The problem of accurately estimating the impacts of wind-turbine-induced wakes on wind power production has become increasingly challenging with the growing size of wind plants. When operating, wind turbines extract kinetic energy from the airflow, leaving the air downstream (i.e. the wake) with reduced speed and static pressure as well as higher turbulence. Currently, there is a reasonably good understanding of wake evolution for single turbines and for small clusters of turbines up to three or four rows deep (defined along the prevailing wind direction). The same cannot be said for much larger arrays consisting of hundreds of turbines or more, where the interactions between multiple turbine wakes and between multiple wakes and the Planetary Boundary Layer (PBL) can become significant. Standard engineering wake models, e.g. Modified Park and Eddy Viscosity, appear to underestimate wake losses in such arrays, and especially offshore, because of their simplified assumptions and empirical derivation from data for small wind plants.

Over the last few years, new codes have been developed to mitigate those issues. For instance, the deep-array wake model (DAWM) within the open-source platform openWind is a hybrid model based on the surface-drag-induced internal boundary layer approach of Frandsen (2007) and the Eddy Viscosity model. DAWM performs better than the standard Eddy Viscosity model for offshore test cases. Such wake models are very attractive for their small runtime on desktop platforms and represent an improvement over standard models, but they are limited in their ability to capture the detailed characteristics of the wakes, and their applicability to much larger projects is uncertain.

Another promising alternative is large eddy simulation (LES), which can theoretically simulate most turbine-atmosphere interactions given an adequate small-scale turbulence closure and sufficient computing power. This approach was developed several decades ago to study the PBL but has been applied only recently to the study of turbine-induced wakes. The benefits of taking into account the influence of unsteadiness and thermal stratification in LES comes with a cost - the runtime for LES simulations is approximately 60 times longer than for computational fluid dynamics (CFD) runs at comparable resolution. LES simulations with the Advance Regional Prediction System (ARPS) are being conducted at resolutions as fine as 80 m in both idealized and real conditions for offshore and onshore wind farms.

In this research, the performance of these two new approaches to wake modeling, DAWM and LES, is compared to that of standard engineering wake models and to observations for an offshore wind project. These results lend insight into wake effects in large arrays and could help improve the accuracy of energy production estimates.