Edited by: Gert-Jan Steeneveld, Wageningen University & Research, Netherlands
Reviewed by: Jeffrey Mirocha, Lawrence Livermore National Laboratory, United States Department of Energy (DOE), United States; Chenghai Wang, Lanzhou University, China
This article was submitted to Atmospheric Science, a section of the journal Frontiers in Earth Science
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The offshore wind market is expected to see massive growth worldwide in the next decade. Recent research in this sector has led to a new generation of wind turbines characterized by larger rotors and higher capacity factors. Furthermore, the trend is to develop larger and denser wind farms which leads to potential deeper interactions with the troposphere. In order to provide an accurate modeling of the processes between the wind turbines, the surface and the atmospheric boundary layer, a new numerical tool has been developed. The ALM has been implemented into the opensource non-hydrostatic mesoscale atmospheric model Meso-NH used in the Large-Eddy Simulation (LES) framework. First, the tool is validated by comparing simulated results with data acquired during the New MEXICO experiments. In particular, good correlation are obtained for normal and tangential loads along the blades and the axial PIV traverse of instantaneous wind velocity components. Once the tool is validated, its ability to reproduce the Horns Rev 1 photo case is evaluated, since it is an emblematic case of wind farm-atmosphere interaction. Simulation output are post-processed using a render algorithm, resulting in synthetic images of cloud radiance. These images show great similarity with the photographs taken. All the results highlight the capacity of this new tool to represent interactions between wind farms and the lower atmosphere with a high level of details.
Considerable deployments of offshore wind farms can be expected in the near future (IEA,
To investigate these interactions, a better understanding of processes occurring between the atmospheric boundary layer, the surface and the wind turbines is necessary. This investigation can be done with numerical tools, among which Navier-Stokes based Computational Fluid Dynamics (CFD) models are frequently used. Several approaches can be implemented to solve the governing flow equations, such as the Reynolds averaged Navier Stokes (RANS), the Large-Eddy Simulation (LES), or the Direct Numerical Simulation (DNS). The RANS method computes mean flow and models an important part of turbulence aspects: it is not the proper tool to explore small scale wake eddies nor atmospheric boundary layer eddies. The DNS method is more precise but computationally too expensive as it resolves all turbulent scales. Because the LES approach consists in solving the large scales and modeling the smaller ones, it can be seen as an intermediate approximation and a good compromise.
In order to introduce wind turbines into a CFD software, different approaches can be used. One of those approaches is to simulate a fully-resolved blade, but this technique is highly resource-consuming for a whole wind farm for the moment. Introduced by Sørensen and Shen (
The present paper introduces a new numerical tool allowing the study of wind farms impacts on local meteorology. This is achieved by coupling the ALM with the meteorological model Meso-NH (Lac et al.,
The study is organized as follows: the numerical methods are described and validated in section 2. Section 3 introduces the Horns Rev 1 photo case, provides the simulation details and presents the results. Concluding remarks are provided in section 4.
Meso-NH is a non-hydrostatic mesoscale atmospheric model (Lac et al.,
The numerical model can be used with a Large-Eddy Simulation framework. The second-order turbulent fluxes are computed through an 1.5-order closure proposed by Redelsperger and Sommeria (
The approach is Eulerian and the domain is discretized using a staggered Arakawa C-grid (Mesinger and Arakawa,
The ALM consists in modeling each blade by one line divided into blade element points (Sørensen and Shen,
2D blade element velocities and forces.
The aerodynamic forces
where ρ and
Meso-NH and the Actuator Line Method are exchanging information with each other. At each time step, the position, the velocity and the orientation matrix of blade element points are evaluated. Then, the ALM can use air properties (as ρ and
where
To avoid numerical instabilities, a commonly used regularization kernel is used (Sørensen and Shen,
To go further in computational cost savings, a time-splitting method was introduced, also known as the Actuator Sector Method (Storey et al.,
where Δ
To assess the accuracy of the results obtained with this coupled system, a validation study based on the New MEXICO (Model Experiments in Controlled Conditions) experiments is carried out (Boorsma and Schepers,
During these experiments, several flow cases were investigated (Boorsma and Schepers,
The wind tunnel wind turbine is a three-bladed 4.5 m diameter rotor. The twisted and tapered blades are based on DU91-W2-250, RISØ-A1-21, and NACA64-418 airfoils from root to tip. The wind turbine is rotating at a constant rotational velocity Ω = 44.55 rad/s and a constant blade pitch angle β = –2.3°. The inflow is laminar and no initial turbulence is introduced. The normal and tangential loads are available at five positions thanks to the integration of pressure sensors measurements.
The dimension of the domain is about
The wind speed, the potential temperature and the pressure are set as inputs according to the experimental data : respectively
The numerical results are compared to the experimental data. The loads along the blade span (shown in
Comparison of computed normal
Comparison of axial PIV traverses for the case 10 m/s, with the New MEXICO experiments (Boorsma and Schepers,
In
The upper plot of
In order to verify the conformity of the time-splitting method, another numerical study is carried out. The loads along the blades and the axial traverses are compared to the experimental data for different time-splitting factors
Time step values used to obtain different time-splitting factors.
Δ |
[s] | 0.0005 | 0.002 | 0.004 | 0.006 | 0.008 | 0.006 |
[–] | 1 | 3 | 5 | 7 | 9 | 1 | |
Computational time | [%] | 100 | 40.4 | 33.3 | 27.4 | 27.0 | 23.9 |
Comparison of predicted normal
Comparison of axial PIV traverses for the case 10 m/s with the New MEXICO experiment (Boorsma and Schepers,
The Horns Rev 1 wind farm is well known for being the first large scale offshore wind farm in the world. It is located in the North Sea, west of Denmark and consists of 80 Vestas V80 wind turbines with 70 m hub height, 80 m rotor diameter and rated power of 2.0 MW. The parallelepiped shaped layout of the wind farm is arranged in 10 vertical and 8 horizontal rows, with a minimal distance of 7 diameters between the closest turbines.
In the morning of February 12, 2008 at around 10:10 UTC, the photographer Christian Steiness took two photos of the Horns Rev 1 wind farm from a helicopter (
A detailed description of the meteorological conditions at the time the photos were taken and the process that would have lead to the clouds formation can be found in Hasager et al. (
Despite the fact that Horns Rev I is one of the most studied wind farms in the world, with three meteorological masts installed on-site, one of the main difficulties to simulate the photo case is the need of very precise information to reproduce the specific conditions that led to the wake fog formation. Hasager et al. (
The wake fog simulation is performed using Meso-NH through two steps. The objective of the first step is to establish the precursor meteorological conditions without the wind turbines. The second one includes the wind farm that leads to the wake clouds development.
In order to start the spin-up simulation, various vertical profiles of the initial state of the atmosphere must be set. In
where
Wind speed, temperature and vapor mixing ration profiles. M6 Data from Hasager et al. (
The initial conditions were deduced from the met mast (M6) and the reanalysis data. Because of its position, the met mast M6 is considered to be not affected by the wake, nor by the fog development. For the wind, several uniform conditions (i.e., intensity and direction) have been tested to initialize the atmospheric state. Indeed, during the simulation, the atmospheric flow will tend to an equilibrium between the geostrophic wind forcing, the surface friction and the Coriolis effect. The chosen value for the initial profile corresponds to the geostrophic wind and allow to obtain the measured intensity and direction provided in Hasager et al. (
The precursor simulation is configured with a domain of 7,500 × 25,000 × 837.5 m, a horizontal resolution of 5 m and a time step of 1 s. The vertical levels are defined with a vertical resolution of 5 m below 200 m and then a 20% stretching is applied until a resolution of 20 m is reached. The boundary conditions are cyclic. The long-wave and short-wave radiation are computed following (Fouquart and Bonnel,
Once the precursor 3D state of the atmosphere is established, the Horns Rev 1 wind farm is placed in the domain and a second simulation is performed for a period of 32 min. The domain configuration is the same as the previous one with 1,500 points in the
Layout of the Horns Rev 1 offshore wind farm. The blue bullets (
The prevailing wind direction is 180° at the hub height (i.e., parallel to
After 32 min of simulation with the wind farm, several variables are studied to analyze the wind farm interaction with the local meteorology.
The relative humidity horizontal cross-sections, at 30 m and 110 m above the sea are presented in
Horizontal cross-sections at 30 m
Horizontal cross-sections at 30 m
Horizontal cross-section at 30 m
From these figures (
In
Vertical cross-section of the cloud mixing ratio through the whole fourth row of the wind farm
Between the photo taken from the southeast (
As wind power production is increasing all over the world and the wind turbines are getting larger and taller over the years, the study of their interaction with the lowest levels of the atmosphere becomes relevant. To address this challenge, a new numerical tool was developed to explore the impacts of wind farms on the local meteorology. This new tool is a coupled model between the atmospheric mesoscale model Meso-NH used in the LES framework and the Actuator Line Method. In order to validate the tool, comparisons against New MEXICO measurements and other numerical models have been made. The predicted loads along the blades and the axial wind intensity at hub height showed good correlation with the measurements for the high tip speed ratio (TSR) value. The time-splitting method has also been validated, allowing to save computation time and obtain a representative wake.
With the goal of evaluating the capability of this new tool to reproduce a real case of interaction between a wind farm and the atmospheric boundary layer, a simulation of an idealized case of the Horns Rev I Photo case was performed. First, a spin-up simulation was carried out to reproduce the existing weather conditions before the photo, based on
It should be underlined that the phenomenon showed by the Horns Rev 1 photo case can only appear under a specific set of atmospheric conditions. In fact, the sensitivity studies carried out without the wind farm (not presented) showed that small changes in temperatures would have led to a global fog formation or no cloud at all.
Based on the results of this study, the capability of this new tool to simulate the wake dynamics and to represent their impact and interaction with the atmospheric quantities was shown. Furthermore, this work demonstrates more generally the ability of coupled weather and wind plant simulation frameworks to capture their complex physical interactions.
Moreover, the atmospheric boundary layer is typically shallower over the sea, while offshore wind turbines are getting larger. It will then become possible for blades to cross the top of the boundary layer, which is often characterized by an abrupt temperature inversion and strong wind shear. At that point, simulations of new generations of wind farms will be conducted with the aim of investigating those interactions.
P-AJ developed the numerical coupling, performed the numerical experiments, and led the analysis of the study. MM gathered all the meteorological data available for the Hors Rev 1 Photo case, launched the firsts Horns Rev simulations and participated to the redaction of the manuscript. VM, FB, QR, MC, and CL substantially contributed to the development and the manipulation of the different models. All authors contributed on the comprehension of the study, as well as on the revision and correction of the manuscript. Agreement to be accountable for all aspects of the work in ensuring that.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
New MEXICO data have been supplied by the consortium which carried out the EU FP5 project MEXICO: Model rotor Experiments In Controlled conditions. We wish to thank the IEA Wind Task 31 WakeBench which provided the description of the Horns Rev 1 wind farm and of the Vestas80 wind turbine model. We would like to acknowledge the Wake Conference 2019 for publishing our preliminary results under the IOP Proceedings Licence (