The Scaled Wind Farm Technology (SWiFT) facility in Lubbock, Texas, is hosted at the Texas Tech University (TTU) National Wind Institute :cite:p:`Hirth2014` and operated by Sandia National Laboratories. This site was previously chosen as an International Energy Agency Task 31 ("Wakebench") benchmark for wind-turbine-wake evolution and dynamics. For MMC research, this site is an attractive validation dataset because it has flat terrain with uniform land cover. Essential validation data include a heavily instrumented 200-m meteorological mast with 10 measurements heights offering high-frequency wind measurements from 0.9 to 200 m above ground level (AGL). These are complemented by a radar wind profiler with radio acoustic sounding system (RASS) for measurements of wind and temperature profiles above 200 m, as well as data from the West Texas Mesonet. Details of the site characterization are provided in :cite:t:`Kelley2016`.
The SWiFT facility with adjacent TTU atmospheric facilities.
A classic diurnal cycle was identified between 8-9 November 2013 based on criteria outlined in :cite:t:`MMCYear2Report`. Conditions were generally clear and considered quiescent. This diurnal cycle featured a morning transition, daytime convective boundary layer, evening transition, and nocturnal low-level jet (LLJ). The convective, neutral, and stable periods were separately assessed for simulations with MYNN and YSU PBL schemes. Despite having a warm bias of up to ~5 K, WRF captures the correct trends in wind speed, direction, and virtual potential temperature :cite:p:`MMCYear2Report`.
A separate study into the effect of the terra incognita on modeling was conducted by :cite:t:`Rai2019`. This study considered 3 cloud free days at SWiFT: 13 July 2013, 22 September 2013, and 6 June 2014. Proper Orthogonal Decomposition analysis clearly indicates that the microscale contains energetic modes that originated from the mesoscale flow. Depending on horizontal grid spacing and turbulence modeling choices, flow from the terra incognita may downscale into unrealistic flow in the microscale.
Relevance to Wind Energy
- Nonstationary conditions result in time-varying hub-height wind speed and direction, wind shear and veer, and turbulence intensity.
- Accurate downscaling of energy from the microscale is important for predicting realistic turbulent flow features in the wind-farm operating environment.
MMC Techniques Demonstrated
- Ensemble mesoscale modeling and assessing best performers + model sensitivity
- Online (WRF/WRF-LES) and offline (WRF/SOWFA) coupling between NWP models and microscale LES
- Internal coupling with two methods: the profile assimilation and mesoscale budget components approaches
The WRF simulation setup contains 3 nested domains with 27 km, 9 km, and 3 km grid spacing, centered at the SWiFT site (:numref:`fig-SWIFT-WRF-doms`). Simulations were started at 00:00 UTC on 8 November 2013. Initial and boundary conditions were set by the final analysis data from the Global Forecast System. Turbulence modeling was provided by the MYNN Level 2.5 PBL scheme. Other physics parameterization are detailed in the WRF namelist.
WRF domain configuration for SWiFT mesoscale simulation.
WRF Setup Available
The WRF setup is available on the WRF-setups repository of the A2e-MMC GitHub.
Two SWiFT datasets were used in the following MMC studies and are freely available on the A2e Data Archive and Portal (DAP).
200-m tower :cite:p:`DAP_TTUtower`: The tower data analysis includes data standardization and sonic tilt correction, and calculates turbulence quantities of interest. Additional calculations include virtual potential temperature, stability (bulk Richardson number and Obukhov stability parameter), and the surface heat flux. A more indepth analysis of the TKE shows anomalous observations around 10:00 UTC on 9 November 2013.
The mean quantities are combined with the radar dataset (described next) to form the input data for the :cite:t:`Allaerts2022` study. The turbulence quantities are used to validate the LES predicted turbulence in :cite:t:`Allaerts2020,Draxl2021,Allaerts2022`.
Radar wind profiler with RASS :cite:p:`DAP_TTUradar`: The radar data analysis shows two wind scanning modes, short range (up to 2 km AGL) and long range (up to 6 km) and corresponding temperature profiles up to a maximum of 800 m AGL. The same notebook also performs the data reconstruction to create the mesoscale forcing dataset for the :cite:t:`Allaerts2022` study.
Preprocessing notebooks available
The SOWFA inputs were generated with the notebooks in the assessment repository: studies/SWiFT/coupling_comparison/preprocessing.
Results from the development and validation of the profile assimilation technique :cite:p:`Allaerts2020`, which couples the WRF mesoscale NWP model to SOWFA LES, are postprocessed in the studies/SWiFT/profile_assimilation_wrf/produce_figures.ipynb notebook. This study demonstrated that simple data assimilation techniques (i.e., direct profile assimilation) can lead to nonphysical shear and turbulence production, due to the algorithm's inability to cope with inaccuracies in the mesoscale data. Applying mesoscale forcing with vertical smoothing (i.e., indirect profile assimilation) improves predictions of turbulence statistics (:numref:`fig-WRFPAT_TKE_comparison`).
Time history of turbulent kinetic energy at 80 m AGL for direct/indirect profile assimilation and budget components coupling, in comparison with the TTU meteorological tower; vertical dashed black lines indicate sunrise and sunset
Results from the evaluation of coupling the WRF mesoscale NWP model to SOWFA through mesoscale budget components :cite:p:`Draxl2021` are postprocessed in the studies/SWiFT/budget_components_coupling/plot_* notebooks. This work shows that mesoscale models can have difficulties predicting profiles of shear and veer. While LES can improve shear and veer predictions, the wind speed and direction are not adjusted. When forcing the LES with the mesoscale budget, spatiotemporal averaging of the forcing terms is not necessary.
Resulting Publications
.. bibliography:: ../all_project_pubs.bib
:filter: mmc_rtd_section % "SWIFT"
Other
.. bibliography:: swift_refs.bib