Method for calibrating a wind turbine controller model

Abstract

The invention relates to a method for calibrating a wind turbine controller model, in which at least the following steps are performed: a) from the measurements taken (Mes), a speed ratio and a blade pitch angle are associated (Ass) with certain wind speeds; b) for each wind speed, a final speed ratio (Ld) and a final blade pitch angle (Bd) are determined to minimize an optimization problem; c) the controller model is calibrated (ContInv) so that it associates the final speed ratio and / or the final blade pitch angle with each wind speed. Figure 1 to be published

Description

Title of the invention: Method for calibrating a wind turbine controller model technical field

[0001] The invention relates to the field of wind power and more particularly to a method for calibrating a wind turbine model, in particular for calibrating the wind turbine controller model, with respect to measurements carried out on the wind turbine. Previous technique

[0002] Wind power is a rapidly growing energy sector. In this sector, wind farm operators purchase wind turbines from turbine manufacturers who are their customers. However, turbine manufacturers often wish to keep certain technical information confidential and are therefore reluctant to share this information about their machines. Consequently, wind farm operators seek to develop their knowledge and understanding of the machines they operate by equipping them with various sensors and through data engineering. This allows them to improve the maintenance and monitoring of their wind turbines, their goal being to reduce the operating costs of their wind farms.To address this need, research teams have developed a technology called a "data-driven approach," which involves identifying non-physical models from measured data on a wind turbine to predict the power the turbine can produce or estimate the current state of mechanical fatigue of various wind turbine components. The following publications mention this type of technology: - Fahhn M., Sharma V., Cao TV, Canberk B., et al. “Machine learning-based digital twin for predictive modeling in wind turbines”, 2022, IEEE Access - De Kooning JDM, Stockman K., De Maeyer J., et al. “Digital twins for wind energy conversion Systems: a literature review of potential modeling techniques focused on model fidelity and computational load”, 2021, MDPI - Dimitrov N., Kelly MC, Vignaroli A. “From wind to loads: wind turbine site-specific load estimation with surrogate models trained on high-fidelity load databases”, 2018, Wind Energy Science

[0003] To accurately reflect reality, these models often require numerous measurements on actual wind turbines. However, on the one hand, instrumenting a wind turbine is expensive, and on the other hand, by collecting data over a limited period, not all operating points of a wind turbine are necessarily visited. By Consequently, simply limiting oneself to measurements of inputs and outputs of an instrumented wind turbine can be an extremely costly process and insufficient to allow a realistic (representative of reality) and accurate model.

[0004] To mitigate these data gap problems, another solution involves developing a simulator of the various physical phenomena (an aero-servo-elastic simulator to simulate aerodynamics, elastic behavior, and control, or an aero-servo-hydro-elastic simulator to simulate aerodynamics, hydrodynamics, elastic behavior, and control) of a realistic wind turbine to generate synthetic numerical data without using measurements on the actual wind turbine. This technology is called a "model-based approach." These solutions are described in particular in: - Fahhn M., Sharma V., Cao TV, Canberk B., et al. “Machine learning-based digital twin for predictive modeling in wind turbines”, 2022, IEEE Access - Dimitrov N., Kelly MC, Vignaroli A. “From wind to loads: wind turbine site-specific load estimation with surrogate models trained on high-fidelity load databases”, 2018, Wind Energy Science

[0005] This synthetic data generation approach is computationally expensive, but it can guarantee a more comprehensive survey of the various operating points of the wind turbine than a physically instrumented wind turbine, in a much shorter time, and without instrumentation. Furthermore, it is possible to create virtual sensors on the model to avoid very expensive or even impractical instrumentation. The drawback of this approach is that the accuracy of the synthetic numerical data depends on the fidelity of the aero-servo-elastic model (or aero-servo-hydro-elastic model when hydrodynamics are also simulated) to the physical wind turbine. Indeed, to achieve sufficient accuracy, it would be necessary to build the aerodynamic and control components of the model from data provided by the wind turbine manufacturer.However, in practice, manufacturers do not provide sufficient data for this.

[0006] Another approach, known as the "hybrid" approach, consists of using minimal instrumentation to reconstruct an aero-servo-elastic model (or aero-servo-hydro-elastic model when hydrodynamics are also simulated) that is sufficiently faithful to the actual turbine when the wind turbine design data is unavailable or incomplete. This approach is notably that of Pimenta F., Pacheco J., Branco CM, et al., "Development of a digital twin of an onshore wind turbine using monitoring data", 2020, Journal of Physics: Conference Series. The hybrid approach makes it possible to guarantee a faithful model with less time-consuming and expensive instrumentation. This model can then be used to supplement data, for example to obtain operating points that have not been measured on the wind turbine. For this, the document Hirvoas A, “Development of a data assimilation method for the calibration and continuous update of wind turbines digital twins”, 2021, Thesis and the document Chetan M., Yao S., Griffith DT “Multi-fidelity digital twin structural model for a sub-scale downwind wind turbine rotor blade”, 2021, Wind Energy allow us to reconstruct the elastic part of the model.

[0007] Other methods are based on scans or blade diagrams that allow the aerodynamic properties of the wind turbine to be determined using fluid mechanics and the Blade Element Momentum (BEM) theory, well known to those skilled in the art. These methods were notably developed in Pimenta F., Pacheco J., Branco CM et al., “Development of a digital twin of an onshore wind turbine using monitoring data”, 2020, Journal of Physics: Conference Series, and Hansen M., “Aerodynamics of wind turbines”, 2015, Taylor-Francis. However, these methods do not allow the use of data from the wind turbine's SCADA (Supervisory Control and Data Acquisition) system.

[0008] Furthermore, the modeled controller is often a standard controller model because wind turbine manufacturers generally do not provide data on the actual controller. Thus, the controller modeled on the wind turbine is usually inaccurate. For example, the publication Abbas NJ, Zalkind DS, Pao L., Wright A. “A reference open-source controller for fixed and floating offshore wind turbines”, 2022, Wind Energy Science, describes the “ROSCO” controller, which is a commonly used standard controller model. Summary of the invention

[0009] The object of the invention is to perform a calibration of at least one part (the "servo" part of the controller model) of an aero-servo-elastic model of wind turbines (preferably an aero-servo-hydro-elastic model for an offshore wind turbine, for example, fixed at sea or floating), improving the model's accuracy by taking into account data measured on the wind turbine. This model can then be used to anticipate maintenance actions on the wind turbine, to modify the wind turbine's control (i.e., its servo system), or to determine the fatigue state of different wind turbine components. In particular, the aim is to improve the accuracy of the controller used in the model.

[0010] The invention relates to a method for calibrating a wind turbine controller model comprising a rotor, the rotor comprising blades, to determine the fatigue state of the wind turbine components and / or the evolution of the wind turbine's performance and / or to adapt the wind turbine control and / or to predict the actions of wind turbine maintenance, based on measurements taken on the wind turbine, including wind speed, blade pitch angle, and rotation speed, in which at least the following steps are carried out: a) Based on the measurements taken, predetermined wind speed values ​​are associated with: - a speed ratio X defined by — with œ the rotational speed of the rotor, R the radius of the rotor and v the wind speed; - and a blade orientation angle; b) For each predetermined value of wind speed, the values ​​of the final speed ratio and the final blade pitch angle are determined as solutions to the following minimization problem: )•#(»») ) ); + A(Cr(z (v,), / l(r,) -^(½) )2*»(n) (•*('a2 aVCC € / .(2(^-), ^(vi) ) a power coefficient map of the wind turbine, Cr(x(vJ, ) a thrust coefficient map of the wind turbine, vi the predetermined wind speed values, N being the number of predetermined wind speed values, Cp mrve{v^ being a power coefficient curve, Cr,cwrve( being a thrust coefficient curve, the blade pitch angle at wind speed vi ; 2( v,) being the speed ratio to wind speed V' ; Pq(vî) being the orientation angle associated with wind speed vi at step a) ; 2()(^-) being the speed ratio associated with wind speed vi at step a), A, and v being first weighting coefficients; c) The controller model is calibrated so that for each predetermined wind speed value, the controller model associates the final speed ratio and / or the final blade pitch angle and preferably controls the wind turbine with the final speed ratio and / or the final blade pitch angle.

[0011] Preferably, in step b), the solutions are constrained with the following inequalities:

[0012] ^>f. .

[0013] ^i< . . n ,vî

[0014] 2(^)^+^2(^-)^ ie{l..., Al]

[0015] with œmin the minimum rotation speed, ^max the maximum rotation speed.

[0016] Advantageously, the wind turbine comprises a generator and preferably a gearbox and the measurements include the generator power, the controller model preferably comprising a generator control means and a blade control means.

[0017] Advantageously, the generator control means controls the generator torque TgenQ of the wind turbine by f with Pgen,o the power * eenfi — » oî, & 'igen ratedJ'£ measured generator, ^gen the generator efficiency, the rated rotor speed, Ng the reduction ratio of the reducer.

[0018] Preferably, in step a), the wind speed range is divided into sub-ranges, each sub-range being between a minimum speed and a maximum speed, each predetermined wind speed value corresponding to the average speed of the wind speeds measured in the relevant sub-range, each associated speed ratio corresponding to the average of the speed ratios calculated from the wind speeds in the relevant sub-range and each associated blade orientation angle corresponding to the average of the blade orientation angles for the wind speeds in the relevant sub-range.

[0019] According to one embodiment of the invention, below a predetermined wind speed, the generator control means controls the rotation speed so as to have the final speed ratio and preferably, below the predetermined wind speed, the blade control means controls the blade orientation angle to the final blade orientation angle.

[0020] According to one aspect of the invention, the controller model includes a saturation means which saturates the orientation angle of the blades.

[0021] Preferably, the controller model includes a regulation system around the predetermined wind speed such as: - if the generator control means is active, the regulation system imposes on the blade control means a rotation speed strictly higher, preferably by a first predetermined deviation, than a predetermined rotation speed; - if the blade control means is active, the regulation system imposes on the generator control means a rotation speed strictly lower, preferably by a second predetermined deviation, than the predetermined rotation speed.

[0022] According to one configuration of the invention, when the wind speed is greater than the predetermined wind speed, the rotor equation is linearized -frotor VGB- Tgenfi ​​■ Ng where Jrotor is the moment of inertia of the rotor, ^gb is the efficiency of the gearbox, Ta is the aerodynamic torque of the rotor, l is the speed ratio, [3 is the blade pitch angle, Ng is the gearbox reduction ratio, and Tgen>Q is the generator torque and co is the rotational speed, around a wind speed conditioning flQ = fl(Vq) and , _ and we seek the solution of J^>r-ù=[Ta^ü)+^

[0023] Advantageously, the fatigue state of the wind turbine components and / or the evolution of the wind turbine's performance and / or the control of the wind turbine are determined and / or maintenance actions are planned on the wind turbine, based on the calibrated controller model.

[0024] The invention also relates to a method for calibrating a wind turbine model, to determine the fatigue state of the wind turbine components and / or the evolution of the wind turbine's performance and / or to adapt the wind turbine's control and / or to predict wind turbine maintenance actions, the wind turbine model comprising a controller model and a rotor model, wherein the controller model calibration method is applied according to one of the variants or combinations of variants described above, and before applying step b) of the controller model calibration method, at least the following steps are carried out: 1) modification curves of the lift and drag coefficients of each blade segment are constructed from input curves of the lift and drag coefficients of each blade segment and sets of modification parameters, and modification parameters are determinedfrom the modification curves of the lift and drag coefficients and for each set of modification parameters, from the first maps of power coefficients and thrust coefficient as a function of the speed ratio and the blade pitch angle; 2) for each pair of speed ratio and blade orientation angle associated with one of the said predetermined wind speed values, and for each set of modification parameters, a power coefficient and a thrust coefficient are determined from the first power coefficient and thrust coefficient maps in order to generate comparative curves of power coefficient and thrust coefficient as a function of wind speed; 3) We determine, as final modification parameters, the sets of modification parameters that allow us to minimize the following function J(p): J ( P ) = ( h ) ( CP^e ( P ) ^P^curve ( ) + ^Tfiurve(.Pp P) ^Tfarve^i)} represents the modification parameter sets, Cp£urve( V-) an input power coefficient curve and .un.e(v-) an input thrust coefficient curve, K'i is a first weighting function dependent on the wind speed vi to weight the curve A as a function of the wind speed, is a second Pçurve Weighting function depending on wind speed y. To weight the curve according to wind speed, 0 is a second weighting coefficient. y. corresponds to each of the said predetermined wind speed values, N is the number of predetermined wind speed values, Cp£tirve ( p) the comparative power coefficient curve, Cpcurve( Vj, p) the comparative thrust coefficient curve; 4) The aerodynamic characteristics of the rotor model are calibrated, the aerodynamic characteristics of the rotor model including - calibrated lift and drag coefficient curves corresponding to the modified lift and drag coefficient curves of each blade segment obtained from the final modification parameters, - calibrated curves of power and thrust coefficients corresponding to the comparative curves of power and thrust coefficients obtained from the final modification parameters; - and preferably calibrated maps of power and thrust coefficients corresponding to the power and thrust coefficient maps obtained from the final modification parameters; said calibrated curves of power and thrust coefficients and of lift and drag coefficients serving respectively as power and thrust coefficient curves and of lift and drag coefficient curves in step b), the calibrated maps of power and thrust coefficients serving as power and thrust coefficient maps respectively in step b).

[0025] According to one aspect of the invention, the blades are composed of blade portions and before step b), input maps of power coefficient and thrust coefficient of the wind turbine are determined as a function of the speed ratio and as a function of the orientation angle of the blades from input curves of lift and drag coefficients of each blade portion.

[0026] Advantageously, between step a) and step 1), for each pair of speed ratio and blade orientation angle associated with one of said predetermined wind speed values, a power coefficient and a thrust coefficient are determined from the input power coefficient and thrust coefficient maps so as to generate first comparative curves of power coefficient and thrust coefficient as a function of wind speed and the following steps are applied only if the difference between the first comparative curves and the input curves of power coefficient and thrust coefficient is greater than a predetermined threshold.

[0027] Preferably, in step 1), the following sub-steps are carried out for each portion of the blade: i) we identify the stall angle of the lift coefficient input curve, preferably by calculating the second derivative of the lift coefficient input curve and taking, as the stall angle, the smallest positive angle of attack that makes the second derivative zero; ii) we shift the value of the stall angle and then we construct the modification curve of the lift coefficient so that the slope before the stall angle is identical to that of the entry curve of the lift coefficient;

[0028]

[0029]

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] iii) the modification curve of the drag coefficient is constructed by shifting the input curve of the drag coefficient around a predetermined value, preferably around half the stall angle. According to one aspect of the invention, in step ii), the value of the stall angle is shifted by applying the following first transformation: CL(f(a))=CL(a) f(a)=a + ke-^ a being the angle of attack, CL(a) being the input curve of the lift coefficient, CL(f(a)) being a modified curve of the lift coefficient, f(a) being the first transformation applied to the stall angle, being the stall angle of the input curve of the lift coefficient, k and are first and second parameters among said modification parameters. Advantageously, once the stall angle value of each blade segment has been shifted, the following second transformation is applied: if k>0, f i-\ „ zx / ।, / । ] \ Ljwwx [1+ y CL(at / y if k<0, z. / y / . / ! , / WHQ 1 a being the angle of attack, CL(a) being the entry curve of the lift coefficient, CL(a) being the modified curve of the lift coefficient, ( a) being the modifying curve of the lift coefficient, being the stall angle of the entry curve of the lift coefficient, astaii being the stall angle of the modified curve of the lift coefficient, a2 being a third parameter among said modification parameters. Advantageously, the first parameter k of each blade segment is determined as follows: k0 being a parameter, mmcv is the minimum thickness of the different portions of blades at the trailing edge, cs is the thickness of the blade at the trailing edge of the portion of blade s considered. According to one embodiment of the invention, in step iii), the following third transformation is applied: =C D (a)-g(a) g(a) =l-CD(a) -e^r- with CpCa) the input curve of the drag coefficient, Ca) the modification curve of the drag coefficient, g( a) the third transformation, being the stall angle of the modifying curve of the lift coefficient, 1 and ^3 of the fourth and fifth parameters among the modification parameters, 1 being preferably greater than or equal to -1 and strictly less than 1.

[0043] Preferably, the fourth parameter1 of each blade portion is determined in the following manner:

[0044] / mine,. / = / ov c »

[0045] Zo being a parameter, nùncx is the minimum thickness of the different portions of blades at the trailing edge, cs is the thickness of the blade at the trailing edge of the portion of blade s considered.

[0046] According to an advantageous aspect of the invention, in step 3), the following change of variable is carried out:

[0047] p = b + atanh ( x )

[0048] where a and b are predefined value vectors, etx is the parameter to be optimized.

[0049] Preferably, the fatigue state of the wind turbine components and / or the evolution of the wind turbine's performance and / or the control of the wind turbine are determined and / or maintenance actions are planned on the wind turbine, based on the calibrated wind turbine model.

[0050] The invention also relates to a computer program product downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor or a server, comprising program code instructions for implementing the controller model calibration method or the wind turbine model calibration method according to one of the variants or combinations of variants described above, when said program is executed on a computer, a mobile phone or a computing device. List of figures

[0051] Other features and advantages of the methods according to the invention will become apparent from the following description of non-limiting examples of embodiments, with reference to the figures attached and described below. [Fig 1]

[0052] Fig. 1 represents a first variant of the controller model calibration method according to the invention. [Fig 2]

[0053] Figure 2 represents a second variant of the calibration method for the model controller according to the invention. [Fig 3]

[0054] Fig. 3 represents a first variant of the calibration of the aerodynamic characteristics of the wind turbine model calibration method according to the invention. [Fig 4]

[0055] Fig. 4 represents a second variant of the calibration of the aerodynamic characteristics of the wind turbine model calibration method according to the invention. [Fig 5]

[0056] Fig. 5 represents a step in the construction of the modification curves of lift and drag of the wind turbine model calibration process according to the invention. [Fig 6]

[0057] Figure 6 represents a variant construction of the lift modification curve of the calibration method of a wind turbine model according to the invention. [Fig 7]

[0058] Fig. 7 represents an example of the operation of a controller for the calibration process of the controller model according to the invention. [Fig 8]

[0059] Fig. 8 represents a comparison of an input lift coefficient curve and a modified lift coefficient curve of the calibration process of a wind turbine model according to the invention. [Fig 9]

[0060] Fig. 9 represents a comparison of an input lift coefficient curve, a modified lift coefficient curve and a modifying lift coefficient curve of the method for calibrating the aerodynamic characteristics of the rotor model of a wind turbine according to the invention. [Fig 10]

[0061] Fig. 10 represents a comparison of an input drag coefficient curve and a modification drag coefficient curve of the aerodynamic characteristics calibration method of the rotor model of a wind turbine according to the invention. [Fig 11]

[0062] Fig. 11 represents the evolution of the coefficient k as a function of the radial position of the relevant portion of the blade in the calibration process of the aerodynamic characteristics of the rotor model of a wind turbine according to the invention. [Fig 12]

[0063] Fig. 12 represents comparisons between the input curves, the modification curves of the invention and the prior art curves of power coefficient and thrust coefficient of the same wind turbine. [Fig 13]

[0064] Fig. 13 represents comparisons of the evolution of the rotation speed over time, between the measurements taken on the wind turbine, the rotation speed determined by the calibration method according to the invention and the rotation speed according to the prior art of the same wind turbine. [Fig 14]

[0065] Fig. 14 represents comparisons of the evolution of the blade orientation angle over time, between measurements taken on the wind turbine, the blade orientation angle determined by the calibration method according to the invention and the blade orientation angle according to the prior art of the same wind turbine. [Fig 15]

[0066] Fig. 15 represents comparisons of the input curves, the modification curves according to the invention and the curves according to the prior art, of the thrust coefficient of the same wind turbine. [Fig 16]

[0067] Fig. 16 represents comparisons of input curves, modification curves according to the invention and curves according to the prior art, of power coefficient of the same wind turbine. [Fig 17]

[0068] Fig. 17 represents comparisons of the rotation speed of the wind turbine with a standard controller model according to the prior art and with a controller model calibrated by the calibration method according to the invention. [Fig 18]

[0069] Fig. 18 represents comparisons of the orientation angle of the wind turbine rotor blades with a standard controller model according to the prior art and with a controller model calibrated by the calibration method according to the invention. [Fig 19]

[0070] Figure 19 illustrates an example of an aerodynamic profile of a blade according to the method of the invention. Description of the implementation methods

[0071] The invention relates to a method for calibrating a controller model (or a controller) of a wind turbine comprising a rotor. The controller model is representative of the controller of the actual wind turbine (i.e., the one actually built at a given location).

[0072] The wind turbine rotor comprises one or more blades, and each of these blades is composed of blade segments. Indeed, each blade has a geometry that evolves in depending on the radial position (i.e., according to its length). A blade becomes thinner, in particular, as it moves away from its axis of rotation.

[0073] Generally, the different blades of the same rotor are identical to each other, except for design / manufacturing defects and wear.

[0074] Each portion of the blade has an aerodynamic profile, corresponding substantially to an aircraft wing profile, so as to generate a lift force (that which makes the aircraft fly) and a drag force (which tends to oppose the forward movement of the aircraft).

[0075] Fig. 19 illustrates an example of a blade aerodynamic profile.

[0076] Aerodynamic profiles are generally defined by a leading edge 10 and a trailing edge 11 separated by an intrados 12 and an extrados 13. These airfoil profiles are generally asymmetrical and have a camber designed to generate overpressure on one side of the blade and underpressure on the other. The leading edge 10 is the edge of the profile where the fluid first enters (air in the case of a wind turbine), the large black arrow indicating the direction in which the fluid enters, in the direction of fluid movement relative to the blade (the radius of curvature at the trailing edge 11 is generally minimal), and the trailing edge 11 is the edge where the fluid (air) exits the blade. The intrados 12 is the face opposite the camber (where the overpressure applied by the fluid is located), while the extrados 13 is the face on the side of the camber (where the underpressure occurs).

[0077] The chord, represented in mixed lines, is a fictitious line which joins the leading edge 10 to the trailing edge 11.

[0078] The lift and drag coefficients are aerodynamic coefficients dependent on the airfoil profile (and therefore on the blade portion) that directly reflect an effect on the lift force and the drag force, respectively. On the wind turbine, the lift force tends to rotate the rotor, and the drag force tends to apply a thrust to the wind turbine in a direction orthogonal to the plane of the rotor.

[0079] These coefficients also depend, for a given airfoil, on the angle of attack of the fluid arriving at the blade (here, air, more explicitly, wind). The angle of attack is the angle between the chord of the airfoil (the line from the leading edge to the trailing edge) and the vector of the relative velocity of the fluid with respect to the blade. This relative velocity vector comprises a first component of the wind velocity arriving at the blade, substantially orthogonal to the plane of the rotor, and a second component due to the rotation of the blade. This second component therefore depends on the radial position of the blade portion. Indeed, the further one moves away from the axis of rotation, the greater the velocity of this second component, this velocity being the product of the rotor's rotational speed and the radial position of the blade portion in question.Furthermore, the angle of attack also depends on the angle of orientation of the blades facing the wind, which can be controlled by a means of blade orientation (a motor).

[0080]

[0081]

[0082]

[0083]

[0084] (for example), generally positioned at the blade root, at the point where the blade connects to the hub. The blade orientation, whether more or less facing the wind, also impacts the wind turbine's energy recovery. Using blade orientation controls, it is possible, for example, to position the blades so that they produce little or no energy. This can be useful during storms to prevent damage to the wind turbine. By dividing each blade into blade segments, we can assign different aerodynamic and mechanical characteristics to each blade segment in the multiphysics model (aero-servo-elastic or aero-servo-hydro-elastic model), thus better representing the aerodynamics with the elasticity of the rotor. We can indeed apply the BEM method (BEM stands for "Blade Element Momentum") to determine the aerodynamic coefficients (particularly power and thrust) of the rotor from the blade segments, as a function of the rotational speed, the blade pitch angle, and / or the wind speed. This BEM method is well known to those skilled in the art. It is explained in detail in the following documents: - the page https: / / en.wikipedia.org / wiki / Blade_element_momentum_theory edited on September 26, 2023, at 22:32 (UTC). - Branlard E., « Wind Turbine Aerodynamics and Vorticity-Based Methods: Fundamentals and recent applications - Chapitre 10 - The blade element Momentum (BEM) method, 2017 Pringer, doi: 10.1007 / 978-3-319-55164-7. To present this BEM method in a simplified manner, it first assumes a steady, axisymmetric, incompressible, and inviscid flow for which it formulates the conservation of mass and momentum. The rotor is viewed as an actuator disk that reduces the flow velocity and modifies the pressure field in its vicinity. To calculate the forces applied to the blades, this method then consists of dividing the blades into segments along their radial length (preferably infinitesimal). Depending on the blade's aerodynamic profile, its radial position, the relative wind speed arriving at the blade (and its possible direction), the air density, and the blade's angle of attack, the following are determined for each blade segment: - a lift force defined by: F port = From J ■ Cportanc^} ' R portion - a drag force defined by: Ftrain ~ Def • Ctrai^a^ • R portion where Fport is the lift force applied to the blade portion; Ftrain is the drag force applied to the blade portion; Def^sl is a coefficient that depends on the air density, apparent speed (i.e., the relative velocity vector defined previously), and the chord of the blade portion, Rproportion is the radial length of the blade portion, Cportmcha} ct Ct ■ (a) are the coefficients, respectively of lift and drag, which depend on the angle of attack a.

[0085] In this document, the operator used in the equations is the multiplication operator. In other words, it corresponds to the operator x to indicate multiplication.

[0086] The Def coefficient generally corresponds to the following equation:

[0087] W = 0.5-pW 2 -c^ t

[0088] With P the density of air, W the apparent speed, cportüm the chord of the portion of blade.

[0089] The lift force and the drag force depend on the direction of the apparent wind speed, the lift force being directed in a direction orthogonal to the direction of the apparent wind speed, and the drag force being directed in the direction of the apparent wind speed.

[0090] We can then make a change of reference frame to determine the force generated by each portion of the blade in the circumferential direction (this force causing the rotation of the blade) and the force generated by the portion of the blade in the direction orthogonal to the plane of the rotor (this force generating a thrust on the blade and on the rotor).

[0091] From the forces in the circumferential direction and in the direction orthogonal to the plane of the rotor of the different portions of the blade of the different blades, it is then possible to determine the torque applied to the rotor (from the forces in the circumferential direction and the radial position of each portion of the blade) and the thrust force (sum of the forces in the direction orthogonal to the plane of the rotor of the different portions of the blade of the different blades) applied to the structure.

[0092] Aerodynamic power can also be determined (aerodynamic power is generally greater than the actual power supplied by the generator due to losses) by summing the product of the force of each portion of blade in the circumferential direction by the speed of the portion of blade (itself being the product of the rotational speed of the rotor by the radial position of the portion of blade).

[0093] The method of the invention preferably comprises at least one aero-servo-elastic model, which includes the elastic and aerodynamic properties of the rotor (rotor model), and at least one controller model for controlling the blade pitch and generator torque as a function of wind speed (and possibly other parameters). In other words, the method of the invention comprises an aero-servo-elastic model of the wind turbine including at least one rotor model (the "aero" and "elastic") and a controller model (part "servo") and we calibrate at least the rotor model.

[0094] The method may also include a hydrodynamic part (the "hydro" part), i.e., a hydrodynamic model when the wind turbine is offshore, and more particularly when it is floating. In this case, the model is an aero-servo-hydro-elastic model.

[0095] The calibration process is based on measurements taken on the wind turbine.

[0096] The measurements include at least the wind speed, the blade pitch angle, and the rotation speed over time. Thus, at any given instant, a wind speed can be related to the blade pitch angle and the rotation speed at the same instant.

[0097] The measurements can be carried out by different means of acquisition (also called means of measurement), for example an anemometer or a LIDAR sensor (for "Light and Detection And Ranging" meaning light, detection and range measurement) for the measurement of speed, an angle measurement sensor for the angle of orientation of the blades and a sensor for measuring the speed of rotation.

[0098] In this description, unless otherwise specified, rotational speed refers to the rotational speed of the rotor. The rotational speed of the generator is related to the rotational speed of the rotor according to a reduction ratio of the gearbox (when the wind turbine does not include a gearbox, the rotational speed of the rotor corresponds to the rotational speed of the generator).

[0099] Preferably, the measurements can be averaged measurements over a certain predetermined period, for example over 10 minutes.

[0100] The measurements may in particular come from the SCADA (for "Supervisory, Control And Data Acquisition" in English meaning system, control, monitoring and data acquisition) of the wind turbine.

[0101] According to one embodiment of the invention, the calibration process may include a preliminary step of data measurements on the (actual) wind turbine, the measurements including at least the wind speed, the angle of orientation of the blades, the speed of rotation over time, the measurements being able to be obtained by different means of measurement, such as an anemometer, a LIDAR sensor, an angle sensor, an angular position sensor etc.

[0102] The blade pitch angle reflects the angle at which the blades are tilted relative to the wind approaching the wind turbine. Indeed, depending on this pitch angle, more or less wind energy can be captured. This blade pitch angle is controlled by a blade orientation mechanism. For example, when the wind is too strong, such as during a storm, the blades can be oriented so that they capture as little energy as possible, thus preventing damage to the wind turbine.

[0103] Alternatively, the measurements can be instantaneous measurements.

[0104] According to the method of the invention, at least the following steps are carried out: a) Based on the measurements taken, predetermined wind speed values ​​are associated with: - a speed ratio X defined by = ~ with œ the rotational speed of the rotor, R the radius of the rotor and v the wind speed; - and a blade orientation angle.

[0105] To associate predetermined wind speed values, each with a speed ratio and a blade orientation angle, consideration is given to the possibly different measurement times between the different acquisition means, so that the associations link the wind speed, the rotation speed and the blade tilt angle at the same instant.

[0106] Thus, from the measurements, a torque can be defined as the ratio of the speed and the angle of inclination of the blades for each wind speed. This torque reflects the measured operation of the wind turbine as a function of wind speed.

[0107] In the context of the invention, the "speed ratio" is defined as the ratio between the blade speed at the outer tip and the wind speed arriving at the blade. The blade speed at the outer tip is the product of the rotational speed (of the rotor) and the rotor radius. This speed ratio is known as TSR, for "Tip Speed ​​Ratio." This speed ratio (TSR) is a key dimensioning parameter of the wind turbine and is well known to those skilled in the art. It can be determined from measured data, as is the case for the associated speed ratio defined in step a), or it can be defined otherwise, particularly using equations to optimize the model or the performance of the wind turbine.

[0108] b) for each predetermined value of wind speed, the values ​​of the final speed ratio and the final blade orientation angle are determined as solutions to the following minimization problem: mm.( CP( / (r,), p(r,) ) -CPm( ,,) )2 + A(CT(i(r,). p () ) - Cr,„( v,) )2 + p(n) ( fl( vt) -(r,) )2 + ,(vf) ( À ( v,)v,- f ,(v,)v()2 a VCC ^(L') ) a power coefficient map of the wind turbine, Cr(x(v,-), pi vJ ) a thrust coefficient map of the wind turbine, the predetermined wind speed values, N being the number of predetermined wind speed values, Cpfiurve(Vj ) being a power coefficient curve, CT£urve(vî) being a thrust coefficient curve, Piy,) the blade pitch angle at wind speed vi ; P(Vi) being the speed ratio to wind speed vi ; being the orientation angle associated with wind speed vi at step a) ; ^o( ) being the speed ratio associated with the wind speed vi at step a), A, and v being the first weighting coefficients.

[0109] The term "mapping" refers to a representation (a map) of the evolution of a parameter (the power coefficient or the thrust coefficient, for example) as a function of two parameters (in our case, as a function of the speed ratio and the blade pitch angle). On this representation, the speed ratio can therefore be on the x-axis and the blade pitch angle on the y-axis (or conversely, the speed ratio on the y-axis and the blade pitch angle on the x-axis), and each point corresponds to the y-coordinate and the x-coordinate to the parameter value (the power coefficient or the thrust coefficient): for example, the different values ​​of the power or thrust coefficients can be represented on the representation by curves with equal values.

[0110] c) The controller model is calibrated so that for each predetermined wind speed value, the controller model associates the final speed ratio and the final blade pitch angle, and preferably, it controls the (actual) wind turbine with the final speed ratio and / or the final blade pitch angle. In other words, the actual wind turbine is controlled using the final speed ratio and / or the final blade pitch angle.

[0111] The controller model often implemented in "digital twin" type models is a standardized (generic) controller model because wind turbine manufacturers generally do not provide the controller details to operators. However, a correct controller model improves the accuracy of the model on the one hand and improves the control of the real wind turbine on the other when attempting to control the real wind turbine using the model's output data.

[0112] Using the measurements taken on the wind turbine allows the standard controller model initially used in the model to be modified, improving it in relation to the measured data. The measurements thus improve the model's accuracy.

[0113] Furthermore, in the minimization problem above, we seek to verify the difference between the solutions found for the speed ratio and the blade angle, relative to these associated values, these associated values ​​being derived from in-situ measurements on the wind turbine. Integrating these differences into the optimization problem above makes it possible to find solutions that are as close as possible to the measured values, in order to improve the accuracy of the model and its consistency with the actual wind turbine.

[0114] Thus, the controller model is representative of the (actual) controller of the real wind turbine with good accuracy, thanks to calibration. This calibrated model can be used to determine the fatigue state of the components of the real wind turbine, the evolution of performance, or to adapt the control of the real wind turbine in order to improve performance, and to predict maintenance actions on the real wind turbine.

[0115] The wind turbine model calibration process can in particular be implemented by computer means, such as a computer or a server.

[0116] Thanks to the method, it is then possible to predict maintenance actions on the actual wind turbine, modify the control of the actual wind turbine (i.e., its servo system), or determine the fatigue state of different components of the actual wind turbine, more precisely. In other words, the method can be used to determine maintenance actions, modify the control of the actual wind turbine, or determine the fatigue state of its components. The method may also include a maintenance step on the actual wind turbine, and / or a control modification step on the actual wind turbine based on the output data of the model calibrated from measurements on the actual wind turbine.

[0117] The wind turbine model includes at least one aero-servo-elastic model, which includes the elastic and aerodynamic properties of the rotor (rotor model), "aero" and "elastic" parts, and a controller model ("servo" part) of the wind turbine to control the pitch of the blades and the torque of the generator as a function of the wind speed (and possible other parameters), so as to faithfully reproduce the behavior of the real wind turbine.

[0118] The model may also include a hydrodynamic "hydro" part (a hydrodynamic model) when the wind turbine is offshore, and more particularly when it is floating. In this case, the wind turbine model is an aero-servo-hydro-elastic model.

[0119] Fig. 1 illustrates, schematically and without limitation, a first variant of the calibration method for a wind turbine controller model according to the invention.

[0120] From measurements Mes taken on the wind turbine, a speed ratio and a blade pitch angle can be associated with different wind speeds. Thus, from the measurements, it is known that for each wind speed, the wind turbine provides a pair of associated data Mes-ass, which includes a speed ratio from the measurements and a blade pitch angle from the measurements.

[0121] From this pair of associated data (Mes-ass), the power coefficient curve Cc_P, the thrust coefficient curve Cc_T, and the power coefficient maps Cart_c_P and thrust maps Cart_c_T, we can find solutions for the speed ratio and blade pitch angle in such a way as to limit both the deviations from the power coefficient curves Cc_P and thrust curves Cc_T, and the deviations from the associated data curves (Mes-ass) derived from the measurements. Thus, we optimize Opt_Cont to minimize these deviations. At the output of the Opt_Cont optimization, we obtain, for different wind speeds, the optimal values ​​for the speed ratio and blade pitch angle, which then become the final speed ratios Ld and final blade pitch angles Bd. We can then use this data to control (servo) the Contlnv wind turbine via the controller model.

[0122] We can then control the orientation angle of the blades BData on the actual wind turbine and / or the speed ratio LData on the actual wind turbine.

[0123] Preferably, in step b), the solutions can be constrained with the following inequalities:

[0124]

[0125] ^L< . . n [°U 6 1 ie{l, .... Nl)

[0127] N being the number of wind speeds considered, 2(Frétant the ratio of speed to wind speed v«, with (Omin the minimum rotation speed, the maximum rotation speed.

[0128] The third inequality ensures a growth in the rotational speed of the rotor and implicitly a continuity of the trajectory which facilitates the resolution of the optimization problem.

[0129] The first and second inequalities make it possible to ensure that the rotation speed is between the minimum rotation speed and the maximum rotation speed, for all wind speeds, so as to ensure the constraints of the generator, in particular in the case of a double induction generator.

[0130] Advantageously, the wind turbine may include a generator capable of converting the energy recovered by the rotor into electrical energy and preferably a gearbox with a reduction ratio. The gearbox thus allows the rotor to operate at an optimal rotational speed for energy recovery and the generator to operate at a generator rotational speed different from the rotor rotational speed. Therefore, the gearbox is a component that connects the rotor to the generator, for example, via drive shafts on either side of the gearbox.

[0131] In such a configuration, the measurements may also include the generator power, and the controller model may preferably include a generator control means (or generator controller) and a blade control means (or blade controller).

[0132] The generator control means allows control of the rotational speed of the generator (or of the rotor via the reducer and the reduction ratio) and / or the torque of the generator.

[0133] The blade control means allows the angle of orientation of the blades facing the wind to be controlled. In other words, it controls (servos) the motor or similar equipment, which positions the blades more or less facing the wind.

[0134] According to one embodiment of the method of the invention, the generator control means can control the torque TgetKQ of the wind turbine generator by T _ p^nf> with Pgenfl the measured power of the generator, ^gen the generator efficiency, rated rotor speed, Ng the reduction ratio of the reducer.

[0135] Thus, the rated power of the generator can be ensured when the rotational speed of the rotor is at its rated rotational speed.

[0136] According to one configuration of the invention, in step a), a range (i.e. a range) of wind speeds can be divided into sub-ranges, each sub-range being between a minimum speed and a maximum speed, each predetermined value of wind speed corresponding to the average speed of the wind speeds measured in the relevant sub-range, each associated speed ratio corresponding to the average of the speed ratios calculated from the wind speeds and rotation speeds measured in the relevant sub-range, and each associated blade orientation angle corresponding to the average of the blade orientation angles measured for the wind speeds in the relevant sub-range.

[0137] For example, each sub-range can be considered as an interval V of wind speeds, for which v is the minimum speed and vi is the maximum speed of the interval considered, i.e., y. — |y J, i varying from 1 to Nl, N is the number of

[0138]

[0139]

[0140]

[0141]

[0142] predetermined wind speed values. For each interval V of wind speeds, the speed ratio y, the wind speed y, and the blade pitch angle are estimated as follows: ^î~ y. — Np vf j Where Vf are the indices of the measurements taken on the wind turbine whose average wind speed is within the interval Vh Wj, and are respectively the rotor speeds, wind speeds and blade pitch angles measured in the relevant interval, Np is the number of measurements taken in the relevant interval and R is the radius of the rotor.

[0143] In other words, the speed ratio, the wind speed jk and the angle The blade orientations correspond to the average of the measured speed ratios. to the average of the measured wind speeds, and to the average of the blade orientation angles measured in each interval.

[0144] This data allows us to have a trajectory of speed ratio and angle of orientation of the blades from the data measured on the wind turbine.

[0145] By projecting this new trajectory onto the power and thrust coefficient maps, comparative curves of the power and thrust coefficients are obtained and these comparative curves are generally significantly different from the input curves of the power and thrust coefficients provided by the manufacturer.

[0146] According to one configuration of the invention, below a predetermined wind speed, the generator control means can control the rotational speed to achieve the final speed ratio. In other words, at a given wind speed, the generator of the (actual) wind turbine is controlled (controlled) at a rotational speed such that the rotor speed corresponds to the product of the final speed ratio and the given wind speed, divided by the radius of the wind turbine. Furthermore, it is known that the rotor speed can be converted to the generator speed via the reduction ratio.

[0147] When the wind turbine does not include a gearbox (for example for a so-called "Direct-drive" wind turbine, i.e. for which the generator is directly connected to the rotor or connected to the rotor only by a shaft, without a gearbox), a reduction ratio of 1 can be taken.

[0148] In this configuration, below the predetermined wind speed, the blade control means can preferably control (servo) the blade orientation angle to the final blade orientation angle.

[0149] The generator control means can be used simultaneously with the blade control means or alternately: for example, it can be arranged that when one of the two is active, the other is inactive.

[0150] Fig. 7 illustrates, schematically and not in a limiting way, an example of controlling the torque Tor of the wind turbine generator as a function of the wind speed V_wind.

[0151] In Part I, when the wind speed is less than a minimum wind speed vn, the wind turbine is not operating (it does not produce electricity). In Part IV, when the wind speed is greater than a maximum wind speed vout, the wind turbine is not operating (it does not produce electricity): it may, for example, rotate at idle speed.

[0152] In Part II, when the wind speed is between the minimum wind speed vin and a predetermined nominal wind speed vnom, the generator is torque-controlled. In other words, the controller model controls the torque of the generator and this is increasing between a minimum torque Tin at the minimum wind speed vin and a maximum torque Tnom at the predetermined nominal wind speed ^nom*

[0153] On part III, when the wind speed is between the predetermined nominal wind speed vnom and the maximum wind speed vout, the generator maintains the maximum torque Tnom and the wind turbine is then controlled by the angle of orientation of the blades.

[0154] According to an advantageous embodiment of the invention, the controller model may include a saturation means that saturates (it is said that it saturates "from below") the blade pitch angle. For a considered wind speed lower than the predetermined wind speed, the controller model can send the saturation means the final blade pitch angle for the considered wind speed, and the saturation means can then saturate the blade pitch angle at this final blade pitch angle (at the considered wind speed), thereby ensuring that this value is maintained.If the wind speed in question is greater than the predetermined wind speed, the blade control system ensures that the rotation speed is regulated to the final rotation speed (i.e., the rotation speed that allows the final speed ratio to be obtained), which results in the blade pitch angle being set to the final blade angle and the generator torque being imposed at its nominal value.

[0155] Preferably, the controller model may also include a regulation system around the predetermined wind speed: - If the generator control system is active, the control system can then fictitiously impose a rotational speed signal on the blade control system that is strictly higher, preferably by a predetermined initial deviation, than the nominal rotational speed. In other words, the control system sends (false) information to the blade control system. Indeed, for the known wind speed, when the rotational speed is too high, the blade control system tends to decrease the blade pitch angle until it reaches zero. This false information avoids controlling both the blade control system and the generator control system simultaneously. - If the blade control means is active, the control system can impose on the generator control means a rotational speed strictly lower, preferably by a predetermined second deviation, than the nominal rotational speed. In other words, the control system sends (false) information to the generator control means. Indeed, for the known wind speed, when the rotational speed is too low, the generator control means tends to increase the rotation speed and thus saturate the generator's torque to its maximum value. This false information avoids controlling both the blade control system and the generator.

[0156] Figure 2 illustrates, schematically and without limitation, a second variant of the calibration process of a wind turbine controller model according to the invention.

[0157] To control the Contlnv wind turbine, the controller model includes a generator control means ContG, a blade control means ContB, a saturation means SatB and a regulation system Regul.

[0158] From the associated Mes-ass data pair, which includes, for each wind speed, a speed ratio derived from measurements and a blade pitch angle derived from measurements, the blade pitch angle can be controlled by the blade controller ContB and / or the generator rotation speed (or torque) can be controlled by the generator controller ContG. The SatB saturation means can saturate (from below) the blade pitch angle.

[0159] The Regul control system can also be used, at the predetermined wind speed and its surroundings, to prevent the blade control means ContB and the generator control means ContG from being active simultaneously. To this end, when the generator control means ContG is active, the Regul control system can send a deliberately erroneous signal indicating a rotational speed higher than the nominal speed for the given wind speed. This has the effect of reducing the blade pitch angle to the predetermined saturation value (referred to as bottom-up saturation), which results in the inhibition of the blade control means ContB. Similarly, when the blade control means ContB is active, the Regul control system can send a deliberately erroneous signal indicating a rotational speed lower than the nominal speed for the given wind speed.This results in the generator torque being at its maximum value. The dashed arrows from the Regul control system to the blade control unit ContB and the generator control unit ContG illustrate the intentionally erroneous information sent by the Regul control system.

[0160] The output data allows control (servo) of the orientation angle of the blades BData of the (real) wind turbine and / or the speed ratio LData of the (real) wind turbine.

[0161] According to one embodiment of the invention, when the wind speed is greater than the predetermined wind speed, the rotor equation can be linearized as Jrotor' = (2, P)nGB-Tgenfl Ng, where Jrotor is the moment of inertia of the rotor, Tg is the efficiency of the gearbox, Ta is the aerodynamic torque of the rotor, Ta is the speed ratio, Tg is the blade pitch angle, and Ng is the reduction ratio of the gearbox. and Tgen$ the nominal torque of the generator and co the rotational speed, around a wind speed conditioning Pn = P ( v0 ) and , œratafR and we seek the solution of A) ~ J rotor • a> = [ Ta ( ) + if ( B ( W ' ^rated ) + B ( V ' V0 ) ) ] ' TgenS ' ' ^inSh 011 It is possible to linearize around different wind speeds (vo) to understand the system dynamics at various operating points and thus find the controller model parameters that allow for optimal system regulation at each operating point. These controller model parameters are then called "gain-scheduled" (meaning programmed gain), that is, they vary according to a setting parameter (the "scheduling" parameter), which could be the wind speed, for example.

[0162] The invention relates to a computer program product downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor or a server, comprising program code instructions for implementing the wind turbine model calibration method according to one of the variants or combinations of variants described later or for implementing the wind turbine controller model calibration method according to one of the variants or combinations of variants described previously, when said program is executed on a computer, a mobile phone or a computing device.

[0163] The invention also relates to a method for calibrating a wind turbine model representative of a real wind turbine, i.e. actually built at a given location (the wind turbine model comprising at least one rotor model and the controller model already described), in which the calibration method for the controller model is applied according to one of the variants or combinations of variants described previously.

[0164] According to the wind turbine calibration method, and before applying step b) of the controller model calibration method, at least the following steps are carried out: 1) Modifying curves of the lift and drag coefficients of each blade segment are constructed from input curves of the lift and drag coefficients of each blade segment and sets of modification parameters. Indeed, the applicant observed that, generally, applying the BEM method to the input lift and drag curves (which may be provided, for example, by the wind turbine manufacturer, or generated by a blade scan or CFD modeling) to reconstruct second curves of the power and thrust coefficients does not allow for an exact retrieval of the input curves of the power and thrust coefficients, which are generally provided by the wind turbine manufacturer, thus generating inaccuracies in the model predictions. The second curves generally differ from the input curves. One can On the contrary, we observe significantly large differences on one or both of the second power and thrust curves, compared to their respective input curves. Then, using the modified lift and drag coefficient curves and for each set of modification parameters, we determine initial power and thrust coefficient maps as a function of the speed ratio and blade pitch angle, for example, using the BEM method. The use of modification parameter sets allows us to test different types of modifications for subsequent stages. 2) for each pair of speed ratio and blade orientation angle associated with one of the said predetermined wind speed values, and for each set of modification parameters, a power coefficient and a thrust coefficient are determined from the first power coefficient and thrust coefficient maps in order to generate comparative curves of power coefficient and thrust coefficient as a function of wind speed.

[0165] By constructing these modifying lift and drag curves, we seek to slightly modify the input lift and drag curves of each blade portion, to generate different comparative power and thrust coefficient curves, that is to say, to reconstruct power and thrust coefficient curves closer to the respective input curves.

[0166] Testing different sets of parameters allows testing different configurations.

[0167] Constructing modification curves from modification parameter sets is very useful because it improves the accuracy of the comparative curves of the power and thrust coefficients to the respective input curves, based on data measured on the wind turbine. This construction therefore improves the accuracy of the model and, where applicable, improves the control of the (real) wind turbine when the control (servo system) is based on the output data of this model. 3) we retain (determine or define) as final modification parameters, the sets of modification parameters allowing minimization of the following function J(p): ■ / (p) “ 1 h) (( h' P) -67^^.(^) ) P) ' ^T.cun:e ( h" ) ) represents the sets of modification parameters; wi is a first weighting function dependent on the wind speed y. To weight the Cp curve as a function of the wind speed, w2 is a second weighting function dependent on the wind speed y. To weight the Crcurv curve as a function of the wind speed, 6 is a second weighting coefficient to weight the deviation of Cp.

[0168] Cp£urve P31 raPP°rt a deviation from Cpmrve corresponds to each of the said predetermined wind speed values, N is the number of predetermined wind speed values, Cp£urve(p) the comparative power coefficient curve obtained in step 2), Cp£urve( p) the comparative thrust coefficient curve obtained in step 2), Cpmrve( v-) an input power coefficient curve and Cpairve(vj an input thrust coefficient curve. Indeed, by searching for the set of modification parameters that minimize the function J(p), we are actually minimizing the differences between the input curves and the comparative curves of the power and thrust coefficients, respectively. The final modification parameters thus allow us to minimize these differences and consequently find the power and thrust modification curves that are closest to their respective input curves. 4) We calibrate the aerodynamic characteristics of the rotor model, the aerodynamic characteristics of the rotor model being defined as the modification curves of the lift and drag coefficient of each portion of the blade obtained from the final modification parameters and / or as the comparative curves of the power and thrust coefficients obtained from the final modification parameters and / or as the first maps of the power and thrust coefficients obtained from the final modification parameters.

[0169] Thus, the aerodynamic characteristics of the rotor model may include: - calibrated lift and drag coefficient curves corresponding to the modification curves of the lift and drag coefficients of each blade portion obtained from the final modification parameters, and / or - calibrated curves of power and thrust coefficients corresponding to the comparative curves of power and thrust coefficients obtained from the final modification parameters; - and preferably calibrated maps of power and thrust coefficients corresponding to the first maps of power and thrust coefficients obtained from the final modification parameters.

[0170] The calibrated curves of power and thrust coefficients and of lift and drag coefficients then serve respectively as power and thrust coefficient curves and of lift and drag coefficient curves in step b). Similarly, the calibrated maps of the power and thrust coefficients serve as power and thrust coefficient maps respectively in step b).

[0171] In other words, the final modification parameters determined in step 3) allow the modification curves of lift and drag, the comparative curves of power and thrust and / or the first maps of the power and thrust coefficients to be retained, which become calibrated curves and calibrated maps.

[0172] The use of modifying lift and drag curves on each blade segment thus improves the accuracy of the calibrated curves and maps of the input curves. This allows the model to be improved while limiting the measurements taken on the wind turbine.

[0173] By calibrating both the aerodynamic characteristics of the rotor model and the controller model, the accuracy of the model can be significantly improved.

[0174] Thus, the wind turbine model accurately represents the actual wind turbine, thanks to the calibration of the aerodynamic characteristics and the controller model. This calibrated model can be used to determine the fatigue state of the actual wind turbine components, the evolution of performance, or to adapt the control of the actual wind turbine in order to improve performance and predict maintenance actions on the actual wind turbine.

[0175] Thanks to the method, it is then possible to predict maintenance actions on the actual wind turbine, modify the control of the actual wind turbine (i.e., its servo system), or determine the fatigue state of different components of the actual wind turbine, more precisely. In other words, the method can be used to determine maintenance actions, modify the control of the actual wind turbine, or determine the fatigue state of its components. The method may also include a maintenance step on the actual wind turbine, and / or a control modification step on the actual wind turbine based on the output data of the model calibrated from measurements on the actual wind turbine.

[0176] The wind turbine model calibration process can in particular be implemented by computer means, such as a computer or a server.

[0177] The wind turbine model calibration process is thus based on measurements taken on the wind turbine and on input curves of the lift and drag coefficients for each blade segment, as well as input curves of the rotor power and thrust coefficients as a function of wind speed. Indeed, each blade segment is defined by a geometry that directly impacts the lift and drag coefficients of each segment. The input curves of rotor power and thrust as a function of wind speed are generally provided by the wind turbine manufacturer and correspond more or less to theoretical curves or curves of the rotor in its "new" state. In fact, time can alter the profile. aerodynamics of the blade sections, which can alter the lift and drag coefficients, and therefore the power and thrust coefficients of the rotor.

[0178] [Fig.3] illustrates, schematically and without limitation, a first variant of the wind turbine model calibration method according to the invention.

[0179] The process has the following input data: - Measurements of rotation speed, wind speed and blade pitch angle taken on the wind turbine, - Inlet curves of lift coefficient C0_port and drag coefficient C0_train of each blade segment; - Input curves of the power coefficient C0_P and thrust coefficient C0_T of the wind turbine.

[0180] The input curves of the power coefficient C0_P and thrust coefficient C0_T are generally provided by the wind turbine manufacturer.

[0181] The input curves of the lift coefficient C0_port and drag coefficient C0_train of each blade portion can be provided by the wind turbine manufacturer or constructed by scanning the blades or by numerical simulations, for example in CFD (CFD for "Computational Fluid Dynamics" in English meaning .numerical fluid dynamics) from the geometries of the blade portions, in particular the aerodynamic profiles of the blade portions.

[0182] From the Mes measurements, Mes-ass data are associated, which include a speed ratio and a blade pitch angle. This reflects the operation of the wind turbine (the control applied to the wind turbine) as a function of the measured wind speed.

[0183] We construct Cons of the modification curves of lift coefficient Cl_port and drag coefficient Cl_train, based on the input curves of the lift coefficients C0_port and drag coefficient C0_train and of the sets Jp of modification parameters.

[0184] These modification curves for the lift coefficient Cl_port and the drag coefficient Cl_train are then used to determine Detl for the power coefficient maps Cart_P and thrust coefficient maps Cart_T. The BEM method can be applied, for example, for this purpose. The maps provide values ​​for the power or thrust coefficients as a function of, on the one hand, the speed ratio and, on the other hand, the blade pitch angle, which are two control parameters of the wind turbine.

[0185] From the maps thus determined, comparative power curves C1_P and thrust curves C1_T, dependent on wind speed, can be established. These curves therefore depend on the choices made regarding the speed ratio control and the blade pitch angle actually applied for each wind speed.

[0186] We then seek to minimize Optl the difference between the comparative power curves C1_P and thrust curves C1_T and the input curves C0_P and C0_T as a function of the modification parameter sets Jp. Finally, we retain the parameter set of modifications minimizing these discrepancies which are defined as the final parameters of Jp_opt modifications.

[0187] When these final Jp_opt modification parameters are defined, the aerodynamic characteristics of the rotor model can be calibrated as corresponding to the following calibrated curves and maps:

[0188] - the calibrated curves of lift coefficients Cc_port and drag coefficients Cc_train of each blade segment corresponds to the modification curves of the lift coefficient Cl_port and drag coefficient Cl_train of each blade segment obtained from the final modification parameters Jp_opt, and / or - the calibrated curves of power coefficients Cc_P and thrust coefficients Cc_T corresponding to the comparative curves of power coefficients C1_P and thrust coefficients C1_T obtained from the final modification parameters Jp_opt; - and preferably the calibrated maps of power coefficients Cart_c_P and thrust coefficients Cart_c_T corresponding to the maps of power coefficients Cart_P and thrust coefficients Cart_T obtained from the final modification parameters Jp_opt.

[0189] The calibrated curves and maps can then be used for the controller model calibration process.

[0190] According to one embodiment of the method of the invention, the blades (each blade) can be composed of several blade segments, and prior to step 1), input maps of the power coefficient and thrust coefficient of the wind turbine can be determined as a function of the speed ratio and the blade pitch angle from input curves of the lift and drag coefficients of each blade segment. Indeed, thanks to the BEM method, these maps can be established directly from the input curves of the lift and drag coefficients of the blade segments for different values ​​of the speed ratio and the blade pitch angle (also called the blade tilt angle).

[0191] Advantageously, between step a) and step 1), for each pair of blade speed ratio and pitch angle associated with one of said predetermined wind speed values, a power coefficient and a thrust coefficient can be determined from the input power coefficient and thrust coefficient maps so as to generate initial comparative curves of power coefficient and thrust coefficient as a function of wind speed, and the subsequent steps of the process can be applied only if the difference between the initial comparative curves and the input power coefficient and thrust coefficient curves is greater than a predetermined threshold. This configuration is advantageous because if the difference between each comparison of curves is less than the predetermined threshold, the construction steps of the Modifying curves are used because the input curves for the lift and drag coefficients are considered sufficiently accurate and reliable. Conversely, if at least one difference between each curve comparison exceeds the predetermined threshold, the difference is considered significant, and the lift and drag curves can be modified to improve the model's accuracy.

[0192] Fig. 4 illustrates, schematically and not in a limiting manner, a second variant of the method for calibrating the aerodynamic characteristics of the rotor model of a wind turbine according to the invention.

[0193] References identical to those in [Fig.3] correspond to the same elements and will not be detailed again.

[0194] Fig.4 differs from Fig.3 in that from the input curves of the lift coefficient C0_port and drag coefficient C0_train of each blade portion, C_ent is determined of the input maps of the power coefficient Cart_E_P and thrust coefficient Cart_E_T depending on the speed ratio and the angle of orientation of the blades.

[0195] Then, from these maps, we determine Det3 the first comparative curves of power coefficients Cl 1_P and thrust coefficients Cl 1_T dependent on wind speed.

[0196] We then compare Comp these first comparative curves of power coefficient Cl 1_P and thrust coefficient Cl 1_T respectively to the input curves of power coefficient C0_P and thrust coefficient C0_T and we construct Cons of the modifying curves of lift coefficient Cl_port and drag coefficient Cl_train only if at least one of these comparisons exceeds a predetermined threshold.

[0197] According to an implementation of the method according to the invention, in step 1), the following sub-steps can be carried out for each portion of the blade: (i) the stall angle of the lift coefficient input curve can be identified, preferably by calculating the second derivative of the lift coefficient input curve and taking, as the stall angle, the smallest positive angle of attack that makes the second derivative zero; ii) the value of the stall angle can be shifted, for example by a first predetermined step, then the modification curve of the lift coefficient can be constructed so that the slope before the stall angle is identical to that of the entry curve of the lift coefficient; iii) the drag coefficient modification curve can be constructed by shifting the input drag coefficient curve, for example by a second predetermined step, around a predetermined value, preferably around half the stall angle.

[0198] The stall angle corresponds to the angle of attack from which the lift coefficient drops: it therefore reflects a loss of lift. This loss of lift generates a Aerodynamic stall, meaning that the air surrounding the blade no longer follows the blade's profile, which generates strong instabilities, resulting in a loss of lift.

[0199] This implementation is particularly advantageous. Indeed, the stall angle is a very influential parameter of the lift coefficient curves. Moreover, this stall angle can be imprecise. In fact, its value can be modified depending on the condition of the blade portion (clean or dirty, presence of scratches or other defects, for example).

[0200] Testing different offsets of the stall angle can therefore make it possible to modify the lift coefficient input curve in a relevant way, while generally preserving the other parameters of the curve.

[0201] Furthermore, maintaining the slope upstream of the stall angle is relevant because this slope is very influential on the curves and on the maps of power and thrust coefficients and because this slope changes little during the service time of the wind turbine.

[0202] Furthermore, by modifying the drag curve around a predetermined value, the most sensitive part of the curve can be modified, for example, the portion of the curve within the main operating range. Half the stall angle is a particularly relevant value because blades are generally operated at an angle of attack between 0° and the stall angle.

[0203] Fig. 5 illustrates, schematically and not in a limiting manner, a step in the construction of the modifying lift and drag curves of the method for calibrating the aerodynamic characteristics of the rotor model of a wind turbine according to the invention.

[0204] From each input curve of lift coefficient C0_port (of each portion of blade), we identify Id a stall angle Ang_d from which we construct Cons2 of the modification curves of the lift coefficient Cl_port (of the portion of blade concerned), each modification curve of the lift coefficient Cl_port depending on a set Jp of modification parameters.

[0205] From each input curve of the drag coefficient C0_train and the stall angle determined Ang_d, one can construct Cons3 modifying curves of the drag coefficient Cl_train, for example by shifting the input curve C0_train around a predetermined value corresponding preferably to half of the stall angle Ang_d.

[0206] The input curves of the power coefficient C0_P and thrust coefficient C0_T can also be used if necessary to assist in these constructions.

[0207] Advantageously, in step ii), the value of the stall angle can be shifted, for example by a predetermined first step, by applying the following first transformation:

[0208] Q( / (a))=Q(«)

[0209] / (a)=a + ^^

[0210] a being the angle of attack of the wind on the portion of the blade, CL(a) being the input curve of the lift coefficient of the portion of the blade concerned, CL ( f ( a ) ) being a modified curve of the lift coefficient of the portion of the blade concerned, f ( a ) being the first transformation applied to the stall angle, being the stall angle of the input curve of the lift coefficient of the portion of the blade concerned, k and are first and second parameters among said modification parameters.

[0211] The first parameter k corresponds to the first predetermined step, that is, to the offset value of the dropout angle. If the first parameter k is negative, the dropout angle resulting from the first transformation is less than the dropout angle of the input curve; conversely, if the first parameter k is positive, the dropout angle resulting from the first transformation is greater than the dropout angle of the input curve.

[0212] The second parameter represents the amplitude over which the angles of attack are modified (to adapt to the offset of the stall angle).

[0213] By this first transformation, we therefore shift the stall angle of each lift curve (of each portion of blade) and we slightly modify the curve around the value of the shifted stall angle to adapt to it.

[0214] Fig. 8 illustrates, schematically and not in a limiting way, an example of the first transformation of the lift coefficient curve of a portion of a blade.

[0215] The curves illustrate the evolution of the lift coefficient C_port as a function of the angle of attack Alpha of the blade portion.

[0216] The stall angle Ang_d is identified on the input curve of the lift coefficient of the relevant portion of the blade C0_port.

[0217] To construct the modified C0_mod_port curve of the lift coefficient, the position of the initial stall angle is shifted to the modified value Ang_d2 and the curve is modified only around this value so that the modified lift coefficient curve C0_mod_port is modified only around the stall angle Ang_d.

[0218] It is noted that during this transformation, the slope of the modified lift coefficient curve C0_mod_port upstream of the stall angle Ang_d2 differs from the slope of the input lift coefficient curve C0_port upstream of the stall angle Ang_d.

[0219] Preferably, once the stall angle value of each blade segment has been shifted, the following second transformation can be applied:

[0220] if k>0, (, xr ( \ \ \ C£^(a)-CL(fl0^1+^^^ ,2 J

[0221] if k<0, p { jp, . / 1 / QùW _ . \ \ LL / mW \ 1+ \ CL(âslM) ' /

[0222] a being the angle of attack of the wind on the portion of the blade, CL(a) being the input curve of the lift coefficient of the portion of the blade concerned, CL ( a) being the modified curve of the lift coefficient of the portion of the blade concerned, Cj JWW ( a ) being the modification curve of the lift coefficient of the portion of the blade concerned, astM being the stall angle of the input curve of the lift coefficient of the portion of the blade concerned, being the stall angle of the modified curve of the lift coefficient of the portion of the blade concerned, ^2 being a third parameter among the said modification parameters.

[0223] This second transformation makes it possible to preserve the slope before the stall angle that the entry curve of the lift coefficient of the relevant blade portion had. Indeed, this slope is very influential and highly dependent on the aerodynamic profile of the relevant blade portion.

[0224] This second transformation, by preserving the slope and taking into account the offset of the stall angle, therefore modifies the value of the lift coefficient at the level of the modified offset angle.

[0225] Fig. 9 illustrates, schematically and not in a limiting way, an example of a second transformation of the lift coefficient curve of a portion of a blade.

[0226] The curves illustrate the evolution of the lift coefficient C_port as a function of the angle of attack Alpha of the blade portion.

[0227] The stall angle Ang_d is identified on the input curve of the lift coefficient of the relevant portion of the blade C0_port.

[0228] The modified C0_mod_port curve of the lift coefficient is that of [Fig. 8] and is obtained as shown in [Fig. 8]. To recover the slope upstream of the stall angle of the input curve of the lift coefficient C0_port, a second transformation is performed where this slope is applied, which has the effect of modifying the value of the lift coefficient at the modified stall angle Ang_d2. The curve resulting from this second transformation corresponds to the modified lift coefficient curve Cl_port.

[0229] According to one aspect of the invention, the first parameter k of each blade portion can be determined in the following manner:

[0230] k — kg^ ~ s

[0231] k0 being first predefined quantity, mincx is the minimum thickness of the different portions of blades at the trailing edge, cs is the thickness of the blade at the trailing edge of the portion of blade s considered.

[0232] Thus, the various first parameters k of the different blade segments can be defined as a function of a single predefined first quantity k^, which limits the number of parameters for the optimization step. Furthermore, along the blade from the inside out (relative to the axis of rotation), the dimensions of the blade segments tend to decrease. In other words, the blade segments become thinner along the radial direction from the inside out. This results in greater uncertainty regarding the aerodynamic properties at the blade tip (at the outermost end).

[0233] Fig. 11 illustrates, schematically and without limitation, an example of application of the determination of the first parameter k from the formula above.

[0234] Depending on the radial position R_portion, of the blade portion, the first parameter k is determined, which limits the number of parameters for the optimization step.

[0235] Fig. 6 illustrates, schematically and not in a limiting manner, a variant of the construction of the lift coefficient modification curve of the method of calibrating the aerodynamic characteristics of the wind turbine rotor model according to the invention.

[0236] From each input curve of lift coefficient C0_port, for each portion of blade, Id is identified as the stall angle Ang_d.

[0237] To construct Cons2 for each lift coefficient modification curve Cl_port for each blade segment, a first transformation T1 is performed where the stall angle Ang_d is shifted to a modified stall angle value Ang_d2. Then, a second transformation T2 is applied, depending on both the initial stall angle Ang_d (that of the input curve) and the modified stall angle Ang_d2. For example, this second transformation T2 could consist of maintaining the slope upstream of the initial stall angle Ang_d to preserve the aerodynamic responses of the rotor model. Following the second transformation T2, the lift coefficient modification curve for each blade segment Cl_port is thus obtained.

[0238] Advantageously, in step iii), the following third transformation can be applied:

[0239] =CD(«)-g(a)

[0240] gçay=i.cD(a)-e^

[0241] with CD{a) the input curve of the drag coefficient of the relevant blade portion, C[)jjew( a ) the modification curve of the drag coefficient of the relevant blade portion, g(a) the third transformation, V-vaH being the stall angle of the modification curve of the lift coefficient of the relevant blade portion,1 and ^3 of the fourth and fifth parameters among the modification parameters.

[0242] By this third transformation, the drag coefficient curve is shifted around the predetermined value corresponding to half of the modified stall angle (of the modifying curve) and the curve is adapted around this predetermined value.

[0243] The fourth parameter z represents the amplitude of the third transformation, i.e. the amplitude of the applied shift, and the fifth parameter ^3 represents the angular amplitude over which the third transformation is applied.

[0244] The third transformation can, for example, be linked to the surface roughness of the relevant portion of the blade, which can change over time: for example, depending on blade wear and / or the level of cleanliness / dirt accumulated on this surface. Preferably, the fourth parameter 1 can be greater than or equal to -1 and strictly less than 1. Indeed, these values ​​are reasonable to allow for a realistic third transformation, that is, a third transformation for which the drag coefficients remain positive.

[0245] Fig. 10 illustrates, schematically and not in a limiting way, an example of a third drag coefficient curve transformation.

[0246] The curves illustrate the evolution of the drag coefficient C_train as a function of the angle of attack Alpha of the blade portion.

[0247] The drag coefficient input curve C0_train of the relevant blade portion is identified, and the drag coefficient modification curve Cl_train is constructed by shifting the drag coefficient input curve around a predetermined value (here, around half the stall angle of the lift coefficient input curve in Figures 8 and 9). Thus, the curve is modified only around this value so that the drag coefficient modification curve Cl_train is only modified in the vicinity of this predetermined value.

[0248] It is noted that on the modification curve of drag coefficient Cl_train, the drag coefficients are always greater than or equal to zero: they are thus realistic.

[0249] Preferably, the fourth parameter1 of each blade portion can be determined in the following manner:

[0250] hninc^

[0251] / 0 being a second predefined quantity, mincx is the minimum thickness of the different portions of blades at the trailing edge, cs is the thickness of the blade at the trailing edge of the portion of blade s considered.

[0252] Thus, the various fourth parameters 1 of the different blade sections can be defined as a function of a single predefined second quantity Iq, which limits the number of parameters for the optimization step. Furthermore, along the blade from the inside out (relative to the axis of rotation), the dimensions of the blade sections tend to decrease. In other words, the blade sections become thinner along the radial direction from the inside out. This results in greater uncertainty regarding the aerodynamic properties at the blade tip (at the outer end).

[0253] In order to further facilitate the optimization step, we can seek to reduce the number of parameters. To do this, we can define that the second, third and fifth parameters ^i, ^2 and ^3 are identical for all blade portions, which makes it possible to limit this number of quantities of the optimization to only three: one for the second parameter, one for the third parameter and one for the fifth parameter, instead of three parameters for each blade portion, which greatly multiplies the number of quantities to be sought in the optimization.

[0254] According to an advantageous implementation of the method of the invention, in step 3), the following change of variable can be carried out:

[0255] p = Z> + «tanh(x)

[0256] where a and b are predefined value vectors, etx is the parameter vector to be optimized.

[0257] Each vector p includes several modification parameters, for example, said first, second parameters of the first transformation, said third parameter of the second transformation, as well as said first and second predefined quantities.

[0258] The values ​​of vector b correspond to the average values ​​of the range of values ​​considered for each modification parameter, and the values ​​of vector a correspond to the permissible deviations around the average values ​​(values ​​of vector b) of each modification parameter. The values ​​of vectors a and b are therefore predetermined according to the acceptable values ​​that we wish to consider for each modification parameter. The minimization problem in step d) then consists of finding the values ​​of vector x. The change of variable thus performed makes it possible to avoid the inequality constraints that could be added to the minimization problem of the function J(p), the inequality constraints then being bounded by the acceptable ranges of each modification parameter. Replace the constraints Addressing inequality through the change of variables makes it easier to solve the optimization problem.

[0259] The invention relates to a computer program product downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor or a server, comprising program code instructions for implementing the wind turbine calibration method according to one of the variants or combinations of variants described above or for implementing the wind turbine controller model calibration method according to one of the variants or combinations of variants described above, when said program is executed on a computer, a mobile phone or a computing device.

[0260] The wind turbine controller model calibration method and the wind turbine model calibration method can each be used to create a digital twin, i.e., a digital model, representative of the wind turbine. The digital twin can interact directly with the wind turbine: it can receive measurements taken on the wind turbine and / or it can send data to control the wind turbine (i.e., to servo-control it) and / or improve its operation.

[0261] The calibration methods according to the invention and the resulting models can be used to determine the fatigue state of various components of the (actual) wind turbine, to improve the control of the (actual) wind turbine to reduce its fatigue and / or improve its performance. It can also be used to define maintenance interventions and schedule them over time, based on data changes. Examples

[0262] Figures 12 to 16 illustrate comparisons of results between a prior art process and a process of the invention for the same 6.2 MW wind turbine.

[0263] The prior art process uses the NREL's (National Renewable Energy Laboratory) OpenFAST simulator described in the following documents:

[0264] - https: / / github.com / OpenFAST / openfast (accessed May 7, 2018);

[0265] - Jonkman, JM “The New Modularization Framework for the FAST Wind Turbine CAE Tool.” 5 Ist AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, 7-10 January 2013, Grapevine (Dallas / Ft. Worth Region), TX [online proceedings]. URL: http: / / arc.aiaa.org / doi / pdf / 10.2514 / 6.2013-202. Astronautics, January 2013; NREL / CP-5000-57228.

[0266] Fig. 12 shows comparisons of power coefficient and thrust curves of the wind turbine.

[0267] The thrust coefficient curve from the prior art C_T_AA is further from the thrust coefficient input curve C0_T than the calibrated thrust coefficient curve Cc_T of the process according to the invention.

[0268] The power coefficient curve from the prior art C_P_AA is further from the input power coefficient curve C0_P than the calibrated power coefficient curve Cc_P of the process according to the invention.

[0269] This demonstrates the quality of the method according to the invention in modifying the power and thrust coefficient curves to approximate the input curves. The accuracy of the associated model, thanks to the calibration of the aerodynamic characteristics of the rotor model, is therefore improved.

[0270] Fig. 13 illustrates the variation of the rotation speed Om as a function of the wind speed V_wind.

[0271] The ComD curve corresponds to the measurements taken on the wind turbine. The ComAA curve corresponds to the curve from the prior art, and the ComI curve corresponds to the wind turbine calibration method according to the invention (which includes calibrating the aerodynamic characteristics of the rotor model and calibrating the controller model). It can be observed that the ComI curve of the invention is much closer to the ComD curve of the measurements than the ComAA curve of the prior art.

[0272] Fig. 14 illustrates the variation of the orientation angle of the blades B1 as a function of the wind speed V_wind.

[0273] Curve CB1D corresponds to measurements taken on the wind turbine. Curve CB1AA corresponds to the curve from the prior art, and curve CBII corresponds to the calibration method according to the invention (which includes calibrating the aerodynamic characteristics of the rotor model and calibrating the controller model). It can be observed that curve CB1I of the invention is much closer to curve CB1D from the measurements than curve CB1AA of the prior art.

[0274] Fig. 15 illustrates the variation of the thrust coefficient as a function of wind speed V_wind.

[0275] The C0_T curve corresponds to the input curve of the thrust coefficient (provided by the wind turbine manufacturer). The CT AA curve corresponds to the thrust coefficient curve from the prior art, and the calibrated Cc_T curve of the thrust coefficient corresponds to the calibration method according to the invention (which includes the calibration of the aerodynamic characteristics of the rotor model and the calibration of the controller model). It can be observed that the calibrated Cc_T curve of the invention is much closer to the input curve of the thrust coefficient C0_T than the CT AA curve of the thrust coefficient from the prior art.

[0276] Fig. 16 illustrates the variation of the power coefficient as a function of wind speed V_wind.

[0277] The C0_P curve corresponds to the input curve of the power coefficient (provided by the wind turbine manufacturer). The CP AA curve corresponds to the power coefficient curve from the prior art, and the calibrated Cc_P curve of the power coefficient corresponds to the calibration method according to the invention (which includes the calibration of the aerodynamic characteristics of the rotor model and the calibration of the controller model). It can be observed that the calibrated Cc_P curve of the invention is much closer to the input curve of the power coefficient C0_P than the CP AA curve of the power coefficient from the prior art.

[0278] Figures 17 and 18 illustrate comparisons of results between a process with a standard controller model of the prior art and a controller model calibrated by the controller model calibration process of the invention for the same 6.2 MW wind turbine.

[0279] The prior art controller model uses the NREL's (National Renewable Energy Laboratory) OpenFAST simulator described in the following documents:

[0280] - https: / / github.com / OpenFAST / openfast (accessed May 7, 2018);

[0281] - Jonkman, JM “The New Modularization Framework for the FAST Wind Turbine CAE Tool.” 5Ist AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, 7-10 January 2013, Grapevine (Dallas / Ft. Worth Region), TX [online proceedings]. URL: http: / / arc.aiaa.org / doi / pdf / 10.2514 / 6.2013-202. Astronautics, January 2013; NREL / CP-5000-57228.

[0282] For Figures 17 and 18, the same aerodynamic characteristics are used for the prior art method (with the standard controller model) and for the controller model calibrated according to the method of the invention. Thus, the only difference between the prior art method and the method of the invention in these figures concerns the calibration (or lack thereof, as in the prior art) of the controller model.

[0283] Figure 17 illustrates the variation of the rotational speed RS as a function of time Ts. The dashed curve RS-Ex represents the curve derived from experimental data measured on the actual wind turbine.

[0284] The RS-St curve corresponds to the prior art process with the standard controller model and the RS_Inv curve corresponds to the process according to the invention with the calibration of the controller model.

[0285] It is observed that the calibration of the controller model of the invention allows a much better accuracy to the experimental data measured than the results obtained by the standard controller model of the prior art.

[0286] Figure 18 illustrates the variation of the blade orientation angle B1_T as a function of time Ts. The dashed curve B1_T-Ex represents the curve derived from experimental data measured on the actual wind turbine.

[0287] The curve B1_T-St corresponds to the prior art process with the standard controller model and the curve Bl_T_Inv corresponds to the process according to the invention with the calibration of the controller model.

[0288] It is observed that the calibration of the controller model of the invention allows a much better accuracy to the experimental data measured than the results obtained by the standard controller model of the prior art, in particular in the first part (before 600 s).

[0289] Figures 17 and 18 show that the wind turbine model, thanks to the calibration of the controller model according to the invention (with the same aerodynamic characteristics of the rotor model), allows better accuracy to the experimental data measured than the prior art model using a standard controller model (the term "standard controller model" means an uncalibrated controller model).

Claims

1. Demands Method for calibrating a wind turbine controller model comprising a rotor, the rotor including blades, to determine the fatigue state of the wind turbine components and / or the evolution of the wind turbine's performance and / or to adapt the control of the wind turbine and / or to predict maintenance actions for the wind turbine, based on measurements (Measurements) performed on the wind turbine, the measurements (Measurements) including wind speed, blade pitch angle, and rotational speed, in which at least the following steps are carried out: a) From the measurements (Mes) taken, we associate (Ass) with predetermined values ​​of wind speeds: - a speed ratio X defined by / — ™ with œ the rotational speed of the rotor, R the radius of the rotor and v the wind speed; - and a blade orientation angle; b) For each predetermined value of wind speed, we determine (Opt_Cont) the values ​​of the final speed ratio (Ld) and the final blade orientation angle (Bd), as the solutions to the following minimization problem: with CP( / 1 ( v; ), P ( vi ) ) a power coefficient map of the wind turbine (Cart_c_P), Cr(2( v;), vj ) a thrust coefficient map (Cart_c_T) of the wind turbine, vi the predetermined wind speed values, N being the number of predetermined wind speed values, Cpcurve( P,-) being a power coefficient curve (Cc_P), Cpcurve ( V; ) being a thrust coefficient curve (Cc_T), ^(^) the blade pitch angle at wind speed vi; AM being the speed ratio at wind speed vi; being the orientation angle associated with wind speed vi at step a); A)(v») being the speed ratio associated with wind speed vi at step a), A, and v being first weighting coefficients; c) We calibrate (Contlnv) the controller model so that for each predetermined wind speed value, the controller model associates the final speed ratio (LData) and / or the final blade pitch angle (BData) and preferably, it controls the wind turbine with the final speed ratio (LData) and / or the final blade orientation angle (BData).

2. Method according to claim 1, wherein in step b), the solutions are constrained with the following inequalities: TU AO “JT tw; { LN] À(vM)vM>À(v{)v- with the minimum rotational speed, œmax the maximum rotational speed.

3. A method according to any one of the preceding claims, wherein the wind turbine comprises a generator and preferably a reducer and wherein the measurements (Mes) comprise the power of the generator, the controller model preferably comprising a generator control means (ContG) and a blade control means (ContB).

4. Method according to claim 3, wherein the generator control means (ContG) controls the generator torque Tgenp of the wind turbine by . _ with Pgen$ the measured power * genü — n , rN„ <> 'Igeir rotor'8 of the generator, ^gen the generator efficiency, ^rated the rated rotor speed, N g the reduction ratio of the reducer.

5. A method according to any one of the preceding claims, wherein in step a), the wind speed range is divided into sub-ranges, each sub-range being between a minimum speed and a maximum speed, each predetermined wind speed value corresponding to the average speed of the wind speeds measured in the relevant sub-range, each associated speed ratio corresponding to the average of the speed ratios calculated from the wind speeds in the relevant sub-range, and each associated blade pitch angle corresponding to the average of the blade pitch angles for the wind speeds in the relevant sub-range.

6. A method according to claim 3 or 4, wherein, below a predetermined wind speed (vnom), the generator control means (ContG) controls the rotational speed so as to have the final speed ratio (LData) and Preferably, below the predetermined wind speed (vnom), the blade control means (ContB) controls the blade orientation angle to the final blade orientation angle (BData).

7. Method according to the preceding claim, wherein the controller model includes a saturation means (SatB) which saturates the blade orientation angle.

8. A method according to any one of claims 6 or 7, wherein the controller model includes a regulation system (Regul) around the predetermined wind speed (vnom) such that: - if the generator control means (ContG) is active, the regulation system (Regul) imposes on the blade control means (ContB) a rotation speed strictly higher, preferably by a first predetermined deviation, than a predetermined rotation speed; - if the blade control means (ContB) is active, the regulation system (Regul) imposes on the generator control means (ContG) a rotation speed strictly lower, preferably by a second predetermined deviation, than the predetermined rotation speed.

9. A method according to any one of claims 6 to 8, wherein, when the wind speed is greater than the predetermined wind speed (vnom), the rotor equation is linearized as Jrotor - TgenQ • Ng, where Jrotor is the moment of inertia of the rotor, ^g is the efficiency of the gearbox, Ta is the aerodynamic torque of the rotor, X is the speed ratio, [3 is the blade pitch angle, Ng is the reduction ratio of the gearbox, and Tgen is the generator torque and co is the rotational speed, around a wind speed ro, conditioning Ai — ) and , _ lihssfA, and the solution of 1 ( V - ) ) ] - is sought.

10. A method according to any one of the preceding claims, wherein the fatigue state of the wind turbine components and / or the evolution of the wind turbine performance and / or the control of the wind turbine are adapted and / or maintenance actions on the wind turbine are planned, from the calibrated controller model.

11. Method for calibrating a wind turbine model, to determine the fatigue state of the wind turbine components and / or the evolution wind turbine performance and / or to adapt the control of the wind turbine and / or to predict wind turbine maintenance actions, the wind turbine model comprising a controller model and a rotor model, wherein the controller model calibration method according to one of the preceding claims is applied, and before applying step b) of the controller model calibration method, at least the following steps are carried out: 1) modification curves of the lift coefficients (Cl_port) and drag coefficients (Cl_train) of each blade segment are constructed (Const) from input curves of the lift coefficients (C0_port) and drag coefficients (C0_train) of each blade segment and sets (Jp) of modification parameters, and modification parameters are determined from the modification curves of the lift coefficients (Cl_port) and drag coefficients (Cl_train) and for each set (Jp) of modification parameters,the first maps of power coefficients (Cart_P) and thrust coefficients (Cart_T) as a function of the speed ratio and the blade pitch angle; 2) for each pair (Mes-ass) of speed ratio and blade orientation angle associated with one of the said predetermined wind speed values, and for each set (Jp) of modification parameters, a power coefficient and a thrust coefficient are determined (Detl) from the first power coefficient (Cart_P) and thrust coefficient (Cart_T) maps in order to generate comparative curves of power coefficient (C1_P) and thrust coefficient (C1_T) as a function of wind speed; 3) We determine as final modification parameters (Jp_opt), the sets of modification parameters allowing us to minimize (Optl) the following function J(p): J(p) = 1^^1(^)(^,^(^ p) -Cpeun.^) ) + w2(h)(Cr.c„r,^ ) WHERE p represents the modification parameter sets, Cpcurve(v) is an input power coefficient curve and Cpcurvc(v~) is an input thrust coefficient curve, is a first weighting function dependent on wind speed to weight the curve as a function of wind speed, w2 is curved a second weighting function depending on the speed of wind y. to weight the curve A as a function of the speed of the 1 1 jcurve wind, 9 is a second weighting coefficient, y. corresponds to each of the said predetermined values ​​of wind speed, N is the number of predetermined values ​​of wind speed, Cpatrve{ vb p) the comparative curve of power coefficient, CycurVe( p) the comparative curve of thrust coefficient;4) we calibrate (Call) the aerodynamic characteristics of the rotor model, the aerodynamic characteristics of the rotor model comprising - calibrated curves of lift coefficients (Cc_port) and drag coefficients (Cc_train) corresponding to the modification curves of the lift coefficients (Cl_port) and drag coefficients (Cl_train) of each blade portion obtained from the final modification parameters (Jp_opt), - calibrated curves of power coefficients (Cc_P) and thrust coefficients (Cc_T) corresponding to the comparative curves of power coefficients (C1_P) and thrust coefficients (C1_T) obtained from the final modification parameters (Jp_opt); - and preferably calibrated maps of power coefficients (Cart_c_P) and thrust coefficients (Cart_c_T) corresponding to the maps of power coefficients (Cart_P) and thrust coefficients (Cart_T) obtained from the final modification parameters (Jp_opt);said calibrated curves of power coefficients (Cc_P) and thrust coefficients (Cc_T) and of lift coefficients (Cc_port) and of drag coefficients (Cc_train) serving respectively as curves of power coefficients and thrust coefficients and of lift coefficients and drag coefficients in step b), the calibrated maps of power coefficients (Cart_c_P) and thrust coefficients (Cart_c_T) serving as maps of power coefficients and thrust coefficients respectively in step b).;

12. A method according to claim 11, wherein the blades are composed of blade portions and, before step b), input maps of the power coefficient (Cart_E_P) and thrust coefficient (Cart_E_T) of the wind turbine are determined (C_ent) as a function of the speed ratio and as a function of the orientation angle blades from input curves of lift coefficients (C0_port) and drag coefficients (C0_train) of each blade portion.

13. A method according to claim 12, wherein, between step a) and step 1), for each pair (Mes-ass) of speed ratio and blade pitch angle associated with one of said predetermined wind speed values, a power coefficient and a thrust coefficient are determined (Det2) from the input power coefficient and thrust coefficient maps so as to generate first comparative curves of power coefficient (Cart_E_P) and thrust coefficient (Cart_E_T) as a function of wind speed and the following steps are applied only if the difference between the first comparative curves and the input curves of power coefficient and thrust coefficient is greater than a predetermined threshold.

14. A method according to any one of claims 12 or 13, wherein, in step 1), the following substeps are carried out for each blade portion: i) the stall angle (Ang_d) of the lift coefficient entry curve (C0_port) is identified, preferably by calculating the second derivative of the lift coefficient entry curve (C0_port) and taking, as the stall angle (Ang_d), the smallest positive angle of attack that makes the second derivative zero; ii) the value of the stall angle is shifted and then the modification curve of the lift coefficient (Cl_port) is constructed so that the slope before the stall angle is identical to that of the lift coefficient entry curve (C0_port);iii) the modification curve of the drag coefficient (Cl_train) is constructed by shifting the input curve of the drag coefficient (C0_train), around a predetermined value, preferably around half of the stall angle (Ang_d).;

15. A method according to claim 14, wherein in step ii), the value of the stall angle is shifted by applying the following first transformation: CL(f(aY)=C^ f (a) = a + ke^Y^ a being the angle of attack, CL{ a) being the input curve of the lift coefficient, CL(f(a) ) being a modified curve of the lift coefficient, f ( a ) being the first transformation applied to the stall angle, ^«zz being the stall angle (Ang_d) of the input curve of the lift coefficient (C0_port), k and are first and second parameters among said modification parameters.

16. A method according to claim 15, wherein once the value of the stall angle of each blade portion has been shifted, the following second transformation is applied: (a) -CL(a) [ -1 )e J CL^(a) - CL(a ) 1 + -1 Je J a being the angle of attack, ^(a) being the input curve of the lift coefficient, ( et) being the modified curve of the lift coefficient, C£hw( a) being the modifying curve of the lift coefficient (Cl_port), being the stall angle (Ang_d) of the input curve of the lift coefficient (C0_port), being the stall angle of the modified curve of the lift coefficient, ^2 being a third parameter among said modification parameters.

17. A method according to any one of claims 15 or 16, wherein the first parameter k of each blade portion is determined in the following manner: / mincr k = kQ^-^r k0 being a parameter, mmcv is the minimum thickness of the different blade portions at the trailing edge, cs is the thickness of the blade at the trailing edge of the blade portion s considered.

18. A method according to any one of claims 14 to 17, wherein, in step iii), the following third transformation is applied: ^D / iew ( *7 ) — C d ( <7 ) - CJ ) g{a) =l-CD(a) -e 4 ; with Cpi a) the input curve of the drag coefficient (C0_train), CDnewt'a the modification curve of the drag coefficient (Cl_train), g( a) the third transformation, ^sttdl being the stall angle (Ang_d2) of the modification curve of the coefficient of lift (Cl_port),1 and the fourth and fifth parameters among the modification parameters, 1 being preferably greater than or equal to -1 and strictly less than i

19. Method according to claim 18, wherein the fourth parameter 1 of each blade portion is determined in the following manner: IminCy l=^hr lo being a parameter, mmcv is the minimum thickness of the different blade portions at the trailing edge, cs is the thickness of the blade at the trailing edge of the blade portion s considered.

20. A method according to any one of claims 11 to 19, wherein, in step 3), the following change of variable is made: p = / ? + «tanh(x) where a and b are predefined value vectors, etx is the parameter to be optimized.

21. A method according to any one of claims 11 to 20, wherein the fatigue state of the wind turbine components and / or the evolution of the wind turbine performance and / or the control of the wind turbine are adapted and / or maintenance actions on the wind turbine are planned, from the calibrated wind turbine model.

22. Product computer program downloadable from a communication network and / or stored on a computer-readable medium and / or executable by a processor or server, comprising program code instructions for implementing the calibration method of a controller model according to any one of claims 1 to 10 or the calibration method of the wind turbine model according to any one of claims 11 to 21, when said program is executed on a computer, mobile phone or computing device.

Images