Understanding the detailed dynamics of wind turbine wakes is critical to predicting the performance and maximizing the efficiency of wind farms. This knowledge requires atmospheric data at a high spatial and temporal resolution, which are not easily obtained from direct measurements. Therefore, research is often based on numerical models, which vary in fidelity and computational cost. The simplest models produce axisymmetric wakes and are only valid beyond the near wake. Higher-fidelity results can be obtained by solving the filtered Navier-Stokes equations at a resolution that is sufficient to resolve the relevant turbulence scales. This work addresses the gap between these two extremes by proposing a stochastic model that produces an unsteady asymmetric wake. The model is developed based on a large-eddy simulation (LES) of an offshore wind farm. Because there are several ways of characterizing wakes, the first part of this work explores different approaches to defining global wake characteristics. From these, a model is developed that captures essential features of a LES-generated wake at a small fraction of the cost. The synthetic wake successfully reproduces the mean characteristics of the original LES wake, including its area and stretching patterns, and statistics of the mean azimuthal radius. The mean and standard deviation of the wake width and height are also reproduced. This preliminary study focuses on reproducing the wake shape, while future work will incorporate velocity deficit and meandering, as well as different stability scenarios.
449A stochastic wind turbine wake model based on new metrics P. Doubrawa et al.equations, using an eddy viscosity model to compute the Reynolds stresses. Both models are only valid beyond the near wake (i.e. approximately 2 rotor diameters downstream). A moving quasi-steady wake can be simulated by adding a stochastic component to the Ainslie model to account for the wake meandering. This dynamic wake meandering (DWM) model 14 treats the wake as a passive tracer that is advected by the large scales in the ambient turbulent flow. Next in the range of fidelity is full three-dimensional computational fluid dynamics using RANS turbulence modeling, either in steady or unsteady form. In the steady form, used with an actuator disk representation of the turbine rotor, a steady wake is formed. This wake need not be axisymmetric and can be affected by shear, terrain and stability. In unsteady form, RANS can be used with a rotating actuator line, and the large-scale unsteady features of wakes can be resolved. Finally, higher fidelity results can be obtained by performing large-eddy simulations (LES), which entail solving the filtered Navier-Stokes equations at a spatial and temporal resolution that is high enough to resolve the relevant turbulence scales, typically O.10 0 / m near the rotor. Different LES codes vary in the treatment of the turbine (e.g. actuator disk versus actuator line 15 ), the turbulence closure used (e.g. Smagorinsky versus mixed-scale models 16 ) and the nu...