Professional Documents
Culture Documents
Munters - 2022 - J. - Phys. - Conf. - Ser. - 2265 - 022049
Munters - 2022 - J. - Phys. - Conf. - Ser. - 2265 - 022049
Abstract. A first assessment of the potential wake impact of the future Princess Elisabeth
wind-farm cluster on the existing Belgian wind farms is performed. We consider 3 different wake
models coupled to a blockage model implemented in FLORIS, and study 15 design scenarios for
the future cluster. Simulations show that, although intra-cluster wake effects are qualitatively
comparable, inter-cluster effects differ strongly among model setups, confirming results in recent
literature. With increasing new-zone capacity, a trend of higher existing-zone AEP loss caused
by the new zone is observed, as well as an incentive to use turbines with higher individual rating.
Quantitatively, AEP loss due to inter-cluster wakes can reach up to 0.8% for the full existing
zone as compared to a reference case without the Princess Elisabeth zone. Further, worst-case
conditions with west-southwesterly winds show the new zone induces an inter-cluster power loss
of 6% for the entire existing zone, with extremes up to 20% for specific turbines.
1. Introduction
Considering efforts for the transition towards a renewable energy system, offshore wind is
expected to take up an increasingly important role in the European electricity landscape. More
specifically, current European ambitions are to deploy up to 400 GW of offshore wind energy
across the continent’s seas by 2050, of which over 200 GW is expected to be installed in the
North Sea [1]. Even though these farms would take up only a limited fraction of the total sea
area, the amount of space with environmental conditions suitable for bottom-fixed wind turbines
is relatively scarce, which can result in closely-spaced wind-farm clusters. For the Belgian case
specifically, the exclusive economic zone of the North Sea is relatively small, further limiting the
amount of space to install offshore wind farms. In 2020, the first Belgian offshore wind energy
zone adjacent to the border with the Netherlands was completed, with an installed capacity of
2.2 GW (zone A on Figure 1a). Over the coming decade, the new Princess Elisabeth zone will
be developed (B, C, D on Figure 1a), where an additional capacity similar to the current zone
will be installed. Considering the dominance of south-westerly winds in the Belgian part of the
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
North Sea as shown in Figure 1b, this new zone will be some tens of kilometers upwind of the
existing zone from the dominant wind direction, with potential wake impact on the latter.
Historically, significantly more focus has been on the wake effects within clusters, where wakes
travel relatively short distances from turbine to turbine. As an example for Belgium, Bulder and
Bot [2] performed a preliminary study on the installation of the 1.4 GW Dutch Borssele farm
adjacent to the Belgian zone on the northeastern side (see Figure 1a), hence forming a single
homogeneous Dutch – Belgian cluster, and report intra-cluster losses in the order of 2 to 3%
for the Belgian farms. In practice however, wakes have also been observed to extend over long
distances void of turbines in between farms or clusters [3, 4]. Therefore, the existing Belgian
zone is expected to experience wake effects also from the new Princess Elisabeth cluster.
Fast engineering models are most often the tool of choice for modeling wake impacts on
annual energy production (AEP), as the necessitated time horizons and different design scenarios
impede the use of high-fidelity techniques. However, studies using these models for inter-cluster
interactions have been relatively scarce to date. Furthermore, the existing studies report mixed
success and strong spread among different models. Hansen et al. [5] compare various engineering
models, as well as RANS and mesoscale tools, to SCADA results for the Rødsand and Nysted
wind farms in the Baltic sea. They conclude that all models under consideration capture to some
extent the wake deficit in the downstream farm, yet a significant spread exists. In a follow-up
study on this case, Fischereit et al. [6] illustrate that especially the standard Jensen [7] and
Gaussian model [8] in PyWake tend to overestimate inter-cluster wake recovery. Nygaard et
al. [9] come to a similar conclusion where Jensen (Park) underestimates cluster wake deficits
for the Westermost Rough and Humber Gateway neighboring wind farms in the English North
Sea, but propose a turbulence-optimized extension of the Park model (TurbOPark) which is
shown to greatly improve the match with SCADA data. This improvement of TurbOPark over
conventional models was recently also observed in canonical large-eddy simulations [10].
The objective of the current study is to perform a first assessment of the potential impact
of the Princess Elisabeth zone on the existing zone in terms of power loss, using wake models
readily available in the open source engineering FLORIS framework1 . We study the influence
of design choices for the new zone, i.e. on individual turbine rating and total capacity, both in
terms of expected losses in AEP and temporary power losses during worst-case wind conditions.
Furthermore, inspired by the reported model spread in literature as mentioned above, we assess
the influence of wake model setup on inter-cluster wake loss.
2. Methodology
The large spatial extent of the Belgian North Sea zone combined with the aim of estimating
effects on full AEP require an efficient wake modeling strategy. In the current study, we
use 3 different wake modeling setups readily available in the open-source FLORIS framework,
and combine them with an induction model to account for blockage effects. Simulations are
performed for 15 design scenarios for the new zone, comprising of 5 total capacities of the new
zones using 3 different turbine ratings. Furthermore, an additional simulation with only the
existing zone serves as a reference scenario. This results in 16 design scenarios, or a total of 48
cases to be simulated, considering the different wake model setups. For each case, first a lookup-
table is generated to cover the full wind rose. Next, in order to estimate AEP, the lookup-table
is linked to long-term meteorological data to account for typical southwesterly wind conditions
in the Belgian North Sea. Note that in the current paper, AEP is defined solely based on the
wind resource and losses related to blockage and wakes. Other loss sources are not taken into
account. Further remark that in the remainder of the text, the term ‘losses’ refers to inter-cluster
losses in the existing zone caused by the impact of the new zone, rather than intra-cluster or
1
FLORIS. Version 2.4 (2021). Available at https://github.com/NREL/floris
2
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
Figure 1. (a) Illustration of the Belgian offshore wind energy zone, including existing wind
farms (zone A), as well as the new Princess Elisabeth concession (zones B, C, and D). Figure
courtesy of the Belgian offshore platform3 . (b) Wind rose of Belgian North Sea conditions,
indicating predominantly southwesterly winds.
intra-farm losses within either of these. Section 2.1 details the meteorological conditions used
as inputs to the modeling framework. Thereafter, Section 2.2 shows the design choices of the
Princess Elisabeth zone considered in this work. Next, Section 2.3 outlines the coupling between
wake and induction models based on the WIZ framework. Finally, 2.4 discusses the 3 different
wake model setups used in this paper.
3
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
1.75 GW capacity, 17MW turbines 2.8 GW capacity, 15MW turbines 3.5 GW capacity, 13MW turbines
40 40 40
Northing [km]
30 30 30
20 20 20
10 10 10
0 0 0
0 20 40 60 0 20 40 60 0 20 40 60
Easting [km] Easting [km] Easting [km]
Figure 2. Examples of wind farm layouts considered in wake studies. Turbine locations are
indicated by circled markers. Marker colors correspond to different wind farms, with new-zone
turbines indicated by dark blue circles.
zone turbines are upscaled extrapolations of the largest turbines in the existing zone. Existing
turbine coordinates are derived from the as-built documentation by the Royal Belgian Institute
of Natural Sciences [12]. Given 5 target capacities as well as 3 individual turbine ratings, 15
wind-farm layouts for the new zone were created from a rough micro-siting exercise respecting
environmental and shipping-route constraints. Considering the southwestern location of the new
concession with respect to the existing zone, cluster-to-cluster wakes and losses due to the new
concession are expected to occur in southwesterly winds. For this reason, we omit the Borssele
wind farm zone adjacent on the northeastern side (see Figure 1a), as it is not expected to directly
influence the new-to-existing zone interactions. Figure 2 illustrates 3 sample layouts.
4
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
Model setup 1: Jensen The first velocity deficit model is the Jensen model [7]. This simple
model assumes wakes develop in a cone shape downstream of the originating turbine, with a
constant expansion coefficient kw , thus resulting in a linear expression for the wake cone diameter
for t = 1, . . . , Nt , with Nt the total amount of turbines considered. Herein Dt is the rotor
diameter of turbine t and x denotes the axial distance from the rotor location. The Jensen
model further assumes a top-hat shape for the velocity in the wake, and states the velocity
deficit ∆Ut for t = 1, . . . , Nt as4
Ut p Dt
∆Ut (x) = U∞ 1 − 1 − CT,t , (2)
U∞ Dw,t (x)
with U∞ the undisturbed freestream velocity, Ut the streamwise velocity at the rotor, and CT,t
the turbine thrust coefficient, evaluated from performance curves at U = Ut . As can be seen,
kw is the sole empirical parameter governing wake development, and thus has to account for a
multitude of flow conditions, e.g. on atmospheric stability and turbulence intensity. For offshore
wind farms, a value between 0.03 and 0.05 is generally recommended. In this study, we employ
kw = 0.04, similar to Nygaard et al. [9]. In the standard Jensen model, a uniform kw is assumed
for all turbines, and the model is thus not further sensitized to the development of turbulence
intensity in the wake and throughout the wind farm. Although several studies have proposed
extensions of Jensen to circumvent this [16], they are not currently available in FLORIS.
Model setup 2: Gaussian Second, we consider the self-similar Gaussian deficit model [8], which
is currently considered the default wake model for use in FLORIS, and has been shown to achieve
a good match with field data [17]. Of the multiple Gaussian models available in FLORIS, we use
the gauss_legacy wake model, which is directly based on Refs. [8, 18]. The Gaussian model
considers conversation of mass and momentum and assumes a Gaussian-like velocity deficit,
with a diffuse wake diameter characterized by its standard deviation σw . Similar to the Jensen
model, a linear wake expansion is postulated as
Further, ε quantifies the scaling between Dt and σw,t (0) at the rotor, and can be expressed as a
non-linear function of CT [8]. The velocity deficit is then computed through
s
2 2
C T,t D t 1 r
∆Ut (x, r) = U∞ 1 − 1 − exp − , (5)
8 σw,t (x) 2 σw,t (x)2
for t = 1, . . . , Nt , where r is the radial distance to the rotor centerline. As shown in Eq. (4),
the Gaussian model is sensitive to total turbulence intensity It at the turbine rotor location.
4
In the FLORIS implementation
√ of the Jensen model, the factor Ut /U∞ is not included, hence deficits are
computed as ∆Ut = U∞ 1 − 1 − CT D/Dw , t(x).
5
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
To determine It for every turbine, the ambient intensity I∞ is combined with wake-generated
turbulence I+ originating from upstream turbines j in quadrature, i.e.
s X
It = I∞2 + 2 ,
I+,j (6)
j∈Ω
where Ω is the set of upstream turbines affecting the current turbine t. Previous studies have
however shown that this effect is dominated strongly by closely neighboring turbines alone [19].
In practice, FLORIS uses a cutoff separation distance of 15D. To model the wake-generated
turbulence terms I+,j , the Crespo-Hernandez model [20] is used, i.e.
I+,j = 0.5aj0.8 I∞
0.0325
(sj /D)−0.32 , (7)
where aj is the induction factor of turbine j, and sj is its axial distance to turbine t. An
important remark is that, once kt∗ is computed from It in Eq. (4), wake decay is no longer
influenced by the evolution of turbulence intensity downstream of the rotor. Within wind
farms, where hub-height turbulence becomes fully-developed after only a few rows (see, e.g.
[21]), this assumption is likely valid. However, when considering wake development over large
distances void of wind turbines, as in farm-farm wake interactions, turbulence will decay
resulting in delayed wake recovery. Hence, for such applications, the Gaussian model is prone
to overestimating recovery, leading to underestimation of farm-farm interactions [6].
with empirical constants c1 = 1.5 and c2 = 0.8. Based on a calibration for Ørsted offshore wind
farms, the current recommendation for the expansion parameter is A = 0.6. Remark the duality
between on the one hand the Gaussian setup, which includes turbulence from upstream wakes
through I+,j in Eq. (6) but does not account for reduced wake expansion due to downstream
turbulence decay, and on the other hand the TurbOPark setup, where downstream decay is
accounted for by the non-linear expansion in Eq. (8) but effects from turbines further upstream
are neglected by the Frandsen turbulence model, which accounts only for the closest turbine.
Note that, in the FLORIS implementation of TurbOPark, the Frandsen model is already included
in the deficit model, and no specific turbulence model is activated. Figure 3 illustrates the flow
fields of a 4 × 4 aligned wind farm simulation for each model setup, both with and without the
induction model coupling.
6
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
Figure 3. Visualization of axial velocity at hub height for each wake model setup (J, G, T)
from an example simulation with a 4 × 4 aligned wind farm. Turbine locations are indicated
with black lines. Left: basic FLORIS wake models. Right: FLORIS wake models coupled with
induction model through WIZ. Contours indicate values of 0.71, 0.8, 0.9, 0.95, 0.99 and 1.
Figure 4. Normalized axial velocity for 3.5 GW – 13 MW scenario in Jensen (left), Gaussian
(center) and TurbOPark (right) model setup, with southwesterly winds of U∞ = 10 m/s.
7
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
Figure 5. (a) Normalized new-zone AEP for the 15 scenarios and the 3 wake models. Top
ticks indicate total new-zone capacity. Bottom ticks indicate turbine rating. New-zone AEP
normalized by Gauss 3.5 GW – 15 MW scenario (case with highest new-zone AEP). (b) Relative
existing-zone AEP loss due to new-zone wake. (c) Relative AEP loss due to new-zone wake for
existing turbines in TurbOPark 3.5 GW – 13 MW scenario (case with highest loss).
8
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
P − Pref Pref − P
(a) NW Pref [%] (b) Pref [%] U/U∞
20.0
10 1.0
40
17.5
35 15.0 0.8
5
0 12.5
U∞
15
[m
25
/s]
20 7.5 0.4
25
−5
30 20 5.0
0.2
SW SE 2.5
−10 15
30 35 40 45 50 55 0.0 0.0
S Eas ing [km]
Figure 6. (a) Relative normalized inter-cluster AEP loss (loss is depicted negative) for
TurbOPark 3.5 GW – 13 MW case. Black circle indicates conditions with highest absolute
loss, for which relative loss is 5.83%. (b) Flow field for worst-case conditions. Turbines colored
by relative inter-cluster AEP loss.
3.2. Inter-cluster wake effects on power extraction during worst-case wind conditions
In addition to the AEP analysis from previous section, here we identify worst-case conditions
resulting in the largest power losses. This is particularly relevant for future electricity systems
reliant on offshore wind, as a momentary drop in wind power due to farm interactions can
potentially cause unexpected grid instability. Figure 6 analyzes these worst-case conditions for
the TurbOPark 3.5 GW – 13 MW scenario. The polar plot in Figure 6a illustrates the relative
power difference of the full existing zone with the reference case as a function of wind speed
and direction. Farm interactions are shown to be present for westerly to southwesterly wind
directions. Noteworthy is the actual power gain at very high wind speeds of 26 m/s, for which
the sheltering effect of the new-zone wake will cause existing-zone turbines to regain operation
as local wind speeds are reduced below cutout. Power losses however concentrate in winds below
rated speed, with the most severe relative loss higher than 10% at a low wind speed of 5 m/s,
where the new-zone wake causes many existing turbines to fall below cutin wind speeds.
An important remark to make is that losses reported in this figure are relative, and different
datapoints are normalized with different power extractions in the reference case. For this reason,
we define the worst case as associated with the highest absolute power loss. This case was
identified to be at a west-southwesterly wind direction at a free-stream wind speed U∞ = 11 m/s,
with a relative loss of 5.83%, as indicated by the black circle in Figure 6a. The corresponding
flow field and turbine power losses are shown in Figure 6b. The figure shows that, for these
conditions, virtually the entire existing zone is immersed in the new-zone wake, and individual
turbine losses can be as high as 20%.
4. Conclusion
In this paper, we perform a first assessment of the potential wake impact of the future Princess
Elisabeth wind-farm cluster on the existing Belgian wind farms. We consider the Jensen,
Gaussian, and TurbOPark implementations in FLORIS coupled to a vortex-cylinder blockage
model, and study 15 design scenarios for the future cluster by varying the turbine type and total
capacity. Simulations show that, although intra-cluster wake effects are qualitatively comparable
between model setups as evidenced by new-zone AEP, the inter-cluster effects are vastly different,
with Jensen and Gauss predicting far lower inter-cluster AEP loss in the existing zone, confirming
9
The Science of Making Torque from Wind (TORQUE 2022) IOP Publishing
Journal of Physics: Conference Series 2265 (2022) 022049 doi:10.1088/1742-6596/2265/2/022049
results from recent literature. A clear trend of higher inter-cluster wake impact with increased
new-zone capacity is observed, as well as an incentive to use turbines with higher rating in
the new zone in balancing new-zone AEP and existing-zone AEP loss. Quantitatively, AEP
loss induced by the new zone can reach up to 0.8% for the entire existing zone, and worst-case
conditions with west-southwesterly winds result in a power loss of almost 6%, with individual
turbine losses up to 20%. However, although clear trends can be observed from the current
study, it is important to highlight that further validation with SCADA data is necessary and
that dedicated follow-up studies, e.g. using mesoscale models, should consider flow physics not
captured in the current model setup, for example the influence of atmospheric stability and waves
on wake recovery [22], and the impact of self-induced gravity waves on wind-farm blockage [23].
References
[1] WindEurope technical report “Our energy, our future - how offshore wind will help Europe go carbon-neutral”,
November 2019
[2] Bulder BH and Bot ET. Quick scan of energy production and wake losses of the Borssele Wind Farm Zone
including losses for nearby Belgian Wind Farms. ECN TNO report 2015
[3] Nygaard NG, Newcombe AC. Wake behind an offshore wind farm observed with dual-Doppler radars.
InJournal of Physics: Conference Series 2018 Jun 1 (Vol. 1037, No. 7, p. 072008). IOP Publishing.
[4] Schneemann J, et al. Cluster wakes impact on a far-distant offshore wind farm’s power. Wind Energy Science.
2020 Jan 3;5(1):29-49.
[5] Hansen KS, et al. Simulation of wake effects between two wind farms. InJournal of Physics: Conference Series
2015 Jun 1 (Vol. 625, No. 1, p. 012008). IOP Publishing.
[6] Fischereit J et al. Comparing and validating intra-farm and farm-to-farm wakes across different mesoscale and
high-resolution wake models. Wind Energy Science Discussions. 2021 Oct 21:1-31.
[7] Jensen NO. A note on wind generator interaction. Riso National Laboratory, 1983.
[8] Bastankhah M and Porté-Agel F. A new analytical model for wind-turbine wakes. Renewable Energy,
70:116–123, 2014.
[9] Nygaard NG, Steen ST, Poulsen L, Pedersen JG. Modelling cluster wakes and wind farm blockage. In Journal
of Physics: Conference Series 2020 Sep 1 (Vol. 1618, No. 6, p. 062072). IOP Publishing.
[10] Stieren A, Stevens RJ. Evaluating wind farm wakes in large eddy simulations and engineering models.
InJournal of Physics: Conference Series 2021 May 1 (Vol. 1934, No. 1, p. 012018). IOP Publishing.
[11] Marek P, Grey T, Hay A. A study of the variation in offshore turbulence intensity around the British Isles.
WindEurope summit 2016, Hamburg.
[12] Royal Belgian institute of Natural Science website, Management of the marine environment.
https://odnature.naturalsciences.be/mumm/en/windfarms/
[13] Branlard E, Meyer Forsting AR. Assessing the blockage effect of wind turbines and wind farms using an
analytical vortex model. Wind Energy. 2020 Nov;23(11):2068-86.
[14] Branlard E et al. Wind farm blockage effects: comparison of different engineering models. InJournal of
Physics: Conference Series 2020 Sep 1 (Vol. 1618, No. 6, p. 062036). IOP Publishing.
[15] Katic I, Højstrup J, Jensen NO. A simple model for cluster efficiency. InEuropean wind energy association
conference and exhibition 1986 Oct 7 (Vol. 1, pp. 407-410).
[16] Duc T, et al. Local turbulence parameterization improves the Jensen wake model and its implementation for
power optimization of an operating wind farm. Wind Energy Science. 2019 May 22;4(2):287-302.
[17] Annoni J et al. Analysis of control-oriented wake modeling tools using lidar field results. Wind Energy
Science. 2018 Nov 1;3(2):819-31.
[18] Niayifar A and Porté-Agel F. Analytical modeling of wind farms: a new approach for power prediction.
Energies, 9(9):741, 2016.
[19] Frandsen S Turbulence and turbulence generated structural loading in wind turbine clusters Risøreport 2003
[20] Crespo A, Hernandez J. Turbulence characteristics in wind-turbine wakes. Journal of wind engineering and
industrial aerodynamics. 1996 Jun 1;61(1):71-85.
[21] Munters W, Meyers J. Towards practical dynamic induction control of wind farms: analysis of optimally
controlled wind-farm boundary layers and sinusoidal induction control of first-row turbines. Wind Energy
Science. 2018 Jun 18;3(1):409-25.
[22] Porchetta S et al. Impact of ocean waves on offshore wind farm power production. Renewable Energy. 2021
Dec 1;180:1179-93.
[23] Allaerts D et al. Annual impact of wind-farm gravity waves on the Belgian–Dutch offshore wind-farm cluster.
In Journal of physics: conference series 2018 Jun 1 (Vol. 1037, No. 7, p. 072006). IOP Publishing.
10