Control methods for wind power plants

Abstract

To provide wind turbine control that maintains efficiency while consuming little to no computing power. [Solution] Each wind turbine has an operating point adjustable by an actuator, a) model a spatial region approximating the wake of each wind turbine with respect to a range of misalignment angles and a predetermined threshold for insufficient wind speed, b) identify wind turbines in the modeled wake of another wind turbine based on the modeled wake, and store identification parameters that associate each wind turbine with the wind turbines in the modeled wake where it is located, c) consider the identification parameters stored in step b) in order to cluster the wind turbines into a plurality of subfarms, d) optimize the power generated by each subfarm by setting at least one operating setpoint for the wind turbines in each subfarm, and e) apply the operating setpoint to the wind turbine.

Description

[Technical Field]

[0001] This invention relates to the field of controlling wind power plants to maximize power generation and reduce wind turbine fatigue. [Background technology]

[0002] A wind farm (also called a wind park or wind power plant) is a facility that generates electricity using multiple wind turbines. These facilities are installed on land or at sea. Therefore, they are distinguished as onshore wind farms and offshore wind farms (i.e., offshore wind power plants).

[0003] The wind turbines in these wind power plants are generally horizontal-axis wind turbines, and they are equipped with a system that aligns the horizontal rotation axis with the wind direction to maximize the energy the wind turbine obtains. Wind turbines can convert the kinetic energy of the wind into electrical or mechanical energy. To convert wind energy into electrical energy, wind turbines consist of the following elements:

[0004] - A tower (required for horizontal-axis wind turbines) that allows the rotor to be mounted at a sufficient height to enable its movement, or a tower that allows the rotor to be mounted at a height where it can be driven by stronger and more regular winds than those at ground level. This tower may house some of the turbine's electrical and electronic components (modulator, controller, gearbox, generator, etc.). - The nacelle, located at the top of the tower, houses the mechanical, pneumatic, and some electrical and electronic components (modulator, controller, gearbox, generator, etc.) necessary for the machine's operation. The nacelle rotates to orient the rotor in the correct direction. - A rotor fixed to a nacelle, consisting of multiple blades (usually three) and a wind turbine nose cone. The rotor is driven by wind energy and connected directly or indirectly (through a system including a gearbox and a mechanical shaft) to an electric motor (such as a generator) that converts the collected energy into electrical energy. The rotor may be equipped with control systems such as variable-angle blades and aerodynamic brakes. -Optionally, a transmission consisting of two axles (the rotor's mechanical shaft and the motor's mechanical shaft) connected by a transmission (gearbox).

[0005] Since the early 1990s, interest in wind energy has grown, with an annual growth rate of approximately 20% in the European Union (EU) in particular. This growth is due to the inherent ability of wind energy to generate electricity without emitting carbon. To maintain this growth rate, it is necessary to continuously improve the efficiency of wind turbines and wind farms. Increasing the amount of electricity generated from wind energy requires the development of effective production tools and advanced control tools to improve mechanical performance. Wind turbines are designed to generate electricity at the lowest possible cost.

[0006] Control devices have been designed for variable-speed wind turbines to regulate the power generated. The purpose of these devices is to maximize power generation, minimize fluctuations in rotor speed, and minimize fatigue and overload on the structure (blades, tower, platform).

[0007] Wind power plants are generally affected by what is called the "wake effect." This is a phenomenon in which turbulence generated by turbines located upstream of a wind power plant prevents other turbines from maintaining optimal power generation conditions. Specifically, a swirling wake is formed downstream of the wind turbine, and in this wake, some of the wind's kinetic energy is absorbed by the wind turbine, increasing the turbulence level and thus reducing the average wind speed. (The terms "upstream" and "downstream" refer to the position of a wind turbine relative to other wind turbines in the direction of the prevailing wind at a particular point in time.)

[0008] A known strategy for maximizing the energy production of a wind turbine is to orient the rotor in the direction of the wind. In this case, the angle between the rotor and the wind direction (called the misalignment angle or yaw angle) is 0°. Figure 1 schematically and non-restrictively illustrates the misalignment angle. Figure 1 is a bird's-eye view of a wind turbine. The wind turbine comprises blades 1 and a nacelle 2 oriented in direction AA. The wind is represented by an arrow U with direction DD. The angle θ between direction AA and direction DD is the misalignment angle. When the turbine of the wind turbine coincides with the wind direction, the angle θ is 0.

[0009] However, when this strategy (misalignment angle θ = zero) is applied to all turbines using the so-called "greedy" method in a wind power plant, it is affected by the so-called "wake effect." The wake effect is a phenomenon in which the downstream wind speed decreases and turbulence increases as the wind turbine extracts energy from the wind. As a result, conditions become less than optimal for energy production by downstream turbines, and the total production loss offshore could reach 40%.

[0010] To mitigate this impact, a certain number of controllable actuators can be used. For example, by controlling the orientation of the blades or the torque of the generator, the capture of energy can be affected, the rotor plane can be tilted to deflect the wake of the turbine under the downstream turbine, or the yaw angle can be changed and deflected to one side using a technique known as wake steering. Wake steering is a control strategy across the entire wind farm and generally maximizes the total energy production by adjusting the interaction between wind turbines. Different from the standard control strategy aimed at maximizing the performance of individual wind turbines, wake steering improves the net performance of the entire wind farm by sacrificing the energy production of a specific wind turbine.

[0011] In addition to maximizing production, it is also possible to suppress or reduce the structural fatigue of wind turbines. This is to find a further compromise between the increase in production, the negative impact on the load on the wind turbine due to misalignment, and the positive impact of keeping the wake away from the rotors of wind turbines located downstream of the misaligned wind turbine. If a favorable compromise point can be found, in addition to an increase in production, the lifespan of the wind turbine is extended and the maintenance cost is reduced, which are favorable results.

[0012] Therefore, one strategy is to use a yaw actuator to shift the position of the turbine relative to the direction of the incident wind. This can limit the impact on downstream turbines using wake steering. Finding the optimal yaw angle (i.e., the yaw angle that maximizes the total power generated in the wind farm while limiting or reducing the fatigue of the wind turbine) is a complex problem. Optimizing the annual power generation of a wind farm requires a huge amount of computing power in real time, and when the wind farm is large, that is, when there are a large number of wind turbines (for example, at least 30, at least 50, or at least 80 wind turbines. This number varies depending on the available computing power.), an exorbitant amount of computing power may be required.

[0013] Many studies focus on considering the wake of a wind turbine when controlling it. For example, a wind turbine control method that implements reinforcement learning methods in a distributed manner (for each wind turbine) is known from patent EP4382743, corresponding to patent application US2024 / 0183337, where the reward is calculated based on the wake propagation delay (this reward is a value generated by the reward function of a reinforcement learning-based machine learning method). Therefore, the reward represents the effect of the previous action (e.g., the previous yaw control).

[0014] More specifically, there are studies that focus on the time-dependent dynamics of wind turbines. For example, Bernardoni et al. (F. Bernardoni, U. Ciri, M. Rotea et S. Leonardi, “Real-time identification of clusters of turbines”, Journal of Physics: Conference Series, v. 1618, p. 022 032, Sept. 2020. doi: 10.1088 / 1742-6596 / 1618 / 2 / 022032) examined the time series of power generated by each wind turbine. The authors first conducted a numerical simulation of a steady-state wind power plant under a standard wind. Next, the authors increased the wind speed in front of wind turbine i and measured the delay τ as this disturbance propagated to wind turbine j. This allowed them to investigate the correlation between the power series of i and j, offset by the delay τ. Finally, wind turbines are classified according to the value of their correlation coefficient.

[0015] GM Starke et al. (GM Starke, P. Stanfel, C. Meneveau, DF Gayme and J. King, “Network based Estimation of wind farm power and velocity data under Changing wind direction”, in 2021 American Control Conference (ACC), IEEE, 2021, pp. 1803-1810) first considered a discrete dynamic system. In this system, the state of the system at time k is represented by a matrix ((Φk)j) that includes the velocity reduction experienced by wind turbine i as a result of wind turbine j.

[0016] The authors determined a coefficient kW that could be used for geometric modeling the wake from the current value of Φk. Specifically, they represented the wind farm in a plane and identified the free-stream wind turbines by geometric arguments (the wind turbines of the farm were represented by Voronoi cells). Next, they defined the wake of i as a planar region enclosed by a half-line originating from wind turbine i and a half-line with gradient kW. All wind turbines whose rotors are in this region were considered to be in the wake of i. [Overview of the project] [Problems that the invention aims to solve]

[0017] Neither of these two solutions is entirely satisfactory in terms of the computing power required, especially for medium- and large-scale power plants.

[0018] Therefore, an object of the present invention is to improve the control of wind power plants by improving the method of optimizing power generation in accordance with wind turbine bins, and more specifically by aiming for optimization that consumes little or no computing power while maintaining efficiency. [Means for solving the problem]

[0019] The first subject of the present invention is a method for controlling a wind power plant, i.e., a wind turbine power plant. The wind turbine consists of a single type or a variety of types, and each wind turbine in the wind power plant has at least one operating point including its misalignment angle θ and / or throttling, and the at least one operating point is adjustable by at least one actuator. The method according to the present invention includes the following steps:

[0020] a) For various wind turbines, a step is to determine a predetermined range of misalignment angle θ and a predetermined wind speed deficiency threshold δ by modeling a spatial region that approximates the wake of each type of wind turbine (this region is called the modeled wake). b) Based on the wake modeled in step a), identify the wind turbines in the modeled wake of another wind turbine in each wind bin, and store identification parameters that associate each wind turbine with the wind turbine in the modeled wake in each wind bin. c) With respect to a wind turbine bin, the step of clustering the wind turbines into a plurality of subfarms, each containing wind turbines located in the modeled wake of the wind turbine bin or the wind turbine bin closest to the measured wind turbine bin, taking into account the identification parameters stored in step b), d) For each subfarm's wind turbine, at least one operating setpoint, including a misalignment angle θ and / or throttling, is determined by a method that optimizes the power generated by each subfarm (e.g., maximizing power under structural fatigue constraints). e) Applying the at least one operating setpoint (including a misalignment angle θ and / or throttling) determined in step d) to the wind turbine of the power plant by acting the at least one actuator.

[0021] It should be noted that the method of the present invention is advantageously performed in real time, and the method of the present invention is also updated in real time.

[0022] At least one actuator may, in particular, be an actuator that controls the bearings of the nacelle, an actuator that controls the orientation of the blades, and / or one or more actuators that control the reverse torque of the electric generator.

[0023] In this patent application, the “type” of wind turbine means either when the wind power plant in question consists of only one type of wind turbine (all wind turbines are identical, operate in the same way, and in particular have the same dimensions) or when it consists of multiple types of wind turbines. In the latter case, in step a), the wake is determined for each type of wind turbine.

[0024] In this patent application, "misalignment angle θ" means the angle between the rotor and the wind direction (this angle is also called "yaw").

[0025] In this patent application, “throttling” means, for example, reducing the aerodynamic efficiency of a wind turbine in operation to limit the converted electrical energy of the wind turbine to a predetermined threshold under certain environmental conditions. This may result in a change in the setpoint of the rotational speed of the wind turbine blades and / or a change in the orientation of the blades, which may cause a change in the angle of attack and wind pressure of the blades. In this patent application, “wind bin” means wind-related data that may include wind speed, wind direction, turbulence level, and other environmental parameters such as atmospheric stability.

[0026] In the present invention, "subfarm" means that the wind turbines of a power plant are clustered into multiple subgroups, and each subgroup includes at least one or at least two wind turbines.

[0027] In this invention, the term "wake" has a known definition in the context of a wind turbine, namely, the region downstream of the wind turbine (with respect to the wind direction) where the airflow slows down and its turbulence level increases in response to the velocity deficiency discussed herein.

[0028] In this invention, "insufficient wind speed δ" means insufficient wind speed caused by the wake of a wind turbine located upstream of the wind turbine in question.

[0029] Therefore, the essence of the present invention is, -By modeling the envelope of the union of all wakes (i.e., regions of insufficient velocity) corresponding to all misalignment setpoints applicable to a given wind condition upstream of the target wind turbine, and -By performing real-time clustering based on the modeled wake and measured wind bins, The solution involves dividing the power plant into multiple sub-farms in real time, each containing wind turbines that interact aerodynamically with one another.

[0030] Various wind bins are acquired, in particular, as in step a), using one or more available databases that are updated / refreshed at predetermined time intervals, specifically every 15 minutes at most, preferably every 10 minutes at most, or every 10 minutes. In particular, a real-time SCADA system is used (SCADA stands for Supervisory Control and Data Acquisition). The wind bin data thus acquired can be supplemented with data acquired in real time by physical sensors.

[0031] This invention optimizes the power that each sub-farm can generate, and then uses the sum of these powers to optimize the entire power plant, rather than optimizing the entire set of wind turbines within the plant. Therefore, this sub-farm optimization significantly reduces the computational power consumed, and consequently, the information technology resources and memory used. It is also performed much more simply and quickly without causing a decrease in performance levels. Because the optimization is performed on the sub-farms rather than the entire power plant, it is fast, and because the calculations for each sub-farm are independently parallelized, the calculations converge quickly.

[0032] Therefore, the present invention makes it possible to consider aerodynamic interactions between wind turbines for the purpose of efficiently controlling wind turbines, and the computational power requirements are, on the one hand, suited to the frequency at which setpoints transmitted to the wind turbines are updated, and on the other hand, suited to the available computational power of the command / control unit of the wind farm, even in the case of a wind farm containing a large number of wind turbines.

[0033] Preferably, at least a portion of steps a), b), c), d), and e) are repeated, and preferably all of steps b), c), d), and e) are repeated for each new wind bin. Thus, according to the present invention, the method of clustering wind turbines into subfarms can be reconfigured for each wind speed measurement, enabling real-time dynamic control.

[0034] The entire control system can be refreshed at intervals of about one minute (in particular, the refresh interval can vary between 30 seconds and 10 minutes), and can be refreshed at longer intervals when considering the widespread impact of large-scale flows within large power plants.

[0035] Step a) determining the modeled wake can be performed within a range of misalignment angles between -θ and +θ, where these values ​​correspond to the limits defined in the wind turbine specifications, in particular where θ is equal to 20°.

[0036] In step a), each wake modeled for a range of misalignment angles θ may be the envelope of the sum of multiple wakes modeled for various misalignment angles within the range of misalignment angles θ, particularly the envelope of wakes modeled for -θ, +θ, and 0°. Specifically, wakes modeled for the maximum misalignment angle and zero misalignment angle tend to partially overlap, and the envelope of these three wake clusters can provide a modeled wake that particularly represents the actual wake of a wind turbine operating within its operating limits.

[0037] Step a) of determining the modeled wake can be performed with a velocity deficit threshold δ between 0.75 m / s and 0.20 m / s, particularly between 0.70 m / s and 0.45 m / s, and especially equal to 0.65 m / s. In fact, the definition of the wake itself corresponds to the region downstream (with respect to the wind direction) of the wind turbine where the wind velocity is lower than the velocity of the free wind (i.e., wind that is not disturbed by the wind turbine). Choosing an appropriate velocity deficit makes it possible to generate a modeled wake that is small enough (particularly in terms of width) to favorably partition subfarms and prevent interaction between the modeled wakes of two wind turbines in two adjacent subfarms.

[0038] Step a) to determine the modeled wake is performed for a threshold of insufficient wind speed δ, taking into account at least one of the following parameters: the number of wind turbines in the power plant, the distance between wind turbines, the configuration of the power plant, the geographical location of the wind turbines in the power plant, and the wind rose.

[0039] Therefore, the threshold δ can be determined using a hyperparameterization method that takes into account many factors related to the environment, operation, and wind turbine design.

[0040] In step a), the modeled wake can be approximated / represented in two-dimensional form, considering the wind bin and wake at a constant height, particularly equal to the height of the rotor axis supporting the blades of the wind turbine. These modeled wakes are trapezoidal in shape, where the smallest base of the trapezoid is the initial width of the wake.

[0041] In step a), the modeled wake can also be approximated / represented in the form of a three-dimensional representation. These representations include a frustocone with the initial size of the wake as the minimum base. Next, we consider the wind bin and wake at a specific height, particularly in the region on both sides of the height of the wind turbine rotor.

[0042] However, the selected 3D representation may have a pseudo-cone with a cross-section that is more elliptical than circular, or a cross-section that is neither elliptical nor circular, which may be particularly true for floating wind turbines.

[0043] The various wind bins considered in step a) where the wake is determined by modeling can be all possible wind bins recorded over a specific period, especially one year, within a geographical area including a wind farm.

[0044] In step b), the wind bin, particularly the distribution of wind speed and direction and / or wind speed and direction, can be measured in real time by at least one lidar sensor and / or at least one (acoustic or mechanical) wind speed measurement system and / or at least one SCADA system that delivers wind speed and direction. These various measuring means can be fixed to the wind turbine, particularly near the nacelle, or installed remotely (for example, on a floating buoy in the case of an offshore wind farm).

[0045] In step d) determining at least one operating setpoint, the power generated for each subfarm may be optimized in parallel and / or sequentially for each subfarm.

[0046] In either case (parallel or sequential computing), optimization is much faster and requires far fewer computing / information technology resources than attempting to optimize all the wind turbines in a power plant at once. This is not only because fewer turbines result in lower computational costs for simulating power generation, but also because the optimization algorithm is applied to a smaller-dimensional space (a smaller space to explore).

[0047] Parallel optimization is even faster than sequential optimization, if possible. In step d) determining at least one operating setpoint, the power generated for each subfarm can be optimized taking into account the structural fatigue constraints of each type of wind turbine in the wind turbine or power plant. These constraints relate particularly to the blades and towers of each type of wind turbine and should generally be considered in the control methods of the power plant.

[0048] It should be noted that considering these fatigue constraints tends to increase the computation time for optimization. Therefore, the present invention is particularly advantageous because it can take these constraints into account while suppressing the increase in computation time.

[0049] In this invention, structural fatigue may relate to all mechanical components of a wind turbine, including the tower, blades, foundation, floating platform, and mooring cables. Furthermore, mechanical forces generated by actuators can also be considered constraints, similar to structural fatigue.

[0050] Step a) determining the modeled wake can be performed offline by one or more arbitrary periodic updates and / or any updates by wind bin measurements performed in real time in step c).

[0051] Another subject of the present invention is a wind power plant, i.e., a wind turbine power plant, wherein the wind turbines are of a single type or various types, and each wind turbine of the wind power plant has at least one adjustable operating point including its misalignment angle θ and / or throttling, wherein the at least one operating point is adjustable by at least one actuator, thereby the wind power plant having or being connected to information technology for implementing the control method described above to apply the operating setpoint, including the misalignment angle θ and / or throttling, to the wind turbines of the power plant by acting the actuators (according to a setpoint determined by the information technology).

[0052] It is desirable to install information technology / computational equipment on the same site as the wind power plant. However, if it can be connected to the wind power plant via a high-performance communication network (such as the internet), it is also possible to install it in a location (completely or partially) away from the wind power plant.

[0053] Another subject of the present invention is any information technology, in particular a PC, server, or computer, configured to perform the method of controlling the wind power plant described above.

[0054] Another subject of the present invention is any computer program product that is downloadable from a communication network and / or stored on a medium readable by a PC, server or computer and / or executable by a processor, wherein, when the program is executed on a PC, server or computer, the program includes program code instructions for carrying out the method for controlling the wind power plant described above.

[0055] Furthermore, the present invention relates to any storage medium that is readable by a PC, server, or computer and, when executed by a PC, server, or computer, stores instructions that cause the PC, server, or computer to execute the method for controlling the wind power plant described above.

[0056] Other features and advantages of the method according to the present invention will become apparent by reading the following description relating to non-limiting examples of embodiments with reference to the accompanying drawings described below. [Brief explanation of the drawing]

[0057] [Figure 1] This diagram shows the misalignment angle of a wind turbine. [Figure 2] This figure shows an example of a histogram of annual wind speed measurements (wind speed and wind direction) over a one-year period, which can be used in the method according to the present invention. [Figure 3] This diagram shows the wake of a wind turbine dependent on a given wind bin, using wind speed ws = 7 m / s and wind direction wd = 270° as an example. The shades of gray represent the velocity shown on the grayscale to the right of the wake display, expressing the wind speed in m / s in the target wake region. [Figure 4] This diagram shows the values ​​of the wind speed field and the output of two wind turbines, with one wind turbine located downstream of the other. The same rules for representing velocity are used as in the previous diagram (and the same rules are used in all subsequent diagrams showing the wake). [Figure 5] This diagram illustrates the deflection of the wake due to misalignment of a wind turbine. The upper part shows the wake of a wind turbine with a misalignment angle θ of zero, and the lower part shows the wake of a wind turbine with a misalignment angle θ of 20°. [Figure 6] A diagram illustrating the interaction between the wakes of three wind turbines clustered in a single subfarm, showing the control of their misalignment through optimization of the power generated by the subfarm according to the present invention. [Figure 7] This graph shows the computation time required to optimize the power output of a simulated wind farm as the number of wind turbines increases, with the x-axis representing the number of wind turbines in the farm and the y-axis representing the optimization time (in hours). [Figure 8]This figure shows the wake of a power plant with nine wind turbines in a predetermined wind bin, without clustering into a subfarm according to the present invention. [Figure 9a] This diagram shows the wake of a power plant with nine wind turbines in a designated wind bin. [Figure 9b] This figure shows a directed graph used to cluster nine wind turbines according to the present invention into a subfarm, corresponding to the wake of Figure 9a. [Figure 10a] This diagram shows the wake of a power plant with 12 wind turbines in a designated wind bin. [Figure 10b] This figure shows a directed graph corresponding to the wake of Figure 10a, used to cluster 12 wind turbines according to the present invention in a subfarm. [Figure 11] This diagram shows the wake of a wind turbine in a specified wind bin (7 m / s, 270°), with the cases shown from top to bottom being when the velocity deficit δ is 0.75 m / s, 0.5 m / s, and 0.25 m / s, respectively. [Figure 12] This diagram shows the wake of a wind turbine in a specified wind bin (7 m / s, 270°), with the cases shown from top to bottom being when the misalignment angle θ is +20°, 0°, and -20°, and the velocity deficit δ is 0.5 m / s. [Figure 13a] This figure shows the so-called additional wake, which corresponds to the envelope obtained by adding the wakes generated when the three misalignment angles θ in Figure 12 are +20°, 0°, and -20°. [Figure 13b] This figure shows how the additional wake in Figure 13a is modeled in a trapezoidal shape according to the present invention. [Figure 14a] This figure shows the additional wake of the 12-wind turbine cluster, depending on whether the velocity threshold δ was set to 0.5 m / s. [Figure 14b] This figure shows the additional wake of the 12-wind turbine cluster, depending on whether the velocity threshold δ was set to 1.5 m / s. [Figure 15]The figure shows a graph for various velocity thresholds δ, where the x-axis represents the number of wind turbines in the power plant and the y-axis represents the ratio of "conventional" optimization time to "clustered" optimization time according to the present invention. All figures, particularly the wind turbines and their wakes, are highly schematic and not necessarily to scale, nor do they represent the actual spatial configuration of the wind turbines at their operating positions (however, the trends, magnitudes, and variations of the quantities of interest represent the facts of the implementation of the present invention). [Modes for carrying out the invention]

[0058] This invention relates to a method for controlling a wind power plant in real time. A wind farm (also called a wind park or wind power plant) is a facility that generates electricity using multiple wind turbines. Each wind turbine (sometimes mistakenly called a windmill) in a wind farm is equipped with at least one actuator to adjust at least one operating point of the wind turbine. An example of an operating point is the misalignment angle θ (or yaw angle) of the wind turbine.

[0059] Other operating points include, in particular, throttling of the wind turbines, or modification of the wind turbine's power curve. The location of the wind turbines within the wind farm (also called the wind turbine arrangement or layout) is known in advance.

[0060] The following description will only include explanations and examples relating to the control of the misalignment angle θ. However, according to the method of the present invention, it is also possible to control other operating points in addition to, or instead of, the misalignment angle. In particular, it is possible to control the level of static, dynamic, or helical induction control that affects the wake range and advection.

[0061] For more details on dynamic induction control, see, for example, the article "Periodic Dynamic Induction Control in Wind Power Plants: Proving its Potential with Simulation and Wind Tunnel Experiments" by JA Frederik et al., published in Wind Energy Science Discussions in August 2019.

[0062] For more details on helical induction control, see, for example, the article "Helix Approach: Enhancing Wake Mixing at Wind Farms Using Dynamic Individual Pitch Control" by JA Frederik et al., published on pages 1739-1751 of Wind Energy, issue 8, August 2020.

[0063] In this patent application, the terms upstream and downstream are defined with respect to the direction of the wind (upstream wind turbines are affected by the wind before downstream wind turbines).

[0064] The method according to the present invention includes the following steps: a) For various wind turbines, a step of determining a given range of misalignment angles (θ) and a given threshold for insufficient wind speed (δ) by modeling a spatial region (called the modeled wake) that approximates the wake of each type of wind turbine, b) Based on the wake modeled in step a), identify the wind turbines in the modeled wake of another wind turbine in each (measured) wind bin, and store identification parameters that associate each wind turbine in each wind bin with the wind turbine in the modeled wake where that wind turbine is located. c) In wind bin measurement, the identification parameters stored in step b) are considered in real time for the wind bin or the wind bin closest to the measured wind bin, and the wind turbines are clustered into multiple subfarms, each subfarm containing wind turbines located in the downstream of the other, d) For each subfarm's wind turbine, a step of determining at least one operating setpoint, including a misalignment angle (θ) and / or throttling, by a method that optimizes the power generated by each subfarm, e) Applying at least one operating setpoint (including a misalignment angle θ and / or throttling) determined in step d) to the wind turbine of the power plant by activating at least one actuator.

[0065] The steps of this method, or at least a portion thereof, can be carried out by information technology, in particular by at least one PC, processor, or computer.

[0066] The procedure will be explained in detail below, with reference to the diagrams. Figure 1 has already been explained. First, let us recall the definitions of the specific terms used in this application.

[0067] -Electricity from wind power plants: The rotor of a wind turbine converts the mechanical energy contained in the wind into electrical energy. The electricity P generated by a wind turbine turbine And the power P included in the wind resistance wind The relationship can be expressed as follows:

number

number

[0068] ρ is the density of air. S represents the area of ​​the wind sector that contacts the blade. U represents the wind speed at the rotor's position.

[0069] A wind power bin is a configuration defined by wind speed (s), wind direction (d), and here the turbulence level (TI). Thus, such a wind power bin has two components (w s , w d ) or three components (w s , w d , TI).

[0070] In the remainder of this description, W = (Ws, Wd, TI) is a random variable representing the wind blowing towards the wind power plant. This is defined in the probability space (Ω, F, P),

Number

[0071] Wind power

Number

Number

[0072] -Annual energy production (also called AEP). The AEP of a wind power plant is defined as the annual energy generated by that wind power plant in a year. Since P represents the amount of energy supplied per hour, the AEP is described using the expected annual production.

Number

[0073] For simplification, the number of wind bins surveyed annually is limited by setting the possible values ​​of W. For years in which energy production is surveyed, W is modeled as a discrete variable because wind speed measurements are available. Based on the measurements, the probability of each possible wind is determined. Thus, it can be written as follows:

[0074] - A discrete set of possible wind speeds

number

number

[0075] - A discrete set of possible wind directions

number

number

[0076] Therefore, the set of possible winds is

number

[0077] Therefore, W's law can be written as follows:

number

[0078] Here, we assume that there exists a set D of N measured values ​​of W in the target year. {p(w si ,w dj The elements of D that are naturally selected as )} are

number

[0079] More precisely,

number

number

[0080] Figure 2 shows a wind rose diagram of wind data collected in 2021 at a specific location, using a hypothetical power plant as an example. Therefore, these various wind bins are collected here over a year for the purpose of defining the corresponding modeled wake in step a) of the method according to the present invention.

[0081] The discretization selected here is, with respect to the wind rose diagram in Figure 2,

number

[0082] This figure is actually a histogram, and the radius of the "box" represents the frequency p(s) of possible winds (si,dj). i d j The graph shows the wind speed (si) and the direction (dj), with shades of gray indicating wind speed (si) and direction (dj). Conventionally, direction w d This indicates the direction from which the wind is blowing. This is measured clockwise, and the angle for a northerly wind is 0°.

[0083] Since the distribution of W was constructed from the data for the target year, the AEP can be rewritten as follows:

number

[0084] In calculating the power output of each wind turbine, various aerodynamic interactions between the turbines are taken into consideration. One of these interaction effects is the wake of the wind turbine, and its simulation is shown in Figure 3. The greater the velocity deficit, the lighter the gray intensity becomes (similar notation is used in the relevant parts of subsequent figures).

[0085] -Swaft The wake of a solid object around which a fluid flows is a turbulent region downstream of the object. In the context of this invention, this is the region downstream of a wind turbine, where the airflow is slow and often turbulent.

[0086] As an example, Figure 4 uses the same type of representation as Figure 3 to show the values ​​of the wind speed field and the power of two wind turbines (one wind turbine is downstream of the other). Therefore, the wind turbine located downstream operates in a region where the wind speed is lower than the speed of free flow (i.e., undisturbed wind). Because the wind turbine receives less energy from the wind flowing behind it, the amount of electricity generated also decreases.

[0087] -Insufficient wind speed δ The wind speed deficit at point x in space R3 is defined as follows:

number

[0088] In the embodiments detailed in this patent application, the wind speed value is measured at the nacelle of the wind turbine. Similarly, all wind turbine models and the values ​​derived therefrom are at a height equal to the height of the wind turbine rotor. Therefore, this study assumes a plane parallel to the ground (height z=z) in 2D space (two-dimensional space). rotor ) is the target.

[0089] However, it should be emphasized that the present invention can also be advantageously applied based on 3D (three-dimensional) measurement and / or wake modeling.

[0090] To convert wind energy into electrical energy as efficiently as possible, the blades of a wind turbine must be directed towards the wind. In fact, when the rotor of a wind turbine is offset by an angle θ relative to the wind (see the comparison of the lower and upper parts in Figure 5), the power coefficient Cp of the wind turbine decreases according to the following relationship.

number

[0091] However, if the position of the wind turbine nacelle shifts, the wake changes. As a result, as shown in Figures 4 and 5, wind turbines located downstream of other wind turbines may become free-flowing due to misalignment.

[0092] In some cases, intentional drift is introduced to reduce the speed deficiency of wind turbines located downstream of the wind, thereby increasing the power generated by the wind turbines.

[0093] - Misalignment angle (or yaw angle) and optimization When the yaw angle θ=0, C p While this may decrease, it could increase the velocity U experienced by the wind turbine. This misalignment could potentially increase the power generated by the wind turbine. Therefore, the power coefficient C due to the increase in wind speed p To compensate for the decrease, calculations are needed to optimize the yaw of each wind turbine.

[0094] In this invention, the power of the power plant is supplied by the yaw angle θ of each wind turbine t. t Consider that it depends on the wind bin w, the selected corresponding yaw angle is

number

number

[0095] The problem of optimizing the annual power generation of a wind power plant as a function of the yaw angle can be formulated as follows: N T Assuming a power plant containing several wind turbines, each wind turbine is index

number

number

number

[0096] Each wind force w i θ is a single yaw vector. i We assume this corresponds to the selection. Specifically, the ground objective is to select the displacement of the wind turbine nacelle at the wind farm in each new wind measurement.

[0097] The following function will be maximized.

number

[0098] In practice, it is desirable to limit the amplitude of the deviation to suppress damage to the rotating wind turbine due to changes in wind. For example, the selection may be limited to a yaw amplitude between θ -= -20° and θ += 20°.

[0099] Therefore, the problem of optimizing the average annual power of a power plant is Constraints

number

number

[0100] Since pwi is positive, it can be written as follows:

number

[0101] The above optimization problem is,

number

[0102] Designated wind bin

number

number

[0103] P is calculated as follows: If we represent the related components of the yaw vector θt and the wind speed U(t) acting on the rotor of the wind turbine t, the power P(t) of each wind turbine in the power plant can be described as follows.

number

[0104] It should be noted that w represents the so-called wind at infinity, that is, the measured value of an ideal wind flow that is not disturbed, while U(t) represents the wind speed actually experienced by the rotor of the wind turbine t.

[0105] However, it was found that calculating the velocity field inside a power plant when the airflow is disturbed by the presence of wind turbines is extremely complex and cannot be analytically represented without approximations (calculated using numerical solutions to the Navier-Stokes equations).

[0106] Therefore, various representations of P(t) in ("conventional" OPT) are inaccessible, resulting in a so-called "black box" optimization problem.

[0107] Furthermore, since even calculating the P(t) value is already costly, numerical optimization quickly becomes prohibitively expensive as the scale of the power plant increases, as shown in Figure 7. Figure 7 is a graph in which the number of wind turbines in the power plant is shown on the x-axis and the optimization time in hours is shown on the y-axis.

[0108] Due to the complexity of these calculations, this study assumes the existence of a ("conventional" OPT) solution, but does not consider its potential uniqueness. Therefore, the following notation is introduced.

[0109] -θ(w) is the maximum value returned after the numerical optimization routine converges. -P conventional (w) := P(θ,w) is the power of the power plant obtained by the numerical solution of ("conventional" OPT). -t conventional (w) is the time it takes for the optimization routine to complete.

[0110] By modeling a power plant equipped with three wind turbines affected by wind w=(10m / s, 270°) (see Figure 6) and solving the ("conventional" OPT), the nominal power P nominal =P(θ=0,w) and P conventional We calculated P(θ,w).

[0111] Considering the computational cost of the ("conventional") OPT solution, it is impossible to optimize the power of a large-scale wind farm in real time as a function of the yaw angle.

[0112] The solution according to the present invention is for this power optimization (step d) of the method according to the present invention (OPT conventional This involves splitting the instance into multiple optimized instances on smaller data clusters.

[0113] The present invention comprises clustering wind turbines in a power plant into sub-farms or "clusters" using appropriate criteria, and then individually considering the power optimization of each sub-farm (the terms sub-farm and cluster are used interchangeably in this patent application). This is step c) of the method of the present invention.

[0114] Therefore, the optimized power of a clustered power plant in a subfarm is the sum of the optimized power of each subfarm.

[0115] From Figure 7 mentioned above, it can be inferred that the time required to optimize the power of the entire power plant will be (much) longer than the sum of the optimization times for each wind turbine cluster. On the other hand, when considering the power of each sub-farm individually, it is desirable to ignore the wake effects between sub-farms.

[0116] Therefore, it is desirable for each subfarm to capture as much of the wake effect between the wind turbines that comprise it as possible. Ideally, it is desirable that two wind turbines in two different subfarms do not enter each other's wake. Figure 8 shows a situation where this is possible, with the first subfarm consisting of wind turbines 1-4 and the second subfarm consisting of wind turbines 6-9.

[0117] wind

number

number

number

number

[0118] It should be noted that maximizing the power of a given cluster C as a function of yaw is merely an example of reducing the size of the power plant in question ("conventional" OPT) where C is the power plant in question.

[0119] each

number

number

[0120] Therefore, the vector θ(C)(w) is proposed as an approximate solution to the "conventional" OPT, and its components are the "clustered" OPT as follows: (C) This is the solution to the problem.

number

[0121] In other words, the vector θ(C) contains the yaw resulting from the optimization of the power of each component cluster C.

[0122] Finally, the optimized power of clustered power plant C

number

number

[0123] The only remaining step is to apply the misalignment setpoint to the wind turbine by appropriately controlling the actuator. This is step d) of the method according to the present invention.

[0124] It should be noted that the optimizations applied to each cluster / subfarm are independent and distributeable operations. Therefore, cluster-based optimizations greatly benefit from the distribution or parallelization of computations, which can further significantly reduce computation execution time. [Examples]

[0125] Here, as shown in Figure 8, N is subjected to a wind w = (10 m / s, 18°). T Let's consider a simulated power plant consisting of nine wind turbines. In Figure 8, the wind turbines are numbered from 1 to 9.

[0126] The possible options for the cluster are:

number

number

[0127] However, considering that wind turbine 7 is almost completely behind wind turbine 6, if we ignore this interaction, partition

number

number

[0128] Numerical analysis yields the following results. -P nominal = 18.87 MW; - Optimization time t conventional If = 2.00 seconds, P conventional =20.11MW; -Calculation time t(C)c clustered If = 1.12 seconds, P(C) clustered =20.11MW; -Calculation time t(C') clustered If =0.96 seconds, then P(C') clustered = 2.10 MW.

[0129] These results first indicate that the partition selection is not unique for a given wind, and that the selected partition accurately reflects the wake interactions between wind turbines of the cluster / subfarm components.

[0130] By constructing appropriate partitions for each wind data set, power calculations can be sped up while bringing P "clustered" closer to P "conventional".

[0131] Construction of a subfarm Now that we have defined all the relevant concepts, if the wind farm is already divided, we will use the method described above to build a subfarm.

[0132] The wind power plant is modeled as a directed graph, where the nodes represent the wind turbines of the power plant, and the edges represent the presence of the wake.

[0133] A directed graph associated with a power plant of wind turbines on an NT platform receiving wind w is,

number

number

number

[0134] The edges of the graph depend entirely on the definition of Σ, which is the spatial region approximating the wake downstream of i.

[0135] Examples of relationship construction are shown below. Here, we assume that ~ has been constructed. A key characteristic of wind turbine cluster C is that it captures all of the wake effects of its components.

[0136] In other words, if there is a path between i and j in the graph, that is,

number

number

[0137] By definition, directed subgraph

number

number

[0138] The problem of determining the weakly connected components of a directed graph is, for example, each node

number

[0139] An example of subfarm clustering according to the present invention is shown below. - In Figures 9a and 9b, a power plant consisting of nine wind turbines is clustered into four sub-farms, each containing two wind turbines, and one sub-farm containing only one wind turbine. - In Figures 10a and 10b, the power plant consisting of 12 wind turbines is clustered into four sub-farms (1, 4, 9, 12) each containing one wind turbine, two sub-farms (2, 7) and (6, 10) each containing two wind turbines, and one sub-farm (3, 5, 8, 11) each containing four wind turbines.

[0140] Approximation of wake As mentioned above, the wake of a wind turbine is the spatial region downstream of the wind turbine where the airflow slows down. The velocity difference ΔU is the difference between the wind speed at infinity (the speed of an ideal, undisturbed wind) and this speed in the wake.

[0141] Wind data

number

[0142] Since the focus is on the influence of the wake on the energy generation of a wind turbine, power loss due to insufficient wind speed can be ignored. Specifically, because the power generated by a wind turbine is proportional to the cube of the wind speed, small wind speed deficiencies do not significantly affect the power generated. Therefore, a parameter δ is introduced to define a threshold for wind speed deficiency at which a wake is judged to exist.

[0143] Wind power bottle w = (w s ,w d In the case of a point in the space within the wake Σ of a wind turbine, the measured velocity is below w s It is a point smaller than -δ.

number

[0144] Depending on the value of this parameter, the shape of the wake changes, as shown in Figure 11. Modeling the wake of a specific cluster group is reminiscent of how it can be part of optimizing the yaw of a wind turbine at a power plant.

[0145] As mentioned above, and as can be seen in Figure 12, the shape of the wake changes due to misalignment caused by yaw. To model the wake region, yaw must be considered.

[0146] However, these values ​​are only calculated by optimization after clustering. At this stage, since these values ​​are unknown, a spatial region containing every other wake is used when the yaw value changes between θ -= -20° and θ += 20°.

[0147] To achieve this, the region arising from the wake coupling at yaw angles 0, θ-, and θ+

number

[0148] By summing not only the extreme yaw angles of the wind turbine but also the wake at zero yaw, it can be seen that a more realistic modeled wake can be obtained. In particular, it is effective to take into account the wake at zero yaw.

[0149] The goal is to approximate it by

Number

[0150] Recall that all velocities and wakes are studied at a constant height equal to the level z of the rotor. Unfortunately, rotor the notation

Number

Number

[0151]

Number

[0152] approximate

Number

[0153] 1. nb_diam: The length of the wake represented as a multiple of the diameter of the wind turbine blade. 2. init_width: The initial width of the wake measured at the rotor. 3. h: The increase in the width of the half wake.

[0154] Once these parameters are determined for the wind bin w, the wake downstream of the wind turbine i is approximated as a trapezoid defined by these parameters. If the wind turbine j belongs to the trapezoid defined in this way, it lies in the wake of i. Figure 13b shows the result of step a) of determining the modeled wake described above.

[0155] Final calculation of adjustments to be made at the operating point of the wind turbine. A two-stage approach will be adopted. 1. Under Development / Offline: Each wind bin w in a discrete set of possible winds fW i For =(si,di) 、 The following steps will be performed.

[0156] -wind speed w si And the wind direction lol di The wake of a single wind turbine receiving a 270° angle

number

number

[0157] 2. Production time / Real-time: When the power plant is exposed to wind, -Node1,...,N T Create a graph with the following properties. this is, Step c) to cluster the subfarm. Specifically, as mentioned above, the power plants with 9 wind turbines are shown in Figures 9a and 9b, and the power plants with 12 wind turbines are shown in Figures 10a and 10b. - Read the trapezoidal dimension parameters that approximate the wake from the lookup table. - For any wind turbine i within the power plant: · Calculate the trapezoidal coordinates downstream of i. · For any wind turbine j ≠ i, if j is within the trapezoid, add an edge (i, j) to the graph representing the power plant. - Calculate the set C of weakly connected components of the graph thus constructed. - For all

Number

Number

[0158] Figures 14a and 14b show the role of the hyperparameter δ that sets the threshold of speed deficiency, and based on this threshold, the wake approximation value

Number

Number

Number

[0159] Figure 15 shows a graph of various speed deficiency thresholds, where the x-axis represents the number of wind turbines in the power plant and the y-axis represents the ratio of "conventional" optimization time to "clustered" optimization time according to the present invention. -Curve C1 corresponds to a velocity threshold δ of 0.5 m / s. - Curve C2 corresponds to a velocity threshold δ of 1.0 m / s. -Curve C3 corresponds to a velocity threshold δ of 1.5 m / s.

[0160] Thus, the present invention significantly reduces the time required to perform optimization, and it can be seen that this reduction increases as the number of wind turbines in the power plant increases.

[0161] In conclusion, the optimization method according to the present invention, which involves clustering into subfarms, makes it possible to optimize the energy production of a power plant at least 10 times faster than the conventional method, which involves optimizing power on all wind turbines in the power plant. Furthermore, it reduces the error rate to a negligible level compared to the conventional method and limits the information technology resources and memory used.

[0162] Therefore, the present invention makes it possible to accelerate the implementation of power plant control strategies and deploy them on a real-time computer, including in the case of large-scale power plants that include tens or hundreds of wind turbines.

Claims

1. A method for controlling a wind turbine in a wind power plant, wherein the wind turbine is of a single type or various types, and each wind turbine in the wind power plant has at least one operating point including a misalignment angle (θ) and / or throttling, and the at least one operating point is adjustable by at least one actuator. The aforementioned method, a) For various wind turbine bins, a range of predetermined misalignment angles (θ) and a predetermined threshold for insufficient wind speed (δ) are determined by modeling a spatial region that approximates the wake of each type of wind turbine (this region is called the modeled wake), b) Based on the wake modeled in step a), identify the wind turbines in the modeled wake of another wind turbine in each wind bin, and store identification parameters that associate each wind turbine with the wind turbine in the modeled wake in each wind bin. c) With respect to a wind turbine bin, the step of clustering the wind turbines into a plurality of subfarms, each containing wind turbines located in the modeled wake of the wind turbine bin or the wind turbine bin closest to the measured wind turbine bin, taking into account the identification parameters stored in step b), d) For each subfarm's wind turbine, the steps of determining at least one operating setpoint, including the misalignment angle (θ) and / or the throttling, by a method that optimizes the power generated by each subfarm, e) A method for controlling a wind turbine, comprising the step of applying the at least one operating setpoint determined in step d) to the wind turbine of the wind turbine by operating the at least one actuator.

2. A method for controlling a wind power plant according to claim 1, characterized in that steps a), b), c), d), and e) are repeated for each new measurement of the wind bin.

3. A method for controlling a wind turbine according to claim 1, characterized in that step a) determining the modeled wake is performed for a range of misalignment angles (θ) between -θ and +θ, where these values ​​correspond to limitations defined in the specifications of the wind turbine, and in particular θ is equal to 20°.

4. A method for controlling a wind power plant according to claim 3, wherein in step a), each wake modeled for the range of misalignment angles (θ) is the envelope of the sum of a plurality of wakes modeled for various misalignment angles within the range of angles (θ), and in particular the envelope of wakes modeled for -θ, +θ, and 0°.

5. A method for controlling a wind turbine according to any one of claims 1 to 4, characterized in that step a) determining the modeled wake is performed for a wind speed deficiency threshold (δ) equal to 0.75 m / s to 0.20 m / s, particularly 0.70 m / s to 0.45 m / s, and particularly 0.65 m / s.

6. A method for controlling a wind turbine according to any one of claims 1 to 4, characterized in that step a) determining the modeled wake is performed for a wind speed deficiency threshold (δ) taking into account at least one of the following: the number of wind turbines in the wind turbine, the distance between wind turbines, the configuration of the wind turbine, the geographical location of the wind turbines in the wind turbine, and the parameters of the wind rose.

7. In step a), the modeled wake is - Considering the wind bin and wake at a certain height, particularly a height equal to the height of the rotor axis supporting the blades of the wind turbine, it is approximated in a two-dimensional representation form with a trapezoidal shape, such that the smallest base of the trapezoid becomes the initial width of the wake. - Or, in particular, a method for controlling a wind power plant according to any one of claims 1 to 4, characterized in that it is approximated in a three-dimensional representation form as a frustum whose smallest base is the initial size of the wake.

8. A method for controlling a wind power plant according to any one of claims 1 to 4, characterized in that the various wind bins considered in step a) determining the wake by modeling are all possible wind bins recorded over a certain period, particularly one year, in the geographical area including the wind power plant.

9. A method for controlling a wind power plant according to any one of claims 1 to 4, characterized in that, in step b), the distribution of wind bins, particularly wind speed and direction, and / or wind speed and direction, are measured in real time by at least one lidar sensor and / or at least one anemometer and / or at least one SCADA (Supervisory Control and Data Acquisition) system.

10. A method for controlling a wind power plant according to any one of claims 1 to 4, characterized in that, in step d) determining at least one operating setpoint, the power generated for each subfarm is optimized in parallel and / or sequentially for each subfarm.

11. A method for controlling a wind power plant according to any one of claims 1 to 4, characterized in that, in step d) determining at least one operating setpoint, the power generated for each subfarm is optimized taking into consideration the constraints of structural fatigue of the wind turbine or each type of wind turbine in the wind power plant.

12. A method for controlling a wind turbine according to any one of claims 1 to 4, characterized in that step a) determining the modeled wake is performed offline by one or more arbitrary periodic updates and / or arbitrary updates by measurements of the wind bin performed in real time in step c).

13. A wind power plant, i.e., a wind turbine power plant, wherein the wind turbines are of one or more types, each wind turbine of the wind power plant has at least one adjustable operating point including its misalignment angle (θ) and / or throttling, the at least one operating point is adjustable by at least one actuator, and the wind power plant is equipped with or connected to information technology for carrying out the control method according to any one of claims 1 to 4 in order to apply the operating set point including its misalignment angle (θ) and / or throttling to the wind turbines of the power plant by acting the actuator.

Images