System for supervising a wind farm

Abstract

The present invention relates to a system (2) for supervising wind turbines (1), each wind turbine (1) comprising a mast (101) accommodating at its upper end a nacelle (102) with an angular orientation in a reference frame (300), the supervising system (2) comprising means for measuring, on site, a main direction (201) over a time period, and control means allowing the angular orientation of the nacelle (102) to be modified, characterized in that it comprises means for detecting the angular orientation, and a module (23) for determining at least one angular differential (400) from the measured main direction (201) and from the detected angular orientation, corresponding to a yaw alignment error that the control means correct by modifying the angular orientation of each nacelle (102).

Description

[0001] WIND FARM MONITORING SYSTEM

[0002] Technical field of the invention

[0003] The present invention falls within the field of managing the production of electrical energy within a wind farm and aims at the supervision of at least one wind turbine.

[0004] As is well known, a wind farm comprises several wind turbines, located on a geographical site and spaced apart, so as to ensure maximum efficiency for each of the wind turbines and thus optimize the overall energy production of the said site. The design and installation of wind turbines within a farm is carried out taking into account numerous factors, in particular depending on the topography of the site, as well as weather patterns recorded locally as well as on a larger scale.

[0005] Indeed, winds at a site have many components, such as speed and direction, but also air shear and density. These components are linked to other meteorological variables, such as temperature, atmospheric pressure, or humidity.

[0006] These variables depend on different temporal factors, mainly depending on the season, but also on a shorter duration, weekly or daily, depending on meteorological phenomena (showers, thunderstorms, storms, etc.), or even shorter, of the order of a second to several hours, such as for gusts or turbulence. Furthermore, the winds are strongly impacted by the topography of the site, in particular the reliefs, likely to generate an orographic lift or a corridor wind. In addition, the temperature differences between the ground and the atmospheric strata generate thermal currents, evolving throughout the day and depending essentially on the sunshine and the densities of said strata.

[0007] More locally at ground level, winds are impacted by the types of soil surfaces and vegetation, and even other infrastructure present, particularly other wind turbines creating a wake effect and turbulence downstream.

[0008] Therefore, the geographical position of each wind turbine on the site and their spacing in relation to each other are essential, in order to optimize the performance of the said park.

[0009] On the other hand, wind turbines also have technical characteristics, linked to their structure (hub height, rotor diameter, type of blades, nominal power of the generator, etc.) inducing operating specifications (power curves, thrust curves, etc.), depending on the components of the winds at a given moment.

[0010] As such, each of the wind turbines in operation is controlled to adapt its operating configuration, in particular the orientation of the nacelle, for example in alignment with the prevailing wind locally at a given time, but also its rotation speed depending on the wind force, in particular to protect its components from excessive loads and mechanical constraints. In addition, the installation of a wind farm is governed by various regulations, related to safety but also to the environment, in particular to limit noise pollution and for the protection of animals, in particular birds and bat species.

[0011] Furthermore, it is advisable to provide regular maintenance plans for wind turbines, preventing their operation during the duration of these interventions.

[0012] In each case, the operation of the wind turbines is restricted, leading to performance losses that must be anticipated and minimized.

[0013] We then understand the need to supervise each wind turbine in the park and the complexity of managing and taking into consideration each of the aforementioned parameters, to optimize the park's energy production.

[0014] In addition, the operator must take into consideration financial aspects linked to the establishment of the park and its operation, to obtain sustainable profitability.

[0015] State of the art

[0016] In this context, it is necessary to carry out, beforehand during a design phase, but especially during the operation of the park, measurements relating to the winds, by means of dedicated sensors positioned at different levels of each wind turbine and the site.

[0017] In particular, in the context of the wind industry, an anemometer type sensor is an instrument for measuring the horizontal direction of the wind, more precisely the component projected in a horizontal plane of a vector representative of the wind speed.

[0018] There are cup anemometers or those using ultrasound, which can measure several components of the wind.

[0019] This anemometer is installed on the nacelle, generally at the rear or downstream, namely behind the blades.

[0020] In addition, sensors, such as weather vanes, can only measure the direction of the wind, reported in the said horizontal plane.

[0021] Thus, these combined sensors make it possible to measure the speed and direction of winds, in particular those of the locally prevailing wind.

[0022] Such positioning is unreliable, subject to wake effects and turbulence generated by the rotation of the blades, but also to gusts and headwinds located at the level of said nacelle. Errors in measuring wind speed are frequently observed, in particular between the free wind measured upstream of the nacelle, and the wind measured on the nacelle, which can be of the order of 1 m / s (meter per second) on average. Regarding the wind direction relative to the nacelle, errors in measuring wind direction are frequently observed, in particular between the free wind measured upstream of the nacelle, and the wind measured on the nacelle, which can be of the order of 10° (degrees) on average. Such deviations generate yaw alignment errors, but also blade pitch errors (i.e. the angle of inclination of the blades relative to the plane of rotation centered on the hub), which cause proportional losses.An alternative is to estimate the wind speed, calculated based on the rotor rotation speed. However, this estimate is biased and remains imprecise, using values ​​related to the operation of the machine and not allowing its operation to be optimized.

[0023] On the other hand, the orientation of the nacelle must be measured, in particular its azimuth in a reference frame, for example relative to geographic north. Regularly, deviations in azimuth referencing are observed, randomly, which can exceed angles of 10°. These deviations can also cause losses when wind turbine restraints are set by the orientation of the nacelle, for example with regard to restraints linked to acoustic regulations or load control.

[0024] In addition, such sensors subject to climatic hazards can fail, causing additional errors and requiring on-site intervention.

[0025] These inconsistent data are detrimental to quantifying the performance of individual wind turbines. Currently, it is difficult to obtain an accuracy of less than 5% in simple wind conditions, and even worse, less than 10% in complex wind conditions, on the performance characteristic between wind and power of a wind turbine.

[0026] Document EP3907402 attempts to solve this problem by proposing to detect a yaw alignment error of the nacelle of a wind turbine through a method for determining a yaw misalignment relative to the geographic north, in particular determined from aerial image captures by satellite or aircraft. The yaw misalignment of the nacelle is determined relative to a correction of the knowledge of the position of the nacelle relative to the geographic north. This technique thus makes it possible to determine a misalignment angle and possibly correct the application of the load control law or acoustic emissions which depend on the orientation of the nacelle relative to the geographic north.

[0027] However, there are still limitations with such a technique, which does not fully take into account the fact that correcting the knowledge of the position of the wind turbine relative to geographic north generally has no impact on its alignment relative to the wind direction.

[0028] Further, the use of the collected data is carried out through a SCADA type supervision system, for "Supervisory Control And Data Acquisition". Such a supervision system ensures management at all levels, as mentioned above.

[0029] More specifically, the said supervision system makes it possible to control the configuration of each of the wind turbines in a farm, in particular to modify the azimuth, the rotation speed, the blade pitch, etc.

[0030] Document CN 112459965 describes a monitoring of several wind turbines on the one hand, in particular to determine a wake effect generated by a first wind turbine located upstream with respect to the wind and impacting a second wind turbine located downstream. Therefore, a realignment of one and / or the other of the wind turbines, in particular the first wind turbine is carried out, with respect to the direction of the wind or a theoretical model, in order to reduce the wake effect and optimize the efficiency of the second downstream wind turbine.

[0031] However, this solution does not take into consideration the relative aspects of winds, as well as the aforementioned yaw misalignment, leading to errors in the data processed by such supervision.

[0032] Furthermore, it is easy to understand that erroneous data inevitably leads to a drop in actual production, but also to a distorted estimate of production, predictively during the design phase, or even during simulations of the operation of an installed wind farm.

[0033] Statement of the invention

[0034] The invention aims to overcome the drawbacks of the state of the art by proposing a system for monitoring at least one wind turbine, in particular several wind turbines in a wind farm, providing a measurement of at least the angular orientation in a reference frame of the wind turbine mast, compared to at least one of the components of one of the measured winds and considered to be locally dominant at a given time, namely the main direction of said dominant wind. Such a system provides for determining a differential between this angular position and the wind direction, to correct any resulting yaw alignment error. In particular, the invention provides for taking into consideration the measurements of the actual dominant wind and the actual angular position of the nacelle, to determine relatively in the reference frames considered a yaw misalignment of said nacelle, in order to correct it.

[0035] To do this, the supervision system provides that each site is swept by winds, each of said winds having components, including at least one main direction of a locally dominant wind in a reference frame and a speed. In addition, each wind turbine comprises a mast receiving at its upper end a nacelle supporting a rotor formed of at least one hub on which blades are mounted, with an electric generator supplying energy according to a power to a production network, each wind turbine having technical characteristics and an operating configuration with at least one angular orientation of each nacelle in said reference frame.

[0036] In this context, the said supervision system includes

[0037] - means for measuring over at least one time period at least said main direction among said components of said winds; said measuring means being installed on said site;

[0038] - means for controlling the configuration of each wind turbine, allowing at least the modification of said angular orientation of said nacelle relative to the main direction measured.

[0039] Said supervision system is characterized in that said supervision system comprises

[0040] - means of detecting said angular orientation relative to geographic north:

[0041] - a module for determining at least one angular differential from said measured main direction and said detected angular orientation, corresponding to a yaw alignment error relative to the main direction of the locally prevailing wind;

[0042] - said control means modifying the angular orientation of each nacelle as a function of said yaw alignment error.

[0043] According to additional, non-limiting characteristics, corresponding to different embodiments, combined or alternative, such a supervision system may comprise a module for analyzing and optimizing the performance of each wind turbine, estimating a production yield for said wind turbine.

[0044] According to one embodiment, said detection means comprise

[0045] - a module for detecting said angular orientation of each wind turbine i) by acquiring at least one photo / video capture, preferably by high-resolution satellite imagery, with a fineness of at least 1 meter, preferably a fineness of less than 0.5 m, more preferably a fineness of less than 0.3 m; ii) then by cutting each image into at least one thumbnail comprising each wind turbine; iii) then by determining the angular orientation of each wind turbine in each thumbnail. According to one embodiment, from the elements visible on an image captured at a given time and a given geographical location and / or on said thumbnail, it is possible to combine and / or compare characteristics of the wind turbine with characteristics of its cast shadow, in order to determine the orientation of the nacelle of a wind turbine.

[0046] According to one embodiment, a first wind turbine and at least one second wind turbine form a wind farm located on said site. Therefore, the system comprises calibration means which ensure the detection j) of a wake effect between the first wind turbine and at least said second wind turbine, and / or jj) of a blocking effect between said first wind turbine and at least said second wind turbine;

[0047] - the angular orientation detection module using the detection of the wake effect and / or the blocking effect of the calibration means.

[0048] According to one embodiment, said supervision system comprises

[0049] - a module for integrating the successive measured values ​​of at least said main direction, over one or more time periods, into at least one time series of data;

[0050] - a module for calibrating the data from each time series; - said analysis and optimization module estimating a predictive yield, generating a predictive simulation of said performance of each wind turbine.

[0051] According to one embodiment, said integration module generates said time series of data prior to the installation of each wind turbine; or said integration module generates said time series of data during the operation of each wind turbine, preferably during a pre-operation phase.

[0052] According to one embodiment, said supervision system comprises a module for generating, from said at least one time series of data, at least one local atmospheric virtual model, corresponding to a three-dimensional mesh of an area of ​​said site comprising at least said wind turbine.

[0053] According to one embodiment, said supervision system comprises a module for calculating transfer functions of the measuring means in the form of an anemometer-type sensor mounted on said nacelle, as a function of said local atmospheric virtual model.

[0054] According to one embodiment, said supervision system comprises a module for parameterizing at least one curve of each wind turbine representative of said power, as a function of said speed of said winds and as a function of operating specifications of said wind turbine.

[0055] Presentation of the drawings

[0056] Other characteristics and advantages of the invention will emerge from the detailed description which follows of the non-limiting embodiments of the invention, with reference to the appended figures, in which:

[0057] [Fig. 1] schematically represents a perspective view of an example of a wind farm, located in a geographical area, showing in enlargement different elements of a wind turbine;

[0058] [Fig. 2] schematically represents a cartographic-type elevation view of a geographical area, highlighting its topography, with a modeling of certain components of the winds, with two main directions;

[0059] [Fig. 3] schematically represents a map view similar to Figure 2, showing the location of several wind turbines in a reference frame centered on geographic north;

[0060] [Fig. 4] schematically represents a side view of a wind turbine, with modeling of certain direction components of several winds and highlighting examples of turbulence generated by said wind turbine and impacting the directions and power among said winds;

[0061] [Fig. 5] schematically represents a top view of Figure 4;

[0062] [Fig. 6] schematically represents a map view similar to Figure 2, highlighting in dotted arrows modifications induced by wind turbines on certain components of direction and power of several winds;

[0063] [Fig. 7] schematically represents a map view similar to Figure 3, highlighting in magnification the detection of the angular position of a wind turbine relative to said reference frame;

[0064] [Fig. 8] schematically represents a top view of a wind turbine with its angular position detected, highlighting an angular differential with respect to the angle of the measured component of the main direction of one of the winds;

[0065] [Fig. 9] schematically represents a view similar to Figure 6, showing a correction of a yaw alignment error of the wind turbine, by modifying its angular position according to said differential; and

[0066] [Fig. 10] schematically represents a simplified architecture of a wind turbine supervision system, showing in particular different data taken into consideration in the context of the installation of said wind turbine on a site.

[0067] Detailed description

[0068] The present invention relates to the management of the production of electrical energy within a wind farm and aims at the supervision of at least one wind turbine 1.

[0069] Usually, such a wind turbine 1 comprises a base 100, secured to foundations made on the ground of a land site, or by suitable means for a maritime site. A mast 101 is connected below said base 100 and erected vertically. The wind turbine 1 comprises, at the upper end of said mast 101, a nacelle 102 supporting an energy generator, preferably electric, by converting the wind power of the winds. Said generator forms an electric generator supplying energy according to a power a production network. Such a generator comprises a rotor in the form of a hub 103 at the upstream or front end of which blades 104 are mounted, at least two in number but preferably three. These blades 104 extend radially around said hub 103, according to regular angular intervals.

[0070] Each of the blades 104 has a specific geometry, with an aerodynamic profile, preferably partially helical and tapered from the proximal end to the distal end. In addition, the proximal end is mounted pivotally articulated with the upstream end of the hub 103, according to a setting angle.

[0071] Furthermore, said nacelle 102 is designed to be orientable according to a rotation centered on said mast 101, in order to align the wind turbine 1 facing the locally prevailing wind.

[0072] Already, within the framework of the present invention, each wind turbine 1 has technical characteristics 110 linked to its structure, essentially dependent on the aforementioned elements, such as for example, but not limited to, the height of the hub 103, the diameter of the rotor, the type of blades 104, or even the nominal power of the generator.

[0073] Each wind turbine 1 also has an operating configuration 111, namely configurable positions of each of the aforementioned elements between them, such as for example, in a non-exhaustive manner, the orientation of the nacelle 102 at the top of the mast 101, preferably in alignment with the direction of the prevailing wind locally at a given time, the rotation speed of the rotor as a function of the strength or power of said prevailing wind, the pitch angle of the blades 104 to optimize wind resistance, but also to create a stall of the bearing surfaces of said blades 104, or even the free rotation of the hub 103, in particular to protect the elements of the wind turbine 1 from excessive loads and mechanical stresses.

[0074] Each wind turbine 1 also induces operating specifications 105, such as, for example, but not limited to, a power curve or a thrust curve. These operating specifications depend on the technical characteristics and the operating configuration, and are also linked to the components 200 of the winds at a given time.

[0075] In this respect, the winds on a site present numerous 200 components, such as, for example, but not limited to, a speed and a direction with a sense.

[0076] In addition, the values ​​of the 200 components depend on different temporal factors, mainly depending on the season, but also on a shorter duration, weekly or daily, depending on meteorological phenomena (showers, thunderstorms, tempests, etc.), or even shorter, of the order of a second to several hours, such as for gusts or turbulence.

[0077] The 200 components of the winds are also strongly impacted by the topography of the site, in particular the reliefs, likely to generate an orographic lift or a corridor wind, or even by the differences in temperatures between the ground and the atmospheric strata generating thermal currents, evolving over the course of the day and depending essentially on the sunshine and the densities of said strata.

[0078] More locally at ground level, winds are impacted by the types of soil surfaces and vegetation, and even other infrastructure present.

[0079] It will also be noted that the rotor of a wind turbine 1 is positioned at a determined height, so that its operating volume (namely the sphere traveled by the rotor when it turns according to the rotation of the nacelle 102 around the mast 101) is located above a zone 105 where the winds are turbulent and unsteady, strongly impacted by the vegetation and surrounding infrastructure.

[0080] Such a zone 105 located under the operating volume of a wind turbine 1 is notably represented in FIG. 4, showing turbulent and unsteady winds with disturbed directions, but also reduced powers, compared to a locally dominant wind. In the context of the present invention, it will be noted that a wind is considered to be “locally dominant”, or also designated “dominant wind”, relative to the scale of said wind farm site, in particular its surface area (and possibly its elevation difference). One or more dominant winds can be determined for a site. Each dominant wind is a combination of the winds, namely that the averages and / or the combination of the components 200 of said winds determine the components 200 of said dominant wind. Figure 2 shows two dominant winds in the form of large arrows, while the winds are represented by smaller arrows.In this example, we see that the winds are impacted by the relief of the site, namely its topography represented in cartographic terms by the contour lines; in particular, the 200 component of the wind direction goes around a high point, modeled by a black triangle. In this example, this results in two dominant winds, with different directions.

[0081] It should be noted that the values ​​of the components 200 of the winds and the prevailing wind on the site are obtained over a period of time, by measurement directly on the site, at different locations, and / or from meteorological models 210, in particular at a macro-meteorological scale which can correspond to an area covered of the order of the country, region or department, but also at a micro-meteorological scale which can correspond to an area covered of the order of one to several square kilometers, corresponding more to the surface area of ​​the site.

[0082] In this context, the presence of wind turbines 1 on a geographical site impacts the winds, converting part of their wind energy into electrical energy, which reduces the power of said winds. The operation of wind turbines 1 also generates turbulence, modifying certain components 200 of the winds, such as the direction.

[0083] Figures 4 and 5 show examples of the impact of a wind turbine 1 on winds passing through it. In particular, turbulence is generated along the blades 104, with drag effects at the distal ends. The nacelle 102 also causes an aerodynamic flow of fluids, impacting their trajectory. Furthermore, in operation, within a cone extending downstream of the rotor, the power of the winds is reduced, due to the conversion of energies.

[0084] It should be noted that the impact of a wind turbine 1 on the winds can be known and determined by the manufacturer, in particular certain modifications of values ​​of the components 200 of said winds.

[0085] Further, the installation of several wind turbines 1 on a site is likely to cause disturbances between them, essentially when the cone of a wind turbine 1 coincides with another wind turbine located downstream, namely when they find themselves in alignment at least partially along an axis passing through the direction of the wind(s).

[0086] Therefore, the geographical position of the wind turbines 1 on the site and their spacing in relation to each other are essential, in order to optimize the yield of the said wind farm.

[0087] Figure 6 takes the example of the geographical site, highlighting in dotted lines the cones impacting the winds downstream, both in power and direction.

[0088] Thus, the forecasts based on the weather models are modified, but the measurements that can be carried out on site are also impacted, with significant errors. In this context, yaw alignment errors of one or more of the wind turbines 1 occur regularly, relative to the main direction 201 of the prevailing wind and / or the impacted winds, and are not detected or corrected. This being the case, the invention aims to limit the losses linked to these yaw alignment errors, with a view to optimizing the efficiency of a wind farm.

[0089] To do this, the invention aims at a system 2 for supervising at least one wind turbine 1, preferably several wind turbines 1 forming one or more wind farms.

[0090] The wind turbine(s) 1 are to be installed and / or installed on at least one production site. In other words, the supervision system 2 makes it possible to simulate the potential production with a view to a theoretical or preferably future installation of one or more wind turbines 1 on a site, conferring a predictive simulation character, or else makes it possible to simulate the future production of an existing site on which one or more wind turbines 1 are already installed.

[0091] As mentioned previously, each site is swept by winds, each of said winds having components 200, including at least one main direction 201 in a reference frame 300 and a speed.

[0092] In addition, each wind turbine 1 has technical characteristics 110 and an operating configuration 11 with at least one angular orientation of each nacelle 102 in said reference frame 300.

[0093] In this respect, said reference frame 300 may be of any type, geo-centered or Cartesian. According to the examples visible in the figures, the reference frame 300 is geographic, namely centered on geographic north, with geographic coordinates according to a latitude and a longitude, as well as an altitude.

[0094] Therefore, the location of a wind turbine 1 corresponds to the coordinates of its mast 101 fixed in said reference frame 300.

[0095] It should be noted that the values ​​of these numerous parameters are recorded digitally in the form of data in corresponding files and accessible by said system 2, in particular through a manager, such as a data server, connected to a suitable communication network. Such a manager may include a SCADA.

[0096] In this respect, it will be noted that the supervision system 2 is implemented by computer and integrates software and algorithms executed on at least one computer terminal. Such a system involves storage within said terminal, as well as data exchanges through at least one communication network, wired or not, in particular with the local or remote manager.

[0097] This being the case, according to the invention, said supervision system 2 firstly comprises means for measuring over at least one time period at least said main direction 201 among said components 200 of said winds. These means therefore make it possible to measure several components 200 of the winds and to record their values ​​for processing, as will be seen later.

[0098] Said measuring means comprise at least one sensor, preferably several sensors. Each sensor may be a measuring instrument of any type, preferably in the form of an anemometer, cup or ultrasonic, but also a lidar, or a wind vane. One or more combined sensors make it possible to measure one or more components 200 of the winds, in particular the speed and direction of the winds.

[0099] In addition, said measuring means are installed on said site.

[0100] As mentioned above, in the case of supervision of a site for the future installation of one or more wind turbines 1, said measuring means are positioned at different locations on the site, for example on the ground or raised, at the upper end of poles or masts.

[0101] When the site is existing, other additional sensors of the measuring means can be embedded on each wind turbine. 1 , in particular at the level of one of these elements, such as the mast 101 , the nacelle 102 or the front end of the hub 103.

[0102] The supervision system 2 also comprises means for controlling the configuration of each wind turbine 1. In other words, the system 2 controls the operating configuration of each wind turbine 1, to adapt the positioning of each of its elements, depending on the winds. These control means allow at least the modification of said angular orientation of said nacelle 102 relative to the main direction 201 measured, namely relative to the direction of the locally dominant wind obtained from an average or a combination of the components 200 of the winds from the means for measuring each of said winds.

[0103] According to one embodiment, the supervision system 2 also comprises a module 20 for analyzing and optimizing the performance of each wind turbine 1, estimating a production yield for said wind turbine 1.

[0104] First of all, such an analysis module 20 makes it possible to determine certain operating specifications 112, such as the power curve, at a given time, as a function of several factors, in particular the components of the winds, especially the direction and the power, with regard to certain operating configurations 111 of each wind turbine 1, such as its angular position, the pitch angle of the blades 104, or even the rotation speed of its hub 103.

[0105] From these elements, said analysis module 20 can then calculate a yield for each wind turbine 1 and update it periodically according to the change in the aforementioned factors.

[0106] Then, said analysis module 20 makes it possible to optimize the yield, in a predictive manner through a simulation for a future period, or in real time. Such optimization is carried out by modifying the values ​​of one or more of the aforementioned factors, or by applying an increase or a reduction, in particular based on the angular orientation of the nacelle 102 of each wind turbine 1.

[0107] It will be noted that the power curve of a wind turbine 1 is a numerical parameterization making it possible to construct a function representative of the power produced by said wind turbine 1, as a function of components 200 of the winds, in particular as a function of the speed of the locally dominant wind, but also in particular the main direction 201. This power curve is provided by the manufacturer of the wind turbine 1, specifying its performance according to a reference power. This reference power constitutes one of the input data with an essential technical and economic character for evaluating the profitability of a wind farm project.

[0108] Therefore, the supervision system 2 plans to use the wind speed from the calibration output to determine a real power curve.

[0109] To do this, according to one embodiment, said system 2 comprises a module 21 for parameterizing at least one curve of each wind turbine 1 representative of said power, as a function of said speed of said winds and as a function of operating specifications 112 of said wind turbine 1.

[0110] Thus, it is possible to directly use the wind measurement resulting from the simulation, as described later, of the measurement means in the form of an anemometer type sensor installed on the nacelle 102, by applying a transfer function as described later, or by iteratively searching for parameters which distort the reference power curve, in order to arrive at a power produced consistent with the observed electrical production.

[0111] Note that if a yaw alignment error has been determined, this error can be taken into account and integrated into the operational power curve, in order to avoid having to account for it additionally later.

[0112] Advantageously, said supervision system 2 comprises means for detecting said angular orientation of said at least one wind turbine 1, preferably of each of the wind turbines 1. In particular, the detection means make it possible to determine the orientation of the nacelle 102 of each wind turbine 1 in the reference frame 300. In particular, the orientation of the nacelle 102 corresponds to an angular position 400 in the reference frame 300. The value of this angular position 400 can be expressed in degrees around a circle, as can be seen in particular in FIGS. 7 to 9.

[0113] In particular, the value of the angular position 400 may correspond to the azimuth of the nacelle 102 of each wind turbine 1 , defined as the orientation of its rotation axis relative to geographic north; a wind turbine 1 with an azimuth of 0° faces geographic north. As mentioned previously, precise knowledge of this orientation is essential and currently leads to deviations of more than 10°, for various reasons during configuration when installing a wind turbine 1 , but also during successive maintenance at the mechanical level and at the software level, tending to increase the recurrence and magnitude of the deviations, which can randomly reach up to 50°.

[0114] Therefore, according to one embodiment, said detection means comprise a module 22 for detecting said angular orientation of each wind turbine 1. This detection module 22 makes it possible to actually observe the orientation of each wind turbine 1, with a view to confirming whether this orientation corresponds to the components 200 of the winds at a given instant, in particular to the component 200 of the main direction 201 of the locally prevailing wind.

[0115] According to a preferred embodiment, the detection of said module 22 is carried out by acquiring at least one photo and / or video capture, preferably by high-resolution satellite imagery, with a fineness of at least 1 m (meter), preferably a fineness of less than 0.5 m, more preferably a fineness of less than 0.3 m.

[0116] In particular, the use of satellite imagery with automated processing eliminates the need for on-site intervention, with the risks of error linked to the human factor.

[0117] In addition, satellites can be selected from a constellation, in heliosynchronous orbit, allowing a bi-daily revisit periodicity above the site, guaranteeing access to several dozen valid images per year.

[0118] Alternatively or additionally, photo / video acquisition can be carried out in any other way, from the ground but preferably at elevation, using a device flying over each wind turbine 1, such as a drone, or using aerial photography.

[0119] According to one embodiment, the aerial imaging can be carried out by interferometry, preferably using a Synthetic Aperture Radar (SAR), which has the advantage of detecting and taking photos / videos of the wind turbine 1 while passing through any possible cloud cover.

[0120] Then, each image is cut into at least one thumbnail including each wind turbine 1.

[0121] Then, a determination of the angular orientation of each wind turbine 1 is carried out in each vignette.

[0122] It will be noted that said determination can preferably be carried out automatically, but also manually by an operator. This determination takes into consideration visual elements, such as the axis of the hub 103, the plane containing the blades 104, or even the shadow cast by the mast 101 related to a timestamp of the position of the sun or to a pre-recorded model of the shadows cast as a function of the year as well as the geographical location of the installation site of the wind turbine 1. In other words, from the elements visible on an image captured at a given time and a given geographical location and / or on said thumbnail, it is possible to combine and / or compare characteristics of the wind turbine 1 with characteristics of its shadow, in order to determine the orientation of the nacelle 102 of a wind turbine 1.

[0123] In combination or alternatively, artificial intelligence can be used to directly estimate the azimuth position from an image naturally containing all of the above-mentioned elements. This intelligence can have semi-automatic and supervised learning, with captured images as input and determined angular positions as output.

[0124] The use of a set of computer-generated images is also possible to form an initial training base.

[0125] In addition, the implementation of an unsupervised learning model is also possible.

[0126] According to another embodiment, alternative or combined, the detection of said module 22 is carried out by means of a compass of the GPS type (for “Global Positioning System”), installed on said nacelle 102 for a determined period of time. The values ​​measured by said compass are recorded and compared to the azimuth recorded at the nacelle 102, due to its mechanical configuration. The difference between the two is determined and becomes an azimuth alignment error of said nacelle 102.

[0127] Using a GPS type compass makes it possible to overcome the strong magnetic interference at the nacelle 102 generated by the electrical components of the generator.

[0128] Such a solution offers increased accuracy, slightly greater than + / - 0.5°. However, the installation of said compass on each nacelle 102 requires alignment precision, which may generate uncertainties due to the manual performance of this operation. According to another embodiment, alternative or combined, the detection of said module 22 is carried out by means of a line of sight, with an aim towards a geographical reference point with a known position located on said site or in the surrounding area, such as a summit or an infrastructure, or even one of the other wind turbines 1 of said farm. This line of sight is fixed and aligned on the nacelle 102 which is then rotated until it points with said reference point.

[0129] When the line of sight points towards the object in question, the azimuth of the nacelle 102 is noted. Using the geographical coordinates of the wind turbine 1 as well as those of the reference point, the value of the true angular position 400 is determined in the reference frame 300. In addition, the deviation between the mechanical azimuth and the true position 400 thus determined is recorded and becomes an azimuth alignment error for each nacelle 102.

[0130] However, the uncertainty of the aim achieved by an operator generates a risk of errors.

[0131] Further, said supervision system 2 comprises a module 23 for determining at least one angular differential 401 from said measured main wind direction 201 and said detected angular orientation, corresponding to a yaw alignment error. In particular, the differential 401 is obtained by comparison with the value of said angular position 400.

[0132] In other words, the yaw alignment errors are defined as the difference between the main direction 201 of the locally prevailing wind at the height of the wind turbine 1, or an average of certain components of the winds 200 over a surface of interest of the site, and the azimuth of the nacelle 102 of said wind turbine 1. For example, if the azimuth of the nacelle 102 at a given instant is 10° and the main direction 201 of the wind is 20°, the yaw alignment error is 10° (or -10° according to an opposite sign convention). It is therefore appropriate to rectify it by increasing the angle of the angular orientation of the nacelle 102. It will be noted that, with respect to the reference frame 300, the invention takes into consideration the passages of angular positions 400 from 0° to 360°, as well as between +180° and -180°.

[0133] According to another embodiment, alternative or combined, the detection of said module 22 is carried out by means of the comparison between observed wake effects and predicted wake effects.

[0134] Wakes indeed cause relative production deficits between different wind turbines 1 . When characterizing the azimuthal position of these deficits with respect to a direction reference, these positions should correspond with the respective azimuthal positions of the masts 101 of the wind turbines 1. For example, a wind turbine 1 oriented towards the east (i.e. at 90°) and located upstream of another wind turbine 1 is likely to cause a production deficit in the east to west direction. Therefore, by using a wake effects position prediction model, assuming that a wind turbine 1 is aligned with the corresponding wind direction, it is possible to determine errors in the azimuth alignments. However, it is known that a misalignment of a wind turbine 1 with the wind causes a deflection of the wake effects.It is therefore necessary to iteratively model the wake effects, the misalignments of the nacelles 102, as well as the misalignments of the wind turbines 1, in order to determine the misalignments of the nacelles 102 from the wake effects.

[0135] Therefore, it is also possible, from the wake effects, to determine the angular orientation of a wind turbine 1. To do this, according to one embodiment, the module 22 for detecting the angular orientation of a wind turbine 1 uses the detection of the wake effect and / or the blocking effect of the calibration means.

[0136] According to one embodiment, conversely, the detection of the wake effect makes it possible to calibrate the yaw orientation of the nacelle 102 of the wind turbine 1 and / or to determine the main direction 201 of a locally prevailing wind. This calibration can be carried out on any time series of data.

[0137] Furthermore, it should be noted that angles are read around a circle in increasing values ​​from a reference point in a clockwise or counterclockwise direction. Usually, the clockwise direction is used by convention. Conversely, Figures 7 to 9 show a reading in a counterclockwise direction.

[0138] According to one embodiment, a series of data relating to yaw alignment errors occurring over a period, for example over a period of 10 minutes.

[0139] Additionally, filtered and calibration data can be used, as described later.

[0140] Then, the determination module 23 can calculate an average of the yaw alignment errors can be calculated for each wind turbine 1, or else break down into groups all the errors observed over a configurable time interval, for example monthly.

[0141] Therefore, said control means modify the angular orientation of each nacelle 102 as a function of said yaw alignment error, to compensate for said differential 401 between the angular position 400 and the main wind direction 201.

[0142] It should be noted that this modification involves the control module to change the orientation of the nacelle 102, either in real life directly on site or virtually within a simulation.

[0143] Said analysis and optimization module 20 then estimates a corrected yield, on the basis of this new orientation, namely an updated yield whose calculation is based on the new orientation.

[0144] Thus, the supervision system 2 makes it possible to adjust the actual or theoretical orientation of the nacelles 102 of each wind turbine 1 dynamically, in particular by checking the presence of an angular differential 401 and correcting it if it exists.

[0145] For example, in the case of a wind turbine 1, the value of the angular position 401 corresponding to the actual orientation of the nacelle 102 is compared to the value of the component 200 of the main direction 201 of the wind. The latter is considered to be correct within the system 2 according to the invention, namely that it is identical to reality. In other words, in the case of a simulation, the predicted direction of the wind corresponds or will correspond to the actual direction of said wind.

[0146] At first glance, over a given time range (i.e. several periods, for example of one minute each), the difference between the actual orientation and the wind direction is constant, due to the time required for the nacelle 102 to align itself during a change in said wind direction. However, this difference has significant deviations.

[0147] Indeed, the actual orientation of the nacelle 102 is considered to be known, in particular within the SCADA, from data originating in particular from the operating specifications 112 of each wind turbine 1, in particular through an angular position sensor located at the junction between the mast 101 and the nacelle 102 of the wind turbine. The sensor located on said nacelle 102 only measures said direction and has measurement inaccuracies, in particular due to the turbulence of the wind turbine 1 and the wake and / or blocking effects between wind turbines 1. Furthermore, said sensor does not take into consideration the actual orientation of the nacelle during its measurements.

[0148] The system 2 therefore provides for verifying that the known orientation corresponds to the actual orientation of the nacelle 102, so as to verify the presence of a differential 401 and to compensate for it, making it possible to minimize the differences between the measurements and reality. In short, by calculating the differential 401, the supervision system 2 indicates to the control means whether or not to intervene on the orientation of the nacelle 102, or whether it is appropriate to add an offset value to be taken into consideration when controlling the rotation of said nacelle 102.

[0149] Thus, the supervision system 2 calculates a possible yaw alignment error and indicates a corresponding correction so that the control means take it into consideration during its actions on the wind turbine 1.

[0150] According to one embodiment, the supervision system 2 takes into consideration the impacts between several wind turbines 1 in the same park.

[0151] To do this, a first wind turbine and at least one second wind turbine form a wind farm located on said site. Said calibration means then ensure the detection of a wake effect between the first wind turbine and at least said second wind turbine.

[0152] According to a combined or alternative embodiment, said calibration means then ensure the detection of a blocking effect between said first wind turbine and at least said second wind turbine.

[0153] To achieve the detection of one or more wake effects (and / or blocking) of at least two wind turbines 1 between them, the supervision system 2 is based on a library of wake models in the form of a collection of numerical models and their parameters, allowing different calculations of the wake effects. The wake effects are defined as the disturbance of the components 200 characterizing the winds, but also the atmosphere, by the aerodynamic effects linked to the presence of the wind turbines 1. As mentioned previously, by extracting energy from the winds, a wind turbine 1 reduces the wind speed and increases the turbulent intensity of the winds in their vicinity, which can affect the other wind turbines 1 of a farm.The term wake effects is used and represents a broader range of interaction phenomena between the wind farm and the atmosphere on site, including for example phenomena of blocking and acceleration of fluid flows, such as air, or even gravity waves. Wakes are significant effects likely to cause between 5% and 30% of production losses.

[0154] On a site, the intensity of the wakes depends on the 200 wind components and atmospheric parameters, initially considered and impacted by several factors related to each wind turbine 1 , such as its operating regime as well as its thrust curve. Although recommendations exist on the use of models to achieve reasonable results, there is no standardized method. A wide variety of models are used in the industry, presenting strong variations in results on the prediction of the resulting efficiency, which can reach deviations of up to 50% on the loss caused by wake effects depending on the models.

[0155] According to one embodiment, the wake models are calculated with a limited number of conditions, in particular after a reduction of the data series over a period with statistical discretization, for example by means of a Weibull type distribution by wind sector.

[0156] Further, the determination of wake effects as a function of time and space is carried out on the same spatial and temporal grid as the unsteady wind field. For each period, notably a period of 10 minutes, a calibrated free wind condition is provided as input by the wind model. The wake model determines as output the disturbance of the wind parameters at each wind turbine 1 and as well as the impact on each wind field. This disturbance essentially results in a reduction in speed and an increase in turbulent intensity. The wakes are therefore materialized by a modification of the components 200 from the wind model.

[0157] Therefore, as input, the wake models receive several data, namely, but not limited to, the free wind mesh, and possibly information on the terrain, the geographical layout and the characteristics 110 of each of the wind turbines 1, in particular their dimensions, the thrust curves of each wind turbine 1, the operating laws of each wind turbine 1, such as the generic restraint or control laws, knowledge of the yaw misalignment of each wind turbine 1, likely to affect the thrust, as well as the redirection of the wakes.

[0158] Wake effects are then calculated by analytical empirical models, notably available in accessible libraries, such as PyWake or FLORIS based on a Python-type computer language. These empirical models include unit wake deficit models, such as the NO Jensen model or the Bastankhah M and Porté-Agel model, as well as blocking models, such as the cylindrical vorticity sheet model (i.e. modeling a vortex in fluid mechanics).

[0159] Then, a superposition model is used to sum the unit deficits of the obtained results.

[0160] The wake effects can then be calculated by CFD (Computational Fluid Dynamics) type tools with an action disk model representing each wind turbine 1 , or WRF (Weather Research Forecasting) type tools, or by machine learning means based on simulated or real data.

[0161] Each of these approaches has a wide variety of options and parameters that lead to different prediction results. The wake model library allows for the grouping of different methods, offering a collection tailored to the site and wind farm and offering a range of configurations to obtain different predictions of wake effects.

[0162] In this context, as wake physics is quite complex and sensitive to the environment, in order to obtain the most accurate prediction of wake effects, the supervision system 2 uses a calibration module, with in particular the operating data from the SCADA as input.

[0163] First, these data are strongly cleaned, accepting a strong filtering, in order to preserve samples where the maximum number of wind turbines 1 are in normal operation. This filtering is carried out using techniques that filter out non-normal operating modes, detecting and removing outliers. The results obtained are processed according to several variables, essentially three variables, in order to build metrics that allow qualifying the performance of the wake model. A first variable can be the power produced by each wind turbine 1. This data has the advantage of being reliable data and of high interest to the operator. In other words, from an operational point of view, a wake model that best predicts the power changes of the wind turbines 1 is all the more relevant and useful.However, a drawback, from a calibration point of view, is that this data requires a power prediction, making an assumption about a downstream element of calculation, which can bias the results. This problem can be solved by carrying out a sensitivity study or successive iterations, to filter and refine the results.

[0164] A second variable can be the speed measured by the sensor of the nacelle 102, namely the anemometer. This measurement has the advantage of being a speed, but the measurements are not necessarily reliable, even when applying its transfer function which can be variable depending on atmospheric conditions. In addition, it is a local sensor, while most wake models provide valid speed variations across the entire rotor of a wind turbine 1. Furthermore, there is the risk that a wake model which is accurate in predicting the wind measured by the anemometer of the nacelle 102, is not valid for predicting the power of the wind turbine 1. This second variable can therefore be used in a combined manner, to refine a result obtained from another variable.

[0165] A third variable can be an estimated equivalent wind speed (EESV) using data from wind turbine 1. This option is attractive and combines the advantages of the other two variables, through an estimation of a speed that reflects the average wind across the entire rotor.

[0166] Several approaches are possible to achieve such a VVEE.

[0167] A first approach consists of having a physical model of the dynamics of the wind turbine 1, on the basis of which a table is constructed relating the power coefficient, the rotation speed normalized by the incident speed and the blade pitch angle 104. It is then possible to invert a series of equations to determine the incident speed.

[0168] A second approach consists of building an empirical model according to an equation of the type VVEE = F (power, blade angle, rotation speed), then calibrating this empirical model using a reference speed data on a part of the data, specifically outside the wake. In this case, the reference target used can be either the anemometer speed of the nacelle 102, or the speed predicted by the model for this wind turbine 1. Then, the next step consists of building metrics, allowing to quantify the performance of the models of the library. For each pair of the model and the parameter to be investigated, a prediction is made and compared to a series of filtered validations from the SCADA. These metrics group errors, in the form of absolute value of the average error, average of the absolute error or squared error, said errors being aggregated on the difference between a prediction and reality.For the validation variable, these metrics are based on the direct comparison between the predicted data set and the observed data, as well as on the comparison between the predicted and observed aggregated form factors. For example, to highlight wind turbines 1 with significant wake effects, a ratio can be made between the time-averaged variable on wind turbine 1 and the time-averaged variable, reported across all wind turbines 1. A ratio can also be made between the time-averaged variable on wind turbine 1 and the time-averaged variable per wind sector, reported across all wind turbines 1. A definition of “out-of-wake” conditions can be defined as a ratio between the averaged variable outside a wake of wind turbine 1 and the variable within said wake.

[0169] A final step is to conduct a series of experiments to determine the best pair(s) of model and parameter. This series of experiments can be based on a search grid to find the best set of parameters. Specific experiments can be conducted to determine the dependence of certain parameters on atmospheric conditions, in order to define a variation law for one or more of these parameters. A selection is made for a pair based on several criteria, or a composite criterion. In addition, the selected model can also include a combination of several pairs. The variation between the results of several pairs provides a basis for estimating the uncertainty in modeling wake effects.

[0170] A similar method can be used for blocking effects.

[0171] Then, said analysis and optimization module 20 estimates a new efficiency or a corresponding corrected efficiency for said second wind turbine, based on the wake effect and / or the blocking effect.

[0172] As such, the electrical power of each wind turbine 1 is expressed as a value in Watt (W), or kiloWatt, or MegaWatt, and the electrical energy production is expressed in kiloWatt-hour (kWh), MegaWatt-hour (MWh), GigaWatt-hour (GWh) or TeraWatt-hour (Twh), depending on the duration of the determined period. This production is therefore represented by a time series, obtained by an electrical production model specific to the types of wind turbines 1 present on the site, namely according to their characteristics 110. As input, these models receive the values ​​of the components 200 of the winds, but also of characteristics of the atmosphere. As output, they generate production values.

[0173] According to one embodiment, a model corresponds only to a power curve, in particular in the form of a table indicating the electrical production as a function of the wind speed. Such tables can be subject to other components 200 of the winds and characteristics of the atmosphere.

[0174] Thus, by interpolating within this table, a wind speed makes it possible to predict the power produced. For the relationship of the power curve, the relationship of the reference curve of the manufacturer of wind turbine 1 makes it possible to give the ideal and theoretical power, while the operational power curve, which represents in a more realistic way the actual operation observed of wind turbine 1, makes it possible in particular to obtain a comparison and an aging state of said wind turbine 1.

[0175] Other variables can be included in the wind prediction series, such as the 200 component of the reference wind speed, air density, turbulent intensity, or shear. Yaw misalignment with respect to wind direction is also a parameter of this model, since misalignment causes a loss of production, which can be quantified, as described later.

[0176] Other empirical, statistical, and data-learned models can also be used.

[0177] Then a predictive model is generated, in particular by learning, to predict the power produced and recorded to be accessible, in particular within the SCADA.

[0178] Finally, the power produced for each wind turbine 1 may depend on the curtailment schemes decided for the wind farm's operating plan. These curtailment schemes are often dependent on certain wind components, such as the main wind speed and direction, as known or predicted for each wind turbine 1.

[0179] Further, as described later, a nacelle transfer function 102 may be used to model said speed and said direction 201, in order to simulate the behavior of each wind turbine 1.

[0180] Thus, using these models, the electrical production is then calculated for each wind turbine 1 , considering or not the wake effects.

[0181] Concerning the measurements carried out, targeting in particular one or more components 200 of the winds, the supervision system 2 provides several treatments, in order to refine and filter the measured values.

[0182] According to one embodiment, said supervision system 2 comprises a module for integrating the successive measured values ​​of at least said main direction 201, over one or more time periods, into at least one time series of data.

[0183] In particular, the integration module can take into consideration measured values ​​of at least said main direction 201 over two time periods, for example a previous period and a current period, in order to compare, combine and / or average them.

[0184] According to one embodiment, the integration comprises data filtering, making it possible to select, for each wind turbine 1 , a portion of the data considered reliable for the performance analysis in normal operation. In other words, the input data are filtered by choosing criteria, such as a production period, events or even the operational statuses of each wind turbine 1 . In short, this involves excluding periods during which the wind turbines 1 are not in normal operation, for example by using the statuses indicated for each wind turbine 1 , then if necessary, by carrying out a series of automatic filtering of the outliers, such as a power peak at start-up. Optionally, the filtering may also include an exclusion of the presence of wakes, by excluding directions specific to each wind turbine 1 when it is subjected to the wake of one or more neighbors 1 .

[0185] In addition, said supervision system 2 includes a module for calibrating the data of each time series.

[0186] In other words, for each time period, a set of values ​​of the measurements carried out on site, before or after installation of the wind turbines 1 , are recorded digitally in the form of uncalibrated data at the input of the calibration module, which processes this data and generates calibrated data at the output, in the form of a predictive digital model of calibrated time series of the wind.

[0187] According to one embodiment, such a model of calibrated time series makes it possible to calculate one or more fields of winds and their components 200, as well as parameters of the atmosphere, preferably of the unsteady atmosphere.

[0188] To do this, said supervision system 2 comprises a module 24 for generating, from said at least one time series of data, at least one local virtual atmospheric model 211, corresponding to a three-dimensional mesh of an area of ​​said site comprising at least said wind turbine 1.

[0189] Thus, as input, a wind field is defined according to a three-dimensional mesh covering the site, with a horizontal resolution ranging from 20 to 200 m, as well as a vertical resolution ranging from 5 to 50 m. The data of the 200 components of the winds in a field of said winds are then summarized in a network or cloud of points, defined according to the position of the wind turbines 1.

[0190] Such a wind field can be obtained from a 210 meteorological model with a sufficiently fine spatial resolution. For example, a 210 meteorological model of the WRF type can be used or coupled with a numerical model of the CFD type, used in fluid mechanics for the simulation of flows in complex configurations, namely within the site depending on its topography and infrastructures, in particular wind turbines 1.

[0191] These models are based on the numerical resolution of the equations of state and conservation of atmospheric physical variables. They require significant computing resources, such as supercomputers, notably based on architectures such as Central Processing Unit (CPU), Graphical Processing Unit (GPU) or CPU-GPU hybrids.

[0192] These models also receive as input data on the topography of the site, such as a digital elevation map of the terrain, on land use, as well as data from global meteorological analysis services by large-scale planetary circulation models. According to one embodiment, these models based on the digital resolution of atmospheric physics equations, for example of the NWP type (for "Numerical Weather Prediction" or numerical weather prediction) can be replaced by artificial intelligence models, preferably with automatic learning, in particular supervised or semi-supervised, on the basis of said NWP models.

[0193] These input meteorological data have a coarse and global spatial resolution, of the order of 30 km (kilometers), which should be refined in relation to the site of a wind farm. Indeed, an uncalibrated model generates prediction uncertainties, particularly on the average wind speed and velocity, of the order of + / - 0.5 m / s, or even more in complex terrain cases. Such uncertainties translate into significant yield deviations, which can be as high as 15%.

[0194] Therefore, said calibration is carried out to refine with sufficient precision the values ​​of the wind components on the site. Such a calibration uses values ​​of the 200 wind components measured on the site during a pre-construction phase, before installation of the wind farm. In particular, the absence of wind turbines 1 in operation ensures that these preliminary measurements are strictly linked to the site, not being impacted by the operation of said wind turbines 1, in particular their wake effect. These preliminary measurements then do not require the installation of additional measuring means after installation, the measurements of which would be likely to be impacted by the operation of the wind turbines 1.

[0195] In a complementary manner, in particular if there is no data relating to wind measurements during the pre-construction phase on the site, namely measurements taken before installation of the wind farm, the calibration is carried out from measurements taken on said site after installation, in particular from sensors installed on the nacelle 102 of each wind turbine 1 or on the site from sensors independent of the wind turbines 1. This step is carried out during a preliminary operation or pre-operation phase, during operation of said wind turbines 1, in order to obtain the necessary data as input to said integration and calibration modules.

[0196] This calibration therefore makes it possible, from preliminary on-site measurements as input, to obtain refined predictions, on the scale of said wind farm, of the order of a meter to a kilometer. Then, the calibration is used during the operating period of said wind farm, to correct the predictions.

[0197] Further, the calibration covers several components of the winds, in particular at least the speed, direction and sense of the winds, as well as the turbulent intensity.

[0198] According to one embodiment, the calibration consists of adjusting the predicted series by an overall multiplicative factor, corresponding to a ratio between the measured speed and the predicted speed during the calibration period, namely during the pre-construction phase or the pre-operation phase. For example, such an adjustment of the calibration can be obtained by matrix methods of the MCP type (for "measure correlate predict"), consisting of linear regressions in least squares by wind sector. These techniques relate to speed and direction.

[0199] Another method is to formulate the calibration as a regression problem according to a function of type Y=F(x), with "Y" equivalent to the time series of the measured target variables, with "x" equivalent to the time series of the descriptive variables coming from the uncalibrated simulation prediction. The variable "x" can also contain spatial dependencies, in the case where several masts 101 would serve as reference in the measurements.

[0200] Furthermore, these regression methods can be implemented with machine learning libraries accessible to said calibration module.

[0201] The calibrated wind model then results from the composition of the function F with the uncalibrated wind model.

[0202] In addition, at the output, the calibrated series can be provided to a supervisor according to several operating scenarios that can be configured and varied, for example through an expected scenario by following the planned operation, an optimized scenario by adjusting in particular the orientation of the wind turbines 1, or even a degraded scenario, depending on climatic hazards or planned maintenance.

[0203] Then, said analysis and optimization module 20 estimates a predictive yield, generating a predictive simulation of said performance of each wind turbine 1. This predictive yield is therefore based on the calibrated data, but also on the differential 401 and the correction of the angular position 400, as previously described (i.e. on said updated or corrected yield), in a real or virtual manner.

[0204] Thus, the use of a chain of several calibrated series predictions makes it possible to determine actions aimed at improving electricity production, but also reducing operating costs.

[0205] Once the chain is ready with calibrated series, predictions can be made to cover a configurable duration, ranging from a few months to several years. In short, several versions of the same simulation are carried out, in order to evaluate potential gains and compare them, allowing to justify the implementation or not of the action to be taken. Based on these series, several results are established, which lead to improvements in the operation of the wind farm.

[0206] According to one embodiment, a gain in production and therefore efficiency can be determined by detecting and correcting yaw misalignment errors, namely that operation of a wind turbine 1 with yaw misalignment leads to a loss, which is proportional to a cosine of the value of the alignment error angle raised to a power of a non-zero real number.

[0207] Therefore, in order to determine the possible actions to be taken, a first simulation is carried out with the current alignment errors for one or more wind turbines 1 , then a second simulation is carried out on the same wind turbine(s) 1 , with said errors corrected. Thus, a gain in said efficiency is determined for said wind turbine(s) 1 and the alignment correction action(s) are transmitted to the operator.

[0208] It is also possible to quantify the efficiency gain, in particular in relation to a defined threshold, for example in percentage, in particular from one to several tenths of a percent. Such a threshold can be determined as a function of the energy required to motorize a wind turbine 1 to correct the alignment error of its nacelle 102, with regard to said gain.

[0209] Therefore, for the wind turbine(s) 1 where the gain is greater than said threshold, one or more corrective actions are transmitted to the operator, for example in the form of a change in a control parameter of each wind turbine 1 concerned, or by maintenance to be carried out on said wind turbine 1, in particular by a modification of its wind vane used to index its angular position 400 in the reference frame 300.

[0210] Furthermore, after a given period, system 2 makes it possible to visualize the actual gains compared to the predicted gains, for each wind turbine 1 on which an action has been carried out or an action has been determined, but not necessarily carried out.

[0211] According to one embodiment, a gain can be determined by improving the restriction schemes of one or more wind turbines 1. Predictions are made on the restriction parameters of the wind turbines 1, then the results are compared to the actual plans already completed, in order to determine potential deviations and to propose actions leading to corresponding gains.

[0212] It is then possible to determine several groups of gain types, for example as described below.

[0213] A first group of gains includes gains resulting from losses related to poor implementation of the throttling plan. To identify this case, the expected and known losses when throttling wind turbine 1 are compared with the predicted losses. If the predicted loss is significantly higher than the expected loss, the throttling plan is verified, for example by comparing the predictions and results during periods when each wind turbine 1 is throttling.

[0214] A second group of gains includes gains from losses linked to poor execution of the initially planned, potentially correctly implemented, containment plan.

[0215] For example, the poor execution of the bridle plan may be linked to a sensor on the site that is not calibrated correctly or that is faulty. In the case of a bridle plan dependent on the azimuth of the nacelle 102 and the wind speed measured on said nacelle 102, but the measurements are biased, then the loss may be greater than the loss incurred. Given that the calibration according to the invention makes it possible to compensate for the angular orientation of the nacelle 102, or even the measurement of the speed at said nacelle 102, the prediction makes it possible to simulate the effect of any deviation or error, by comparing the losses linked to the bridle, with respect to the losses predicted if said deviations or errors were corrected. In particular, if the amplitude of these deviations exceeds a certain threshold, the supervision system 2 transmits an action to the operator, recommending the calibration of the corresponding sensors, in order to achieve the predicted gains.Such calibration of sensors can be carried out by remote actions or on site.

[0216] A third group of gains includes supposed gains in the case of correct implementation and execution of the throttling plans, with in addition an optimization of the throttling plan, taking into account the real effects of this throttling on the environment.

[0217] According to one embodiment, conversely, a loss can be determined from the determined wake effects of several wind turbines 1 between them.

[0218] Therefore, for each wind turbine concerned, one of the deficits linked to wake effects makes it possible to calculate aggregated averages, for example over a monthly period. These averages allow the operator in particular to explain the monthly production, in order to financially optimize the operating budget of the wind farm.

[0219] According to one embodiment, it is possible to perform detection and quantification of changes in performance of one or more wind turbines 1.

[0220] Therefore, for each wind turbine 1 concerned, a prediction is made with a predicted yield and is compared to a past actual yield, determining a deviation. Aggregated averages of the deviations make it possible to identify performance groups between several wind turbines 1. Wind turbines 1 with an unfavorable deviation, in particular compared to a defined threshold, are considered to be in relative underperformance. Specific investigations can be carried out, remotely or on said site, in order to resolve a potential problem common to several wind turbines 1, in relation to the software of the control means, a construction or installation defect, premature aging of an element, etc. Knowledge of other components 200 of the winds also makes it possible to diagnose part of the problem.

[0221] Thus, it is possible to rank each wind turbine 1 in a park, allowing performance improvements or degradations to be quantified. Such a ranking makes it possible, in particular, to verify that a correction of yaw alignment error to be made is likely to generate the predicted benefits.

[0222] According to one embodiment, it is possible to carry out an absolute measurement of the performance of one or more wind turbines 1.

[0223] First, a predicted power curve is obtained from the predicted wind speed, and then compared with the actual power produced. This predicted power curve is also compared with the reference power curve of each wind turbine 1 .

[0224] Therefore, if the predicted power curve is lower than the actual curve, the wind turbine 1 product is likely to produce less than according to the reference power curve. This verification to confirm an absolute underperformance by an independent measurement, and possibly to detect a construction or manufacturing defect, with a view to solving such a problem.

[0225] According to one embodiment, it is possible to perform a prediction of long-term yields, affected with various of the aforementioned scenarios. These predicted yields can be aggregated over a determined period, for example monthly, then a long-term extrapolation is carried out, in particular by using a long-term reanalysis source, for example by following an Operational Energy Yield Assessment approach (for evaluation of the operational energy resource). Such a method can be repeated by simulating variations on the specification of each wind turbine 1 , for example with significant updates to the control software of certain wind turbines 1 which impact their power curve, updates to the bridling scheme, the replacement of one or more wind turbines 1 on the corresponding location, the installation of new wind turbines 1 within the wind farm, and / or the dismantling of certain wind turbines 1 .

[0226] Further, according to one embodiment, said integration module generates said time series of data prior to the installation of each wind turbine 1. The supervision system 2 then has a fully predictive virtual simulation character.

[0227] According to an alternative embodiment, said integration module generates said time series of data during the operation of each wind turbine 1, preferably during a pre-operation phase. The supervision system 2 then makes it possible to verify proper operation, over a past, present, but also future period, with a predictive simulation character.

[0228] Furthermore, it will be noted that the predictive nature can be considered recursively, with successive iterations or loops of the output simulations on the input of said supervision system 2, depending on one or more of said scenarios.

[0229] According to one embodiment, said system 2 comprises a module for calculating transfer functions of the measuring means in the form of an anemometer type sensor mounted on said nacelle 102, as a function of said virtual atmospheric model 211.

[0230] Such transfer functions make it possible to predict the “free” wind from an altered wind, taking into account the turbulence and any modification of the components 200 impacted by the operation of the wind turbine 1 , in particular the turbulence generated by the passage of the blades 104, which impact the measurements made by the anemometer positioned downstream on the nacelle 102 of said wind turbine 1 , but also on the nacelle 102 of another wind turbine 1 undergoing a wake effect and / or a blocking effect. Indeed, as mentioned previously, the winds are impacted near a wind turbine 1 , distorting the measurements. In particular, the winds undergo a slowdown downstream, induced by the energy conversion, which can be as much as a 30% drop.

[0231] To determine these transfer functions, preliminary tests are carried out, in particular by positioning a sensor at the top of a mast in front of a wind turbine 1. Thus, the difference in measurements between this upstream sensor and the downstream sensor installed on the nacelle 102 makes it possible to obtain a function of the type mast_speed = F(nacelle_speed). Such a function can be linear or polynomial, obtained by a regression, in particular a least squares regression.

[0232] According to a preferred embodiment, the transfer function is adjusted, by replacing the “mat_speed” by the speed of the 200 component of the winds within the wind mesh generated at the output of the calibration module. Such an adjusted transfer function makes it possible to adapt the conversion of the winds in relation to the reality of the site, before or after installation, by doing away with a sensor positioned in front of each wind turbine 1.

[0233] As previously described, the time period can be predefined to a duration ranging from a few seconds, for example ten seconds, to one or more minutes, or one or more hours, or one or more days.

[0234] Several time periods can be combined, in particular by concatenation, to obtain a longer period, of one or more days, or one or more weeks, or one or more months, or even extending over one or more years.

[0235] For example, the angular position 400 of each nacelle 102 can be updated automatically by the controller and the actuators of the wind turbine 1, which seeks to follow the direction of the wind by changes in orientation occurring every minute. The supervision system 2 then performs a closed control loop, over a time period of one minute.

[0236] Thus, the supervision system 2 according to the invention makes it possible to compensate for degradations and errors which occur over time, in particular to compensate for a shift in the angular orientation of the wind turbines which can reach several degrees.

[0237] The supervision system 2 then makes it possible to optimize the performance of a wind farm, existing or future, by implementing a digital twin generating calibrated time series of winds, wakes and electrical production, for each wind turbine 1 , during an operating period. The time series are produced by implementing numerical and virtual models of the winds, the interactions of the farm with the atmosphere and the electrical production. The calibration is used to adjust the virtual representations of the winds, with calibrated measurements of their components 200, with an adjustment of the wake models. The time series generated in a calibrated manner makes it possible to quantify and improve the performance of the wind turbines 1 , by carrying out actions which will increase the electricity production of the wind farm.In particular, the specification 112 for operating the angular orientation of each wind turbine 1 makes it possible to detect and correct yaw misalignments of the wind turbines 1.

[0238] The supervision system 2 also allows the improvement of the definition of wind farm restriction plans.

[0239] As mentioned above, the supervision system 2 is implemented as a software service, in the form of software and data access, which is made available to wind farm operators. The data is regularly routed to an interface, enabling supervision and monitoring, as well as decision-making, providing visibility to operators.

[0240] As input, from the user's point of view, the system 2 allows to select a time range of interest during operation, for example the last month or the last year. As output, the user can visualize for the park as a whole but also for each wind turbine 1, a series of data distributed temporally over periods of interest, typically at a time step of 10 minutes, cutting out said time range and corresponding to a part of the operating phase.

[0241] In particular, the supervision system 2 makes it possible to view and choose a time-indexed series for each wind turbine 1. This indexed series may correspond to one or more operating configurations of each wind turbine 1, in particular one or more angular orientations of the nacelle 102. It is also possible for the user to vary this factor, in order to simulate yields in a predictive manner.

[0242] Other information is accessible to the user in this indexed data series, such as the predicted power and the 200 components of the winds and the atmosphere, provided at different heights above the ground, in particular at the height of the hub of wind turbine 1, and at other heights between 20 m and 300 m, allowing the winds to be quickly visualized locally at each wind turbine 1.

[0243] In addition, the wind components 200, such as the main speed and direction 201, and the atmospheric variables, such as air density, temperature, pressure and humidity, and atmospheric stability, inclination angle, are complemented by impacting parameters, such as turbulence intensity and shear. These parameters can be provided with the presence of a wake effect and without the presence of a wake from one or more wind turbines 1.

[0244] Thus, by comparing the predicted time series and the time series obtained, for example from the SCADA, the supervision system 2 allows a user to obtain different results, such as a correction of the yaw alignment errors of the nacelles 102, with an estimation of the gains, or a correction of the transfer functions of the anemometers installed on the nacelles 102, with gains estimated via an update of the bridling diagrams. This comparison also makes it possible to obtain an analysis of the correct implementation of the bridling plans, with a proposal for corrective action.

[0245] The supervision system 2 allows an estimation of absolute performance, making it possible to compare current production with expected production, according to the wind turbine operating specifications 112, and to highlight deviations and some causes, to adjust the results, in particular by providing an estimation of wake losses during the operating period.

Claims

CLAIMS 1. System (2) for supervising at least one wind turbine (1) to be installed and / or installed on at least one production site, - each site being swept by winds, each of said winds having components (200), including at least one main direction (201) of a locally dominant wind in a reference frame (300) and a speed; - each wind turbine (1) comprising a mast (101) receiving at its upper end a nacelle (102) supporting a rotor formed of at least one hub (103) on which blades (104) are mounted, with an electric generator supplying energy according to a power to a production network, each wind turbine (1) having technical characteristics (119) and an operating configuration (111) with at least one angular orientation of each nacelle (102) in said reference frame (300); said supervision system (2) comprising - means for measuring over at least one time period at least said main direction (201) among said components (200) of said winds; said measuring means being installed on said site; - means for controlling the configuration of each wind turbine (1), allowing at least the modification of said angular orientation of said nacelle (102) relative to the main direction (201) measured; characterized in that said supervision system (2) comprises - means of detecting said angular orientation relative to geographic north: - a module (23) for determining at least one angular differential (401) from said measured main direction (201) and said detected angular orientation, corresponding to a yaw alignment error relative to the main direction (201) of the locally prevailing wind; - said control means modifying the angular orientation of each nacelle (102) as a function of said yaw alignment error.

2. Supervision system (2) according to the preceding claim, characterized in that it comprises - a module (20) for analyzing and optimizing the performance of each wind turbine (1), estimating a production yield for said wind turbine (1).

3. Supervision system (2) according to any one of the preceding claims, characterized in that said detection means comprise - a module (22) for detecting said angular orientation of each wind turbine (1) i) by acquiring at least one photo / video capture, preferably by high-resolution satellite imagery, with a fineness of at least 1 meter, preferably a fineness of less than 0.5 m, more preferably a fineness of less than 0.3 m; ii) then by cutting each image into at least one thumbnail comprising each wind turbine (1); iii) then by determining the angular orientation of each wind turbine (1) in each thumbnail.

4. Supervision system (2) according to the preceding claim, characterized in that - from the elements visible on an image captured at a given time and a given geographical location and / or on said thumbnail, it is possible to combine and / or compare characteristics of the wind turbine (1) with characteristics of its cast shadow, in order to determine the orientation of the nacelle (102) of a wind turbine (1).

5. Supervision system (2) according to any one of the preceding claims, characterized in that - a first wind turbine (1) and at least one second wind turbine (1) form a wind farm installed on said site; the system comprising calibration means which ensure the detection j) of a wake effect between the first wind turbine (1) and at least of said second wind turbine (1), and / or jj) of a blocking effect between said first wind turbine (1) and at least of said second wind turbine (1); - the module (22) for detecting the angular orientation using the detection of the wake effect and / or the blocking effect of the calibration means.

6. Supervision system (2) according to any one of the preceding claims, characterized in that it comprises - a module for integrating the successive measured values ​​of at least said main direction (201), over one or more time periods, into at least one time series of data; - a module for calibrating the data of each time series; - said analysis and optimization module (20) estimating a predictive yield, generating a predictive simulation of said performance of each wind turbine (1).

7. Supervision system (2) according to the preceding claim, characterized in that k) said integration module generates said time series of data prior to the installation of each wind turbine (1); or kk) said integration module generates said time series of data during the operation of each wind turbine (1), preferably during a pre-operation phase.

8. Supervision system (2) according to any one of the preceding claims 6 or 7, characterized in that it comprises - a module (24) for generating, from said at least one time series of data, at least one local virtual atmospheric model (211), corresponding to a three-dimensional mesh of an area of ​​said site comprising at least said wind turbine (1).

9. Supervision system (2) according to the preceding claim, characterized in that it comprises - a module for calculating transfer functions of the measuring means in the form of an anemometer type sensor mounted on said nacelle (102), as a function of said local virtual atmospheric model (211).

10. Supervision system (2) according to the preceding claim, characterized in that it comprises - a parameterization module for at least one curve of each wind turbine (1) representative of said power, as a function of said speed of said winds and as a function of operating specifications (112) of said wind turbine (1).

Images