the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
On the power and control of a misaligned rotor – beyond the cosine law
Simone Tamaro
Filippo Campagnolo
We present a new model to estimate the performance of a wind turbine operating in misaligned conditions. The model is based on the classic momentum and liftingline theories, considering a misaligned rotor as a lifting wing of finite span, and accounts for the combined effects of both yaw and uptilt angles.
Improving on the classical empirical cosine law in widespread use, the new model reveals the dependency of power not only on the misalignment angle, but also on some rotor design parameters and – crucially – on the way a rotor is governed when it is yawed out of the wind. We show how the model can be readily integrated with arbitrary control laws below, above, and around the rated wind speed. Additionally, the model also shows that a sheared inflow is responsible for the observed lack of symmetry for positive and negative misalignment angles. Notwithstanding its simplicity and insignificant computational cost, the new proposed approach is in excellent agreement with large eddy simulations (LESs) and wind tunnel experiments.
Building on the new model, we derive the optimal control strategy for maximizing power on a misaligned rotor. Additionally, we maximize the total power of a cluster of two turbines by wake steering, improving on the solution based on the cosine law.
 Article
(6324 KB)  Fulltext XML
 BibTeX
 EndNote
Wind farm control by wake steering consists of deflecting the wake away from downstream rotors to boost the total power of a plant (Meyers et al., 2022). The effectiveness of this control strategy has been proven numerically (Jiménez et al., 2010), experimentally in the wind tunnel (Campagnolo et al., 2016), and in field tests (Fleming et al., 2019; Doekemeijer et al., 2021). At the core of the powerboosting ability of wake steering is a tradeoff: on the one hand, there is an enhanced momentum of the inflow at a downstream turbine when a wake is shifted laterally away from it; on the other hand, some power is lost at the upstream misaligned rotor, because it does not point into the wind anymore. In general the tradeoff budget is positive, in the sense that the power that is gained downstream is larger than the power upstream. The problem is, however, highly complex: downstream, power capture is determined by the interaction of the impinged rotor with the wake that, in turn, is influenced by the ambient conditions and those of the wakeshedding turbine; upstream, power losses depend on the inflow characteristics but also on the rotor and on the way it is governed. Understanding and controlling this delicate balance between upstream and downstream behavior is clearly of paramount importance for improving the power capture of wind farms by wake steering. Great progress has been made in recent years to understand, model, and control wakes (see, for example, the review in Meyers et al., 2022), i.e., on the downstream aspect of the problem. However, the upstream aspect remains much less explored and understood. How much power does a yawed turbine really lose? And what inflow, rotor, and rotorcontrol parameters influence its behavior? It is a major aim of this paper to try and answer these questions.
The aerodynamic power P of a wind turbine is customarily written as $P=\frac{\mathrm{1}}{\mathrm{2}}\mathit{\rho}A{u}_{\mathrm{\infty}}^{\mathrm{3}}{C}_{\mathrm{P}}$, where ρ is the air density, A is the rotor swept area, u_{∞} is the ambient freestream wind speed, and C_{P} is the power coefficient. When a turbine is misaligned with respect to the wind vector by an angle γ, the rotororthogonal velocity component becomes u_{∞}cos γ. Accordingly, one would expect the yawinduced power loss to be ${\mathit{\eta}}_{\mathrm{P}}=P/{P}_{\mathrm{0}}={\mathrm{cos}}^{\mathrm{3}}\mathit{\gamma}$, where P_{0} is the aerodynamic power produced for γ=0. Unfortunately, this is only a naive interpretation of the true behavior of a misaligned rotor, and its predictions are not confirmed by experimental and numerical observations (Liew et al., 2020). To reflect this fact, a pragmatic solution has been adopted by most of the literature, where power losses due to misalignment are assumed to obey the simple law ${\mathit{\eta}}_{\mathrm{P}}\approx {\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}\mathit{\gamma}$, where p_{p} is a tunable parameter.
Unsurprisingly, since such a model is not based on actual physics, a large spread of values for p_{p} has been reported in the literature. In wind tunnel experiments with scaled models, Campagnolo et al. (2020) measured p_{p}=2.1, Krogstad and Adaramola (2012) and Bartl et al. (2018) reported p_{p}≈3, and Medici (2005) found a value p_{p}=2. Numerically, Fleming et al. (2015) measured p_{p}=1.88 on the NREL 5 MW wind turbine (Jonkman et al., 2009), whereas Draper et al. (2018) obtained values between 1.3 and 2.5 for scaled wind turbine models operating in waked inflow conditions. The power production in misaligned conditions has also been measured in multiple field tests. For example, Fleming et al. (2017) reported a value of 1.41 for an Envision 4 MW turbine, and Dahlberg and Montgomerie (2005) published a range of values for p_{p} between 1.9 and 5.1 at an offshore plant. More recently, Hulsman et al. (2022) observed $\mathrm{2}<{p}_{\mathrm{p}}<\mathrm{2.5}$ at an onshore wind farm in the north of Germany.
The large scatter characterizing the p_{p} coefficients reported in the literature is a relevant source of uncertainty, creating a significant hindrance to the development of wind farm control strategies and suggesting that some relevant phenomena are not captured by the ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ law. In hindsight, this is to be expected, because this simple model fails to explicitly represent how the power coefficient C_{P} changes when a turbine is misaligned, and it somehow absorbs this effect into the tunable exponent. Some indications that there is more to this problem than a simple power cosine law have already been reported by various authors. Based on experiments and numerical simulations, Campagnolo et al. (2023), Cossu (2021a, b), and Heck et al. (2023) suggested that the power of a misaligned rotor strongly depends on its loading, in the form of the thrust coefficient C_{T}; clearly, in turn this has a strong effect on the behavior of the wake (Cossu, 2021a, b). Other variables that have been shown to play a role on power losses are related to the inflow. Recently, Draper et al. (2018) and Liew et al. (2020) have observed that power losses in misaligned conditions differ depending on whether a rotor is waked or not. Howland et al. (2020) observed a significant influence of shear and veer, while Simley et al. (2021) measured a strong dependency on inflow speed. The behavior of power losses has also been shown to depend on the direction of yaw misalignment and not only on its magnitude as implied by the power cosine law. This asymmetric behavior of yaw misalignment has been observed by Fleming et al. (2015), Schottler et al. (2017), Fleming et al. (2018), and Campagnolo et al. (2020), among others. However, an agreement on which misalignment direction yields more or less power has not been reached yet.
In this paper, we present a new analytical model for misaligned wind turbine rotors. The proposed approach combines the classic momentum and liftingline theories, considering a misaligned rotor as a lifting wing of finite span, in close parallel to the analysis conducted for helicopter rotors in forward flight (Johnson, 1995). While existing approaches do not explain the lack of symmetry with respect to yaw direction, the present model includes the effects of wind shear, which is shown to be the culprit for the observed break of symmetry with respect to the misalignment direction. For improved accuracy, the model also includes the effects of the uptilt angle, as it contributes to the overall misalignment of the rotor with respect to the wind vector. Thanks to this feature, the proposed methodology is also readily applicable to vertical wakesteering control, which could be implemented with floating wind turbines (Nanos et al., 2022) or downwind teetering rotors. The resulting model equations are integrated over the blade span and averaged over one rotor revolution, leading to a semianalytical formulation of negligible computational cost that can be readily coupled with engineering wake models such as FLORIS (NREL, 2023b) or PyWake (Pedersen et al., 2019). However, the model governing equations could also be integrated numerically and embedded into blade element momentum (BEM) codes (Hansen, 2015), such as the AeroDyn package (NREL, 2023a) implemented in OpenFAST (NREL, 2023c).
Very recently, Heck et al. (2023) published a misaligned rotor model based on similar arguments. However, their approach does not include the effects of shear and therefore fails to capture the asymmetric behavior of yaw direction. More importantly, their formulation uses a modified thrust coefficient ${C}_{\mathrm{T}}^{\prime}$, which is assumed to remain constant between aligned and misaligned conditions. This hypothesis is indeed verified when the turbine operates in the partialload region. Departing from this approach, the method proposed here is based on a completely general dependency of the thrust coefficient on the misalignment angle and therefore can readily accommodate arbitrary regulation strategies in the partial, full, and intermediate regulation regions, including thrust clipping and derating (Campagnolo et al., 2023). Moreover, the model of Heck et al. (2023) cannot predict power losses higher than cos ^{3}γ, which have been reported in the literature. A detailed comparison of the new proposed model and the one of Heck et al. (2023) is developed in the following.
The proposed semianalytical model shows that the behavior of a misaligned rotor does not follow the ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ law, contradicting this empirical formula in widespread use. Additionally, the new model clarifies the behavior of power capture with respect to some rotor design parameters and – even more importantly – with respect to the way a rotor is governed when it is misaligned. This is an effect that has been neglected so far, but that – as already noted by Howland et al. (2020) – most probably explains the large scatter observed by various authors. Building on the unique ability of the proposed method to handle arbitrary control policies, we derive the optimal strategy for maximizing power capture when pointing a rotor away from the wind. Finally, we implement the semianalytical model in FLORIS, and we optimize the power of a cluster of two turbines. We obtain set points that differ from those that can be computed with the empirical ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ law and that lead to a slight improvement of the cluster power.
The new models exhibits an excellent match with highfidelity LESs obtained with a TUMmodified version of NREL's large eddy simulator actuator line model (LESALM) SOWFA (Fleming et al., 2014; Wang et al., 2019, 2018). Additionally, the model is further validated with wind tunnel data from experiments conducted with the TUM G1 scaled wind turbines (Bottasso and Campagnolo, 2022; Campagnolo et al., 2020).
The paper is organized as follows: Sect. 2 presents the new formulation, and Sect. 3 explains its implementation in an engineering wake model, including the integration with arbitrary control strategies. Next, Sect. 4 considers its validation with respect to simulated and experimental data, while Sect. 5 analyzes the effects of the new model on wake steering. Finally, Sect. 6 draws conclusions and offers an outlook towards future work.
2.1 Frames of reference
Three reference frames are necessary to completely characterize a misaligned rotor interacting with the wind, as shown in Fig. 1: a groundfixed reference frame and a nacellefixed reference frame, which together describe the relative orientation of the rotor with respect to the ground, and a wakedeflection intrinsic frame, which describes the relative orientation of the rotor with respect to the incoming wind vector.
The groundfixed windaligned frame of reference is indicated with a subscript “g” and is defined by the righthanded triad of unit vectors ${\mathcal{F}}_{\mathrm{g}}=\mathit{\{}{\mathit{x}}_{\mathrm{g}},{\mathit{y}}_{\mathrm{g}},{\mathit{z}}_{\mathrm{g}}\mathit{\}}$. z_{g} points vertically down towards the ground; x_{g} is parallel to the terrain pointing downstream and is contained in the plane formed by the wind vector u_{∞} and z_{g}; finally, y_{g} completes a righthanded triad. In the following, for simplicity we consider the wind vector to be parallel to the terrain, i.e., u_{∞}∥x_{g}, although this is not strictly necessary.
The nacellefixed frame of reference is indicated with a subscript “n” and is defined by the triad of unit vectors ${\mathcal{F}}_{\mathrm{n}}=\mathit{\{}{\mathit{x}}_{\mathrm{n}},{\mathit{y}}_{\mathrm{n}},{\mathit{z}}_{\mathrm{n}}\mathit{\}}$. ℱ_{n} is obtained from ℱ_{g} by two successive rotations: a first rotation by the tilt angle δ about the horizontal axis y_{g}, followed by a second rotation by the yaw angle γ about the vertical axis z_{g}. Both rotations are positive about their respective axes according to the righthand rule (notice that, according to this definition, the typical uptilt of an upwind turbine results in a negative value for δ).
However, the interaction of the rotor with the flow depends only on their mutual orientation and not on how they are oriented with respect to the ground, which is a fundamental principle of fluid mechanics known as Galilean relativity. Therefore, a third frame is necessary, which is termed here wakedeflection intrinsic frame and is indicated with a subscript “d”. The frame is formed by a righthanded triad of unit vectors ${\mathcal{F}}_{\mathrm{d}}=\mathit{\{}{\mathit{x}}_{\mathrm{d}},{\mathit{y}}_{\mathrm{d}},{\mathit{z}}_{\mathrm{d}}\mathit{\}}$. Vector x_{d} is parallel to the rotor axis, i.e., x_{d}=x_{n}, while vectors y_{d} and z_{d} are contained in the rotor disk plane Ψ. Together, the rotor axis x_{d} and the wind velocity vector u_{∞} define the Π plane. The angle in the Π plane between these two vectors is the true misalignment angle μ:
where ${u}_{\mathrm{\infty}}=\left{\mathit{u}}_{\mathrm{\infty}}\right$ is the scalar ambient wind speed. The unit vector z_{d} is orthogonal to the Π plane, i.e., ${\mathit{z}}_{\mathrm{d}}={\mathit{x}}_{\mathrm{d}}\times {\mathit{u}}_{\mathrm{\infty}}/\left({u}_{\mathrm{\infty}}\mathrm{sin}\mathit{\mu}\right)$, while unit vector y_{d} is finally chosen to form a righthanded triad. Using the coordinate transformations in Appendix A, it can be readily shown that cos μ=cos δcos γ; i.e., the total misalignment is caused by both the tilt and yaw angles, with the former typically being neglected in most wake models. Notice that, given its definition, the misalignment angle μ is always positive, because z_{d} flips from one side of the Π plane to the other, depending on the relative orientation of the wind velocity and rotor axis vectors. When the wind comes from the right looking upstream in the Π plane, z_{d} points downwards (see Fig. 1d), whereas it points upwards when the wind comes from the left.
Figure 2 shows a visualization of the wakes developing behind a wind turbine rotor for two different pairs of tilt and yaw values: δ=0°, $\mathit{\gamma}=\mathrm{30}$° and $\mathit{\delta}=\mathrm{28.43}$°, $\mathit{\gamma}=\mathrm{10}$°. Both pairs correspond to a same true misalignment angle μ=30°. The figure confirms that the wake is invariant for an observer on the ℱ_{d} frame. This is particularly evident in the images of the longitudinal speed on the Π plane (marked with a solid black border), which are identical in the two cases. Clearly, for large values of tilt the interaction of the wake with the ground or with a sheared inflow would break the Πframe invariance.
As a result, in the following the wake analysis is developed in the Π plane instead of the horizontal one as customarily done. Transformation matrices that map vector components from one frame to the other are reported in Appendix A.
2.2 Sheared inflow
Considering a linear vertical shear of the inflow, the ambient wind speed is written as
Here u_{∞,hub} is the ambient wind speed at hub height; k is the vertical linear shear coefficient; z_{g} is the vertical coordinate in the ground frame of reference, centered at the hub; and R is the rotor radius. The choice of a linear shear distribution was made just to simplify the derivations, and other choices are clearly possible, for example, to model the more common power law or the presence of lowlevel jets. Additionally, it would be interesting to also include the effects of a horizontal shear, to account for waked conditions, and of veer. These further model improvements are, however, deferred to a continuation of this study.
By applying the coordinate transformation of Appendix A, the ambient wind speed of Eq. (2) can be written in terms of the radial r and azimuthal ψ coordinates on the rotor plane, yielding
Here ψ is positive about x_{d} according to the righthand rule (i.e., clockwise looking downstream), and it is measured starting from the z_{d} unit vector (which flips from one side of the Π plane to other depending on whether the wind blows from the right or left looking upstream, as explained in Sect. 2.1; see also Fig. 1).
2.3 Force and velocity components at a blade section
With reference to Fig. 3, the tangential F_{t} and normal F_{n} components of the aerodynamic force at a blade section are
where $\mathit{\phi}={\mathrm{tan}}^{\mathrm{1}}{u}_{\mathrm{n}}/{u}_{\mathrm{t}}$ is the inflow angle, $u=\sqrt{{u}_{\mathrm{t}}^{\mathrm{2}}+{u}_{\mathrm{n}}^{\mathrm{2}}}$ is the total flow speed at the blade section, c is the sectional chord length, and finally C_{L} and C_{D} are the lift and drag coefficients, respectively. Using the coordinate transformations of Appendix A, the tangential u_{t} and normal u_{n} velocity components are written as
where Ω is the angular speed of the rotor and a is the axial induction factor, which expresses how much the rotororthogonal component of the freestream speed u_{∞} is slowed down at the rotor disk.
2.4 Induction model
It is well known that a nonuniform description of the induction is necessary in order to accurately capture the azimuthal variation of loads on a rotor operating in nonaxial conditions (Johnson, 1995). However, it appears that this is not necessary when computing integral rotor quantities such as power, torque, and thrust, as in the present case. To show this, the induction is modeled here with an expansion limited to oneperrevolution (1P) harmonics, i.e.,
where a_{0} is the constantovertherotor (0P) induction, while a_{0}κ_{1s} and a_{0}κ_{1c} are the 1P sine and cosine harmonic amplitudes, respectively.
Following the classical approach used for helicopter rotors in forward flight (Johnson, 1995), the sine term accounts for the tilting of the induction plane caused by the misalignment μ of the rotor with the incoming wind. As such, it is written in terms of ψ, which is measured starting from the z_{d} unit vector, and therefore it expresses a rotation of the induction plane about the axis normal to the wakedeflection intrinsic frame Π. The coefficient κ_{1s} can be modeled according to Coleman et al. (1945) and Pitt and Peters (1981), resulting in the expression
where the initial wake skew angle is $\mathit{\chi}=\mathit{\mu}+\mathrm{sin}\mathit{\mu}{C}_{\mathrm{T}}/\mathrm{2}$ (Jiménez et al., 2010), and ${C}_{\mathrm{T}}=\mathrm{2}T/\left(\mathit{\rho}A{u}_{\mathrm{\infty},\mathrm{hub}}^{\mathrm{2}}\right)$ is the thrust coefficient. Notice that the definition of the skew angle differs from the one given by Eq. (20) of Jiménez et al. (2010), because of the different definition of the thrust coefficient used in that publication.
The cosine term is introduced to account for the effects on the induction caused by vertical shear. As such, it is written as a function of the azimuthal angle ψ_{g}, which is measured from the (vertical) z_{g} unit vector, and therefore it expresses a rotation of the induction plane about the (horizontal) unit vector y_{g}. Using Eq. (A4b), it is readily found that ψ_{g}=ψ cos μ. Following Meyer Forsting et al. (2018), the cosine term is proportional to both the shear k and the thrust C_{T} coefficients, i.e.,
The cosine term significantly complicates the analytical derivations of power, torque, and thrust, which must now be expressed in terms of Bessel functions (Abramowitz et al., 1988) because of the term cos (ψcos μ). Before attempting the modeling of the proportionality coefficient ${\mathit{\kappa}}_{\mathrm{1}\mathrm{c}}^{*}$, this term was numerically optimized to best fit the numerical simulations and experimental measurements, as explained later in Sect. 3.
The inclusion of the sine and cosine induction terms has only an extremely modest effect on the quality of the results. In fact the match of C_{P} improves by 0.35 % when the sine term is included and by 0.60 % when both terms are used, as more precisely shown in Appendix B. Because of their modest effects, these terms are dropped from the following discussion, to simplify the resulting expressions, and they were not used in the results reported later in this article. However, these terms are retained in the software implementation of the model (Tamaro et al., 2024a) and can be switched on if desired by the user.
It should also be noted that the present model neglects the effects of the tangential induction, which in fact does not appear in the tangential velocity component expressed by Eq. (5a). This is justified by the fact that the rotor swirl is concentrated close to the hub, and it is small for a large extent of the blade span (Burton et al., 2011), where most of the thrust and power are generated.
More in general, there are several other effects that are present in a rotor and that are not modeled here, such as radial drag, tip and root losses, blade sweep, prebend and cone, and others. All these effects can be taken into account in detailed BEM models (Hansen, 2015; Burton et al., 2011) but would significantly complicate the present simplified analytical method. However, notwithstanding these limitations, the results of Sect. 4 show a remarkable ability of the proposed approach in predicting the trends of power and thrust as functions of various operating and inflow conditions. Additionally, the ability of the model in predicting actual power and thrust values (instead of trends) can be improved by the use of loss functions, as explained in Sect. 3.2.
2.5 Streamtube model
An expression for the axial induction can be derived using the concept of a streamtube (Hansen, 2015), as shown in Fig. 4 with reference to the present case of a misaligned rotor. Four stations are considered along the stream tube: inlet i, outlet o, section r^{−} located immediately in front of the rotor, and section r^{+} located immediately behind the rotor.
The principle of impulse and momentum applied to the streamtube is written as
where T is the thrust force; u_{o} is the longitudinal flow speed at the streamtube outlet; and $\dot{m}$ is the mass flux,
By using the thrust coefficient, Eq. (9) yields the nondimensional longitudinal flow speed at the streamtube outlet:
Next, Bernoulli's energy conservation theorem is applied between the streamtube inlet and the section immediately upstream of the rotor (stations i and r^{−} in Fig. 4) and between the section immediately downstream of the rotor and the streamtube outlet (stations r^{+} and o in the same figure):
where p is pressure and ${p}_{\mathrm{i}}={p}_{\mathrm{o}}={p}_{\mathrm{\infty}}$, while p_{∞} is the ambient value. Additionally, ${V}_{{\mathrm{r}}^{}}={V}_{{\mathrm{r}}^{+}}$ for continuity, and furthermore ${V}_{\mathrm{o}}^{\mathrm{2}}={u}_{\mathrm{o}}^{\mathrm{2}}+{v}_{\mathrm{o}}^{\mathrm{2}}$ at the outlet section, where v_{o} is the lateral (sidewash) speed component.
Following a customarily textbook assumption used for helicopters' rotors in forward flight (Johnson, 1995), only more recently adopted also for wind turbines by Shapiro et al. (2018), the misaligned rotor can be seen as a lifting wing of finite span (albeit of a small aspect ratio $\mathrm{AR}={D}^{\mathrm{2}}/A=\mathrm{4}/\mathit{\pi}$) operating at an angle of attack μ^{1}. The chord C(y_{d}) of the wing in the streamwise direction has an elliptic distribution: $\left(C\right({y}_{\mathrm{d}})/\mathrm{2}{)}^{\mathrm{2}}+{y}_{\mathrm{d}}^{\mathrm{2}}={R}^{\mathrm{2}}$. According to Prandtl's liftingline theory (Tietjens and Prandtl, 1957; Katz and Plotkin, 2001), the wing has consequently an elliptic lift distribution, which induces a spanwiseconstant downwash (in this case, sidewash) ${v}_{\mathrm{o}}=\mathrm{\Gamma}/\mathrm{4}R$. $\mathrm{\Gamma}=\stackrel{\mathrm{\u203e}}{L}/\mathit{\rho}{u}_{\mathrm{\infty},\mathrm{hub}}$ is the circulation at the wing midsection y_{d}=0, and $\stackrel{\mathrm{\u203e}}{L}=LC\left(\mathrm{0}\right)/A$ is the lift per unit span at that same location. Since the wing lift is the rotor side force, i.e., L=Tsin μ, it follows that the nondimensional sidewash at the streamtube outlet can be expressed as (Heck et al., 2023)
Combining the previous equations yields an expression for the 0P axial induction a_{0} as a function of the misalignment μ and thrust coefficient C_{T}:
2.6 Thrust force
Equation (14) furnishes an expression for the 0P axial induction as a function of the thrust coefficient. To close the problem, an expression for the thrust coefficient in terms of the operating conditions of the turbine is necessary. To this end, the thrust force T is expressed in terms of the normal sectional force F_{n} as
where B indicates the number of blades. Using Eq. (4b) under the assumption of a small inflow angle (i.e., sin φ≈φ and cos φ≈1) yields
The lift coefficient is written as ${C}_{\mathrm{L}}={C}_{\mathrm{L},\mathit{\alpha}}\mathit{\alpha}={C}_{\mathrm{L},\mathit{\alpha}}({\mathit{\alpha}}_{\mathrm{g}}{\mathit{\alpha}}_{\mathrm{0}})$, where C_{L,α} is the lift slope, and $\mathit{\alpha}=\mathit{\phi}\mathit{\theta}$ is the angle of attack measured with respect to the zerolift direction, whereas α_{g} is the angle of attack measured with respect to a generic direction. Without any loss of generality, the use of α is preferred in the following to avoid carrying along in the derivations the unnecessary extra term α_{0}. Furthermore, $\mathit{\theta}={\mathit{\theta}}_{\mathrm{p}}+\mathit{\beta}$ is the local pitch angle (see Fig. 3), where θ_{p} is the blade pitch rotation at the pitch bearing, and β indicates the blade twist referred to the zerolift direction.
Neglecting swirl induction, the inflow angle is $\mathrm{tan}\mathit{\phi}=(\mathrm{1}{a}_{\mathrm{0}})/(\mathit{\lambda}r/R)$, where $\mathit{\lambda}=\mathrm{\Omega}R/{u}_{\mathrm{\infty},\mathrm{hub}}$ is the tip speed ratio. For a turbine operating close to optimal induction (i.e., ${a}_{\mathrm{0}}=\mathrm{1}/\mathrm{3}$) and a typical tip speed ratio of 8.5, the inflow angle at threequarter span is less than 6^{∘}, justifying the smallangle assumption. This assumption clearly becomes less accurate for small inductions and tip speed ratios or close to the blade root, where only a modest contribution to the thrust is generated.
Using again the small inflow angle assumption, it follows that $\mathit{\phi}\approx {u}_{\mathrm{n}}/{u}_{\mathrm{t}}$ and u≈u_{t}, and the thrust T becomes
Using Eqs. (3) and (5), solving the double integral, and expressing T through the thrust coefficient ${C}_{\mathrm{T}}={C}_{{\mathrm{T}}_{\mathrm{1}}}+{C}_{{\mathrm{T}}_{\mathrm{2}}}$ finally gives
where $\mathit{\sigma}=BcR/A$ is the rotor solidity.
For null shear, i.e., k=0, this expression simplifies to
Notice that the terms in Eq. (18) depending on shear k also depend on the angles γ and δ, whereas Eq. (19) only depends on the total misalignment angle μ. This is because the wind shear is defined with respect to the ground frame, which is mapped into the nacelle frame by the γ and δ angles, whereas μ only depends on the relative orientation of the wind vector with the rotor axis, as explained in Sect. 2.1.
Furthermore, we note that – differently from the approach of Heck et al. (2023) – these expressions for thrust are applicable to any desired control policy, as they depend explicitly on the tip speed ratio λ and the pitch setting θ.
2.7 Power
The aerodynamic power P generated by a wind turbine is P=QΩ, where the aerodynamic torque Q is written as
Considering small angles, power can be written as
Using Eqs. (3) and (5), expressing the angular velocity as $\mathrm{\Omega}=\mathit{\lambda}{u}_{\mathrm{\infty},\mathrm{hub}}/R$, and solving the double integral yields the power coefficient C_{P}:
For null shear, i.e., k=0, the power coefficient simplifies to
Due to the explicit dependency of C_{P} on cos μ and of 1−a_{0} on sin ^{2}μ (see Eq. 14), it follows that – in an unsheared inflow – the aerodynamic power does not depend on the misalignment direction.
2.8 Dependency on misalignment direction
The power model reveals that vertical shear is the culprit for the observed lack of symmetry with respect to yaw misalignment. In fact, the term responsible for the asymmetry in Eq. (22) is
This term shows that yawing a rotor out of the wind in a sheared inflow will produce a nonsymmetric behavior with respect to positive and negative yaw angles γ, i.e., $P(\mathit{\gamma})\ne P(+\mathit{\gamma})$. In fact, the following can be noted.

For small θ, the two terms within the parenthesis of Eq. (24) are small. Consequently, the asymmetry is small. Furthermore, either positive or negative yaw angles could produce more power, depending on the balance of these two terms.

On the other hand, for larger θ the thrust coefficient C_{T} decreases, in turn increasing the term (1−a_{0}). As a consequence, the first of the two terms of Eq. (24) prevails. The origin of this prevailing term can be traced to the contribution to power of the fraction of lift that depends on the pitch angle, which is proportional to C_{L,α}θu_{t}u_{n} (see Eq. 21). When the rotor is misaligned, u_{t} exhibits a cos ψ variation (see Eq. 5a) that is in phase with or in opposition to a similar cos ψ variation of u_{n} caused by shear (see Eqs. 5b and 3). The integral of the product of these two terms over one revolution is different from zero, and it has the sign of the misalignment, resulting in some extra power for positive yaw angles and some losses for negative ones, when the thrust is low.
This complex balance of effects is probably the cause for the lack of agreement in the literature on which misalignment direction yields more. As shown by the model, there is no simple answer, and the behavior depends on the rotor design and on how it is operated.
Very similar conclusions apply also to the thrust coefficient. According to Eq. (18), ${C}_{{\mathrm{T}}_{\mathrm{1}}}(\mathit{\gamma})>{C}_{{\mathrm{T}}_{\mathrm{1}}}(+\mathit{\gamma})$, because ${C}_{{\mathrm{T}}_{\mathrm{1}}}$ depends on −ksin γ, whereas ${C}_{{\mathrm{T}}_{\mathrm{2}}}(\mathit{\gamma})<{C}_{{\mathrm{T}}_{\mathrm{2}}}(+\mathit{\gamma})$ (when θ>0), because ${C}_{{\mathrm{T}}_{\mathrm{2}}}$ depends on +θksin γ. Therefore, one can expect a slightly higher thrust for negative yaw angles at low pitch settings and vice versa at the higher pitch values (with the effect being more pronounced at larger tip speed ratios).
To illustrate these findings in an example case, Fig. 5 shows the thrust and power coefficients as functions of the misalignment angle γ, for different shear coefficients k and blade pitch angles θ_{p}. All coefficients have been normalized by their respective value in aligned conditions.
The lack of symmetry of the rotor with respect to misalignment direction is in general rather small. In a typical field implementation of wake steering, various uncertainties – due to limits in the knowledge of the ambient conditions, actual yaw orientation of the rotor, asymmetric behavior of the onboard anemometry, etc. – and other model errors probably dominate the problem, making the asymmetric behavior of misalignment a negligible effect, especially for small pitch values and moderate tip speed ratios.
3.1 Integration with a wake model
The analytical model derived in the previous pages can be readily implemented in engineering flow models (NREL, 2023b; Pedersen et al., 2019). An opensource implementation in FLORIS (NREL, 2023b) is available on GitHub (Tamaro et al., 2024a). As a summary, we report here for convenience the governing Eqs. (14) and (18):
This represents a closed system of equations that can be solved for the axial induction a_{0} and thrust coefficient C_{T}, given the tip speed ratio λ, the pitch setting ${\mathit{\theta}}_{\mathrm{p}}=\mathit{\theta}\mathit{\beta}$, and the yaw misalignment γ. Crucially, the presence of λ and θ_{p} enable using any desired control policy when misaligning the turbine. Having obtained a_{0} and C_{T}, the power coefficient ${C}_{\mathrm{P}}(\mathit{\gamma},{\mathit{\theta}}_{\mathrm{p}},\mathit{\lambda})$ is obtained by Eq. (22), and, finally, the rotor power is computed as
3.2 Improved accuracy by the use of loss functions
The analytical derivation of the equations implies that the model lacks many of the features that are present in more sophisticated BEM implementations, such as radially and azimuthally nonuniform induction, swirl induction, tip and root losses, radial flow, spanwise varying geometric characteristics, prebend, and others. Clearly, this lack of accuracy could be resolved by numerically implementing the same model in a BEM code (Hansen, 2015). However, this way the use of the misaligned rotor model in combination with an engineering wake model would become much more complex and numerically expensive.
To address this problem, the following implementation is recommended, which was used in the validation and the examples reported in the following section. A power loss function η_{P} is computed by using Eq. (22) to yield
Next, a refined power coefficient ${C}_{\mathrm{P}}^{\mathrm{rf}}$ is obtained as
where ${C}_{\mathrm{P}}^{\mathrm{hf}}$ is the power coefficient computed in aligned conditions through a higherfidelity model, for example, based on a sophisticated BEM implementation or even on experimental measurements, when available. In other words, the analytical model is used not to predict the actual power coefficient but only the fraction of it that is lost by misalignment. The actual power coefficient is obtained by applying the loss model to a more accurate power coefficient model in aligned conditions.
The same approach is adopted for thrust. First, a thrust coefficient loss factor is computed by using the proposed model as ${\mathit{\eta}}_{\mathrm{T}}(\mathit{\lambda},{\mathit{\theta}}_{\mathrm{p}},\mathit{\gamma})={C}_{\mathrm{T}}(\mathit{\lambda},{\mathit{\theta}}_{\mathrm{p}},\mathit{\gamma})/{C}_{\mathrm{T}}(\mathit{\lambda},{\mathit{\theta}}_{\mathrm{p}},\mathrm{0})$. Next, a refined estimate in misaligned conditions is obtained by computing ${C}_{\mathrm{T}}^{\mathrm{rf}}$ from a higherfidelity aligned value ${C}_{\mathrm{T}}^{\mathrm{hf}}$, i.e.,
In the following, we always adopt this approach for the power and thrust coefficients. However, to simplify the notation, we drop the superscript (⋅)^{rf}. Hence, for example, when we write ${C}_{\mathrm{P}}(\mathit{\lambda},{\mathit{\theta}}_{\mathrm{p}},\mathit{\gamma})$, we in reality imply that Eq. (28) is used; the same holds for C_{T}.
3.3 Implementation of arbitrary control strategies
When a wind turbine yaws out of the wind, the inflow seen by the rotor changes with respect to the aligned condition. The controller reacts to the changed inflow, modifying the set point, which in turn affects the power captured by the rotor and its loading. Hence, the problem is implicit, in the sense that the misalignment model has to be solved together with the controller. This general implicit approach should be contrasted with the explicit one proposed in Heck et al. (2023), which assumes a rotor performance parameter, ${C}_{\mathrm{T}}^{\prime}=\mathrm{2}T/\left(\mathit{\rho}A{u}_{\mathrm{n}}^{\mathrm{2}}\right)$, remains constant even in misaligned conditions. This section explains how arbitrary control laws can be integrated with the present more general model.
The control of a modern variablespeed wind turbine is typically based on the definition of two or three main operational regions.
In region II (also called the belowrated or partialload regime), the turbine should maximize its power output. This is achieved by operating at the maximum power coefficient ${C}_{\mathrm{P}}^{*}({\mathit{\lambda}}^{*},{\mathit{\theta}}_{\mathrm{p}}^{*})$, which corresponds to the optimal tip speed ratio λ^{*} and pitch setting ${\mathit{\theta}}_{\mathrm{p}}^{*}$. As the tip speed ratio must remain constant at its value λ^{*} throughout this control region, the rotor speed increases linearly with wind speed, i.e., $\mathrm{\Omega}={\mathit{\lambda}}^{*}{u}_{\mathrm{\infty}}/R$. The aerodynamic torque Q_{a} is readily computed as Q_{a}=K(ρ)Ω^{2}, with $K\left(\mathit{\rho}\right)=\frac{\mathrm{1}}{\mathrm{2}}\mathit{\rho}{\mathrm{AR}}^{\mathrm{3}}{C}_{\mathrm{P}}^{*}/{\mathit{\lambda}}^{{*}^{\mathrm{3}}}$. Once the aerodynamic torque is known, the torque provided by the generator Q_{g} is obtained from the expression Q_{a}=η_{m}η_{e}Q_{g}, where η_{m} and η_{e} are the mechanical and electrical efficiencies, respectively.
When the ambient wind speed is above the rated value ${u}_{{\mathrm{\infty}}_{\mathrm{r}}}={\mathrm{\Omega}}_{\mathrm{r}}R/{\mathit{\lambda}}^{*}$, there is enough power carried by the wind for the turbine to produce its maximum (rated) output P_{r}. This is called region III (also termed the aboverated or fullload regime), and the turbine operates at the constant (rated) rotor speed Ω_{r}. Hence, the aerodynamic torque is constant, i.e., ${Q}_{\mathrm{a}}={P}_{\mathrm{r}}/{\mathrm{\Omega}}_{\mathrm{r}}$, whereas blades are progressively pitched into the wind to reduce C_{P} as wind speed increases.
To implement these standard region II and III control strategies in the proposed model, the following power equation is introduced:
The equation has three unknowns: the rotor speed Ω, the blade pitch θ_{p}, and the aerodynamic torque Q_{a}. Given ambient conditions u_{∞} and ρ, two additional conditions are necessary before the three unknowns can be computed. To this end, one can first assume that the machine operates in region II. Hence, Eq. (30) is solved by appending to it the following two constraints:
If the computed rotor speed exceeds the rated value, i.e., Ω>Ω_{r}, then it means that – for the given ambient conditions and misalignment angle – the turbine operates in region III and not region II. Hence, the solution is discarded, and Eq. (30) is solved again by appending this time the following two constraints:
This same approach can be used for curtailment and derating strategies (Juangarcia et al., 2018).
Figure 6 shows the application of this approach to the IEA 3.4 MW reference wind turbine, a typical onshore machine with contemporary design characteristics (Bortolotti et al., 2019). In this example the turbine is exposed to an inflow characterized by a wind speed u_{∞}=10.5 m s^{−1}, an air density ρ=1.22 kg m^{−3}, and a linear vertical shear coefficient k=0.2. For these ambient conditions, the turbine operates in region III when it is aligned with the wind (γ=0°). As the turbine starts yawing out of the wind, it initially keeps operating in region III. Accordingly, the tip speed ratio λ (Fig. 6d) and the power coefficient C_{P} (Fig. 6b) remain constant, while the thrust coefficient increases (Fig. 6a) and the blades pitch back (Fig. 6c). However, at around $\left\mathit{\gamma}\right\approx \mathrm{15}$°, the turbine enters into region II, because the rotororthogonal component of the wind speed is no longer large enough to maintain the rated power output. As the misalignment keeps increasing, the pitch angle remains fixed at its optimal value (Fig. 6c), whereas the tip speed ratio drops on account of the slowing rotor speed (Fig. 6d), in accordance with the region II policy.
Often turbines present an additional intermediate operating regime, called region II$\mathrm{1}/\mathrm{2}$, which occupies a wind speed interval across the rated value ${u}_{\mathrm{\infty}}^{\mathrm{lb}}\le {u}_{{\mathrm{\infty}}_{\mathrm{r}}}\le {u}_{\mathrm{\infty}}^{\mathrm{ub}}$. In such cases, the turbine operates in region II when ${u}_{\mathrm{\infty}}<{u}_{\mathrm{\infty}}^{\mathrm{lb}}$, in region III when ${u}_{\mathrm{\infty}}>{u}_{\mathrm{\infty}}^{\mathrm{ub}}$, and in region II$\mathrm{1}/\mathrm{2}$ when ${u}_{\mathrm{\infty}}^{\mathrm{lb}}\le {u}_{\mathrm{\infty}}\le {u}_{\mathrm{\infty}}^{\mathrm{ub}}$.
Differently from regions II and III, where controllers only require knowledge of the rotor speed, the control policy in region II$\mathrm{1}/\mathrm{2}$ typically requires prescribing the desired pitch and torque settings as functions of wind speed. In other words, one has to provide the schedules θ_{p}=θ_{p}(u_{∞}) and Q_{a}=Q_{a}(u_{∞}) in the desired range ${u}_{\mathrm{\infty}}^{\mathrm{lb}}\le {u}_{\mathrm{\infty}}\le {u}_{\mathrm{\infty}}^{\mathrm{ub}}$. Two common examples of region II$\mathrm{1}/\mathrm{2}$ control policies are provided by load and noise alleviation techniques.
Load alleviation is often necessary because thrust reaches a sharp maximum at the rated wind speed, ${T}_{\mathrm{r}}=\frac{\mathrm{1}}{\mathrm{2}}\mathit{\rho}A{u}_{{\mathrm{\infty}}_{\mathrm{r}}}^{\mathrm{2}}{C}_{\mathrm{T}}({\mathit{\theta}}_{\mathrm{p}}^{*},{\mathit{\lambda}}^{*})$. To reduce the effects of this large load on the sizing of various turbine components, thrust clipping (or peak shaving) is used, where blades are pitched to feather according to a desired schedule θ_{p}=θ_{p}(u_{∞}). This has the effect of reducing the angle of attack and hence the thrust (Zalkind et al., 2022), at the cost of some reduced power. To minimize power losses for a given pitch schedule, the optimal power coefficient schedule can be computed as ${C}_{\mathrm{P}}\left({u}_{\mathrm{\infty}}\right)={\mathrm{max}}_{\mathrm{\Omega}}{C}_{\mathrm{P}}(\mathrm{\Omega}R/{u}_{\mathrm{\infty}},{\mathit{\theta}}_{\mathrm{p}}({u}_{\mathrm{\infty}}\left)\right)$ under the constraint Ω≤Ω_{r}, which also returns the rotor speed schedule Ω(u_{∞}). Consequently, the torque schedule becomes ${Q}_{\mathrm{a}}\left({u}_{\mathrm{\infty}}\right)=\frac{\mathrm{1}}{\mathrm{2}}\mathit{\rho}A{u}_{\mathrm{\infty}}^{\mathrm{3}}{C}_{\mathrm{P}}\left({u}_{\mathrm{\infty}}\right)/\mathrm{\Omega}\left({u}_{\mathrm{\infty}}\right)$.
One effective way of constraining noise emissions is to limit the rotor speed (Leloudas et al., 2007; Bottasso et al., 2012) to a maximum noiseacceptable value Ω_{n}. When Ω_{n}<Ω_{r}, the increase in rotor speed as a function of wind speed that characterizes region II is stopped before the machine reaches the rated power. In this case, region II$\mathrm{1}/\mathrm{2}$ is entered when ${u}_{\mathrm{\infty}}>{u}_{\mathrm{\infty}}^{\mathrm{lb}}={\mathrm{\Omega}}_{\mathrm{n}}R/{\mathit{\lambda}}^{*}$, and, above this wind speed, the rotor operates at the constant speed Ω_{n}. To minimize power losses, the blade pitch setting can be computed for each wind speed u_{∞} as ${\mathit{\theta}}_{\mathrm{p}}\left({u}_{\mathrm{\infty}}\right)={\mathrm{argmax}}_{{\mathit{\theta}}_{\mathrm{p}}}{C}_{\mathrm{P}}\left(\mathit{\lambda}\right({u}_{\mathrm{\infty}}),{\mathit{\theta}}_{\mathrm{p}})$, where $\mathit{\lambda}\left({u}_{\mathrm{\infty}}\right)={\mathrm{\Omega}}_{\mathrm{n}}R/{u}_{\mathrm{\infty}}$. The corresponding aerodynamic torque schedule is readily obtained as ${Q}_{\mathrm{a}}\left({u}_{\mathrm{\infty}}\right)=\frac{\mathrm{1}}{\mathrm{2}}\mathit{\rho}\mathrm{AR}{C}_{\mathrm{P}}\left(\mathit{\lambda}\right({u}_{\mathrm{\infty}}),{\mathit{\theta}}_{\mathrm{p}}({u}_{\mathrm{\infty}}\left)\right)/\mathit{\lambda}\left({u}_{\mathrm{\infty}}\right)$. The end of region II$\mathrm{1}/\mathrm{2}$ is reached when, for sufficiently high u_{∞}, the turbine reaches the rated power, finally entering into region III.
In order to implement a given control policy for region II$\mathrm{1}/\mathrm{2}$ with the proposed model, we assume that the controller will implement the desired schedules θ_{p}(u_{∞}) and Q_{a}(u_{∞}) (whether computed as explained above or according to different criteria) by reacting to the rotororthogonal wind speed component u_{∞}cos γ. In the absence of specific details on the implementation, this is a reasonable assumption, as region II$\mathrm{1}/\mathrm{2}$ controllers are typically based on rotoreffective wind speed estimates (Bottasso et al., 2012; Zalkind et al., 2022) that, in a misaligned condition, will sense u_{∞}cos γ and not u_{∞}. Therefore, when ${u}_{\mathrm{\infty}}^{\mathrm{lb}}\le {u}_{\mathrm{\infty}}\mathrm{cos}\mathit{\gamma}\le {u}_{\mathrm{\infty}}^{\mathrm{ub}}$, Eq. (30) is solved by appending to it the following two constraints:
In summary, using Eq. (30) in combination with the constraint Eqs. (31), (32), or (33) yields the set point achieved by the turbine for given ambient conditions and a given misalignment, no matter what region it corresponds to and what control strategy is implemented by the controller.
3.4 Model calibration
Through Eq. (25b) and (26), the analytical model depends on C_{D}, C_{L,α}, and β. These are average parameters, which represent in the model the equivalent effect caused by corresponding quantities that in reality exhibit a spanwise variability.
When numerical or experimental measurements are available, the parameters C_{D}, C_{L,α}, and β can be calibrated to minimize the error produced by the model in the prediction of the power loss factor η_{P} and of the thrust coefficient C_{T}. Notice that C_{T} is preferred to η_{T} for this scope, because it was found that the informational content of η_{T} is very similar to the one of η_{P}, reducing the quality of the tuning. Calibration is here performed by numerically solving the following minimization problem:
where $(\cdot {)}_{i}^{\mathrm{obs}}$ represents N numerical or experimental observations, and $(\cdot {)}_{i}^{\mathrm{mod}}$ represents the corresponding model predictions. For each data set, tuning was performed solving $N/\mathrm{2}$ times the problem expressed by Eq. (34) using a gradientbased optimization, each time with a different random 50 % subset of the available data, and finally averaging the resulting parameters.
3.5 Simplified choice of model parameters in the absence of calibration data
When calibration is not possible, the equivalent model parameters C_{D}, C_{L,α}, and β must be estimated from the corresponding actual spanwise distributions ${C}_{\mathrm{D}}(r/R)$, ${C}_{\mathrm{L},\mathit{\alpha}}(r/R)$, and $\mathit{\beta}(r/R)$.
Examining the expression for thrust given by Eq. (16), neglecting drag, it appears that the lift force has roughly a spanwise triangular distribution. In fact, inspecting Eq. (5), u_{t} is proportional to r, whereas u_{n} does not depend on r. Additionally, the leading term of the Taylor series of the optimal twist distribution is $\mathrm{1}/r$ (Burton et al., 2011). Similarly, inspecting the expression for power given by Eq. (21), again neglecting the contribution of drag, it appears that – for the same reasons – also the spanwise power capture has a triangular distribution. This suggests the need to evaluate the spanwise integrals at the centroid of the triangle, which is located at $r/R=\mathrm{2}/\mathrm{3}$. Adopting this approach, the model parameters are then set to the following values:
Coefficient f_{d} is a correction factor for drag, while f_{l} is a knockdown factor for lift, which accounts for the finite span of the blades. Based on comparisons with calibrated values (see Sect. 4.2), we recommend a drag correction factor f_{d}=1 for moderate yaw (up to 20°) and pitch values and a smaller value of 0.45 if the model has to be used also for large yaw and pitch settings. This smaller value is probably due to the approximation of a small inflow angle used in the model, which is partially corrected by a smaller drag coefficient. For lift, we recommend the value ${f}_{\mathrm{l}}=\mathrm{2}/\mathrm{3}$.
The performance of this simplified choice of model parameters is demonstrated later in Sect. 4.2.
4.1 Validation with respect to LESALM simulations
LESALM simulations are used for testing the accuracy of the model in representing misaligned conditions, similarly to what has been done by other authors (Gebraad et al., 2016; Liew et al., 2020; Nanos et al., 2022). The effects of the rotor on the flow are modeled with the filtered ALM of Troldborg et al. (2007) and MartínezTossas and Meneveau (2019) by projecting forces computed along the lifting lines onto the LES grid. The Cartesian mesh consists of approximately 3.5 million cells and uses four refinement levels. The smallest cells measure 1 m and are used to discretize the area surrounding the rotor.
Simulations were conducted for the IEA 3.4 MW reference wind turbine, whose complete technical specifications are reported in Bortolotti et al. (2019). Here we only note that the turbine has a 5° uptilt angle; i.e., $\mathit{\delta}=\mathrm{5}$°. The parameters of the proposed model were calibrated as explained in Sect. 3 based on the LESALM simulations described in Sect. 4.1.2 but not the ones of Sect. 4.1.1, obtaining the values ${C}_{\mathrm{D}}=\mathrm{0.0052}\pm \mathrm{0.0001}$, ${C}_{\mathrm{L},\mathit{\alpha}}=\mathrm{4.759}\pm \mathrm{0.007}$ rad^{−1}, and $\mathit{\beta}=\mathrm{3.345}\pm \mathrm{0.007}$°, for a 95 % confidence level.
4.1.1 Simulations in control regions II and III
First we demonstrate the integration with a standard controller, including operations in region II, III, and the transition between the two as the turbine is progressively yawed out of the wind. For the LESALM simulations, set points were computed using a controller in the loop, based on an implementation similar to the one of Bortolotti et al. (2019). For the proposed model, set points were obtained from Eq. (30) in combination with the constraint Eqs. (31) and (32).
We consider laminar inflows with four wind speeds: one belowrated speed of 8.5 m s^{−1} and three aboverated speeds of 10.5, 11, and 13 m s^{−1}. Figure 7 reports the results in term of the thrust loss factor η_{T} (Fig 7a), power loss factor η_{P} (Fig. 7b), blade pitch θ_{p} (Fig. 7c), and tip speed ratio λ (Fig. 7d), all plotted as functions of the misalignment angle γ. The solid markers are the results of the LESALM simulations, while the lines represent predictions of the proposed model.
For the lowest wind speed of 8.5 m s^{−1}, the turbine always operates in region II. This is the only case where the method of Heck et al. (2023) is strictly applicable. In fact, their method does not contain a generic thrust model, but rather it is formulated in terms of the modified thrust coefficient ${C}_{\mathrm{T}}^{\prime}$, which is constant when a turbine yaws out of the wind in region II. The results of this alternative model are shown with a dashed orange line in the figure. The benchmark LESs feature a negative uptilt $\mathit{\delta}=\mathrm{5}$°, which is not modeled in the approach of Heck et al. (2023). To avoid cluttering the results with this additional effect, here and in the following examples the total true misalignment angle μ (instead of γ) is provided as input to the model of Heck et al. (2023). The figure shows that both methods are in excellent agreement with the CFD results.
For the highest wind speed of 13 m s^{−1}, the turbine operates in region III for all misalignment angles. On the other hand, for a wind speed of 11 m s^{−1} the machine enters region II around γ=27°, and for a wind speed of 10.5 m s^{−1} the machine enters region II at about γ=16°.
In general, there is a very good agreement of the model with the higherfidelity CFD results, not only in terms of loss factors, but also on the calculation of the set points.
4.1.2 Simulations with fixed tip speed ratio and pitch setting
Next, we present a second set of results obtained by varying the misalignment while keeping the tip speed ratio and pitch constant, for different inflows. These conditions are meant to provide a more general view of the performance of the proposed method in a variety of conditions, although – as explained in Sect. 3.3 – tip speed ratio and pitch in general do not both remain constant when a turbine yaws out of the wind.
The four operational scenarios of Table 1 are considered. The flow is laminar and steady in all scenarios. Cases 1 and 2 have no shear and different tip speed ratios, whereas cases 3 and 4 are sheared and have the same λ.
Figures 8 and 9 report the power loss factor η_{P} in the range of yaw misalignment angles $\mathrm{30}\mathit{\xb0}<\mathit{\gamma}<\mathrm{30}\mathit{\xb0}$ for different pitch settings, each corresponding to a different thrust coefficient C_{T,0} in aligned conditions. Figure 8 corresponds to scenarios 1 and 2 of Table 1, i.e., no shear, while Fig. 9 reports the solution for the sheared cases 3 and 4 of that same table.
In the figures, LESALM results (shown with circular black markers) are compared with the proposed approach (shown with solid blue lines) and the method proposed by Heck et al. (2023) (shown with dashed orange lines). For the latter, the modified thrust coefficient ${C}_{\mathrm{T}}^{\prime}$ was obtained directly from each LES at the corresponding γ value.
Overall, there is a very good match between the predictions of the proposed model and numerical simulations. For the null shear cases of Fig. 8, results are reported only for positive yaw angles, as power is symmetric. On the other hand, power is not symmetric for the sheared inflow cases of Fig. 9, which shows clear evidence of the complex behavior described in Sect. 2.8. At high C_{T} (low pitch), curves are very nearly symmetric with respect to γ. However, as thrust is decreased (and pitch increased), power capture is larger at positive γ values compared to negative γ values.
The model of Heck et al. (2023) performs similarly well at high and moderate rotor loading, when ${C}_{\mathrm{T}}^{\prime}$ is roughly constant. However, as the C_{T} is reduced, the model tends towards the solution cos ^{3}γ, and therefore its accuracy is compromised. Moreover, the model fails to predict the shearinduced asymmetry (see Fig. 9).
As predicted by the proposed model, the power asymmetry increases with shear (see Table 1 and the explanation given in Sect. 2.8). To facilitate the visualization of this effect, Fig. 10 shows the difference $\mathrm{\Delta}{\mathit{\eta}}_{\mathrm{P},\mathit{\gamma}=\pm \mathrm{30}\mathit{\xb0}}$ between the two values of η_{P} at $\mathit{\gamma}=\pm \mathrm{30}$° as a function of shear for varying thrust coefficients. The asymmetry also exhibits a noticeable dependency on the thrust coefficient, with larger asymmetries being observed for lower values of C_{T}, as predicted in Sect. 2.8 by examining Eq. (22).
Figures 11 and 12 report the thrust loss factor η_{T} as a function of yaw misalignment for the same four scenarios and different thrust settings. Here again model predictions are indicated with lines and LESALM results with markers. There is a consistently good match for all scenarios and for all yaw and pitch values. The lack of symmetry is again consistent with the model, similarly to the case of power discussed above. Figure 12a–d show a higher thrust for positive yaw angles at low thrust coefficients (high pitch values) because of the high tip speed ratio of scenario 3, indicating that term ${C}_{{\mathrm{T}}_{\mathrm{2}}}$ prevails over ${C}_{{\mathrm{T}}_{\mathrm{1}}}$. The opposite happens in Fig. 12e–h due to the lower λ of scenario 4.
Overall, it appears that the performance of the rotor is strongly dependent on the thrust coefficient and tip speed ratio and hence on the way it is controlled when it yaws out of the wind. Therefore, the standard power law ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ may oversimplify the complex aerodynamics that are typical of this problem. On the other hand, notwithstanding its simplicity, the proposed model is in very good agreement with sophisticated CFD simulations, and it is capable of describing even relatively minor effects of the complex behavior of a misaligned wind turbine rotor in a sheared inflow.
4.2 Validation of the simplified choice of model parameters
The simplified choice of model parameters described in Sect. 3.5 is based on Eq. (35), which includes the correction factors f_{d} and f_{l}. To verify the existence of typical values for these factors, we considered four different wind turbines: IEA 3.4 MW (Bortolotti et al., 2019); NREL 5 MW (Jonkman et al., 2009); G178, which is a modified version of the DTU 10 MW (Bak et al., 2013; Wang et al., 2021); and the smallscale G1 turbine (Bottasso and Campagnolo, 2022; Campagnolo et al., 2020). For the three fullscale machines, the model parameters were first calibrated using the LESALM simulation results of Sect. 4.1.2, whereas for the G1 model the calibration was performed using wind tunnel measurements (see later Sect. 4.3).
The parameters calibrated this way were then compared to the ones based on the simplified approach of Eq. (35), leading to the recommended values reported in Sect. 3.5. The ${C}_{\mathrm{D}}(r/R)$ and ${C}_{\mathrm{L},\mathit{\alpha}}(r/R)$ coefficients were obtained by averaging over the interval of angles of attack 2° below the negative and positive stall limits. In all cases, the calibrated value of the twist corresponded remarkably well with the actual twist at $\mathrm{2}/\mathrm{3}$ span, i.e., $\mathit{\beta}(\mathrm{2}/\mathrm{3})$.
The simplified choice of the model parameters was then applied to the NREL 5 MW and G178 10 MW wind turbines. Simulations were performed with a steady inflow, at different misalignments, and for two different blade pitch settings. The proposed model was calculated with the parameters based on Eq. (35), using the default correction coefficients f_{d}=0.45 and ${f}_{\mathrm{l}}=\mathrm{2}/\mathrm{3}$ (in other words, without using LESALM calibrated values, replicating what one could do in the absence of suitable tuning data).
The results in terms of η_{T} and η_{P} for the two turbines are reported in Fig. 13 and compared with LESALM simulations. For both turbines there is an excellent match between model predictions and CFD results. This seems to indicate that the even a simplified choice of the model parameters is sufficient for a good performance of the model.
4.3 Validation with respect to wind tunnel measurements
Next, the model is compared to data recorded during wind tunnel experimental campaigns performed with a G1 wind turbine (Campagnolo et al., 2016). This scaled machine has a diameter of 1.1 m, a rated rotor speed of 850 rpm, and null tilt. The design of the G1 is described in Bottasso and Campagnolo (2022), and its rotor aerodynamic and wake characteristics have been reported in Wang et al. (2021) and references therein.
Tests were performed in a boundary layer wind tunnel (Bottasso et al., 2014) with three different inflows: the first one, termed LowTI, has no shear and a very low turbulence intensity (approximatively equal to 1 %); the other two, termed ModTI and HighTI, have TIs of about 6 % and 13 % at hub height and vertical linear shears in the rotor region equal to k=0.11 and k=0.15, respectively. Figure 14a reports the vertical profiles of the longitudinal wind speed component u measured by means of CTA probes (Bottasso et al., 2014), normalized by the wind speed u_{pitot} measured by a pitot tube placed at hub height. Figure 14b shows the vertical profiles of the turbulence intensity, as measured with the same instrumentation.
The experimental characterization of power losses in misaligned conditions was performed based on the three campaigns of Table 2, totaling 119 observations of a duration of 2 min each. The measured average θ_{p} and λ are reported in Appendix C for campaigns 1 and 2 and in Fig. 19 for campaign 3.
The first campaign was conducted in region II (i.e., the belowrated partialload regime), using the classical variablerotorspeed maximumpowertracking strategy. Tests were conducted in all three inflow conditions – LowTI, ModTI, and HighTI – with hubheight wind speeds of 5.86, 5.69, and 5.40 m s^{−1}, respectively.
The second campaign was also conducted in region II, but in this case the turbine was derated in the range ${P}_{\mathrm{d}}\in [\mathrm{50},\mathrm{100}]$ % while adopting two different strategies: isoλ, where the tip speed ratio is held constant (Campagnolo et al., 2023), and minC_{T}, where the thrust coefficient is minimized (Juangarcia et al., 2018). Tests were conducted only in the ModTI inflow, with a hubheight wind speed of 5.62 m s^{−1},
The third campaign was conducted in region III (i.e., the aboverated fullload regime) in the LowTI and ModTI inflows, with hubheight wind speeds of 6.97 and 6.11 m s^{−1}, respectively.
Various sources of error affect the experimental observations. These include measurements of the wind speed u_{pitot} upstream of the model (obtained by a pitot tube placed at hub height 3 D in front of the turbine), of the air density ρ, of the rotor speed Ω, of the shaft torque Q, of the bending moment at tower base (which is used to estimate thrust), of the blade pitch angle, and of the nacelle orientation with respect to the wind tunnel (i.e., of the misalignment angle). The error in u_{pitot} is related to the uncertainty associated with the measurements of flow density and dynamic pressure. This latter quantity is measured with a MKS Baratrontype 226A transducer (MKS Instruments Inc., 2022) with full span equal to 1 Torr, characterized by an accuracy of ±0.4 Pa. Density is instead derived from measurements of air pressure, temperature, and humidity, and it is affected by an error equal to ±0.01 kg m^{3} (Wang et al., 2020). Torque is measured with a load cell installed on the rotor shaft, and it is affected by an uncertainty of ±0.005 Nm. The rotor speed measurement, provided by an optical incremental encoder, is instead affected by an error equal to ±1.5 rpm. The measurement uncertainty on power P=QΩ is derived by adding in quadrature the uncertainties on Q and Ω. Thrust T is obtained by correcting the measurements of the bending moments at tower base by the effects induced by the drag of the tower, nacelle, and hub spinner (Wang et al., 2020). The calibration of the load cell at tower base revealed an uncertainty in the thrust of ±0.14 N. Blade pitch and nacelle orientation are measured by optical encoders, affected by uncertainties of ±0.2°. In turn, all these effects are used to quantify uncertainties in the tip speed ratio λ, as well as yawinduced power and thrust losses η_{P} and η_{T}, again by adding errors in quadrature. In the following, the resulting uncertainties are reported for a 95 % confidence level.
Uncertainties in some experimental measurements affect also the predictions of the proposed model. The uncertainties of the four model inputs – tip speed ratio, blade pitch, rotor speed, and yaw misalignment – were propagated forward throughout the model by Latin hypercube sampling with 10 000 sample points, using the UQLab software (Marelli and Sudret, 2014).
Tuning of the model parameters C_{D}, C_{L,α}, and β was performed with Eq. (34), using a total of 94 observations from experimental campaigns 1 and 2.
Given the small size of the G1 wind turbine, the Reynolds number at its blade sections is particularly low. Although special lowReynolds airfoils are used in the design of the G1 blades (Bottasso and Campagnolo, 2022), their aerodynamic characteristics are particularly sensitive to the operating conditions of the turbine (Wang et al., 2020). In fact, the Reynolds number has a significant effect on the drag and on the zerolift direction, which in turn affects the parameter β, whereas the effect on the lift slope C_{L,α} is negligible (Wang et al., 2020). Accordingly, the model parameters C_{D} and β are assumed to depend on the rotational speed Ω, since the relative speed at the airfoils is close to the tangential speed (u≈u_{t}; see Fig. 3). The values of C_{D} and β at $\mathrm{\Omega}=[\mathrm{850},\mathrm{625},\mathrm{400}]$ rpm were assumed as unknowns, and a piecewise linear interpolation was used at other intermediate values of the rotor speed.
Figure 15 reports the tuned C_{D} and β parameters, with whiskers representing the corresponding 95 % confidence intervals. As expected, drag decreases for increasing rotor speed, i.e., for increasing sectional Reynolds number. The twist β also exhibits the same trend, since the zerolift direction rotates nose up as the Reynolds number increases (Wang et al., 2020). The tuned parameter C_{L,α} is equal to 4.5033±0.0459.
For maximumpower tracking operation in region II (test campaign 1), Fig. 16 reports a comparison between modelpredicted and measured power and thrust losses. The present model results are indicated with solid blue lines, while measurements are indicated by black circles. Whiskers indicate the respective 95 % confidence intervals. There is a very good match between experimental measurements and the present model, with the latter falling within the uncertainty range of the measurements in most cases. As predicted by the model, the sheared inflow conditions ModTI and HighTI exhibit the expected nonsymmetric behavior with respect to positive and negative yaw angles.
In principle, the model of Heck et al. (2023) would be applicable to these tests in region II. However, their method does not directly consider the aerodynamic characteristics of the blades, as it expresses their behavior through the single parameter represented by ${C}_{\mathrm{T}}^{\prime}$. Therefore, it is blind to the variability of twist and drag with respect to the Reynolds number, which drops significantly as the misalignment angle increases. Since this strong Reynolds dependency is specific to the small scale of wind tunnel models, the results of the method of Heck et al. (2023) are not shown here, because its poor match with the measurements would be misleading, as these effects would not be present at full scale. The present method is not affected by this issue, because it uses a liftingline approach and specifically includes the blade aerodynamic characteristics in the governing Eqs. (22) and (25b).
The loss factors are reported for derated operation (test campaign 2) in Fig. 17. Here again the match between experimental data and predictions of the present model is very good, with the latter being mostly within the uncertainty band of the measurements. Slightly larger deviations are observed for the minC_{T} case at P_{d}=50 %. This can be explained by the fact that the machine operates at significantly low λ values, with consequent low rotational speeds. This results in particularly high angles of attack (Juangarcia et al., 2018) and very low chordbased Reynolds numbers. Both have significant impacts on the airfoil performance, which are likely not properly captured by the analytical model. Overall, it appears that the model is capable of capturing the reduction in the thrust coefficient as derating P_{d} increases, as well as the lack of symmetry with respect to the misalignment angle.
The effects of thrust and shear are visualized in Fig. 18 in terms of the average and the difference of the power loss factors at $\mathit{\gamma}\pm {\mathrm{20}}^{\circ}$, respectively denoted ${\stackrel{\mathrm{\u203e}}{\mathit{\eta}}}_{\mathrm{P},\mathit{\gamma}=\pm \mathrm{20}\mathit{\xb0}}$ and $\mathrm{\Delta}{\mathit{\eta}}_{P,\mathit{\gamma}=\pm {\mathrm{20}}^{\circ}}$. It appears that power losses tend to decrease with increasing thrust coefficients (Fig. 18a), whereas there is no significant dependency on shear (Fig. 18b). The power loss asymmetry grows with increasing shear (Fig. 18d). On the other hand, the asymmetry is roughly constant with respect to the thrust coefficient (Fig. 18c).
In the third test campaign, the wind turbine is operated above rated conditions. Figure 19 reports the 2 min average tip speed ratio and pitch angles measured during the experiment and plotted as functions of the misalignment angle γ. For the LowTI case, the turbine operates in region III for all misalignment angles. Recalling that the tip speed ratio is defined as $\mathit{\lambda}=\mathrm{\Omega}R/{u}_{\mathrm{\infty},\mathrm{hub}}$, since both the ambient wind speed u_{∞,hub} and rotor speed Ω are constant, when the turbine yaws away from the wind λ (indicated by red circles in Fig. 19a) remains constant, while the blades are pitched back (red circles in Fig. 19b) in order to keep the power output equal to the rated value. The same happens for the ModTI case. However, when $\mathit{\gamma}<\mathrm{25}$°, the turbine exits region III and enters region II. Therefore, as blade pitch (purple squares in Fig. 19b) reaches the value for the maximum power coefficient, λ starts decreasing (purple squares in Fig. 19a).
Figure 20 shows the results for the power and thrust loss factors in this scenario. Once again the proposed model exhibits a very good match with the experiments, falling within the uncertainty bands in most cases.
These results confirm the ability of the method to correctly represent the effects of different control approaches, covering both regions II and III, including derating. This is crucially important because, as shown, control laws have a strong impact on the behavior of power and trust in misaligned conditions.
The insight provided by the new model suggests two questions.

What is the poweroptimal way to yaw a single turbine out of the wind?

And does the new model affect the way wake steering should be conducted?
We try to give some initial answers to these questions in the following two sections.
5.1 Optimal power capture of a single misaligned turbine
The new model was used to compute the optimal power of a wind turbine when it is misaligned with respect to the wind. The analysis was conducted for the same IEA 3.4 MW wind turbine used for the previous numerical validation of the model.
The optimal control strategy was computed by numerical optimization using an adaptive Nelder–Mead algorithm (Gao and Han, 2012), and results are shown in Fig. 21. The figure also reports the standard region II control approach, which consists in holding the pitch angle fixed while the generator torque is varied proportionally to the square of the rotor speed, i.e., Q∼Ω^{2}, as explained in Sect. 3.3. Assuming as a first approximation that $P=\mathrm{cos}{\mathit{\gamma}}^{{\mathrm{p}}_{\mathrm{p}}}$ and considering that $Q=P/\mathrm{\Omega}$, it follows that $\mathrm{\Omega}\sim \mathrm{cos}{\mathit{\gamma}}^{{\mathrm{p}}_{\mathrm{p}}/\mathrm{3}}$. The figure reports the solution computed for a coefficient p_{p}=1.88, following Fleming et al. (2015).
As the yaw misalignment increases, the tip speed ratio drops for the standard control strategy, driven by the reduced rotororthogonal component of the wind. Since the pitch angle remains fixed, the reduced λ also leads to a decreased thrust coefficient. Although this might be beneficial for reducing loading on the yawed turbine, the resulting drop in power is significant. On the other hand, the optimal strategy governs the turbine to keep a much more constant tip speed ratio and thrust coefficient, while the blade pitches back a little. This results in some power boost, which is small for moderate angles but reaches above 3 % around ±30°. These findings are in line with results presented by Cossu (2021a) and Heck et al. (2023). While the higher C_{T} implies that the turbine is loaded more than in the standard case, it also has an effect on the wake that will be felt downstream, as explored in the next section.
5.2 Optimal power capture of two turbines
The previous section showed that a single turbine can extract more energy from the wind in misaligned conditions when its C_{T} is increased compared to a standard region II control approach. This socalled overinductive yaw control (Cossu, 2021a) increases the velocity deficit in the wake, but it also affects its recovery and enhances its deflection. It is therefore necessary to find the optimal tradeoff among these complex effects when considering wakesteering wind farm control (Meyers et al., 2022).
FLORIS v3 (NREL, 2023b), modified with the present model, was used to optimize the power capture of a cluster of two IEA 3.4 MW wind turbines placed at a distance of 5 D. The wake was modeled with the Gauss–curlhybrid model (King et al., 2021). The inflow is characterized by an ambient wind speed ${u}_{\mathrm{\infty},\mathrm{hub}}=\mathrm{9.7}$ m s^{−1}, a shear of 0.12, and a turbulence intensity of 6 %. A 60° range of wind directions Φ was considered in order to realize different degrees of overlap between the wake and the downstream rotor. The optimal wind farm control strategy was computed by numerically maximizing the cluster power with the same adaptive Nelder–Mead algorithm used for the singleturbine case of the previous section.
Results are shown in Fig. 22. The plots report in dotted green lines the results obtained with greedy control (i.e., each turbine maximizes its own power capture), in solid blue lines the solution obtained with wakesteering control based on the ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ law using p_{p}=1.88, and in dashed red lines with wakesteering control based on the present model. For the three control strategies, results were validated with LESALM simulations run for five different wind directions, namely $\mathrm{\Phi}=\mathit{\{}\mathrm{270}\pm \mathrm{5.74},\mathrm{270}\pm \mathrm{2.5},\mathrm{270}\mathit{\}}$°, corresponding to rotor overlaps of 50 %, 78.2 %, and 100 %, respectively. The LESALM results are indicated in the figure with markers, where the colors correspond to the control strategy.
Figures 22a–c show the front turbine tip speed ratio λ, pitch angle θ_{p}, and absolute misalignment angle $\left\mathit{\gamma}\right$, respectively, all plotted as functions of wind direction Φ. The solution for the present model is characterized by a fairly constant tip speed ratio that, in conjunction with some pitchback of the blades at the highest misalignments, also results in a roughly constant thrust coefficient (not shown for brevity). This is in contrast to the solution based on the ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ law, where both the tip speed ratio and the thrust coefficient drop at the higher misalignments that correspond to the strongest wake overlap conditions. In addition, the present model also results in slightly larger misalignment angles, as shown by Fig. 22c.
The bottom three plots show the effects of the various control strategies on power as a function of wind direction Φ. Figure 22d and e report the power changes with respect to the greedy strategy for the front and back turbines, respectively. It appears that the upstream machine, due to a larger misalignment, looses more power than in the ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ case. Conversely, it also appears that the second machine gains more power with the strategy based on the new model, thanks to the larger misalignment of the upstream turbine but also due to its larger thrust coefficient. Finally, Fig. 22f shows the overall gain at the cluster level. Results indicate a fairly consistent improvement, in excess of roughly 1 %, for almost the entire wake overlap range.
The LESALM results confirm the findings based on the FLORIS engineering wake model: more power losses for the front turbine and more gains for the downstream one, resulting in a positive net gain for the cluster.
We have presented a new model to estimate the power performance of a misaligned wind turbine rotor. The model is a modified version of the classical blade element momentum theory, where the rotor is considered a lifting wing of finite span operating at an angle of attack.
The new model reveals the following characteristics of the behavior of a misaligned rotor.

Power does not depend on the misalignment angle according to the ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ law, a formula in widespread use in the literature.

The true effective misalignment angle that drives wake behavior is a combination of both yaw and tilt. Therefore, a twodimensional wake model should be described in the plane formed by the rotor axis and wind vectors, not on the horizontal plane as commonly assumed.

Power depends on the true misalignment angle but – crucially – also on the way the rotor is governed as it is pointed out of the wind, a fact that probably explains the widely different performance observed by various authors. This fact also means that power losses due to misalignment can be mitigated by using a suitable control strategy.

According to the model, the observed lack of symmetry between positive and negative misalignment angles is caused by the interaction with a sheared inflow. In these conditions, there is a complex interplay of various effects that may lead to different outcomes in terms of which yaw sign yields more or less power. In general, one can expect a small asymmetry at high thrust coefficients, while a more pronounced asymmetry emerges for low thrust and high tip speed ratios, where a higher power is generated for positive yaw angles. However, in general the behavior of power (but of thrust too, which also exhibits an asymmetric behavior) depends not only on the rotor design characteristics but also on the way it is governed, through the values of the pitch setting and of the tip speed ratio. Additionally, in the field other effects may be present (e.g., due to an asymmetric behavior of the onboard wind vane), which may add to the phenomena described by the model.

A constantovertherotor induction is sufficient to accurately describe the power and thrust behavior of a misaligned rotor in a sheared inflow. In fact, under classical smallangle assumptions, the tilting of the inflow due to misalignment and shear has only a negligible effect on the quality of the results.
The model was derived in a semianalytical form, leading to a closed system of equations that can be directly integrated with engineering wake models, at an irrelevant computational cost. To improve its accuracy, we proposed a specific implementation that overcomes the intrinsic limits brought by the analytical solution of some model integrals. The proposed implementation corrects for the effects of misalignment a higherfidelity power model obtained in aligned conditions, and it calibrates the model parameters based on measurements.
The model was validated in a broad range of cases, considering LESALM numerical simulations of various multiMW machines as well as experimental observations on a scaled wind tunnel model, in different inflows (from unsheared laminar to sheared highly turbulent conditions), operating with controllers in the loop in regions II and III, as well as in derated conditions. In all cases, the model achieved a very satisfactory agreement with the numerical and experimental reference power and thrust values. Additionally, we demonstrated how the model can be integrated with given control laws, achieving an excellent match in the calculation of the set points. The model was also compared with a similar model recently developed by Heck et al. (2023), limitedly to the control region II where it is applicable, consistently improving on its predictions and exhibiting a wider applicability to arbitrary control strategies.
Using the proposed model, we maximized the power capture of a wind turbine for a range of misalignment angles, obtaining the optimal power strategy in terms of pitch setting and tip speed ratio. Results indicate that the maximum power extraction is obtained by keeping an almost constant tip speed ratio and by slightly reducing the blade pitch as the turbine yaws out of the wind. This also implies a roughly constant thrust coefficient, which will increase the loading on the yawed turbine, but will also have an effect on its wake.
Next, we applied the new model to the maximization of power by wake steering for a cluster of two turbines. The resulting control strategy was compared to the one obtained by the classical ${\mathrm{cos}}^{{\mathrm{p}}_{\mathrm{p}}}$ power loss model and validated by means of LESALM simulations for a few selected cases. Results indicate that the proposed model results in slightly greater power losses for the wakesteering turbine, which are more than compensated for by greater power gains for the wakeaffected one, achieving a small but consistent gain in power at the cluster level for the full range of possible wake overlaps.
Future work should investigate the effects of the new model and its resulting control strategy in more complex conditions. Of particular interest is the analysis of the effects on loads, which might increase because of the eliminated drop in thrust coefficient as the turbine is yawed out of the wind.
A1 Transformation from ground to nacelle frame of reference
The nacellefixed frame of reference is obtained from the groundfixed frame by a first rotation δ about the horizontal axis y_{g}, followed by a rotation γ about the vertical axis z_{g}. The components of a generic vector v are denoted ${\underset{\mathrm{\xaf}}{v}}_{\mathrm{g}}$ when measured in the ground frame ℱ_{g} and ${\underset{\mathrm{\xaf}}{v}}_{\mathrm{n}}$ when measured in the nacelle frame ℱ_{n}. Combining the two successive rotations, one obtains the transformation of components from one frame to the other as
The inverse transformation is simply given by the matrix transpose.
Using Eq. (A1), the components of the ambient velocity vector u_{∞} in the nacelleattached frame are readily found to be ${\underset{\mathrm{\xaf}}{u}}_{{\mathrm{\infty}}_{\mathrm{n}}}={u}_{\mathrm{\infty}}\mathit{\{}\mathrm{cos}\mathit{\delta}\mathrm{cos}\mathit{\gamma},\mathrm{sin}\mathit{\delta},\mathrm{sin}\mathit{\delta}\mathrm{cos}\mathit{\gamma}{\mathit{\}}}^{T}$, where the scalar wind speed is ${u}_{\mathrm{\infty}}=\left{\mathit{u}}_{\mathrm{\infty}}\right$, while ${\underset{\mathrm{\xaf}}{x}}_{{\mathrm{n}}_{\mathrm{n}}}=\mathit{\{}\mathrm{1},\mathrm{0},\mathrm{0}{\mathit{\}}}^{T}$. Hence, it follows that
A2 Transformation from wakedeflection to ground frame of reference
The nacelle and wakedeflection frames share the same unit vector x_{n}=x_{d}, which corresponds to the rotor axis of rotation. The z_{d} unit vector is orthogonal to the plane composed by u_{∞} and x_{d}, and therefore it can be written as ${\mathit{z}}_{\mathrm{d}}=z{\mathit{x}}_{\mathrm{d}}\times {\mathit{u}}_{\mathrm{\infty}}$, where z is a normalization scalar such that $\mathit{z}\cdot \mathit{z}=\mathrm{1}$. Performing the cross product and the normalization, one finds ${\underset{\mathrm{\xaf}}{z}}_{{\mathrm{d}}_{\mathrm{n}}}=z\mathit{\{}\mathrm{0},\mathrm{sin}\mathit{\delta}\mathrm{cos}\mathit{\gamma},\mathrm{sin}\mathit{\gamma}{\mathit{\}}}^{T}$ and $z=\mathrm{1}/\sqrt{{\mathrm{cos}}^{\mathrm{2}}\mathit{\gamma}{\mathrm{sin}}^{\mathrm{2}}\mathit{\delta}+{\mathrm{sin}}^{\mathrm{2}}\mathit{\gamma}}=\mathrm{1}/\mathrm{sin}\mathit{\mu}$. A righthanded triad is completed by setting ${\mathit{y}}_{\mathrm{d}}={\mathit{x}}_{\mathrm{d}}\times {\mathit{z}}_{\mathrm{d}}$, which yields ${\underset{\mathrm{\xaf}}{y}}_{{\mathrm{d}}_{\mathrm{n}}}=z\mathit{\{}\mathrm{0},\mathrm{sin}\mathit{\gamma},\mathrm{sin}\mathit{\delta}\mathrm{cos}\mathit{\gamma}{\mathit{\}}}^{T}$. The transformation matrix between the wakedeflection and nacellefixed components is therefore readily obtained as
Finally, the transformation between wakedeflection and groundfixed components follows by using Eqs. (A3) and (A1), which yields
The inverse transformation is simply given by the matrix transpose.
Using Eq. (A4b), the longitudinal (given by Eq. 11) and lateral (sidewash, given by Eq. 13) flow velocity components at the streamtube outlet can be transformed into the corresponding longitudinal and lateral components in the ground frame:
As mentioned in Sect. 2.4, misalignment and shear cause a nonuniform distribution of the induction over the rotor disk. Following the classical approach used for helicopter rotors in forward flight (Johnson, 1995), the simplest model of nonuniform axial induction is the one expressed by Eq. (6), based on the 1P harmonics κ_{1s} and κ_{1c}. For the present application, however, it appears that the inclusion of these terms is not necessary. To show this, we consider the κ_{1s} term, which is triggered by the misalignment μ and results in the largest induction in the most downstream portion of the rotor disk. Figure B1 presents the loss factors η_{T} and η_{P} predicted by the model with (dotted red line) and without (solid blue line) the 1P sine term κ_{1s}. Differences appear to be negligible, especially for η_{T}. Neglecting κ_{1s} leads to a slight decrease in η_{P} as the misalignment increases, reaching a maximum difference of 0.21 % at $\mathit{\gamma}=\mathrm{30}$°.
A  Rotor disk area 
a  Axial induction 
B  Number of blades 
C_{D}  Drag coefficient 
C_{L}  Lift coefficient 
C_{L,α}  Lift slope 
C_{P}  Power coefficient 
C_{T}  Thrust coefficient 
${C}_{\mathrm{T}}^{\prime}$  Modified thrust coefficient of Heck et al. (2023) 
D  Rotor diameter 
F_{n}  Normal force 
F_{t}  Tangential force 
K  Coefficient relating aerodynamic torque 
and squared rotor speed in control region II  
k  Linear vertical wind shear coefficient 
$\dot{m}$  Mass flux 
p  Pressure 
P  Power 
P_{d}  Power demand (derating) 
Q  Rotor torque 
R  Rotor radius 
r  Spanwise coordinate 
T  Thrust force 
u_{∞}  Freestream wind speed 
u_{∞,hub}  Freestream wind speed at hub height 
u  Longitudinal velocity component 
u_{n}  Rotororthogonal velocity component 
u_{t}  Rotortangential velocity component 
v  Lateral velocity component 
x  Cartesian coordinate 
y  Cartesian coordinate 
z  Cartesian coordinate 
α  Angle of attack 
β  Blade twist angle 
δ  Rotor tilt angle 
η_{P}  Power loss factor 
η_{T}  Thrust loss factor 
γ  Rotor yaw angle 
λ  Tip speed ratio 
μ  Rotor total (true) misalignment angle 
Ω  Rotor angular speed 
φ  Inflow angle 
Φ  Wind direction 
ψ  Rotor azimuth angle 
ρ  Air density 
θ  Local pitch angle 
θ_{p}  Blade pitch rotation at the pitch bearing 
${\underset{\mathrm{\xaf}}{v}}_{\mathrm{f}}$  Components of vector v in frame f 
(⋅)_{d}  Quantity evaluated in the wakedeflection 
intrinsic frame of reference  
(⋅)_{g}  Quantity evaluated in the ground frame of 
reference 
(⋅)_{n}  Quantity evaluated in the nacelle frame of 
reference  
(⋅)_{1c}  Onceperrevolution cosine harmonic 
(⋅)_{1s}  Onceperrevolution sine harmonic 
0P  Zeroth (constant) harmonic 
1P  Oneperrevolution harmonic 
ALM  Actuator line method 
BEM  Blade element momentum 
CFD  Computational fluid dynamics 
CTA  Constant temperature anemometry 
FLORIS  FLOw Redirection and Induction in Steady 
State  
LES  Large eddy simulation 
rpm  Revolutions per minute 
SOWFA  Simulator fOr Wind Farm Applications 
TI  Turbulence intensity 
An implementation of the model described in this article in the FLORIS framework is available on Zenodo at https://doi.org/10.5281/zenodo.10974493 (Tamaro et al., 2024a). The repository also contains all the data and the Jupyter notebooks used to generate the figures. The code and the scripts to reproduce the figures can be run on Binder at the link https://tinyurl.com/btclfigs (Tamaro et al., 2024b). The notebook of Fig. 5 can be used to interactively plot the thrust and power coefficients for other userdefined values of the model parameters, while the one of Fig. 6 can be similarly used to visualize different control trajectories. The notebooks of Figs. 21 and 22 contain also the code used for computing the optimal control policies. The complete data sets from the LESs and wind tunnel experiments are available upon request.
CLB developed the formulation of the misalignment model and supervised the overall research. ST implemented the model, performed the LESs and the corresponding model validation, and conducted the wakesteering analyses with FLORIS. FC performed the validation with respect to the experimental measurements. All authors contributed to the interpretation of the results. CLB and ST wrote the manuscript, with contributions by FC in the experimental section. All authors provided important input to this research work through discussions and feedback and improved the manuscript.
At least one of the (co)authors is a member of the editorial board of Wind Energy Science. The peerreview process was guided by an independent editor, and the authors also have no other competing interests to declare.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors.
The authors express their gratitude to the Leibniz Supercomputing Centre (LRZ) for providing access and computing time on the SuperMUC Petascale System under project ID pr84be “Largeeddy Simulation for Wind Farm Control”. The authors would like to thank Kirby Heck for the valuable discussion and Adrien Guilloré and Franz Mühle for help in the measurement campaign.
This work has been supported in part by the PowerTracker project, which receives funding from the German Federal Ministry for Economic Affairs and Climate Action (FKZ 03EE2036A). This work has also been partially supported by the MERIDIONAL project, which receives funding from the European Union’s Horizon Europe program under the grant agreement no. 101084216.
This paper was edited by Johan Meyers and reviewed by two anonymous referees.
Abramowitz, M., Stegun, I. A., and Romer, R. H.: Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, Am. J. Phys., 56, 958–958, https://doi.org/10.1119/1.15378, 1988. a
Bak, C., Zahle, F., Bitsche, R., Kim, T., Yde, A., Henriksen, L. C., Natarajan, A., and Hansen, M.: Description of the DTU 10 MW reference wind turbine, Tech. Rep. I0092 5, DTU Wind Energy, https://backend.orbit.dtu.dk/ws/portalfiles/portal/55645274/The_DTU_10MW_Reference_Turbine_Christian_Bak.pdf (last access: 12 July 2024), 2013. a, b
Bartl, J., Mühle, F., Schottler, J., Sætran, L., Peinke, J., Adaramola, M., and Hölling, M.: Wind tunnel experiments on wind turbine wakes in yaw: effects of inflow turbulence and shear, Wind Energ. Sci., 3, 329–343, https://doi.org/10.5194/wes33292018, 2018. a
Bortolotti, P., Tarres, H. C., Dykes, K. L., Merz, K., Sethuraman, L., Verelst, D., and Zahle, F.: IEA Wind TCP Task 37: Systems Engineering in Wind Energy – WP2.1 Reference Wind Turbines, Tech. rep., NREL – National Renewable Energy Lab., https://doi.org/10.2172/1529216, 2019. a, b, c, d
Bottasso, C., Croce, A., Nam, Y., and Riboldi, C.: Power curve tracking in the presence of a tip speed constraint, Renew. Energy, 40, 1–12, https://doi.org/10.1016/j.renene.2011.07.045, 2012. a, b
Bottasso, C. L. and Campagnolo, F.: Wind Tunnel Testing of Wind Turbines and Farms, in: Handbook of Wind Energy Aerodynamics, edited by: Stoevesandt, B., Schepers, G., Fuglsang, P., and Sun, Y., Springer International Publishing, Cham, 1077–1126, ISBN 9783030313074, https://doi.org/10.1007/9783030313074_54, 2022. a, b, c, d
Bottasso, C. L., Campagnolo, F., and Petrović, V.: Wind tunnel testing of scaled wind turbine models: Beyond aerodynamics, J. Wind Eng. Indust. Aerodynam., 127, 11–28, https://doi.org/10.1016/j.jweia.2014.01.009, 2014. a, b
Burton, T., Jenkins, N., Sharpe, D., and Bossanyi, E.: Wind Energy Handbook, John Wiley & Sons, Ltd, ISBN 9780470699751, https://doi.org/10.1002/9781119992714, 2011. a, b, c
Campagnolo, F., Petrović, V., Schreiber, J., Nanos, E. M., Croce, A., and Bottasso, C. L.: Wind tunnel testing of a closedloop wake deflection controller for wind farm power maximization, J. Phys.: Conf. Ser., 753, 032006, https://doi.org/10.1088/17426596/753/3/032006, 2016. a, b
Campagnolo, F., Weber, R., Schreiber, J., and Bottasso, C. L.: Wind tunnel testing of wake steering with dynamic wind direction changes, Wind Energ. Sci., 5, 1273–1295, https://doi.org/10.5194/wes512732020, 2020. a, b, c, d
Campagnolo, F., Tamaro, S., Mühle, F., and Bottasso, C. L.: Wind Tunnel Testing of Combined Derating and Wake Steering, IFACPapersOnLine, 56, 8400–8405, https://doi.org/10.1016/j.ifacol.2023.10.1034, 2023. a, b, c
Coleman, R. P., Feingold, A. M., and Stempin, C. W.: Evaluation of the InducedVelocity Field of an Idealized Helicopter Rotor, Wartime report, National Advisory Committee for Aeronautics, Langley Memorial Aeronautical Laboratory, https://apps.dtic.mil/sti/pdfs/ADA801123.pdf (last access: 12 July 2024), 1945. a
Cossu, C.: Wake redirection at higher axial induction, Wind Energ. Sci., 6, 377–388, https://doi.org/10.5194/wes63772021, 2021a. a, b, c, d
Cossu, C.: Evaluation of tilt control for windturbine arrays in the atmospheric boundary layer, Wind Energ. Sci., 6, 663–675, https://doi.org/10.5194/wes66632021, 2021b. a, b
Dahlberg, J. and Montgomerie, B.: Research program of the Utgrunden demonstration offshore wind farm, final report part 2, wake effects and other loads, Technical report 0217, Swedish Defense Research Agency (FOI), Kista, Sweden, 2005. a
Doekemeijer, B. M., Kern, S., Maturu, S., Kanev, S., Salbert, B., Schreiber, J., Campagnolo, F., Bottasso, C. L., Schuler, S., Wilts, F., Neumann, T., Potenza, G., Calabretta, F., Fioretti, F., and van Wingerden, J.W.: Field experiment for openloop yawbased wake steering at a commercial onshore wind farm in Italy, Wind Energ. Sci., 6, 159–176, https://doi.org/10.5194/wes61592021, 2021. a
Draper, M., Guggeri, A., López, B., Díaz, A., Campagnolo, F., and Usera, G.: A Large Eddy Simulation framework to assess wind farm power maximization strategies: Validation of maximization by yawing, J. Phys.: Conf. Ser., 1037, 072051, https://doi.org/10.1088/17426596/1037/7/072051, 2018. a, b
Fleming, P., Gebraad, P. M., Lee, S., van Wingerden, J.W., Johnson, K., Churchfield, M., Michalakes, J., Spalart, P., and Moriarty, P.: Simulation comparison of wake mitigation control strategies for a twoturbine case, Wind Energy, 18, 2135–2143, https://doi.org/10.1002/we.1810, 2015. a, b, c
Fleming, P., Annoni, J., Shah, J. J., Wang, L., Ananthan, S., Zhang, Z., Hutchings, K., Wang, P., Chen, W., and Chen, L.: Field test of wake steering at an offshore wind farm, Wind Energ. Sci., 2, 229–239, https://doi.org/10.5194/wes22292017, 2017. a
Fleming, P., Annoni, J., Churchfield, M., MartinezTossas, L. A., Gruchalla, K., Lawson, M., and Moriarty, P.: A simulation study demonstrating the importance of largescale trailing vortices in wake steering, Wind Energ. Sci., 3, 243–255, https://doi.org/10.5194/wes32432018, 2018. a
Fleming, P., King, J., Dykes, K., Simley, E., Roadman, J., Scholbrock, A., Murphy, P., Lundquist, J. K., Moriarty, P., Fleming, K., van Dam, J., Bay, C., Mudafort, R., Lopez, H., Skopek, J., Scott, M., Ryan, B., Guernsey, C., and Brake, D.: Initial results from a field campaign of wake steering applied at a commercial wind farm – Part 1, Wind Energ. Sci., 4, 273–285, https://doi.org/10.5194/wes42732019, 2019. a
Fleming, P. A., Gebraad, P. M., Lee, S., van Wingerden, J.W., Johnson, K., Churchfield, M., Michalakes, J., Spalart, P., and Moriarty, P.: Evaluating techniques for redirecting turbine wakes using SOWFA, Renew. Energy, 70, 211–218, https://doi.org/10.1016/j.renene.2014.02.015, 2014. a
Gao, F. and Han, L.: Implementing the NelderMead simplex algorithm with adaptive parameters, Comput. Optimiz. Appl., 51, 259–277, https://doi.org/10.1007/s1058901093293, 2012. a
Gebraad, P. M. O., Teeuwisse, F. W., van Wingerden, J. W., Fleming, P. A., Ruben, S. D., Marden, J. R., and Pao, L. Y.: Wind plant power optimization through yaw control using a parametric model for wake effects – a CFD simulation study, Wind Energy, 19, 95–114, https://doi.org/10.1002/we.1822, 2016. a
Hansen, M.: Aerodynamics of wind turbines: Third edition, Routledge, ISBN 9781315769981, https://doi.org/10.4324/9781315769981, 2015. a, b, c, d
Heck, K., Johlas, H., and Howland, M.: Modelling the induction, thrust and power of a yawmisaligned actuator disk, J. Fluid Mech., 959, A9, https://doi.org/10.1017/jfm.2023.129, 2023. a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r
Howland, M. F., González, C. M., Martínez, J. J. P., Quesada, J. B., Larrañaga, F. P., Yadav, N. K., Chawla, J. S., and Dabiri, J. O.: Influence of atmospheric conditions on the power production of utilityscale wind turbines in yaw misalignment, J. Renew. Sustain. Energ., 12, 063307, https://doi.org/10.1063/5.0023746, 2020. a, b
Hulsman, P., Sucameli, C., Petrović, V., Rott, A., Gerds, A., and Kühn, M.: Turbine power loss during yawmisaligned free field tests at different atmospheric conditions, J. Phys.: Conf. Ser., 2265, 032074, https://doi.org/10.1088/17426596/2265/3/032074, 2022. a
Jiménez, A., Crespo, A., and Migoya, E.: Application of a LES technique to characterize the wake deflection of a wind turbine in yaw, Wind Energy, 13, 559–572, https://doi.org/10.1002/we.380, 2010. a, b, c
Johnson, W.: Helicopter Theory, Dover Books on Aeronautical Engineering, Dover Publications, ISBN 9780486682303, 1995. a, b, c, d, e
Jonkman, J., Butterfield, S., Musial, W., and Scott, G.: Definition of a 5MW Reference Wind Turbine for Offshore System Development, Tech. rep., National Renewable Energy Laboratory, https://doi.org/10.2172/947422, 2009. a, b, c
Juangarcia, D. A., Eguinoa, I., and Knudsen, T.: Derating a single wind farm turbine for reducing its wake and fatigue, J. Phys. Conf. Ser., 1037, 032039, https://doi.org/10.1088/17426596/1037/3/032039, 2018. a, b, c
Katz, J. and Plotkin, A.: LowSpeed Aerodynamics, in: Cambridge Aerospace Series, 2nd Edn., Cambridge University Press, https://doi.org/10.1017/CBO9780511810329, 2001. a
King, J., Fleming, P., King, R., MartínezTossas, L. A., Bay, C. J., Mudafort, R., and Simley, E.: Controloriented model for secondary effects of wake steering, Wind Energ. Sci., 6, 701–714, https://doi.org/10.5194/wes67012021, 2021. a
Krogstad, P.A. and Adaramola, M. S.: Performance and near wake measurements of a model horizontal axis wind turbine, Wind Energy, 15, 743–756, https://doi.org/10.1002/we.502, 2012. a
Leloudas, G., Zhu, W. J., Sørensen, J. N., Shen, W. Z., and Hjort, S.: Prediction and Reduction of Noise from a 2.3 MW Wind Turbine, J. Phys.: Conf. Ser., 75, 012083, https://doi.org/10.1088/17426596/75/1/012083, 2007. a
Liew, J., Urbán, A. M., and Andersen, S. J.: Analytical model for the power–yaw sensitivity of wind turbines operating in full wake, Wind Energ. Sci., 5, 427–437, https://doi.org/10.5194/wes54272020, 2020. a, b, c
Marelli, S. and Sudret, B.: UQLab: A Framework for Uncertainty Quantification in Matlab, in: The 2nd International Conference on Vulnerability and Risk Analysis and Management (ICVRAM 2014), American Society of Civil Engineers, 2554–2563, https://doi.org/10.1061/9780784413609.257, 2014. a
MartínezTossas, L. A. and Meneveau, C.: Filtered lifting line theory and application to the actuator line model, J. Fluid Mech., 863, 269–292, https://doi.org/10.1017/jfm.2018.994, 2019. a
MartínezTossas, L. A., King, J., Quon, E., Bay, C. J., Mudafort, R., Hamilton, N., Howland, M. F., and Fleming, P. A.: The curled wake model: a threedimensional and extremely fast steadystate wake solver for wind plant flows, Wind Energ. Sci., 6, 555—570, https://doi.org/10.5194/wes65552021, 2021. a
Medici, D.: Experimental studies of wind turbine wakes : power optimisation and meandering, PhD thesis, Royal Institute of Technology, Stockholm, Sweden, https://www.mech.kth.se/thesis/2006/phd/phd_2006_davide_medici.pdf (last access: 12 July 2024), 2005. a
Meyer Forsting, A., van der Laan, M., and Troldborg, N.: The induction zone/factor and sheared inflow: A linear connection?, J. Phys.: Conf. Ser., 1037, 072031, https://doi.org/10.1088/17426596/1037/7/072031, 2018. a
Meyers, J., Bottasso, C., Dykes, K., Fleming, P., Gebraad, P., Giebel, G., Göçmen, T., and van Wingerden, J.W.: Wind farm flow control: prospects and challenges, Wind Energ. Sci., 7, 2271–2306, https://doi.org/10.5194/wes722712022, 2022. a, b, c
MKS Instruments Inc.: 226A Baratron Differential Capacitance Manometer, https://www.mks.com/ (last access: 12 July 2024), 2022. a
Nanos, E. M., Bottasso, C. L., Tamaro, S., Manolas, D. I., and Riziotis, V. A.: Vertical wake deflection for floating wind turbines by differential ballast control, Wind Energ. Sci., 7, 1641–1660, https://doi.org/10.5194/wes716412022, 2022. a, b
NREL: AeroDyn Users Guide and Theory Manual, https://github.com/OpenFAST/OpenFAST/blob/main/docs/source/user/aerodyn/index.rst (last access: 19 March 2023), 2023a. a
NREL: FLORIS, Version 3, https://github.com/NREL/floris (last access: 19 March 2023), 2023b. a, b, c, d
NREL: OpenFAST, https://github.com/openfast (last access: 19 March 2023), 2023c. a
Pedersen, M. M., van der Laan, P., FriisMøller, M., Rinker, J., and Réthoré, P.E.: DTUWindEnergy/PyWake: PyWake, Zenodo [code], https://doi.org/10.5281/zenodo.2562662, 2019. a, b
Pitt, D. M. and Peters, D. A.: Theoretical prediction of dynamic inflow derivatives, Vertica, 5, 1981. a
Schottler, J., Hölling, A., Peinke, J., and Hölling, M.: Brief communication: On the influence of vertical wind shear on the combined power output of two model wind turbines in yaw, Wind Energ. Sci., 2, 439–442, https://doi.org/10.5194/wes24392017, 2017. a
Shapiro, C. R., Gayme, D. F., and Meneveau, C.: Modelling yawed wind turbine wakes: a lifting line approach, J. Fluid Mech., 841, R1, https://doi.org/10.1017/jfm.2018.75, 2018. a
Simley, E., Fleming, P., Girard, N., Alloin, L., Godefroy, E., and Duc, T.: Results from a wakesteering experiment at a commercial wind plant: investigating the wind speed dependence of wakesteering performance, Wind Energ. Sci., 6, 1427–1453, https://doi.org/10.5194/wes614272021, 2021. a
Tamaro, S., Campagnolo, F., and Bottasso, C. L.: On the power and control of a misaligned rotor – Beyond the cosine law, Zenodo [source code and data set], https://doi.org/10.5281/zenodo.10974493, 2024a. a, b, c
Tamaro, S., Campagnolo, F., and Bottasso, C. L.: On the power and control of a misaligned rotor – Beyond the cosine law, Binder [source code and data set for Figs. 2, 5, 6, 21, and 22], https://tinyurl.com/btclfigs (last access: 12 July 2024), 2024b. a
Tietjens, O. and Prandtl, L.: Applied Hydro and Aeromechanics: Based on Lectures of L. Prandtl, in: no. Bd. 2 in Applied hydro and aeromechanics: based on lectures of L. Prandtl, Dover Publications, ISBN 9780486603759, https://books.google.de/books?id=Dsbd0zAwIYC (last access: 12 July 2024), 1957. a
Troldborg, N., Sørensen, J. N., and Mikkelsen, R.: Actuator Line Simulation of Wake of Wind Turbine Operating in Turbulent Inflow, J. Phys.: Conf. Ser., 75, 012063, https://doi.org/10.1088/17426596/75/1/012063, 2007. a
Wang, C., Wang, J., Campagnolo, F., Carraón, D. B., and Bottasso, C. L.: Validation of largeeddy simulation of scaled waked wind turbines in different yaw misalignment conditions, J. Phys.: Conf. Ser., 1037, 062007, https://doi.org/10.1088/17426596/1037/6/062007, 2018. a
Wang, C., Campagnolo, F., and Bottasso, C. L.: Identification of airfoil polars from uncertain experimental measurements, Wind Energ. Sci., 5, 1537–1550, https://doi.org/10.5194/wes515372020, 2020. a, b, c, d, e
Wang, C., Campagnolo, F., Canet, H., Barreiro, D. J., and Bottasso, C. L.: How realistic are the wakes of scaled wind turbine models?, Wind Energ. Sci., 6, 961–981, https://doi.org/10.5194/wes69612021, 2021. a, b, c
Wang, J., Wang, C., Campagnolo, F., and Bottasso, C. L.: Wake behavior and control: comparison of LES simulations and wind tunnel measurements, Wind Energ. Sci., 4, 71–88, https://doi.org/10.5194/wes4712019, 2019. a
Zalkind, D., Nicotra, M., and Pao, L.: Constrained power reference control for wind turbines, Wind Energy, 25, 914–934, https://doi.org/10.1002/we.2705, 2022. a, b
This interpretation also reveals that the socalled curled shape of laterally deflected wakes (see, e.g., MartínezTossas et al., 2021, and references therein) is nothing else than the effect of the horseshoe vortex structure generated behind a lifting wing, albeit with the addition of the swirl caused by the rotor rotation.
 Abstract
 Introduction
 Misalignment model
 Implementation
 Model validation
 Optimal wake steering
 Conclusions
 Appendix A: Transformation matrices
 Appendix B: Effect of a nonuniform axial induction on the rotor disk
 Appendix C: Experimental data set
 Appendix D: Nomenclature
 Code and data availability
 Author contributions
 Competing interests
 Disclaimer
 Acknowledgements
 Financial support
 Review statement
 References
 Abstract
 Introduction
 Misalignment model
 Implementation
 Model validation
 Optimal wake steering
 Conclusions
 Appendix A: Transformation matrices
 Appendix B: Effect of a nonuniform axial induction on the rotor disk
 Appendix C: Experimental data set
 Appendix D: Nomenclature
 Code and data availability
 Author contributions
 Competing interests
 Disclaimer
 Acknowledgements
 Financial support
 Review statement
 References