As a consequence of the rapid growth of the globally installed offshore wind energy capacity, the size of individual wind farms is increasing. This poses a challenge to models that predict energy production. For instance, the current generation of wake models has mostly been calibrated on existing wind farms of much smaller size. This work analyzes annual energy production and wake losses for future, multi-gigawatt wind farms with atmospheric large-eddy simulation. To that end, 1 year of actual weather has been simulated for a suite of hypothetical 4 GW offshore wind farm scenarios. The scenarios differ in terms of applied turbine type, installed capacity density, and layout. 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. Even for turbine types with similar rated power but slightly different power curves, significant differences in production were found. Although wind speed was identified as the most dominant factor determining the aerodynamic losses, a clear impact of atmospheric stability and boundary layer height has been identified. By analyzing losses of the first-row turbines, the yearly average global-blockage effect is estimated to between 2 and 3 %, but it can reach levels over 10 % for stably stratified conditions and wind speeds around 8 m s−1. Using a high-fidelity modeling technique, the present work provides insights into the performance of future, multi-gigawatt wind farms for a full year of realistic weather conditions.

In the next few decades climate mitigation efforts will transform the North Sea into one of the most important energy sources. The present wind energy capacity on the North Sea is expected to increase by almost a factor 5 in 2030 and almost a factor 10 in 2050. It is therefore of paramount importance to know how wind farms influence the atmosphere.

Wind farms extract kinetic energy from the atmosphere and in doing so decrease the wind speed and increase turbulence levels. More turbulence means more mixing of vertical layers in the atmosphere and a change in humidity and temperature profiles. This may lead to cloud forming or dissipation. Wind farms are also an obstacle to the flow, which is what is called the blockage effect, as opposed to the wake effect behind the wind farm. This report is about the wake effect, mainly on wind, but we also analysed temperature and humidity profiles.

In order to assess and quantify the wake effect, we compared two high resolution re-analyses for the year 2019 on a 2000 by 2000 km North Sea domain. The high resolution re-analyses with a 2.5 km horizontal grid spacing is based on global re-analysis ERA5 and downscaled with mesoscale weather model HARMONIE-AROME which is used operationally at KNMI. One of the re-analyses is without the effect of wind farms (referred to as control or HarmCY43-CTL in this report) and one with the Fitch wind farm parametrization that was recently incorporated in HARMONIE-AROME (HarmCY43-WFP). From the differences between the two we can isolate the wind speed deficits, or wakes, from the wind farms.

Earlier validation studies have shown that a previous version of the HARMONIE-AROME model (HarmCY40) produces accurate wind climatology for undisturbed wind fields (period 2008-2018) and validates well against disturbed tower, aircraft and lidar measurements from 2016. In these studies the wind climatology is not validated for different stability regimes. In this study we do make that distinction and use measurements from 2019 for validation of HarmCY43-CTL and HarmCY43-WFP.

* Generally HarmCY43-WFP outperforms HarmCY43-CTL in wake areas. HarmCY43-WFP even seems to capture the wind in wind farms reasonably well, although the WFP is not designed for that.
* The selection criterion that we used to select disturbed (in wakes) and undisturbed wind directions (outside wakes) seems to work well: the WFP reduces the wind speed bias for disturbed winds significantly, but hardly affects undisturbed winds.
* Our results confirm earlier studies that wakes are strongest for situations with stable stratification: we observed wake lengths as long as about 50 km. We can conclude that HarmCY43-CTL tends to underestimate the wind speed for stable stratification and overestimate the wind speed for weakly stable and unstable stratification, mainly for the lidar measurements. As expected HarmCY43-WFP reduces the wind speed in the wake. This means that HarmCY43-WFP validates better against measurements for weakly stable and unstable stratification. However, for stable stratification HarmCY43-WFP makes the underestimation of the measurements worse (note that this does not imply the wake deficits are biased). This could even become worse if wind turbines are not performing according to the power curve or are not turning at all because of maintenance or legislation, the WFP will not be aware of that and will extract too much energy, overestimate the wake effect and underestimate the wind speed.
* Earlier studies have shown that HarmCY40-CTL captures the diurnal cycle well. HarmCY43-CTL does as well and including the WFP does not seem to affect that.

The results of this study give us confidence that the present HARMONIE-AROME model configuration, including the Fitch WFP, can be used to assess the influence of the anticipated wind farm infrastructure in 2050 on the wind climatology.