weather Prediction (NWP) models, running at ~l km grid sizes, simulate the Boundary Layer processes using column PBL parameterizations that assume equilibrium adjustment for given temperature, moisture, wind and surface fluxes. The observed rich and fast evolving microscale flow features invalidate the PBL parameterization assumptions and thus cause large wind prediction errors. In this paper, the National Center for Atmospheric Research (NCAR) WRF based Real-Time Four-Dimensional Data Assimilation (RTFDDA) and forecasting system was used to study the microscale weather flows at a large wind farm in the Northern Colorado, where there are 274 wind turbines. The model was run for a two-day period using simultaneous nested-grid domains that simulate the weather processes from synoptic scale to large eddy simulation (LES) scale. Six nested domains having grid sizes of 30, 10, 3.333, 1.111, 0.369 and 0.123 km were set up for WRF-RTFDDA. The four coarser domains were run in mesoscale modeling mode, in which a PBL scheme and four dimensional data assimilation were activated, while the two finest grid domains run in the LES mode, with sub-grid PBL turbulence-mixing simulated with 3D TKE. The model results are verified against the measurements of two meteorological towers and the wind speeds measured by the nacelle anemometers of the wind turbines in the farm. We demonstrated 1) the WRF-RTFDDA-LES model capability in resolving many observed microscale flow structures, 2) the deficiency of mesoscale model PBL parameterization for the simulation of the wind shear in the lowest 300m layer, 3) the impact of the model grid resolutions, and 4) the more accurate farm-wide wind simulation through the upscaling of the LES modeled circulations.