Using a large eddy simulation model to simulate wakes from large wind projects

Accurate prediction of the atmospheric flow through a wind farm with several hundred turbines (i.e. a large array) is very important for power production estimates as much larger farms are planned or under construction. Wind turbines extract kinetic energy from the airflow leaving the air downstream of a turbine (i.e. the wake) with reduced speed and static pressure as well as higher levels of turbulence. Currently, there is a limited understanding of the interaction between multiple turbine wakes as well as between wakes and the planetary boundary layer (PBL), both of which become relatively more important in large arrays. Almost all standard wake models are underestimating energy losses caused by turbine wakes in large wind farm arrays due to their empirical assumptions derived from actual data at smaller wind farms. On the other hand, stand-alone computational fluid dynamical (CFD) models are more robust and can capture some of the two-way interactions between wind turbines and the PBL. However, they do not typically account for the time-varying thermal structure of the boundary layer that may substantially alter the results. Numerical weather prediction (NWP) models especially those capable of large eddy simulations (LES) can theoretically simulate most turbine-atmosphere interactions, with the exception of the near wake (two to five rotor diameters downstream). Some research with mesoscale NWP models in this area has already been done. However, the spatial resolution was too coarse to resolve individual turbines or the spaces between them, a seeming prerequisite for a realistic simulation.

AWS Truewind is conducting research using customized NWP models in LES mode to explore turbine-atmosphere interactions with more specific attention to the turbine wakes. A parameterization was developed for these models to simulate the impact of wind turbines on atmospheric flow. The wind turbines are parameterized explicitly as sinks of momentum (i.e. drag force) and sources of turbulent kinetic energy. Simulations are being conducted at resolutions as fine as 100 m in both idealized and real conditions for onshore and offshore wind farms. Simulation results will be compared against real data from turbine-mounted instruments and separate meteorological masts on site. The focus will be on validating wind speed deficits and increased levels of turbulence within large arrays as well as the influence of large arrays on atmospheric properties. The main goal is to develop an LES modeling approach for predicting wakes in large arrays that will be substantially more accurate compared with standard wake models. The results could also help to refine and tune standard wake models that are much less computationally intensive than CFD or NWP models and therefore most practical for current applications involving turbine layout optimizations.