Research

Within WINS50, a wind atlas for the North Sea area has been produced that includes the effects of both current and future wind farms. With the model data that has been made publicly available. These data enable researchers from the meteorological and energy communities, as well as policy makers, to make detailed analysis of, for instance, future wind farm wake effects and energy production.

A few examples of currently relevant research topics are briefly discussed here. In addition, backgound information on the two weather models that are used in WINS50 (HARMONIE and ASPIRE) is included.



Validation

Model output from the WINS50 simulations have been compared with observations. As an example, below scatter plots show modeled versus observed wind speed and direction at 116 m height for the LiDAR operated at the K13 offshore platform. It includes three years of data from the undisturbed HARMONIE-AROME simulation. For both wind speed and direction, the agreement between model and observations is excellent.


Scatterplots of modeled versus observed wind speed (left) and wind direction (right) at 116 m height for K13.
Image: WINS50.


Futher validation of the HARMONIE-AROME simulations can be found in (1). Focusing on the year 2019, they demonstrate that including wind farms in the simulation gives a better representation in wake-affected conditions. They report that wakes are strongest in stably-stratified conditions. Additional validation of the Fitch wind farm parameterization within HARMONIE-AROME is presented in (2).

For a complex coastal site, (3) showed the benefits of using high-resolution LES simulations. In a validation against LiDAR measurements, they showed that ASPIRE accurately captures nearby wake effects.



ASPIRE validation around the Eemshaven wind farm site. The top-panels show directional bias (left) and mean vertical profiles (right) at the Strekdammen location. Blue lines represent the observations, yellow lines a control simulation without wind turbines, and red lines the simulation with all wind turbines included. The bottom-panel shows a map of the simulated area.
Image: WINS50.


Impact future capacity scenario on wind climate

With both HARMONIE-AROME and ASPIRE a present-day and a hypothetical 2050 energy capacity scenario have been simulated. For both models, the Figure below shows the yearly averaged difference in 100-m wind speed. Differences are largest within the hypothetical future wind farms. For a significant fraction of the North Sea the mean wind speed redution is 0.5 m/s or larger.


Mean 100 m wind speed velocity deficits (W50-W20) from the HARMONIE-AROME (left) and the ASPIRE (right) 2050 scenario simulations.
Image: WINS50.


Wind farm wakes extend tens of kilometers downstream of a wind farm. This is shown in the following Figure, which shows the average 100-m velocity deficit for an area in front of the Dutch coast for northeasterly winds only. Both modes, HARMONIE-AROME and ASPIRE, show the same large-scale wake effect, but the difference in detail between the 2.5 km HARMONIE-AROME grid and the 128 m ASPIRE grid is clearly visible.


Mean 100 m wind speed deficit (W50-W20) from the HARMONIE-AROME (left) and the ASPIRE (right) 2050 scenario simulations (northeasterly winds only) for an area in front of the Dutch coast. In the ASPIRE deficit field the locations of individual turbines can be recognized.
Image: WINS50.


Obviously, the affected areas are much larger in the hypotheical 2050 scenario than in the present-day (2020) scenario. In the 2050 scenario on average 1/3 of the North Sea has a velocity deficit larger than 0.5 m/s, which is already a significant value. For larger threshold values the size of the affected area decreases. In stably-stratified conditions the wake-affected area is roughly twice as large as in unstable conditions.


Offshore area fraction for which the 100 m wind speed velocity deficit of the HARMONIE-AROME 2020 and 2050 scenarios is larger than a certain threshold value. Different colors represent different stability classes. Solid lines give the 2050 fractions, dashed lines the 2020 fractions. The horizontal lines indicate the area fractions occupied by the wind farms themselves.
Image: WINS50.

The 2050 scenario and energy production

What will be the impact a tenfold increase of installed energy capacity on power production? How much will the efficiency go down as a result of farm-to-farm wake interactions? To answer these questions, we inspect the capacity factors (energy production divided by installed capacity) of the 2020 and 2050 scenarios. In the 2050 scenario capacity factors are lower than in the 2020 scenario. In the 2050 scenario (right panel) the interior of the wind farms have the lowest capacity factors.


Capacity factors for the HARMONIE-AROME 2020 (left) and 2050 (right) scenarios.
Image: WINS50.


Because all wind farms of the 2020 scenario are also present in the 2050 scenario, we can compare the energy production of these wind farms for both scenarios. For each wind farm, the difference indicates the reduction of energy production due to the hypothetical future wind farms. On average, the production goes down by 6 to 7% but differences among wind farms are large. Wind farms which remain relatively isolated in the 2050 scenario experience much lower production losses than wind farms that are facing large additional wind farms in their surroundings (e.g. in the German Bight or the Belgian wind farm cluster).


Production decrease of the 2020 wind farms in the 2050 scenario simulation in ASPIRE (left). Map of wind farms (right) with the color coding of the 2020 wind farms corresponding to the decrease in production as indicated in the left panel.
Image: WINS50.

Sensitivity studies

Using ASPIRE, (4) investigated the energy production and wake losses for multi-gigawatt offshore wind farms. To that end, one year of actual weather from 2015 was simulated for a suite of hypothetical 4 GW offshore wind farm scenarios. The scenarios differed in terms of applied turbine type, installed capacity density, and layout. Free-stream production numbers were obtained from an additional simulation that contained drag-free turbines that do produce power but do not exert drag on the flow.

The results suggest that production numbers increase significantly when the rated power of the individual turbines is larger while keeping the total installed capacity the same (e.g. IEA10 -> IEA15 -> Scaled21). Even for turbine types with similar rated power but slightly different power curves, significant differences in production were found (DTU10 vs IEA10). The Figure below shows power production and aerodynamic losses (i.e. the sum of losses related to global blockage and internal wake losses) for each of the six scenarios.


Free-stream and actual power production (left) and aerodynamic losses (right) for six 4 GW offshore wind farm scenarios.
Image: WINS50.


With HARMONIE-AROME, an explorative 3-months simulation was performed with the entire North Sea filled with 15 MW turbines with a capacity density of 8 MW/km2. Below Figure shows the 10 m wind speed for the HARMONIE-AROME WINS50 2050 scenario simulation and the simulation with the North Sea filled-up with wind turbines compared to the WINS50 control simulation. Obviously, the wind speed is seriously reduced in case of such a hypothetically large wind farm. Even over land a significant decrease in 10 m wind speed occurs (in contrast to the 2050 scenario).


Mean HARMONIE-AROME 10 m wind speed reduction for the WINS50 2050 scenario simulation (left) and for the simulation with the North Sea filled-up with turbines (right) compared to the WINS50 control simulation.
Image: WINS50.


To explore the limits of the energy production further, additional simulations were performed with ASPIRE for different capacity densities. Periodic boundary conditions were applied so that wakes leaving the domain entered the domain at the opposite site, thus mimicking effectively an infinite wind farm.

Production numbers and aerodynamic losses from the three ASPIRE 'infinite' wind farm simulations are shown below. As expected, the free-stream production per turbine is the same for each of the capacity density scenarios (around 9.5 MW). At the same time the actual production decreases with increasing density capacity. Converted to production per square kilometer, the free stream production increases proportionate with the installed capacity density. Also the actual production per square kilometer increases with capacity density, but less then proportionate. This is reflected in the aerodynamic losses, which increase from around 20% in case of the 3.6 MW/km2 scenario to 55% in case of 12.0 MW/km2 scenario.


Free-stream (blue) and actual (orange) production numbers for 'infinite' wind farms with varying capacity density (left and middle panel). Corresponding aerodynamic (wake) losses (green) are presented as well (right panel).
Image: WINS50.




WINS50 Weather models

HARMONIE-AROME

HARMONIE-AROME (HIRLAM ALADIN Research On Mesoscale Operational NWP in Europe) is the operational Numerical Weather Prediction model of KNMI since 2012. It is a limited area model, that was developed by a consortium involving many European countries. HARMONIE is intended to run on a high grid resolution of typically 2.5 km, but allows for grid resolutions well under the km scale. HARMONIE is a so-called non-hydrostatic model, meaning that instead of employing the hydrostatic approximation, which often breaks down in severe-weather events, the vertical momentum equation is solved explicitly. More details on HARMONIE/AROME are given in (5).

HARMONIE-AROME has been used to produced high-resolution wind atlases before. The KNMI North Sea Wind (KNW) atlas contains hourly undisturbed wind data (without wind turbine wake effects) based on more than 40 years (1979-2019) of re-analysed data. However, due to the applied method to create the KNW-atlas, the hourly correlation with measurements (e.g. daily cycle) is rather poor. A substantial improvement was achieved in the Dutch Offshore Wind Atlas (DOWA), both over sea and over land (6, 7), by introducing a new turbulence scheme and assimilating scatterometer (10m wind) and Mode-S EHS aircraft observations (wind). The DOWA is a 10 year climatology (2008-2017) with a better hourly correlation than the KNW-atlas. It has information up to 600 m height and covers a larger area compared to the KNW atlas. The DOWA-data and all information on the DOWA-project can be found on the Dutch Offhore Wind Atlas (DOWA) website.

In the WINS50 project the DOWA will be extended by three years (2019, 2020, 2021). In order to create a homogeneous dataset, changes in method should be avoided, but it turned out that the HARMONIE-AROME Cycle 40 that was used for DOWA could no longer be used for WINS50. Instead, for WINS50 an updated version of HARMONIE was used: HARMONIE-AROME Cycle 43. The consequences for the continuity between the the DOWA and WINS50 dataset is discussed in the memo Comparison DOWA and WINS50-data.

Within WINS50, also a model simulation that includes wind turbine wakes will be performed (current and future capactity). For modeling wind farm wakes, HARMONIE uses the Fitch et. al, 2012 (8) wind farm parameterization. In this parameterization wind turbines act as a sink for momentum leading to a reduction of the wind speed. The released energy is distributed between increased levels of turbulent kinetic energy and power production. The Fitch 2012 parameterization is commonly used in numerical weather prediction models and has the advantage of very little tuning parameters.


ASPIRE

ASPIRE (Atmospheric Simulation Platform for Innovation, Research, and Education) is a large-eddy simulation (LES) model operated by Whiffle. ASPIRE originates from the Dutch Atmosperic Large-Eddy Simulation (DALES) model, which is extensively described in (9). DALES has been and is still widely used in the boundary-layer meteorology community. To overcome the barrier of the large computational costs that have long prohibited the use of LES in operational weather forecasting, the DALES model was translated to a code that runs most of its computational routines on GPUs (10). ASPIRE is used both as an operation weather model (to provide high-resultion wind and power forecasts) and for making wind resource assessments.

In WINS50, Whiffle's LES model ASPIRE is used to perform LES simulations at a domain of roughly 300 by 500 km. In order to enable sensible simulations at a domain of this size, major model development steps had been carried out. The most fundamental development concerns the transition from a model with periodic boundary conditions to a model with open/prescribed boundary conditions. In the prescribed boundary conditions setting, ASPIRE has a one-way nesting structure as applied in most limited-area NWP models. The model state of the outer domains is prescribed at the boundaries of the inner domains. In the inner domain of ASPIRE, which has a typical grid spacing of 10 to 100 m, the LES equations are solved.

ASPIRE uses an actuator disk parametrisation as described by (11). This parametrisation only needs information about the power curve, thrust curve, rotor diameter and hub height. This information is publicly available or can be estimated with good accuracy. The parametrisation calculates the drag forces (using the thrust curve) and rotational forces (using the power curve) based on the local wind speed, taking the actual induction into account. Individual yaw control based on the local wind direction is applied to the turbines.




  1. M Dirksen, I Wijnant, P Siebesma, P Baas, NE Theeuwes (2022): Validation of wind farm parameterisation in Weather Forecast Model HARMONIE-AROME - Analysis of 2019, TU Delft Report.
  2. B van Stratum, NE Theeuwes, J Barkmeijer, B van Ulft, I Wijnant (2022): A year-long evaluation of a wind-farm parameterisation in HARMONIE-AROME, J. Adv. Mod. Earth Sys., 4, e2021MS002947.
  3. M Dirksen, M Rijntalder, P Baas, I Wijnant (2022): Wake analysis of coastal wind turbines in GRASP on the Dutch shoreline, Wins50 Report.
  4. P Baas, R Verzijlbergh, P van Dorp, H Jonker (2023): Investigating energy production and wake losses of multi-gigawatt offshore wind farms with atmospheric large-eddy simulation, Wind Energy Sci., 8, 787-805.
  5. L Bengtsson, U Andrae, T Aspelien, Y Batrak, J Calvo , W de Rooy, E Gleeson, B Hansen-Sass, M Homleid, M Hortal, KI Ivarsson, G Lenderink, S Niemelä, K Pagh Nielsen, J Onvlee, L Rontu, P Samuelsson, D Santos Muñoz, A Subias, S Tijm, V Toll, X Yang, M Odegaard Koltzow (2017): The HARMONIE-AROME Model Configuration in the ALADIN-HIRLAM NWP System, Monthly Weather Review, 145, 1919-1935.
  6. JB Duncan, IL Wijnant, S Knoop (2019): DOWA validation against offshore mast and LiDAR measurements, TNO report 2019 R10062.
  7. S Knoop, P Ramakrishnan, I Wijnant (2020): Dutch Offshore Wind Atlas Validation against Cabauw Meteomast Wind Measurements, Energies, 13, 6558.
  8. AC Fitch, JB Olson, JK Lundquist, J Dudhia, AK Gupta, J Michalakes, I Barstad (2012): Local and Mesoscale Impacts of Wind Farms as Parameterized in a Mesoscale NWP Model, Monthly Weather Review, 140, 3017-3038.
  9. T Heus, CC van Heerwaarden, HJJ Jonker, AP Siebesma, S Axelsen, K van den Dries, O Geoffroy, AF Moene, D Pino, SR de Roode, J Vilà-Guerau de Arellano (2010): Formulation of the Dutch Atmospheric Large-Eddy Simulation (DALES) and overview of its applications, Geosci. Model Dev., 3, 415-444.
  10. J Schalkwijk, EJ Griffith, FH Post, HJJ Jonker (2012): High-Performance Simulations of Turbulent Clouds on a Desktop PC: Exploiting the GPU, Bull. Am. Met. Soc., 93, 307-314.
  11. J Meyers, C Meneveau (2012): Large Eddy Simulations of large wind-turbine arrays in the atmospheric boundary layer, in 48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition.