A method and system for coordinated restoration of wind speed in wind turbine nacelles

Abstract

This application discloses a method and system for collaboratively restoring wind speed in a wind turbine nacelle. The method includes: constructing a local flow field response model for the nacelle and a wind field restoration model, denoted as the G model and F model, respectively; acquiring historical operating datasets of the wind farm and preprocessing them; performing parameter calibration of the G model and joint consistency calibration of the G-F model parameters based on the preprocessed historical operating datasets; using the calibrated G model and F model to restore the free-flow wind speed and direction based on real-time collected nacelle wind speeds; performing uncertainty quantification assessment on the restored free-flow wind speed, and adaptively updating the parameters of the G model and F model based on the assessment results. Using the scheme of this application, the error accumulation caused by local nacelle disturbances and global station wake effects in the joint derivation process is effectively eliminated, significantly improving the accuracy of restoring free-flow wind speed and direction relying solely on basic nacelle wind measurement data.

Description

Technical Field

[0001] This application generally relates to the field of wind power generation technology. More specifically, this application relates to a method and system for coordinated restoration of wind speed in a wind turbine nacelle. Background Technology

[0002] Accurately obtaining the free-flow wind speed at the height of the wind turbine hub is fundamental for wind farm power prediction, performance evaluation, and advanced control. Currently, wind farms mainly rely on nacelle anemometers for measurement, but their readings are severely affected by a dual physical effect: First, the local flow field distortion effect in the nacelle: When the impeller extracts energy, it causes significant acceleration, deceleration and deflection of the airflow, resulting in a systematic deviation between the local flow velocity at the nacelle anemometer and the actual incoming flow.

[0003] Secondly, the wake interference effect between fans: the downstream fans are affected by the attenuation and disturbance of the upstream wake, and the measured values ​​reflect the information of the contaminated flow field, rather than the original free flow.

[0004] Existing technologies typically address these issues in isolation, exhibiting the following fundamental limitations: First, the manufacturer calibration curve method only provides static nacelle effect correction, failing to consider turbine aging and wake effects; second, correction methods based on empirical wake models often implicitly assume "accurate anemometer measurements," using data containing sensor errors to correct flow field errors, leading to confusion of error sources; third, data-driven black-box models lack physical interpretability, exhibit poor generalization ability in uncovered operating conditions, and cannot distinguish error sources; fourth, existing high-fidelity simulations (such as FLORIS) and inverse optimization methods typically directly input raw data contaminated by nacelle distortion as "true observation values" into the model, resulting in a "garbage in, garbage out" defect, making it difficult to guarantee restoration accuracy.

[0005] In view of this, there is an urgent need to provide a wind turbine nacelle wind speed collaborative restoration scheme to solve the above problems and restore high-precision free flow information from the disturbed nacelle wind speed data. Summary of the Invention

[0006] In order to at least solve one or more of the technical problems mentioned above, this application proposes a wind turbine nacelle wind speed coordinated restoration scheme in several aspects.

[0007] In the first aspect, this application provides a method for collaborative restoration of wind turbine nacelle wind speed, comprising: constructing a local flow field response model for the nacelle and a wind field restoration model, denoted as the G model and the F model respectively; acquiring a historical operation dataset of the wind farm and preprocessing it; performing G model parameter calibration and GF model parameter joint consistency calibration sequentially based on the preprocessed historical operation dataset of the wind farm; using the calibrated G model and F model to restore the free-flow wind speed and direction based on the real-time collected nacelle wind speed; performing uncertainty quantification assessment on the restored free-flow wind speed, and performing adaptive updates of the G model parameters and F model parameters according to the assessment results.

[0008] In some embodiments, the G-model is used to predict nacelle wind speed measurements based on the local effective incoming wind speed in front of the wind turbine rotor, the yaw error angle, and the turbulence intensity; the expression for the G-model is: V nacelle =G(V local ,γ,T I ;θ G )=α(γ,T I )×V local +β(γ,T I ), where V nacelle Here is the measured wind speed in the cabin, G(·) is the G model, and V local The effective inflow velocity in front of the fan rotor is γ, where γ is the yaw error angle and T is the yaw rate. I For turbulence intensity, θ G Here are the parameters of the G model, α(·) is the gain function, and β(·) is the offset function.

[0009] In some embodiments, the F model includes a forward wake simulation module and a reverse optimization solution module. The forward wake simulation module is a high-fidelity wake model based on physical mechanisms. It simulates and calculates the wind speed at any point within the wind farm based on the input free-flow wind speed, free-flow wind direction, turbulence intensity, and wind farm layout. The reverse optimization solution module uses the free-flow wind speed and free-flow wind direction as decision variables, and the matching error between the wind speed at the corresponding location calculated by the forward wake simulation module and the corresponding observed value as the objective function. It iteratively calls the wind speed at the corresponding location to perform reverse optimization to obtain the free-flow wind speed and free-flow wind direction that minimize the objective function.

[0010] In some embodiments, during the calibration of G-model parameters, the following steps are performed: a subset of data from the preprocessed historical wind farm operation dataset is selected where the wind turbines are not affected by the wake of the upstream wind turbines; the initial parameters of the G-model are obtained by fitting the data in the subset using weighted least squares; wherein, during the selection of the subset of data where the wind turbines are not affected by the wake of the upstream wind turbines, the following steps are performed: an empty subset of data is constructed; a multi-dimensional constraint score is obtained for each data sample from the preprocessed historical wind farm operation dataset; a weighted summation of the multi-dimensional constraint scores for each data sample is performed to obtain a comprehensive score for each data sample; samples with comprehensive scores higher than a preset threshold are stored in the subset of data.

[0011] In some embodiments, the multi-dimensional constraint scoring includes free-flow direction geometric constraint scoring, aerodynamic performance constraint scoring, and adjacent wind turbine comparison constraint scoring; wherein, in the process of acquiring the free-flow direction geometric constraint scoring corresponding to each data sample, the following steps are performed: based on the turbine layout coordinates of the wind farm and the free-flow direction data at the current moment, construct the wake influence sector of all upstream wind turbines, and determine whether the target wind turbine is located within the wake influence sector of at least one upstream wind turbine; in response to the target wind turbine not being located within the wake influence sector of at least one upstream wind turbine, it is determined that there is no wake conflict, and the pre- The highest score is set as the free-flow direction geometric constraint score. In response to the target turbine being located within the wake influence sector of at least one upstream turbine, a wake conflict is determined. The free-flow direction geometric constraint score is obtained based on the spatial distance and azimuth difference between the target turbine and its corresponding upstream turbine. During this process, an attenuation ratio is determined based on the spatial distance and azimuth difference, and points are deducted from the preset highest score according to this attenuation ratio. The score after deduction is used as the free-flow direction geometric constraint score. The aerodynamic performance constraint scoring process for each data sample involves the following steps: obtaining the nacelle wind speed measurement value of the target wind turbine in the current data sample and substituting it into the standard power curve corresponding to the target wind turbine to calculate the theoretical expected power generation; obtaining the actual power generation of the target wind turbine in the current data sample and calculating the ratio between the actual power generation and the theoretical expected power generation, denoted as the power ratio; comparing the power ratio with multiple preset intervals to determine which preset interval it falls into, and determining the aerodynamic constraint based on the preset interval the power ratio falls into according to the scoring rules. The performance constraint scoring, wherein the scoring rule is: the larger the value corresponding to the power ratio falling into the preset interval, the higher the aerodynamic performance constraint score is assigned; in the process of obtaining the adjacent comparison constraint score corresponding to each data sample, the following steps are performed: based on the wind farm layout and the free flow direction data at the current moment, determine the potential downstream wind turbines affected by the wake of the target wind turbine in the airflow direction; extract the actual power generation of the target wind turbine in the current data sample, and the actual power generation of the downstream wind turbine at the corresponding moment, and calculate the power difference characteristics between the two; obtain the adjacent comparison constraint score based on the power difference characteristics.

[0012] In some embodiments, during the joint consistency calibration of GF model parameters, the following steps are performed: Selecting time periods with stable free-flow wind direction from the preprocessed historical operation dataset of the wind farm, constructing a multivariate joint optimization problem, wherein the decision variables of the joint optimization problem include: free-flow wind speed sequence, free-flow wind direction sequence, G model parameters, and corresponding parameters in the F model; setting the objective function of the joint optimization problem to minimize the overall error between the predicted and actual measured values ​​of the nacelle wind speed of all wind turbine units in the entire wind farm; using the G model parameters obtained after calibration as initial values, solving the objective function of the joint optimization problem through a phased optimization algorithm to obtain the parameter solutions of the G model and the F model, and using them as the optimal parameter combination.

[0013] In some embodiments, the following steps are performed during the process of restoring the free-flow wind speed and direction: The real-time measured values ​​of the nacelle wind speed, yaw error angle, and turbulence intensity are acquired and input into the calibrated G model for inverse calculation to obtain the corrected local effective wind speed in front of the turbine rotor after excluding nacelle disturbances; the local effective wind speed in front of the turbine rotor is used as the target wind speed reference value for the corresponding turbine observation point in the F model, and an inverse optimization problem is constructed with the free-flow wind speed and direction as the variables to be optimized; the inverse optimization solution module in the F model is called to drive the forward wake simulation module to perform multiple rounds of iterative calculations based on the variables to be optimized, automatically solving the inverse optimization problem to obtain the optimal free-flow wind speed and direction estimates at the current moment.

[0014] In some embodiments, the following formula is used in obtaining the locally effective inflow velocity in front of the fan rotor after nacelle disturbance is eliminated: ,in, To correct for the local effective inflow velocity in front of the fan rotor at time t after eliminating nacelle disturbances, G(·) is the G model, V nacelle (t) represents the cabin wind speed measurement at time t, γ represents the yaw error angle at time t, and T I Let t be the turbulence intensity. These are the parameters of the calibrated G model.

[0015] In some embodiments, the expression for the inverse optimization problem is: , Here is the estimated free-flow wind speed at time t. Here is the estimated free-flow wind direction at time t. The effective inflow velocity in front of the fan rotor at time t after eliminating nacelle disturbances. Here, WD represents the assumed free-flow wind speed, TI represents the assumed free-flow wind direction, Layout represents the turbulence intensity at time t, and Layout represents the wind farm turbine location layout. These are the parameters of the calibrated G model. This represents the calculation results of the forward wake simulation module at the corresponding wind turbine observation point in the F model.

[0016] In a second aspect, this application provides a wind turbine nacelle wind speed collaborative restoration system, which uses the wind turbine nacelle wind speed collaborative restoration method as described in any embodiment of the first aspect to perform wind turbine nacelle wind speed collaborative restoration. The system includes: a model building module, used to build a nacelle local flow field response model and a wind field restoration model respectively, denoted as the G model and the F model; a data acquisition and preprocessing module, used to acquire historical operating datasets of the wind farm and preprocess them; a model calibration module, used to perform G model parameter calibration and GF model parameter joint consistency calibration sequentially based on the preprocessed historical operating datasets of the wind farm; a restoration module, used to restore the free-flow wind speed and free-flow wind direction based on the real-time acquired nacelle wind speed using the calibrated G model and F model; and an uncertainty quantification module, used to perform uncertainty quantification evaluation on the restored free-flow wind speed and perform adaptive updates of the G model parameters and F model parameters according to the evaluation results.

[0017] By employing the wind turbine nacelle wind speed collaborative restoration scheme provided above, this embodiment of the application effectively eliminates the error accumulation caused by local nacelle disturbances and global site wake effects during the joint derivation process by separately constructing a local flow field response model and a wind field restoration model for the nacelle, and adopting a strategy of sequential calibration and joint consistency calibration. This significantly improves the accuracy of restoring free-flow wind speed and direction based solely on basic nacelle anemometer data. Simultaneously, by introducing uncertainty quantification assessment and an adaptive update mechanism for model parameters, the system can keenly perceive the dynamic evolution of the external environment and the long-term degradation of the turbine's aerodynamic characteristics, endowing it with strong robustness and self-calibration capabilities throughout its entire lifecycle. Thus, without the need for additional expensive hardware such as anemometer towers or lidar, it provides high-precision, low-cost, and sustainably evolving underlying wind condition data support for refined yaw-wind control and overall power collaborative optimization of wind farms.

[0018] Furthermore, in some embodiments, a multi-dimensional constraint scoring mechanism is introduced to achieve high-precision intelligent screening of massive wind farm operation data. This scheme breaks through the limitations of traditional screening that relies solely on wind direction geometry. By deeply integrating spatial layout characteristics (sector area judgment), turbine operating mechanisms (power curve comparison), and spatial coupling relationships between turbines (upstream and downstream power difference characteristics), a multi-level cross-validation mechanism is constructed. This mechanism can accurately identify and eliminate hidden wake interference in complex wind farm environments, thereby obtaining a subset of free-flow data with extremely high purity. A quantitative evaluation strategy combining "preset high score and attenuation deduction" and "interval scoring" scientifically characterizes the strength of wake interference and the confidence level of data samples, greatly improving the precision and fault tolerance of sample screening. This high-quality sample, selected based on multi-dimensional scoring, provides an accurate physical benchmark for parameter calibration of the nacelle local flow field response model (G model), eliminating model distortion caused by sample bias at the source, and laying a solid data quality foundation for subsequent high-precision restoration of the entire wind speed.

[0019] Furthermore, in some embodiments, by performing joint consistency calibration of the G-model and F-model parameters, the error accumulation and coupling distortion barriers caused by the isolated calibration of traditional local aerodynamic models and global wake models are broken. First, by selectively screening high-quality time-period data with stable wind direction to construct a multivariate joint optimization problem, the interference of transient complex wind condition fluctuations on the calibration of underlying parameters is effectively filtered, improving the reliability of the baseline data. Second, with the minimum overall error of all wind turbines in the field as the global objective function, the free flow sequence and the parameters of the two sets of models are incorporated into the same optimization framework for collaborative solution, forcing the G-model and F-model to achieve a high degree of consistency and deep integration in physical space and data logic. Finally, the previously separately calibrated G-model parameters are introduced as high-quality prior initial values, and a phased optimization algorithm is used for dimensionality reduction solution, which not only significantly narrows the optimization blind spot in the high-dimensional nonlinear parameter space and effectively avoids the risk of the algorithm getting trapped in local optima, but also greatly accelerates the convergence efficiency of the joint solution.

[0020] Furthermore, in some embodiments, a step-by-step real-time collaborative solution architecture is constructed, from fine correction of local nacelle data to global free-flow inversion of the wind farm. First, the multi-dimensional operating parameters collected in real time are precisely decoupled using a pre-calibrated G model, effectively eliminating complex aerodynamic disturbances caused by rotor sweep and nacelle structure obstruction, and restoring a high-confidence local effective wind speed in front of the rotor, providing a reliable physical benchmark for global inversion. Second, this local effective wind speed is innovatively used as the observation target constraint of the macroscopic wake F model, cleverly constructing an inverse optimization problem, and transforming the extremely complex nonlinear aerodynamics site-level calculation into an efficient mathematical optimization process through a closed-loop iterative optimization mechanism. This mechanism not only significantly improves the anti-interference capability and calculation accuracy of real-time wind field reconstruction, but also enables the wind farm to achieve dynamic, high-precision, and accurate restoration of the macroscopic free-flow wind speed and direction of the entire field using only conventional sensor data from a single unit, without relying on external hardware such as anemometer towers. Attached Figure Description

[0021] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts, wherein: Figure 1 An exemplary flowchart of the wind turbine nacelle wind speed collaborative restoration method according to an embodiment of this application is shown; Figure 2 An exemplary flowchart illustrating an embodiment of this application is provided for filtering out a subset of data from a wind turbine that is not affected by the wake of the upstream wind turbine. Figure 3 An exemplary flowchart illustrating the acquisition of the free-flow wind direction geometric constraint score corresponding to each data sample is shown in an embodiment of this application. Figure 4 An embodiment of this application is shown. Figure 4 An exemplary flowchart illustrating the acquisition of aerodynamic performance constraint scores corresponding to each data sample is shown in an embodiment of this application. Figure 5 An exemplary flowchart illustrating the acquisition of the adjacent comparison constraint score corresponding to each data sample in an embodiment of this application is shown; Figure 6 An exemplary flowchart illustrating the joint consistency calibration of GF model parameters according to an embodiment of this application is shown; Figure 7 An exemplary flowchart illustrating the restoration of free-flow velocity and free-flow direction according to an embodiment of this application is shown; Figure 8 An exemplary structural block diagram of the wind turbine nacelle wind speed collaborative restoration system according to an embodiment of this application is shown. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] It should be understood that the terms "comprising" and "including" used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0025] Figure 1 An exemplary flowchart of the wind turbine nacelle wind speed collaborative restoration method 100 according to an embodiment of this application is shown.

[0026] like Figure 1 As shown, in step S110, a local flow field response model and a wind field reconstruction model are constructed respectively. The local flow field response model is denoted as model G, and the wind field reconstruction model is denoted as model F. Specifically, the wind field reconstruction model is a wind field reconstruction model based on high-fidelity true wake simulation and inverse optimization.

[0027] In the embodiments of this application, the G-model is used to predict nacelle wind speed measurements based on the effective local inflow velocity in front of the wind turbine rotor, the yaw error angle, and the turbulence intensity. By constructing an independent nacelle local flow field response model (G-model), an anemometer measurement distortion caused by the wind turbine's own structure is specifically characterized and corrected. This allows the subsequent processing to focus on the relatively clean local inflow velocity, which is only affected by the wake.

[0028] Specifically, the expression for the G model is: V nacelle =G(V local ,γ,T I ;θ G )=α(γ,T I)×V local +β(γ,T I ), where V nacelle Here is the measured wind speed in the cabin, G(·) is the G model, and V local The effective inflow velocity in front of the fan rotor is γ, where γ is the yaw error angle and T is the yaw rate. I For turbulence intensity, θ G Here are the parameters of the G model, α(·) is the gain function, and β(·) is the offset function.

[0029] In the embodiments of this application, the F-model includes a forward wake simulation module and a reverse optimization solution module. The corrected local inflow wind speed is used as the input observation value of the F-model. The F-model uses a physics-based wake simulator to accurately characterize the flow field interaction between the fans, and uses an intelligent optimization algorithm to solve for the free flow conditions in reverse.

[0030] Specifically, the forward wake simulation module is a high-fidelity wake model based on physical mechanisms. It simulates and calculates the wind speed at any point in the wind farm based on the input free-flow wind speed, free-flow wind direction, turbulence intensity, and wind farm layout.

[0031] Specifically, the inverse optimization solution module uses the free-flow wind speed and free-flow wind direction as decision variables, and the matching error between the wind speed at the corresponding location calculated by the forward wake simulation module and the corresponding observed value as the objective function. It iteratively calls the wind speed at the corresponding location to perform inverse optimization to obtain the free-flow wind speed and free-flow wind direction that minimize the objective function.

[0032] After completing step S110, in step S120, the historical operation dataset of the wind farm is obtained and preprocessed.

[0033] In the embodiments of this application, continuous historical operation datasets of wind farms are obtained from the wind farm SCADA system and preprocessed, including quality checks, outlier removal, time alignment, and air density correction.

[0034] After completing step S120, in step S130, the G model parameters are calibrated and the GF model parameters are jointly calibrated based on the preprocessed historical operation dataset of the wind farm.

[0035] In the embodiments of this application, during the calibration of G-model parameters, firstly, a subset of data from the preprocessed historical operation dataset of the wind farm is selected, where the wind turbines are not affected by the wake of the upstream wind turbines. Then, the initial parameters of the G-model are obtained by performing weighted least squares fitting on the data in the subset.

[0036] In the embodiments of this application, an islanding condition identification algorithm is executed during the process of filtering out the data subset of wind turbines that are not affected by the wake of upstream wind turbines. The specific process can be found in [reference needed]. Figure 2 .

[0037] Figure 2 An exemplary flowchart illustrating an embodiment of this application is shown to filter out a subset of data from wind turbines that are not affected by the wake of upstream wind turbines.

[0038] like Figure 2 As shown, in step S210, an empty data subset is constructed. In step S220, the multi-dimensional constraint score corresponding to each data sample is obtained through the preprocessed historical operation dataset of the wind farm. In step S230, the multi-dimensional constraint scores corresponding to each data sample are weighted and summed to obtain the comprehensive score corresponding to each data sample. In step S240, samples with comprehensive scores higher than a preset threshold are stored in the data subset.

[0039] In the embodiments of this application, the multi-dimensional constraint score includes the free-flow direction geometric constraint score, the aerodynamic performance constraint score, and the adjacent wind turbine comparison constraint score.

[0040] In the embodiments of this application, the formula S=w1S is used in the process of weighted summation of the multi-dimensional constraint scores corresponding to each data sample. dir +w2S pwr +w3S nei , of which S dir S is used to score the geometric constraints of the free-flowing wind direction. pwr For aerodynamic performance constraint scoring, S nei The adjacent wind turbines are compared and constrained for scoring, with w1, w2, and w3 being preset weights, and w1 > w2, w3.

[0041] In the embodiments of this application, the specific process involved in obtaining the free-flow wind direction geometric constraint score corresponding to each data sample can be found in [reference needed]. Figure 3 .

[0042] Figure 3 An exemplary flowchart illustrating the process of obtaining the free-flowing wind direction geometric constraint score for each data sample according to an embodiment of this application is shown.

[0043] like Figure 3As shown, in step S310, based on the unit layout coordinates of the wind farm and the free incoming wind direction data at the current moment, the wake influence sector areas of all upstream wind turbines are constructed. In step S320, it is determined whether the target wind turbine is located within the wake influence sector area of at least one upstream wind turbine. In response to the target wind turbine not being located within the wake influence sector area of at least one upstream wind turbine, in step S330, it is determined that there is no wake conflict, and the preset highest score is used as the geometric constraint score for the free incoming wind direction. In response to the target wind turbine being located within the wake influence sector area of at least one upstream wind turbine, in step S340, it is determined that there is a wake conflict, and the geometric constraint score for the free incoming wind direction is obtained based on the spatial distance and azimuth difference between the target wind turbine and the corresponding upstream wind turbine.

[0044] In an embodiment of the present application, during the execution of step S310, first, a global planar coordinate system of the wind farm is established. The exact geographical or relative coordinates of all wind turbines in the wind farm are obtained. Denote the coordinates of the i-th wind turbine as (X i , Y i ), and the coordinates of the target wind turbine (i.e., the wind turbine whose wake influence needs to be judged) are denoted as (X j , Y j ). The free incoming wind direction angle at the current moment is obtained and denoted as θ wd (usually taking due north as 0° and calculating clockwise). To simplify geometric judgment, usually, the coordinate system of the measured wind farm is rotated along the current wind direction so that the X-axis (or Y-axis) is parallel to the wind direction. The new coordinates of the wind turbine after coordinate rotation can be expressed as Xi′ = X i × cos(θ wd =) - Y i × sin(θ wd ), Yi′ = X i × sin(θ wd ) + Y i × cos(θ wd ). By comparing the downwind coordinates after rotation, those wind turbines that belong to "upstream wind turbines" can be quickly screened out (for example, if the wind blows along the positive X′ direction, the wind turbine i with Xi′ < Xj′ is the upstream wind turbine of the target wind turbine j).

[0045] Next, according to aerodynamic principles (such as the Jensen wake model), the wind turbine wake diffuses as it propagates downstream. A wake expansion half-angle, denoted as α, needs to be defined. The value of α can be calculated using the wake expansion coefficient k: tan(α) = k (for example, k is often taken as 0.075 for onshore wind farms and 0.04 for offshore wind farms, and can also be dynamically adjusted based on turbulence intensity). Using the wake centerline as the axis of symmetry, angles α are expanded to both sides, forming an infinitely extending region with a apex angle of 2α. Adding the initial influence range of the wind turbine rotor diameter D, the apex of the actual wake sector is often located a certain distance upstream of the rotor, or directly with the wind turbine as the apex and an initial width of D for the sector region. For ease of calculation, it is usually abstracted as (X... i (i, y) is a vertex with direction θ. wd A sector with an angle of [-α, +α].

[0046] In the embodiments of this application, during the execution of step S340, the attenuation ratio is determined based on the spatial distance and azimuth difference, and points are deducted from the preset highest score according to the attenuation ratio. The score after deduction is used as the geometric constraint score of the free flow direction.

[0047] In the embodiments of this application, the specific process involved in obtaining the aerodynamic performance constraint score corresponding to each data sample can be found in [reference needed]. Figure 4 .

[0048] Figure 4 An exemplary flowchart illustrating the process of obtaining the aerodynamic performance constraint score corresponding to each data sample according to an embodiment of this application is shown.

[0049] like Figure 4 As shown, in step S410, the nacelle wind speed measurement value of the target wind turbine in the current data sample is obtained and substituted into the standard power curve corresponding to the target wind turbine to calculate the theoretical expected power generation. In step S420, the actual power generation of the target wind turbine in the current data sample is obtained, and the ratio between the actual power generation and the theoretical expected power generation is calculated and recorded as the power ratio. In step S430, the power ratio is compared with multiple preset intervals to determine which preset interval it falls into, and the aerodynamic performance constraint score is determined based on the scoring rules according to the preset interval into which the power ratio falls.

[0050] In the embodiments of this application, the scoring rule is as follows: the larger the value corresponding to the preset interval into which the power ratio falls, the higher the aerodynamic performance constraint score is assigned. Specifically, the possible range of power ratio values ​​is divided into N non-overlapping, increasing preset intervals, and a corresponding aerodynamic performance constraint score is assigned to each interval. This satisfies a monotonically increasing relationship where "the larger the interval value, the higher the score." When the preset interval corresponding to the power ratio is obtained, its corresponding aerodynamic performance constraint score is directly acquired.

[0051] In the embodiments of this application, the specific process of obtaining the adjacent comparison constraint score corresponding to each data sample can be found in [reference needed]. Figure 5 .

[0052] Figure 5 An exemplary flowchart illustrating the process of obtaining the adjacent comparison constraint score corresponding to each data sample according to an embodiment of this application is shown.

[0053] like Figure 5 As shown, in step S510, based on the wind farm layout and the current free-flow direction data, potential downstream wind turbines affected by the wake of the target wind turbine in the airflow direction are identified. In step S520, the actual power generation of the target wind turbine in the current data sample and the actual power generation of the downstream wind turbine at the corresponding time are extracted, and the power difference characteristics between the two are calculated. In step S530, the adjacent comparison constraint score is obtained based on the power difference characteristics.

[0054] In the embodiments of this application, during step S510, the global turbine coordinate layout data of the wind farm (i.e., the three-dimensional or two-dimensional spatial coordinates (X,Y) of each turbine) and the free-flow wind direction angle θ from the current meteorological data are read. Using the currently evaluated "target turbine" as a reference point, a wake influence area (typically a fan-shaped or conical diffusion area with a certain wake expansion angle) is established along the wind direction θ. Through spatial geometric projection and coordinate calculation, the system automatically traverses other turbines within the wind farm, filters out turbines whose physical locations fall within the wake expansion area, and marks them as "potential downstream turbines" affected by the wake interference of the target turbine.

[0055] In the embodiments of this application, during step S520, after determining the upstream and downstream topological relationship, the system synchronously extracts the real-time actual power generation P of the target wind turbine based on the data sample at the current moment. target And the real-time actual power generation P of the potential downstream wind turbines locked in step S510. downstream Then, the power difference characteristic ΔP between the two is calculated using mathematical formulas. This characteristic is typically expressed as the absolute difference (i.e., ΔP = P). target -P downstream It exists in the form of a relative attenuation ratio, and is used to quantify the degree of energy loss between upstream and downstream units.

[0056] In the embodiments of this application, during step S530, the power difference feature ΔP calculated in step S520 is substituted into a preset evaluation rule or mapping function to transform it into the final adjacent comparison constraint score. According to the wake aerodynamics mechanism, if the target wind turbine is in an undisturbed free flow, the wake generated after it fully absorbs wind energy will cause a significant decrease in the power of the downstream wind turbine. In this case, ΔP will show a positive and significant difference consistent with the wake deficit law. The more the difference feature conforms to theoretical physical expectations, the higher the adjacent comparison constraint score assigned by the system. Conversely, if the power of the downstream wind turbine is close to or even greater than that of the target wind turbine (e.g., ΔP≤0), it indicates that the target wind turbine is likely not in an advantageous free flow position, and the system will assign it an extremely low constraint score.

[0057] For details regarding the specific process of joint consistency calibration of GF model parameters in the embodiments of this application, please refer to [link / reference]. Figure 6 .

[0058] Figure 6 An exemplary flowchart of joint consistency calibration of GF model parameters is shown in an embodiment of this application.

[0059] like Figure 6 As shown, in step S610, data from periods with stable free-flow wind direction are selected from the preprocessed historical operation dataset of the wind farm to construct a multivariate joint optimization problem. In step S620, the objective function of the joint optimization problem is set to minimize the overall error between the predicted and actual measured wind speed values ​​of all wind turbine units in the entire wind farm. In step S630, the G-model parameters obtained after calibration are used as initial values, and the objective function of the joint optimization problem is obtained through a phased optimization algorithm to obtain the parameter solutions of the G-model and F-model, which are then used as the optimal parameter combination.

[0060] In the embodiments of this application, the decision variables of the joint optimization problem include: the free-flow wind speed sequence, the free-flow wind direction sequence, the parameters of the G model, and the parameters of the forward wake simulation module in the F model.

[0061] In the embodiments of this application, the staged optimization algorithm specifically includes: In the first stage, the parameters of the forward wake simulation module of the F model are fixed at the default value, and the wind direction is fixed as the wind direction of the wind tower or numerical weather prediction. The free flow wind speed sequence and the parameters of the G model are used as optimization variables to minimize the cabin wind speed prediction error.

[0062] In the second stage, the parameters of the G model are fixed as the optimization results of the first stage, and the parameters of the free-flow wind speed sequence and the positive wake simulation module of the F model are used as optimization variables to continue optimization.

[0063] In the third stage, the results of the first two stages are used as the starting point, and all decision variables are optimized at the same time, and global fine-tuning is performed until convergence.

[0064] After completing step S130, in step S140, the calibrated G model and F model are used to reconstruct the free-flow wind speed and free-flow wind direction based on the real-time collected cabin wind speed.

[0065] The specific process involved in step S140 in the embodiments of this application can be found in [reference needed]. Figure 7 .

[0066] Figure 7 An exemplary flowchart illustrating the restoration of free-flow velocity and free-flow direction according to an embodiment of this application is shown.

[0067] like Figure 7 As shown, in step S710, the real-time measured values ​​of the nacelle wind speed, yaw error angle, and turbulence intensity are acquired and input into the calibrated G model for inverse calculation to obtain the local effective inflow wind speed in front of the turbine rotor after excluding nacelle disturbances. In step S720, the local effective inflow wind speed in front of the turbine rotor is used as the target wind speed benchmark value for the corresponding turbine observation point in the F model, and an inverse optimization problem is constructed with the free inflow wind speed and free inflow wind direction as the variables to be optimized. In step S730, the inverse optimization solution module in the F model is called to drive the forward wake simulation module to perform multiple rounds of iterative calculations based on the variables to be optimized, automatically solving the inverse optimization problem to obtain the optimal free inflow wind speed estimate and free inflow wind direction estimate at the current moment.

[0068] In the embodiments of this application, the following formula is used in the process of obtaining the locally effective incoming air velocity in front of the fan rotor after eliminating nacelle disturbances: ,in, To correct for the local effective inflow velocity in front of the fan rotor at time t after eliminating nacelle disturbances, G(·) is the G model, V nacelle (t) represents the cabin wind speed measurement at time t, γ represents the yaw error angle at time t, and T I Let t be the turbulence intensity. These are the parameters of the calibrated G model.

[0069] In the embodiments of this application, the expression for the inverse optimization problem is: , Here is the estimated free-flow wind speed at time t. Here is the estimated free-flow wind direction at time t. The effective inflow velocity in front of the fan rotor at time t after eliminating nacelle disturbances. Here, WD represents the assumed free-flow wind speed, TI represents the assumed free-flow wind direction, Layout represents the turbulence intensity at time t, and Layout represents the wind farm turbine location layout. These are the parameters of the calibrated G model. This represents the calculation results of the forward wake simulation module at the corresponding wind turbine observation point in the F model.

[0070] After step S140 is completed, in step S150, the uncertainty of the restored free-flow wind speed is quantitatively evaluated, and the G model parameters and F model parameters are adaptively updated according to the evaluation results.

[0071] In the embodiments of this application, the impact of uncertainties in model parameters and input measurements on the reconstructed free-flow wind speed is quantified based on Bayesian inference or Monte Carlo simulation methods. Specifically, firstly, it is assumed that the previously calibrated G-model parameters and F-model parameters follow a multidimensional Gaussian distribution centered at their calibration values. Multiple sets of parameter samples are randomly sampled from the constructed multidimensional Gaussian distribution using Bayesian inference or Monte Carlo simulation methods. For each sampled set of parameters, the free-flow wind speed reconstruction process is repeated to obtain a series of reconstructed wind speed sample sets. Statistical analysis is performed on this series of reconstructed wind speed samples to calculate the mean and standard deviation of the samples. Then, confidence intervals at a specified confidence level are calculated and output, ultimately providing the reconstructed wind speed result with confidence intervals.

[0072] In the embodiments of this application, during the process of obtaining the restored wind speed with confidence intervals, real-time operational measurement data is continuously collected, and positive prediction calculations are performed using the current G-model parameters and F-model parameters. The system compares the "predicted value calculated by the model" with the "actual value measured by the sensor" in real time, continuously monitoring and statistically analyzing the prediction residuals between the two. The system internally presets a "performance degradation threshold" to measure model accuracy. As the wind turbine operates for a long time (potentially due to blade aging, environmental changes, etc., leading to changes in aerodynamic characteristics), if the evaluation results show that the monitored prediction residuals continuously increase and exceed the set threshold, the system determines that the currently used G-model and F-model have undergone performance degradation and are no longer fully adapted to the current actual operating conditions. Once the above triggering condition is met, the system will no longer rely on outdated static parameters but will automatically trigger an "online update mechanism for model parameters based on incremental learning." By introducing the latest real-time measurement data as incremental samples, online iteration and fine-tuning are performed on the original parameters, thereby achieving adaptive updates of the G-model parameters and F-model parameters, enabling the model to regain high-precision prediction and restoration capabilities.

[0073] In summary, the embodiments of this application, through the wind turbine nacelle wind speed collaborative restoration scheme provided above, effectively eliminate the error accumulation of local nacelle disturbances and global site wake effects in the joint derivation process by constructing a local flow field response model and a wind field restoration model for the nacelle separately, and adopting a strategy of sequential calibration and joint consistency calibration. This significantly improves the accuracy of restoring free-flow wind speed and direction based solely on basic nacelle wind measurement data. Simultaneously, by introducing uncertainty quantification assessment and an adaptive update mechanism for model parameters, the system can keenly perceive the dynamic evolution of the external environment and the long-term degradation of the turbine's aerodynamic characteristics, endowing the system with strong robustness and self-calibration capabilities throughout its entire lifecycle. Thus, without the need for additional expensive hardware such as wind measurement towers or lidar, it provides high-precision, low-cost, and sustainably evolving underlying wind condition data support for refined yaw wind control and overall power collaborative optimization of wind farms.

[0074] Furthermore, in some embodiments, a multi-dimensional constraint scoring mechanism is introduced to achieve high-precision intelligent screening of massive wind farm operation data. This scheme breaks through the limitations of traditional screening that relies solely on wind direction geometry. By deeply integrating spatial layout characteristics (sector area judgment), turbine operating mechanisms (power curve comparison), and spatial coupling relationships between turbines (upstream and downstream power difference characteristics), a multi-level cross-validation mechanism is constructed. This mechanism can accurately identify and eliminate hidden wake interference in complex wind farm environments, thereby obtaining a subset of free-flow data with extremely high purity. A quantitative evaluation strategy combining "preset high score and attenuation deduction" and "interval scoring" scientifically characterizes the strength of wake interference and the confidence level of data samples, greatly improving the precision and fault tolerance of sample screening. This high-quality sample, selected based on multi-dimensional scoring, provides an accurate physical benchmark for parameter calibration of the nacelle local flow field response model (G model), eliminating model distortion caused by sample bias at the source, and laying a solid data quality foundation for subsequent high-precision restoration of the entire wind speed.

[0075] Furthermore, in some embodiments, by performing joint consistency calibration of the G-model and F-model parameters, the error accumulation and coupling distortion barriers caused by the isolated calibration of traditional local aerodynamic models and global wake models are broken. First, by selectively screening high-quality time-period data with stable wind direction to construct a multivariate joint optimization problem, the interference of transient complex wind condition fluctuations on the calibration of underlying parameters is effectively filtered, improving the reliability of the baseline data. Second, with the minimum overall error of all wind turbines in the field as the global objective function, the free flow sequence and the parameters of the two sets of models are incorporated into the same optimization framework for collaborative solution, forcing the G-model and F-model to achieve a high degree of consistency and deep integration in physical space and data logic. Finally, the previously separately calibrated G-model parameters are introduced as high-quality prior initial values, and a phased optimization algorithm is used for dimensionality reduction solution, which not only significantly narrows the optimization blind spot in the high-dimensional nonlinear parameter space and effectively avoids the risk of the algorithm getting trapped in local optima, but also greatly accelerates the convergence efficiency of the joint solution.

[0076] Furthermore, in some embodiments, a step-by-step real-time collaborative solution architecture is constructed, from fine correction of local nacelle data to global free-flow inversion of the wind farm. First, the multi-dimensional operating parameters collected in real time are precisely decoupled through a pre-calibrated G model, effectively eliminating complex aerodynamic disturbances caused by rotor sweep and nacelle structure obstruction, and restoring a high-confidence local effective wind speed in front of the rotor, providing a reliable physical benchmark for global inversion. Second, this local effective wind speed is innovatively used as the observation target constraint of the macroscopic wake F model, cleverly constructing an inverse optimization problem, and transforming the extremely complex nonlinear aerodynamics site-level calculation into an efficient mathematical optimization process through a closed-loop iterative optimization mechanism. This mechanism not only significantly improves the anti-interference capability and calculation accuracy of real-time wind field reconstruction, but also enables the wind farm to achieve dynamic, high-precision restoration of the macroscopic free-flow wind speed and direction of the entire field without relying on external hardware such as anemometer towers, using only conventional sensor data from a single unit.

[0077] This application also provides a wind turbine nacelle wind speed collaborative restoration system, which can use the aforementioned wind turbine nacelle wind speed collaborative restoration method 100 to perform wind turbine nacelle wind speed collaborative restoration, or other methods to perform wind turbine nacelle wind speed collaborative restoration, and this application does not limit it here.

[0078] Figure 8 An exemplary structural block diagram of the wind turbine nacelle wind speed collaborative restoration system according to an embodiment of this application is shown.

[0079] like Figure 8 As shown, the system 800 includes a model building module 810, a data acquisition and preprocessing module 820, a model calibration module 830, a restoration module 840, and an uncertainty quantification module 850.

[0080] Specifically, the model building module 810 is used to build the local flow field response model and the wind field restoration model of the nacelle respectively. The local flow field response model of the nacelle is denoted as the G model, and the wind field restoration model is denoted as the F model.

[0081] Specifically, the data acquisition and preprocessing module 820 is used to acquire historical operation datasets of wind farms and preprocess them.

[0082] Specifically, the model calibration module 830 is used to perform G model parameter calibration and GF model parameter joint consistency calibration sequentially based on the preprocessed historical operation dataset of the wind farm.

[0083] Specifically, the restoration module 840 is used to restore the free-flow wind speed and direction based on the real-time collected cabin wind speed using the calibrated G model and F model.

[0084] In the embodiments of this application, the restoration module 840 adopts a microservice architecture, wherein the G-model correction service and the F-model optimization solution service are decoupled and deployed, communicating through a message middleware or RPC framework. The F-model optimization solution service can dynamically scale up or down according to the load to meet the needs of real-time processing of large-scale wind farm data.

[0085] Specifically, the uncertainty quantification module 850 is used to perform uncertainty quantification assessment on the restored free-flow wind speed and to perform adaptive updates of the G model parameters and F model parameters based on the assessment results.

[0086] In the embodiments of this application, a visualization and service interface module is also provided to display the restoration results, model performance indicators and flow field visualization graphics, and to provide a standard data interface for external systems to call.

[0087] When system 800 performs wind turbine nacelle wind speed collaborative restoration using the aforementioned wind turbine nacelle wind speed collaborative restoration method 100, the aforementioned steps S110 are executed through model building module 810, S120 through data acquisition and preprocessing module 820, S130 through model calibration module 830, S140 through restoration module 840, and S150 through uncertainty quantification module 850. The specific execution process can be found above and will not be repeated here.

[0088] While numerous embodiments of this application have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will arise for those skilled in the art without departing from the spirit and intent of this application. It should be understood that various alternatives to the embodiments of this application described herein may be employed in the practice of this application. The appended claims are intended to define the scope of protection of this application and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A method for coordinated restoration of wind speed in a wind turbine nacelle, characterized in that, include: A local flow field response model and a wind field reconstruction model were constructed separately. The local flow field response model was denoted as model G, and the wind field reconstruction model was denoted as model F. Obtain historical operation datasets of wind farms and preprocess them; Based on the preprocessed historical operation dataset of the wind farm, the parameters of the G model and the joint consistency calibration of the parameters of the GF model are performed sequentially. The calibrated G model and F model are used to reconstruct the free-flow wind speed and direction based on the real-time collected cabin wind speed; The uncertainty of the restored free-flow wind speed is quantitatively evaluated, and the parameters of the G model and the F model are adaptively updated based on the evaluation results.

2. The wind turbine nacelle wind speed collaborative restoration method according to claim 1, characterized in that, The G model is used to predict the nacelle wind speed measurement based on the local effective incoming wind speed in front of the wind turbine rotor, the yaw error angle, and the turbulence intensity. The expression for the G model is: V nacelle =G(V local ,γ,T I ;θ G )=α(γ,T I )×V local +β(γ,T I ), where V nacelle Here is the measured wind speed in the cabin, G(·) is the G model, and V local The effective inflow velocity in front of the fan rotor is γ, where γ is the yaw error angle and T is the yaw rate. I For turbulence intensity, θ G Here are the parameters of the G model, α(·) is the gain function, and β(·) is the offset function.

3. The wind turbine nacelle wind speed collaborative restoration method according to claim 1, characterized in that, The F model includes a forward wake simulation module and a reverse optimization solution module; The forward wake simulation module is a high-fidelity wake model based on physical mechanisms. It simulates and calculates the wind speed at any point in the wind field based on the input free-flow wind speed, free-flow wind direction, turbulence intensity and wind farm layout. The reverse optimization solution module uses the free-flow wind speed and free-flow wind direction as decision variables, and the matching error between the wind speed at the corresponding location calculated by the forward wake simulation module and the corresponding observed value as the objective function. It iteratively calls the wind speed at the corresponding location to perform reverse optimization to obtain the free-flow wind speed and free-flow wind direction that minimize the objective function.

4. The wind turbine nacelle wind speed coordinated restoration method according to claim 1, characterized in that, The following steps are performed during the calibration of G-model parameters: From the preprocessed historical operation dataset of the wind farm, a subset of data is selected that is not affected by the wake of the upstream wind turbines; The initial parameters of the G model are obtained by performing weighted least squares fitting on the data in the aforementioned subset. In the process of selecting the subset of data for wind turbine units that are not affected by the wake of upstream wind turbines, the following steps are performed: Construct an empty subset of data; The multi-dimensional constraint score corresponding to each data sample is obtained by preprocessing the historical operation dataset of the wind farm. The multi-dimensional constraint scores corresponding to each data sample are weighted and summed to obtain the comprehensive score for each data sample. Samples with a comprehensive score higher than a preset threshold are stored in a subset of the data.

5. The wind turbine nacelle wind speed coordinated restoration method according to claim 4, characterized in that, The multi-dimensional constraint score includes free-flow direction geometric constraint score, aerodynamic performance constraint score, and adjacent wind turbine comparison constraint score. In the process of obtaining the free-flow wind direction geometric constraint score for each data sample, the following steps are performed: Based on the wind farm's turbine layout coordinates and the current free-flow wind direction data, construct the wake influence sector of all upstream wind turbines, and determine whether the target wind turbine is located within the wake influence sector of at least one upstream wind turbine. In response to the fact that the target wind turbine is not located within the wake influence sector of at least one upstream wind turbine, it is determined that there is no wake conflict, and the preset highest score is used as the geometric constraint score for the free flow direction. In response to the fact that the target wind turbine is located within the wake influence sector of at least one upstream wind turbine, a wake conflict is determined. A free-flow wind direction geometric constraint score is obtained based on the spatial distance and azimuth difference between the target wind turbine and the corresponding upstream wind turbine. In the process of obtaining the free-flow wind direction geometric constraint score based on the spatial distance and azimuth difference between the target wind turbine and the corresponding upstream wind turbine, an attenuation ratio is determined based on the spatial distance and azimuth difference. Points are deducted from the preset maximum score according to the attenuation ratio, and the score after deduction is used as the free-flow wind direction geometric constraint score. In the process of obtaining the aerodynamic performance constraint score corresponding to each data sample, the following steps are performed: Obtain the nacelle wind speed measurement value of the target wind turbine in the current data sample, and substitute it into the standard power curve corresponding to the target wind turbine to calculate the theoretical expected power generation. Obtain the actual power generation of the target wind turbine in the current data sample, and calculate the ratio between the actual power generation and the theoretical expected power generation, which is denoted as the power ratio. The power ratio is compared with multiple preset intervals to determine which preset interval it falls into. Based on the scoring rules, the aerodynamic performance constraint score is determined according to the preset interval into which the power ratio falls. The scoring rules are as follows: the larger the value corresponding to the preset interval into which the power ratio falls, the higher the aerodynamic performance constraint score is assigned. In the process of obtaining the adjacent comparison constraint score for each data sample, the following steps are performed: Based on the wind farm layout and the current free-flow wind direction data, potential downstream wind turbines affected by the wake of the target wind turbine in the airflow direction are identified. Extract the actual power generation of the target wind turbine in the current data sample, and the actual power generation of the downstream wind turbine at the corresponding time, and calculate the power difference characteristics between the two. The adjacent comparison constraint score is obtained based on the power difference feature.

6. The wind turbine nacelle wind speed collaborative restoration method according to claim 1, characterized in that, During the joint consistency calibration of GF model parameters, the following steps are performed: Selecting time periods with stable free-flow wind direction from the preprocessed historical operation dataset of the wind farm, a multivariate joint optimization problem is constructed. The decision variables of the joint optimization problem include: free-flow wind speed sequence, free-flow wind direction sequence, G model parameters, and corresponding parameters in the F model. The objective function of the joint optimization problem is set to minimize the overall error between the predicted and actual measured wind speed values ​​of the nacelles of all wind turbines in the entire wind farm. The parameters of the G model obtained after calibration are used as initial values. The objective function of the joint optimization problem is obtained through a phased optimization algorithm, and the parameter solutions of the G model and F model are obtained, which are then used as the optimal parameter combination.

7. The wind turbine nacelle wind speed coordinated restoration method according to claim 3, characterized in that, In the process of reconstructing the free-flow wind speed and direction, the following steps are performed: The system acquires the real-time wind speed measurement value of the nacelle, yaw error angle and turbulence intensity at the current moment, and inputs them into the calibrated G model for reverse calculation to obtain the local effective incoming wind speed in front of the wind turbine rotor after eliminating nacelle disturbances. The effective inflow velocity in front of the wind turbine rotor is used as the target wind speed benchmark value of the corresponding wind turbine observation point in the F model. An inverse optimization problem is constructed with the free inflow velocity and free inflow direction as the variables to be optimized. The inverse optimization solution module in the F model is invoked to drive the forward wake simulation module to perform multiple rounds of iterative calculations based on the variables to be optimized, and automatically solve the inverse optimization problem to obtain the optimal free-flow wind speed estimate and free-flow wind direction estimate at the current moment.

8. The wind turbine nacelle wind speed coordinated restoration method according to claim 7, characterized in that, The following formula is used to obtain the locally effective inflow velocity in front of the fan rotor after eliminating nacelle disturbances: ,in, To correct for the local effective inflow velocity in front of the fan rotor at time t after eliminating nacelle disturbances, G(·) is the G model, V nacelle (t) represents the cabin wind speed measurement at time t, γ represents the yaw error angle at time t, and T I Let t be the turbulence intensity. These are the parameters of the calibrated G model.

9. The wind turbine nacelle wind speed collaborative restoration method according to claim 7, characterized in that, The expression for the inverse optimization problem is: , Here is the estimated free-flow wind speed at time t. Here is the estimated free-flow wind direction at time t. The effective inflow velocity in front of the fan rotor at time t after eliminating nacelle disturbances. Here, WD represents the assumed free-flow wind speed, TI represents the assumed free-flow wind direction, Layout represents the turbulence intensity at time t, and Layout represents the wind farm turbine location layout. These are the parameters of the calibrated G model. This represents the calculation results of the forward wake simulation module at the corresponding wind turbine observation point in the F model.

10. A wind turbine nacelle wind speed coordinated restoration system, characterized in that, The wind turbine nacelle wind speed is restored collaboratively using the wind turbine nacelle wind speed restoration method as described in any one of claims 1-9, wherein the system comprises: The model building module is used to build the local flow field response model and the wind field reconstruction model of the nacelle respectively. The local flow field response model of the nacelle is denoted as model G, and the wind field reconstruction model is denoted as model F. The data acquisition and preprocessing module is used to acquire historical operation datasets of wind farms and preprocess them; The model calibration module is used to perform G model parameter calibration and GF model parameter joint consistency calibration sequentially based on the preprocessed historical operation dataset of the wind farm. The restoration module is used to restore the free-flow wind speed and direction based on the real-time collected cabin wind speed using the calibrated G model and F model. The uncertainty quantification module is used to perform uncertainty quantification assessment on the restored free-flow wind speed and to perform adaptive updates of the G model parameters and F model parameters based on the assessment results.

Images