Search

Wind farm simulation tools are used for a multitude of purposes, including energy yield calculations, wind farm control optimization, layout optimization, structural load analysis, and many more. However, the vast majority of farm software either fails to capture the dynamic nature of both the flow and the turbine structural response or demands a high computational cost such as large-eddy simulations (LES). In this study, we present a new mid-fidelity aeroelastic wind farm simulation software, HAWC2Farm, which can perform dynamic wind farm computations at the same temporal and spatial resolution as LES using a fraction of the computational resources. Wind farm simulations are performed with wake steering, and statistics are verified with results from LES performed using Ellipsys3D.

Enhancing wake recovery behind wind turbines has the potential to significantly improve the power production and efficiency of large wind farms. Rather than investigating turbine control strategies, floater motion or global turbulent quantities such as turbulence intensity, this work aims to study wake stability and recovery through a focus on the turbulent scales of the inflow. Using Large Eddy Simulations of a single turbine, sinusoidal streamwise forcing is applied to the inflow with a constant amplitude and mean flow velocity, but differing time scales between 80s and 140s. For all applied time scales the turbine wake is characterised by the rolling-up of the near wake into the periodic shedding of vortex rings, and an excitation of the applied forcing frequency resulting in velocity fluctuations in the wake several times larger than that at the inflow. For shorter time scales (80s - 90s) a more aggressive and earlier wake roll-up led to a shorter near wake region, faster overall recovery and significantly improved the expected power output from 6R downstream onwards. An inflow time period of 80s gave rise to more than a 50% increase in power output of a fictive downstream turbine placed at 14R downstream, compared to an inflow time scale of 140s.

Wind farm Annual Energy Production (AEP) calculations are required to design energy efficient wind farm layouts. We investigate methods that can reduce the computational effort of AEP calculations using Reynolds-averaged Navier-Stokes simulations of an idealized atmospheric wind farm setup. In addition, we introduce a generalized wind turbine model that compares well with wind turbine aerodynamic data covering a large range of wind turbine sizes. We apply the general wind turbine model to reduce the computational effort of the AEP calculations by decreasing the number of independent wind speed flow cases. Furthermore, we apply Reynolds-number similarity to compute the wind speed flow cases faster and we show how wind farm layout mirror- and rotational-symmetry can reduce the number of independent wind direction flow cases.

Actuator disc and actuator line techniques are widely used for modelling wind turbines operating in wind farms. These techniques essentially replace the blade geometry with applied body forces, which reduce the resolved length scales significantly and hence the required grid resolution. This work is a verification of the coupling between the flow solver EllipSys3D and the aeroelastic tool Flex5, through a quantitative comparison of coupled actuator line, coupled actuator disc, and standalone Flex5. Steady state performance predictions, instantaneous reaction to turbulence and damage equivalent load analyses all show a very good agreement between the three methods. Differences can be explained primarily by the higher fidelity modelling of the coupled simulations; this is particularly in regard to the influence of blade flexibility, as the actuators deflect and interact with the modelled flow. Additionally, some overpredictions of loading at the blade tip and root below rated velocity for the actuator methods can be attributed to the Gaussian smearing used to apply the body forces.

Fast engineering wake models are the backbone of wind farm annual energy production (AEP) estimators, whereas the addition of induction zone models in existing tools is a more recent response to rising concerns over wind farm blockage associated losses. Here, “blockage” describes the combined induction fields of all wind turbines inside a farm. Unlike the term might suggest, blockage not only reduces flow speeds, but also increases them; for instance along the outer edges of wind farms. Evaluating the overall impact on AEP of these gains and losses necessitates accurate wind farm induction models. Whilst engineering wake models are tuned to predict wind farm performance, existing induction zone models are all derived for accurately predicting the near field induction not the far field value—important for wind farm simulations. This paper presents the induction models implemented in the wind farm simulation tool PyWake, as well as results from novel analytical models and compares their far field predictions to RANS-AD simulations of different turbines. We demonstrate that when including blockage in AEP simulations, the downstream speed-ups need to be included to avoid an unrealistic bias toward AEP loss and that wake expansion significantly impacts induction at rated thrust levels.

Please see Article PDF for corrigendum text.

The actuator line (AL) is a lifting line (LL) representation of aerodynamic surfaces in computational fluid dynamics (CFD) applications. The AL blade forces are computed from 2D airfoil polars and the CFD velocity vector extracted at the line position, as the self-induction at the very centre of the bound vortex should, following vortex theory, be nil. Yet, this is not the case in CFD, which leads to errors in the angle-of-attack computation. We derive an expression for the error in the lift force from vortex considerations and show it to be a function of chord, the smearing length scale used in distributing the AL forces over the numerical domain and the number of grid cells per smearing length scale. Thereby demonstrating that the required number of grid cells-contrary to current belief-needs to grow faster than the inverse of the smearing length scale refinement to maintain the error level. We additionally show that the error can be large for the commonly used ratio of 2 grid cells per length scale, especially if the latter is relatively small with respect to the rotor radius. Ultimately, the recommendation is to always run with the largest smearing length scale possible for the specific application in conjunction with a smearing correction, as this minimizes the error in the blade forces whilst reducing the computational resources required.

The Actuator Line Method exists for more than a decade and has become a well established choice for simulating wind rotors in computational fluid dynamics. Numerous implementations exist and are used in the wind energy research community. These codes were verified by experimental data such as the MEXICO experiment. Often the verification against other codes were made on a very broad scale. Therefore this study attempts first a validation by comparing two different implementations, namely an adapted version of SOWFA/OpenFOAM and EllipSys3D and also a verification by comparing against experimental results from the MEXICO and NEW MEXICO experiments.

The flow around a wind turbine airfoil equipped with Vortex Generators (VGs) is examined. Predictions from three different Reynolds Averaged Navier Stokes (RANS) solvers with two different turbulence models and two different VG modelling approaches are compared between them and with experimental data. The best results are obtained with the more expensive fully resolved VG approach. The cost efficient BAY model can also provide acceptable results, if grid related numerical diffusion is minimized and only force coefficient polars are considered.

Sheared inflow causes significant periodic load variations in wind turbine blades, but has only limited impact on the mean wake deficit. Following these findings the wind speed reduction upstream of the turbine - referred to as the induction zone - might also show little difference to uniform inflow. Using the local free-stream velocity to normalise the upstream flow-field should then render uniform and sheared inflow induced velocity profiles indiscernible, hinting towards wind shear acting solely as a linear addition. This has great implications in BEM methods for determining the velocity at the blades and also for near-rotor lidar measurements. The latter being applied in for power/loads assessment and turbine control. LES simulations with an actuator line representation of the rotor confirm the linearity assumption for moderate wind shear. To estimate the normal velocities at the disc the annularly averaged thrust coefficient is best suited, when the induction is imposed on the inflow profile. A strictly local relationship breaks down in strongly sheared flow. A simple induction zone model devised for uniform inflow estimates the velocity upstream within ±0.5% even at extreme shear in the upper half of the rotor and at least three rotor radii away from the turbine.

Design loads on turbines are normally simulated with an aeroelastic model using an engineering BEM type model with the turbulent inflow generated with a turbulence model like the Mann model. There are several fundamental uncertainties in this approach, e.g. how well the unsteady induction in response to the turbulent flow is computed. However, within the last few years full 3D CFD rotor computations with turbulent inflow have been performed which can provide detailed insight into this complex load response. In the present work we present computations with the EllipSys3D solver on the 80m diameter NM80 turbine used in the DANAERO project where surface pressure measurements at four radial positions were conducted. The aerodynamic loads integrated from the pressure distributions have been derived and compared with computations by the aeroelastic code HAWC2. Overall a very good correlation is found by comparing PSD spectra of the measured sectional blade forces with HAWC2 simulations using specific flow input from the meteorology mast at six heights. In another comparison using purely turbulent inflow for the simulations on the NM80 rotor some deviations between the force spectra are found between EllipSys3D results and HAWC2 simulations at the inboard part of the blade and at high frequencies.

We present Large-Eddy Simulation (LES) results of flow over the double ridge complex site at Perdigao in Portugal. The focus is to compare simulated flow features from two LES codes with different discretization techniques. We compare a finite volume discretization with a pseudo spectral approach in two different terrains. Mean wind properties and turbulent kinetic energy from the two codes are to a large degree in agreement. The largest discrepancy we observe is attributed to the different effective resolution in the two codes which results from the numerical discretizations. Comparison with measured data from three installed meteorological masts inside the simulated domain show that many of the main flow features have been captured by the LES simulations despite its relatively simple setup.

Pairs of synchronously scanning Doppler lidars measure the average wind speed of flows crossing the parallel ridges at Perdigao, Portugal, with the ultimate purpose of wind resource estimation. The availability of the data from the lidars when they are running is quite low (50–70%). Furthermore, the instruments did only run less than half the time of the experimental period. These figures have to be improved in order for scanning lidars to be a viable option for wind resource estimation. The variations along the ridges are compared to neutral LES calculations making a good match at the upstream ridge but a significantly worse prediction at the downstream ridge. One reason could be an insufficient representation of the terrain. Another unknown is the influence of the atmospheric stability on the flow which is clearly seen by the scanning lidars.

We present Large-Eddy Simulation results of a turbine wake in realistic complex terrain with slopes above 0.5. By comparing simulations including and without the wind turbine we can estimate the induction factor, a, and we show how the presence of a strong recirculation zone in the terrain dictates the positioning of the wake. This last finding is in contrast to what would happen in gentle terrain with no substantial increase of turbulent kinetic energy in the terrain induced wakes.

Lidar velocity measurements need to be interpreted differently than conventional in-situ readings. A commonly ignored factor is "volume-averaging", which refers to lidars not sampling in a single, distinct point but along its entire beam length. However, especially in regions with large velocity gradients, like the rotor wake, can it be detrimental. Hence, an efficient algorithm mimicking lidar flow sampling is presented, which considers both pulsed and continous-wave lidar weighting functions. The flow-field around a 2.3 MW turbine is simulated using Detached Eddy Simulation in combination with an actuator line to test the algorithm and investigate the potential impact of volume-averaging. Even with very few points discretising the lidar beam is volume-averaging captured accurately. The difference in a lidar compared to a point measurement is greatest at the wake edges and increases from 30% one rotor diameter (D) downstream of the rotor to 60% at 3D.

The Actuator Line Method exists for more than a decade and has become a well established choice for simulating wind rotors in computational fluid dynamics. Numerous implementations exist and are used in the wind energy research community. These codes were verified by experimental data such as the MEXICO experiment. Often the verification against other codes were made on a very broad scale. Therefore this study attempts first a validation by comparing two different implementations, namely an adapted version of SOWFA/OpenFOAM and EllipSys3D and also a verification by comparing against experimental results from the MEXICO and NEW MEXICO experiments.

The interaction of a wind turbine with the upstream flow-field in complex and flat terrain is studied using Reynolds-averaged Navier-Stokes (RANS) simulations with a two equation turbulence closure. The complex site modelled is Perdigao (Portugal), where a turbine is located on one of two parallel running ridges. Simulating various wind directions with and without rotor, the impact of the rotor on the flow-field upstream is determined. This is compared and related to simulations with sheared and uniform inflow. The induction zones forming for these two inflows agree to such an extent, that shear could be interpreted as linear perturbation to the uniform inflow solution. However, for complex terrain this is not the case, as it is highly dependant on flow features caused by the topography and their interaction with the rotor. Separation in the lee of the ridge plays a crucial role, as it dictates the wind turbine wake trajectory which in turn governs the orientation of the induction zone.