The key to achieving the wide-scale deployment of wind energy is enabling a new understanding of, and the ability to predict, the fundamental flow physics and coupled structural dynamics that govern whole wind plant performance, including wake formation, complex-terrain impacts, and turbine-turbine interactions through wakes. Current methods for modeling wind plant performance fall short due to insufficient model fidelity and the inadequate treatment of key phenomena combined with the lack of computational power needed to address the wide range of relevant length scales associated with wind plants. Thus, the ExaWind challenge problem is a predictive simulation of a wind farm with tens of megawatt-scale wind turbines dispersed over an area of many square kilometers with complex terrain that involves simulations with tens to hundreds of billions of grid points. These predictive, physics-based, high-fidelity computational models validated with targeted experiments will drive innovation in the blade, turbine, and wind plant design processes by providing a validated “ground truth” foundation for new turbine design models, wind plant siting, and operational controls, as well as by reliably integrating wind energy into the grid.

Project Details

The ExaWind project’s scientific goal is to advance fundamental understanding of the flow physics that govern whole wind plant performance, including wake formation, complex terrain impacts, and turbine-turbine interaction effects. The greater use of the nation’s abundant wind resources for electric power generation—reaching 30% of US electrical supply—will have profound societal and economic impacts, such as strengthening US energy security through greater diversity in its energy supply, providing cost-competitive electricity to key regions across the country, reducing greenhouse gas emissions, and reducing water used in thermoelectric power generation.

This multidisciplinary project embodies the systematic development of the modeling capability and the computational performance and scalability required for effective exascale simulations. The project plan builds progressively from predictive petascale simulations of a single turbine for which the detailed blade geometry is resolved, meshes rotate and deform with blade and tower motions, and atmospheric turbulence is realistically modeled to a multi-turbine array in complex terrain. This new modeling and simulation capability will establish a virtual wind farm test bed that will revolutionize the design and control of wind farms and significantly advance the scientific community’s ability to predict the response of wind farms to a wide range of atmospheric conditions.

The project objective is to capture crucial phenomena that are under-resolved in today’s models, including wake formation, complex-terrain impacts, wake-atmosphere interaction, turbine-turbine interaction, and blade-boundary-layer dynamics. This target requires a modeling and simulation capability that resolves turbine geometry and uses adequate grid resolution (down to micron scales within the blade boundary layers). The resolution must capture the upstream chord-scale atmospheric turbulent eddies, the generation of near-blade vorticity, and the propagation and breakdown of this vorticity within the turbine wake to a distance of many rotor diameters downstream. This application uses the ExaWind software stack, which has three physics-solver components: Nalu-Wind, an unstructured-grid computational fluid dynamics (CFD) code; AMR-Wind, a structured-grid CFD background solver built on AMReX; and OpenFAST, a turbine simulation code. The challenge-problem simulation will require a hybrid Reynolds-averaged-Navier-Stokes/large-eddy-simulation (RANS/LES) turbulence model, fluid-structure interaction, and turbulent atmospheric flow.

The challenge simulation will contain at least nine megawatt-scale turbines (e.g., NREL 5 MW reference turbines) organized in a 3 × 3 array and residing in a 4 × 4 km2 domain with a height of at least 1 km. A hybrid-RANS/LES model will be employed, wherein an unsteady RANS model will be used near turbine surfaces and an LES model will be used in the wake region. The simulation will have a mean wind speed at the turbines’ rated speed (e.g., 11.4 m/s for the NREL 5 MW reference turbine). The model will require at least 30 billion grid points and 150 billion degrees of freedom to resolve the system, and near-blade grid spacing will be such that the viscous sublayer within the RANS region is resolved. A successful simulation will require an optimized solver stack that minimizes time per time step. A scientifically meaningful simulation duration will be for at least one domain transit time (about 500 s for the 4 × 4 km domain at 11.4 m/s). The project will demonstrate that such a simulation is feasible within 4 weeks of system time.

Principal Investigator(s):

Mike Sprague, National Renewable Energy Laboratory; Paul Crozier, Sandia National Laboratories; John Turner, Oak Ridge National Laboratory


National Renewable Energy Laboratory, Sandia National Laboratories, Oak Ridge National Laboratory, University of Texas, Parallel Geometric Algorithms LLC, Scientific Simulations LLC

Progress to date

  • Key CFD computational physics kernels in Nalu-Wind were transferred to a highly portable, high-performance Kokkos-based design paradigm.
  • In collaboration with the US Department of Energy Wind Energy Technologies Office High-Fidelity Modeling project, a full set of baseline physics models were implemented and tested in the ExaWind modeling and simulation environment, including hybrid-RANS/LES models and fluid-structure interaction coupling.
  • A new hybrid solver capability was established in which the near-turbine fluid was simulated with the unstructured-grid Nalu-Wind solver that captures the blade boundary layers and wake formation. Nalu-Wind is tightly coupled to the AMR-Wind structured-grid solver that captures the background turbulent atmospheric flow and turbine wake propagation.
  • A simulation of the atmospheric boundary layer was demonstrated with the AMR-Wind solver on 90% of the Summit supercomputer, using over 24,000 GPUs.

Exascale computing will help drive innovation in the design of wind farms, resulting in increased efficiency and reduced cost per megawatt-hour of energy production.

National Nuclear Security Administration logo U.S. Department of Energy Office of Science logo