Previous attempts to describe the structure of wind turbine wakes and their mutual interaction were mostly limited to large-eddy and Reynolds-averaged Navier-Stokes simulations using finite-volume solvers. We employ the higher-order spectral-element code Nek5000 to study the influence of numerical aspects on the prediction of the wind turbine wake structure and the wake interaction between two turbines. The spectral-element method enables an accurate representation of the vortical structures, with lower numerical dissipation than the more commonly used finite-volume codes. The wind-turbine blades are modeled as body forces using the actuator-line method (ACL) in the incompressible Navier-Stokes equations. Both tower and nacelle are represented with appropriate body forces. An inflow boundary condition is used which emulates homogeneous isotropic turbulence of wind-tunnel flows. We validate the implementation with results from experimental campaigns undertaken at the Norwegian University of Science and Technology (NTNU Blind Tests), investigate parametric influences and compare computational aspects with existing numerical simulations. In general the results show good agreement between the experiments and the numerical simulations both for a single-turbine setup as well as a two-turbine setup where the turbines are offset in the spanwise direction. A shift in the wake center caused by the tower wake is detected similar to experiments. The additional velocity deficit caused by the tower agrees well with the experimental data. The wake is captured well by Nek5000 in comparison with experiments both for the single wind turbine and in the two-turbine setup. The blade loading however shows large discrepancies for the high-turbulence, two-turbine case. While the experiments predicted higher thrust for the downstream turbine than for the upstream turbine, the opposite case was observed in Nek5000.

The wake structure behind a wind turbine, generated by the spectral element code Nek5000, is compared with that from the finite volume code EllipSys3D. The wind turbine blades are modeled using the actuator line method. We conduct the comparison on two different setups. One is based on an idealized rotor approximation with constant circulation imposed along the blades corresponding to Glauert's optimal operating condition, and the other is the Tjffireborg wind turbine. The focus lies on analyzing the differences in the wake structures entailed by the different codes and corresponding setups. The comparisons show good agreement for the defining parameters of the wake such as the wake expansion, helix pitch and circulation of the helical vortices. Differences can be related to the lower numerical dissipation in Nek5000 and to the domain differences at the rotor center. At comparable resolution Nek5000 yields more accurate results. It is observed that in the spectral element method the helical vortices, both at the tip and root of the actuator lines, retain their initial swirl velocity distribution for a longer distance in the near wake. This results in a lower vortex core growth and larger maximum vorticity along the wake. Additionally, it is observed that the break down process of the spiral tip vortices is significantly different between the two methods, with vortex merging occurring immediately after the onset of instability in the finite volume code, while Nek5000 simulations exhibit a 2-3 radii period of vortex pairing before merging.

The flow around an isolated horizontal-axis wind turbine is estimated by means of a new vortex code based on the Biot-Savart law with constant circulation along the blades. The results have been compared with numerical simulations where the wind turbine blades are replaced with actuator lines. Two different wind turbines have been simulated: one with constant circulation along the blades, to replicate the vortex method approximations, and the other with a realistic circulation distribution, to compare the outcomes of the vortex model with real operative wind-turbine conditions (Tjaereborg wind turbine). The vortex model matched the numerical simulation of the turbine with constant blade circulation in terms of the near-wake structure and local forces along the blade. The results from the Tjaereborg turbine case showed some discrepancies between the two approaches, but overall, the agreement is qualitatively good, validating the analytical method for more general conditions. The present results show that a simple vortex code is able to provide an estimation of the flow around the wind turbine similar to the actuator-line approach but with a negligible computational effort.

Actuator line and disc techniques are employed to analyse the wake obtained in the New Mexico wind turbine experiment. The New Mexico measurement campaign done in 2014 is a follow-up to the MEXICO campaign, which was completed in 2006. Three flow configurations in axial flow condition are simulated and both computed loads and velocity fields around the rotor are compared with detailed PIV measurements. The comparisons show that the computed loadings are generally in agreement with the measurements under the rotor's design condition. Both actuator approaches under-predicted the loading in the inboard part of blade in stall condition as only 2D airfoil data were used in the simulations. The predicted wake velocities generally agree well with the PIV measurements. In the experiment, PIV measurements are also provided close to the hub and nacelle. To study the effect of hub and nacelle, numerical simulations are performed both in the presence and absence of the hub geometry. This study shows that the large hub used in the experiment has only small effects on overall wake behaviour.

Results from Large-Eddy Simulations using the actuator line technique have been validated against experimental results. The experimental rotor wake, which forms the basis for the comparison, was studied in a recirculating free-surface water channel, where a helical vortex was generated by a single-bladed rotor mounted on a shaft. An investigation of how the experimental blade geometry and aerofoil characteristics affect the results was performed. Based on this, an adjustment of the pitch setting was introduced, which is still well within the limits of the experimental uncertainty. Excellent agreement between the experimental and the numerical results was achieved concerning the circulation, wake expansion and pitch of the helical tip vortex. A disagreement was found regarding the root vortex position and the axial velocity along the centre line of the tip vortex. This work establishes a good base for further studies of more fundamental stability parameters of helical rotor wakes.

The power production of the Lillgrund wind farm is determined numerically using large-eddy simulations and compared with measurements. In order to simulate realistic atmospheric conditions, pre-generated turbulence and wind shear are imposed in the computational domain. The atmospheric conditions are determined from data extracted from a met mast, which was erected prior to the establishment of the farm. In order to allocate most of the computational power to the simulations of the wake flow, the turbines are modeled using an actuator disc method where the discs are imposed in the computational domain as body forces which for every time step are calculated from tabulated airfoil data. A study of the influence of imposed upstream ambient turbulence is performed and shows that higher levels of turbulence results in slightly increased total power production and that it is of great importance to include ambient turbulence in the simulations. By introducing ambient atmospheric turbulence, the simulations compare very well with measurements at the studied inflow angles. A final study aiming at increasing the farm production by curtailing the power output of the front row turbines and thus letting more kinetic energy pass downstream is performed. The results, however, show that manipulating only the front row turbines has no positive effect on the farm production, and therefore, more complex curtailment strategies are needed to be tested.

The future development of offshore wind power will include many wind farms built in the same areas. It is known that wind farms produce long distance wakes, which means that we will see more occasions of farm to farm interaction, namely one wind farm operating in the wake of another wind farm. This study investigates how to perform accurate power predictions on large wind farms and how to assess the long distance wakes generated by these farms. The focus of this paper is the production's and wake's sensitivity to the extension of the grid as well as the turbulence when using Large-eddy simulations (LES) with pregenerated Mann turbulence. The aim is to determine an optimal grid which minimizes blockage effects and ensures constant resolution in the entire wake region at the lowest computational cost. The simulations are first performed in the absence of wind turbines in order to assess how the atmospheric turbulence and wind profile are evolving downstream (up to 12,000 m behind the position where the turbulence is imposed). In the second step, 10 turbines are added in the domain (using an actuator disc method) and their production is analyzed alongside the mean velocities in the domain. The blockage effects are tested using grids with different vertical extents. An equidistant region is used in order to ensure high resolution in the wake region. The importance of covering the entire wake structure inside the equidistant region is analyzed by decreasing the size of this region. In this step, the importance of the lateral size of the Mann turbulence box is also analyzed. In the results it can be seen that the flow is acceptably preserved through the empty domain if a larger turbulence box is used. The relative production is increased (due to blockage effects) for the last turbines using a smaller vertical domain, increased for a lower or narrower equidistant region (due to the smearing of the wake in the stretched area) and decreased when using a smaller turbulence box (due to decreased inmixing) The long distance wake behind the row is most impacted by the use of a smaller turbulence box, while the other simulation setups have less influence on these results. In summary, the results show the importance of having relatively large extensions of the domain, large extensions of the equidistant region and especially large extensions of the turbulence box.

The actuator line technique was introduced as a numerical tool to be employed in combination with large eddy simulations to enable the study of wakes and wake interaction in wind farms. The technique is today largely used for studying basic features of wakes as well as for making performance predictions of wind farms. In this paper, we give a short introduction to the wake problem and the actuator line methodology and present a study in which the technique is employed to determine the near-wake properties of wind turbines. The presented results include a comparison of experimental results of the wake characteristics of the flow around a three-bladed model wind turbine, the development of a simple analytical formula for determining the near-wake length behind a wind turbine and a detailed investigation of wake structures based on proper orthogonal decomposition analysis of numerically generated snapshots of the wake.