Wind turbine aerodynamic load dynamic correction method, device, medium and equipment

Abstract

The present disclosure relates to the field of wind power generation technology, and provides a wind turbine aerodynamic load dynamic correction method, device, medium and equipment, comprising: collecting sensor data about the target wind turbine based on the sensors prepositioned in the target wind turbine; using the blade finite element model and the blade natural frequency of the target wind turbine in combination with the structural load data to inversely calculate the aerodynamic thrust and torque of each blade element, and generate an aerodynamic force time series; inputting the aerodynamic force time series, environmental parameter data and operating state data into the trained correction coefficient prediction model, predicting the correction coefficient set based on the model, and obtaining the predicted correction coefficient set; configuring the predicted correction coefficient set to the standard blade element momentum theory model, and determining the dynamic corrected aerodynamic load result based on the model after the coefficient configuration and the sensor data. The present disclosure realizes dynamic correction of aerodynamic load under complex working conditions, and improves the operation safety and power generation efficiency of the wind turbine.

Description

Technical Field

[0001] This disclosure relates to the field of wind power generation technology, and more specifically, to a method, apparatus, medium, and equipment for dynamic correction of aerodynamic loads on wind turbine generators. Background Technology

[0002] Momentum-Blade Element Theory (BEMT) is the core model for predicting aerodynamic loads on wind turbines. It predicts aerodynamic loads by calculating the velocity-induced factors of the turbine blades. Since its inception, this theory has become an important tool for the design and operation analysis of wind turbine generators, particularly in the aerodynamic performance evaluation and load prediction stages of wind turbine design. The BEMT method iteratively determines the velocity-induced factors of the turbine blades through repeated calculations, thereby calculating the wind turbine loads.

[0003] While BEMT provides a theoretical basis for wind turbine aerodynamic load prediction, its accuracy heavily relies on numerous empirical correction coefficients, such as tip loss correction, hub loss correction, and Glauert correction. These correction coefficients are typically determined through wind tunnel testing or calibration under static conditions. However, due to the influence of factors such as turbulence, yaw, and stall in complex wind farm environments, traditional correction methods struggle to adapt to dynamic changes during actual operation, leading to significant deviations in load prediction. Furthermore, most correction coefficients are set to fixed values ​​or piecewise functions, failing to respond in real-time to dynamic factors such as sudden wind speed changes and blade aging. Moreover, correction models based on theoretical assumptions lack data-driven capabilities, failing to fully utilize the large amounts of data generated during wind turbine operation for real-time optimization and adjustment. Summary of the Invention

[0004] This disclosure provides at least one method, apparatus, medium, and device for dynamic correction of aerodynamic loads on wind turbine units, which enables dynamic correction of aerodynamic loads under complex operating conditions and improves the operational safety and power generation efficiency of wind turbine units.

[0005] This disclosure provides a method for dynamically correcting the aerodynamic load of a wind turbine, including:

[0006] Sensor data about the target wind turbine is collected based on sensors pre-installed on the target wind turbine; wherein, the sensor data includes structural load data, environmental parameter data and operating status data;

[0007] Using the finite element model of the blade and the natural frequency of the target wind turbine blade, combined with the structural load data, the aerodynamic thrust and torque of each blade element are calculated in reverse to generate an aerodynamic time series.

[0008] The aerodynamic time series, the environmental parameter data, and the operating status data are input into the trained correction coefficient prediction model. Based on the trained correction coefficient prediction model, the set of correction coefficients is predicted to obtain the predicted correction coefficient set.

[0009] The set of predicted correction coefficients is configured into the standard blade element momentum theory model, and the dynamically corrected aerodynamic load results are determined based on the standard blade element momentum theory model with the coefficients configured, the sensing data, and the aerodynamic time series.

[0010] This disclosure provides a dynamic correction device for the aerodynamic load of a wind turbine generator, comprising:

[0011] The data acquisition module is used to acquire sensing data about the target wind turbine based on sensors pre-installed on the target wind turbine; wherein, the sensing data includes structural load data, environmental parameter data, and operating status data;

[0012] The sequence generation module is used to perform inverse calculations on the aerodynamic thrust and torque of each blade element using the blade finite element model and the natural frequency of the target wind turbine, combined with the structural load data, to generate an aerodynamic time series.

[0013] The coefficient prediction module is used to input the aerodynamic time series, the environmental parameter data and the operating status data into the trained correction coefficient prediction model, and predict the correction coefficient set based on the trained correction coefficient prediction model to obtain the predicted correction coefficient set.

[0014] The result correction module is used to configure the set of prediction correction coefficients into the standard blade element momentum theory model, and determine the dynamically corrected aerodynamic load result based on the standard blade element momentum theory model after coefficient configuration, the sensing data and the aerodynamic time series.

[0015] This disclosure provides a computer device, including a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the computer device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, a dynamic correction method for aerodynamic loads of a wind turbine generator as described in any of the above possible embodiments is performed.

[0016] This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the dynamic correction method for aerodynamic loads of a wind turbine as described in any of the above possible embodiments.

[0017] The dynamic correction method, apparatus, medium, and equipment for aerodynamic loads of wind turbines provided in this disclosure specifically involve: collecting sensor data about the target wind turbine based on sensors pre-installed on the target wind turbine; using a finite element model of the blades and the natural frequencies of the target wind turbine blades, combined with structural load data, performing inverse calculations of the aerodynamic thrust and torque of each blade element to generate an aerodynamic time series; inputting the aerodynamic time series, environmental parameter data, and operating status data into a trained correction coefficient prediction model; predicting the set of correction coefficients based on the trained correction coefficient prediction model to obtain a predicted correction coefficient set; configuring the predicted correction coefficient set into a standard blade element momentum theory model; and determining the dynamically corrected aerodynamic load result based on the standard blade element momentum theory model with configured coefficients and the sensor data.

[0018] In this way, the present disclosure obtains a set of correction coefficients by using sensor data and a trained correction coefficient prediction model, and realizes dynamic correction of the aerodynamic load of the wind turbine based on the set of correction coefficients and the standard blade element momentum theory model, so as to continuously optimize the performance of the unit in complex environment and under different operating conditions, and improve the operating safety and power generation efficiency of the wind turbine.

[0019] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings referenced in the embodiments will be briefly described below. These drawings are incorporated in and constitute a part of this specification. They illustrate embodiments conforming to this disclosure and, together with the specification, serve to explain the technical solutions of this disclosure. It should be understood that the following drawings only show some embodiments of this disclosure and should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without creative effort.

[0021] Figure 1 A flowchart of a dynamic correction method for aerodynamic loads of a wind turbine provided in an embodiment of this disclosure is shown;

[0022] Figure 2 A flowchart of a method for calculating aerodynamic time series provided by an embodiment of this disclosure is shown;

[0023] Figure 3 A flowchart is shown below illustrating a method for determining a trained correction coefficient prediction model provided in an embodiment of this disclosure.

[0024] Figure 4A flowchart is shown for a training method of a modified coefficient prediction model provided in an embodiment of this disclosure;

[0025] Figure 5 A flowchart of a model optimization method provided by an embodiment of this disclosure is shown;

[0026] Figure 6 A schematic diagram of the structure of a dynamic correction device for aerodynamic loads of a wind turbine provided in an embodiment of this disclosure is shown.

[0027] Figure 7 A schematic diagram of another dynamic correction device for aerodynamic loads of a wind turbine provided in an embodiment of this disclosure is shown;

[0028] Figure 8 A schematic diagram of the structure of a computer device provided in an embodiment of this disclosure is shown. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0030] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0031] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0032] Momentum-Blade Element Theory (BEMT) is a core method widely used for predicting aerodynamic loads on wind turbines. This theory analyzes the interaction between the rotational motion of wind turbine blades and airflow, predicting the aerodynamic loads on the blades in a wind field based on the momentum theorem and energy balance equations. In this process, BEMT determines the velocity-induced factor of the wind turbine blades through iterative calculations, thereby solving for the wind turbine's load results. The advantages of this method lie in its simplicity and relatively high computational efficiency, making it suitable for wind turbine design and preliminary aerodynamic performance evaluation.

[0033] However, with the increase in the scale of wind turbines and the increasing complexity of the working environment, the BEMT method faces the problems of slow calculation speed and large amount of computation. Especially in the actual working conditions of wind turbines with complex and ever-changing operating conditions, repeated iterations of the calculation process will significantly reduce real-time performance.

[0034] Research has revealed that the forecast accuracy of BEMT relies on a series of empirical correction coefficients, such as tip loss correction, hub loss correction, and Glauert correction. These correction coefficients typically need to be set through wind tunnel experiments or static condition calibration, thus limiting their application in complex wind farm environments. Dynamic factors in wind farms, such as turbulence, yaw, and stall, affect blade aerodynamic performance, rendering traditional correction coefficients ineffective in addressing sudden wind speed changes and blade aging. Furthermore, correction coefficients are generally set to fixed values ​​or piecewise functions, failing to respond in real-time to wind speed variations and equipment aging, leading to deviations in wind turbine load predictions. This not only affects the lifespan of the turbine but may also impact the power generation efficiency of the wind farm. Since existing models primarily rely on assumptions and empirical corrections, lacking real-time data-driven dynamic optimization capabilities, they cannot effectively utilize real-time data from wind turbine operation for parameter optimization.

[0035] Based on the above research, this disclosure provides a method, apparatus, medium, and device for dynamic correction of aerodynamic loads of wind turbine units. Specifically, it collects sensing data about the target wind turbine unit based on sensors pre-installed on the target wind turbine unit; using the blade finite element model and the natural frequency of the target wind turbine unit's blades, combined with structural load data, it performs inverse calculations of the aerodynamic thrust and torque of each blade element to generate an aerodynamic time series; it inputs the aerodynamic time series, environmental parameter data, and operating status data into a trained correction coefficient prediction model, and predicts the correction coefficient set based on the trained correction coefficient prediction model to obtain a predicted correction coefficient set; it configures the predicted correction coefficient set into a standard blade element momentum theory model, and determines the dynamically corrected aerodynamic load result based on the standard blade element momentum theory model with configured coefficients and the sensing data.

[0036] In this embodiment of the disclosure, a set of correction coefficients is obtained by using sensor data and a trained correction coefficient prediction model. Based on this set of correction coefficients and the standard blade element momentum theory model, the aerodynamic load of the wind turbine is dynamically corrected to continuously optimize the performance of the unit in complex environments and under different operating conditions, thereby improving the operational safety and power generation efficiency of the wind turbine.

[0037] To facilitate understanding of this embodiment, the executing entity of the dynamic correction method for aerodynamic loads of wind turbine generators provided in this disclosure will first be described in detail. The executing entity of the dynamic correction method for aerodynamic loads of wind turbine generators provided in this disclosure is a computer device. This computer device can be a terminal device or a server. The terminal device can also be a mobile device, a user terminal, a terminal, a handheld device, a computing device, an in-vehicle device, a wearable device, etc. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data, and artificial intelligence platforms. Optionally, this method can also be applied to an implementation environment composed of computer devices and servers.

[0038] The dynamic correction method for aerodynamic loads of wind turbine generators provided in this application will be described in detail below with reference to the accompanying drawings. See also Figure 1 The diagram shows a flowchart of a dynamic correction method for aerodynamic loads of a wind turbine provided in an embodiment of this disclosure. The method includes the following steps S101 to S104:

[0039] S101, based on sensors pre-installed on the target wind turbine, collects sensing data about the target wind turbine.

[0040] Understandably, the various sensors pre-installed on the target wind turbine are primarily used to monitor the turbine's operation in real time, ensuring accurate real-time data. This sensor data includes structural load data, environmental parameter data, and operational status data. Structural load data may include blade root bending moment and tower vibration acceleration; environmental parameter data may include wind speed and turbulence intensity; and operational status data may include yaw angle.

[0041] Specifically, structural load data reflects the magnitude and direction of the forces borne by various components of the wind turbine. For example, strain gauge sensors installed at the blade root are common devices used to measure the bending moment at the blade root. When the blade bends under wind force, the strain gauge generates a corresponding change in resistance. By measuring this resistance change and performing a series of conversions and calculations, the magnitude of the bending moment at the blade root can be obtained. The bending moment at the blade root is an important indicator of the stress condition of the blade; excessive bending moment may lead to cracks or even breakage of the blade. As another example, the tower, as the supporting structure of the wind turbine, vibrates under wind force. Accelerometers at the bottom of the tower can measure the acceleration of the tower in different directions in real time. By analyzing this vibration acceleration data, the tower's vibration acceleration can be indirectly calculated. The magnitude and frequency of the tower's vibration acceleration reflect the stress state and stability of the tower. If the vibration acceleration is too high or the frequency is abnormal, it may indicate a structural problem with the tower or that it has been subjected to excessive external forces, requiring timely intervention to avoid serious accidents such as collapse. Structural load data may also include hub torque, blade torque, and other data.

[0042] Specifically, environmental parameter data describes the external environmental conditions of the wind turbine, which directly affect its operating performance and aerodynamic loads. Wind speed sensors measure wind speed in real time; it is one of the most important environmental parameters in wind turbine operation, as wind speed determines the amount of wind energy the turbine can capture, thus affecting its power generation. Wind direction sensors determine the direction of the wind, which is crucial for yaw control. By adjusting the yaw angle to ensure the blades are always facing the wind, wind energy capture efficiency can be maximized. Furthermore, lidar sensors can be used to obtain turbulence intensity, which refers to the degree of irregular variation in wind speed over time and space. High turbulence intensity makes the aerodynamic loads on the wind turbine more complex and unstable, increasing the risk of fatigue damage. LiDAR sensors accurately measure the three-dimensional distribution of wind speed by emitting a laser beam and measuring the time and intensity changes of the reflected light, thereby calculating turbulence intensity. Environmental parameter data can also include temperature, humidity, and air pressure data.

[0043] Specifically, operational status data records the wind turbine's own operating status, reflecting its performance under different operating conditions. Yaw angle is a crucial parameter in the operational status data; it represents the angle at which the wind turbine nacelle deflects relative to the wind direction. Real-time monitoring of the yaw angle allows us to understand whether the turbine's yaw system is functioning correctly and whether the turbine can accurately and promptly track changes in wind direction. Abnormal yaw angles may prevent the blades from consistently facing the wind, reducing wind energy capture efficiency and potentially damaging the turbine's structure. Besides yaw angle, operational status data may also include parameters such as blade speed and pitch angle.

[0044] S102, using the finite element model of the blade and the natural frequency of the target wind turbine blade, combined with the structural load data, the aerodynamic thrust and torque of each blade element are calculated in reverse to generate an aerodynamic time series.

[0045] Here, the finite element model of a blade is a mathematical model that discretizes the blade into a finite number of elements. By dividing the blade into many small elements and performing mechanical analysis on each element, the stress conditions of the blade during actual operation can be simulated more accurately. For example, the blade can be divided into hundreds of triangular or quadrilateral elements, each with its specific material properties, geometry, and boundary conditions. By calculating the mechanical behavior of these elements, the mechanical response of the entire blade can be obtained. The natural frequency of the blade is an inherent property of the blade itself. When the blade is subjected to external forces, it will vibrate near its natural frequency. When the external excitation frequency is close to the blade's natural frequency, resonance may occur, leading to a greater load on the blade.

[0046] Understandably, given the structural load data, the aerodynamic thrust and torque of each blade element can be calculated backwards using the finite element model and natural frequencies of the blade. A blade element is a small segment on the blade, serving as the basic unit for analyzing its aerodynamic performance. Each blade element is subjected to aerodynamic forces generated by the pressure distribution of wind on the blade surface, including lift and drag. Lift is perpendicular to the airflow direction, causing the blade to generate a rotational torque that drives the wind turbine to generate electricity; drag, on the other hand, is opposite to the airflow direction, consuming some wind energy and reducing the efficiency of the wind turbine. Specifically, by measuring the load data at the blade root (i.e., the bending moment at the blade root and the tower vibration acceleration), and combining this with the finite element model and natural frequencies, the aerodynamic thrust and torque experienced by the blade element near that location can be calculated. Arranging these calculation results in chronological order generates an aerodynamic time series. Here, the aerodynamic time series reflects the changes in the aerodynamic forces on the blades over time. At different points in time, due to the continuous changes in environmental factors such as wind speed and direction, as well as the operating status of the wind turbine, the aerodynamic thrust and torque experienced by each blade element will also change. For example, when the wind speed suddenly increases, the aerodynamic forces experienced by the blades will also increase accordingly; when the wind direction changes, the aerodynamic force distribution at different positions on the blades will also change.

[0047] For example, refer to Figure 2 As shown, this disclosure presents a method for calculating aerodynamic time series, which may include the following steps S201 to S202:

[0048] S201, Establish a finite element model of the blades of the target wind turbine, and perform modal analysis on the finite element model of the blades to obtain the natural frequencies of each blade of the target wind turbine.

[0049] Here, when establishing the finite element model of the blade, it is necessary to comprehensively consider the actual structural characteristics and material properties of the blade. First, the three-dimensional geometric data of the blade is acquired, including information such as the blade's profile shape and thickness variation. Then, based on the type of material used in the blade, its elastic modulus, Poisson's ratio, density, and other material property parameters are determined. Next, an appropriate element type is selected to discretize the blade, such as using high-precision triangular or quadrilateral elements, and the element mesh is reasonably divided according to the blade's geometry and stress characteristics to ensure that the quality and density of the mesh can meet the requirements of computational accuracy.

[0050] Furthermore, after the model is established, modal analysis is performed. Modal analysis is a method used to determine the natural frequencies and mode shapes of a structure. It obtains the natural frequencies and corresponding mode shapes of the blades at different orders by solving the vibration equations of the finite element model. During modal analysis, the boundary conditions of the blades need to be considered, such as whether the blade root is fixed. Through modal analysis, the natural frequencies of each blade in different vibration modes can be obtained.

[0051] S202, based on the natural frequency of each blade, the bending moment at the root of the blade, and the vibration acceleration of the tower, the aerodynamic thrust and torque of each blade element are calculated in reverse to generate the aerodynamic time series.

[0052] Specifically, after obtaining the natural frequencies of each blade, reverse calculations can be performed by combining structural load data such as blade root bending moments and tower vibration acceleration measured by sensors. Reverse calculation is a complex iterative process. First, based on initial aerodynamic assumptions, the blade response, including blade deformation, stress distribution, and vibration, needs to be calculated using a finite element model. Then, the calculated blade response is compared with the actually measured blade root bending moments and tower vibration accelerations. If discrepancies exist, the aerodynamic assumptions need to be adjusted, and the calculation recalculated. Optimization algorithms, such as gradient descent and genetic algorithms, can be used to improve computational efficiency and accuracy when adjusting the aerodynamic assumptions. This process continues iterating and adjusting until the calculated blade response matches the actual measured data. At this point, the obtained aerodynamic values ​​can be considered as the actual aerodynamic thrust and torque experienced by each blade element at different time points. Arranging these aerodynamic values ​​in chronological order generates an aerodynamic time series.

[0053] In this way, the aerodynamic changes of wind turbines under different operating conditions can be accurately obtained, providing strong support for the performance evaluation and optimization of wind turbines.

[0054] In some possible embodiments, since the blades are subject to various vibration and noise interferences during actual wind turbine operation, such as random vibrations caused by wind turbulence and vibration transmission from mechanical components, these noises can affect the accuracy of subsequent aerodynamic calculations. Therefore, modal analysis results can be used to filter out these vibration and noise interferences. By analyzing the blade's vibration response at different frequencies, signals caused by inherent vibrations and those caused by noise can be distinguished, thereby filtering the measurement data and improving the signal-to-noise ratio. After filtering out vibration and noise interference, the aerodynamic thrust and torque of each blade element can be inferred using the relationship between strain and bending moment.

[0055] Here, when using the relationship between strain and bending moment to inversely deduce the aerodynamic thrust and torque of each blade element, a mathematical relationship can be established between the bending moment at the blade root and the aerodynamic thrust and torque of each blade element based on the principles of mechanics of materials and structural mechanics. This requires considering multiple factors such as blade geometry, material properties, stress distribution, and vibration. For example, different airfoils at different positions of the blade result in different responses to aerodynamic forces, affecting the distribution of bending moment. Simultaneously, blade vibration alters the effect of aerodynamic forces, thus influencing the magnitude and direction of the bending moment. Furthermore, by combining this relationship with measured bending moment data at the blade root, the aerodynamic thrust and torque experienced by each blade element can be inversely deduced. During this inverse deduction process, multiple iterative calculations may be necessary to continuously refine the estimated values ​​of aerodynamic thrust and torque, ensuring that the calculated blade root bending moment matches the measured values.

[0056] In some possible embodiments, since the aerodynamic thrust and torque data of each blade element obtained through back-reaming are raw load data, these data may contain some high-frequency noise components. This high-frequency noise may be caused by factors such as measurement errors and environmental interference. In order to make these data compatible with the subsequent BEMT (Blade Element Momentum Theory) model input, the converted load data needs to be low-pass filtered. Here, low-pass filtering is a signal processing technique that allows low-frequency signals to pass through while suppressing high-frequency signals. By selecting an appropriate cutoff frequency, high-frequency noise in the load data can be effectively removed, while retaining low-frequency signals that reflect the true trend of aerodynamic force changes. The aerodynamic data after low-pass filtering forms an aerodynamic time series that matches the input requirements of the BEMT model.

[0057] S103, the aerodynamic time series, the environmental parameter data and the operating status data are input into the trained correction coefficient prediction model, and the correction coefficient set is predicted based on the trained correction coefficient prediction model to obtain the predicted correction coefficient set.

[0058] Understandably, the set of prediction correction coefficients contains several key parameters that play an indispensable role in accurately describing the aerodynamic characteristics and operating status of wind turbines. Among them, the tip loss correction factor compensates for aerodynamic losses at the blade tip caused by airflow around the blade. In actual operation, the airflow velocity and pressure distribution at the blade tip differs from that at the center of the blade, and this correction factor effectively corrects the impact of this difference on overall aerodynamic performance. The critical induction factor threshold for Glauert correction is an important parameter in blade element momentum theory. It determines under what induction factor conditions Glauert correction is required to avoid unreasonable predictions from the theoretical model under high load conditions. The rotational enhancement coefficient considers the aerodynamic performance enhancement effect of factors such as centrifugal force and Coriolis force during blade rotation, and is of great significance for accurately evaluating the power output and load characteristics of wind turbines.

[0059] Here, aerodynamic time series, environmental parameter data, and operational status data are input into a trained correction coefficient prediction model, which can predict a set of correction coefficients based on the input data.

[0060] Specifically, refer to Figure 3 As shown, the trained correction coefficient prediction model can be obtained through the following steps S301 to S303:

[0061] S301, construct a standard leaf element momentum theory model.

[0062] Understandably, the standard blade element momentum theory model is the fundamental theoretical model for the aerodynamic performance analysis of wind turbines. It divides the blade along its span into multiple blade elements, calculates the aerodynamic forces on each element separately, and then considers the influence of the wake through momentum theory to obtain the aerodynamic performance of the entire blade. Constructing a standard blade element momentum theory model requires specifying the blade's geometric parameters (such as airfoil, chord length, and twist angle), aerodynamic parameters (such as lift coefficient and drag coefficient), and operating parameters (such as wind speed and rotational speed). The model's computational functions are then implemented through mathematical formulas and algorithms.

[0063] S302, Obtain the correction coefficient prediction model to be trained and the training dataset.

[0064] Specifically, the correction coefficient prediction model to be trained is typically constructed using machine learning or deep learning algorithms, such as neural networks and support vector machines. This model can learn the complex relationship between input data and correction coefficients from a large amount of data. Here, this disclosure employs a Long Short-Term Memory (LSTM) network. This algorithm can effectively process time-series data and capture long-term dependencies within the data, making it suitable for modeling and predicting time-series data such as aerodynamic forces, environmental parameters, and operational states. Through training the LSTM model, the correction coefficient can be predicted more accurately, thereby optimizing the system's control strategy.

[0065] The training dataset comprises multiple training data subsets, each of which contains multiple time-series data sets. Each time-series data set covers aerodynamic training data, environmental parameter training data, and operational status training data at the corresponding time point.

[0066] S303, the modified coefficient prediction model to be trained is trained based on the training dataset and the standard leaf element momentum theory model to obtain the trained modified coefficient prediction model.

[0067] Here, refer to Figure 4 As shown, the training method for the modified coefficient prediction model may include the following steps S3031 to S3033:

[0068] S3031, for each subset of training data, determine a set of training prediction correction coefficients corresponding to the subset of training data based on the prediction model of correction coefficients to be trained and the subset of training data; configure the standard blade element momentum theory model based on the set of training prediction correction coefficients; and determine the training aerodynamic load result corresponding to the subset of training data based on the configured standard blade element momentum theory model and the subset of training data.

[0069] Specifically, for each subset of training data, firstly, based on the training correction coefficient prediction model to be trained and the subset of training data, a set of training prediction correction coefficients corresponding to that subset is determined. The training correction coefficient prediction model to be trained outputs a set of predicted correction coefficients based on the input aerodynamic training data, environmental parameter training data, and operational status training data. These correction coefficients constitute the training prediction correction coefficient set. Then, the standard blade element momentum theory model is configured based on the training prediction correction coefficient set. That is, the predicted correction coefficients are substituted into the standard blade element momentum theory model to adjust the relevant parameters of the model, so that the model can better adapt to the working conditions represented by the current subset of training data. Finally, based on the configured standard blade element momentum theory model and the subset of training data, the training aerodynamic load results corresponding to that subset are determined. Specifically, the input parameters in the subset of training data are input into the configured standard blade element momentum theory model, and the model calculates the corresponding aerodynamic load results, such as blade thrust and torque.

[0070] S3032, determine the loss value between the training aerodynamic load result corresponding to the training data subset and the sample label corresponding to the training data subset based on a preset loss function, and adjust the model parameters of the correction coefficient prediction model to be trained based on the loss value.

[0071] Specifically, based on a preset loss function, the loss value between the predicted aerodynamic load result corresponding to a subset of training data and the sample label corresponding to that subset can be determined. Here, the sample label is the actual measured aerodynamic load result, which serves as the true value to evaluate the model's prediction accuracy. The loss function measures the degree of difference between the predicted aerodynamic load result and the sample label. Common loss functions include mean squared error and mean absolute error. By calculating the loss value, the model's prediction error can be quantified.

[0072] Furthermore, the model parameters of the prediction model with corrected coefficients can be adjusted based on the loss value. The purpose of adjusting the model parameters is to make the model's prediction results closer to the true values, thereby reducing the loss value. In machine learning and deep learning, commonly used parameter tuning methods include gradient descent, stochastic gradient descent, and the Adam optimization algorithm. These methods update the model parameters according to the gradient information of the loss function, following the direction that reduces the loss value, and gradually optimize the model's performance through multiple iterations.

[0073] S3033, Repeat steps S3031 to S3032 until the training results meet the preset requirements, and obtain the trained correction coefficient prediction model.

[0074] Here, steps S3031 and S3032 are repeated to continuously adjust the model parameters of the correction coefficient prediction model to be trained. During the training process, certain preset requirements need to be set, such as the loss value being less than a certain threshold and the model's accuracy on the validation set reaching a certain level. When the training results meet the preset requirements, it indicates that the model has learned a reasonable relationship between the input data and the correction coefficients, at which point the trained correction coefficient prediction model is obtained.

[0075] For example, in view of the variability of complex wind field environments, in order to further improve the accuracy and adaptability of the correction coefficient prediction model, this disclosure also proposes a working condition model optimization method, which may include the following steps (1) to (4):

[0076] (1) When the wind speed corresponding to the training data subset is higher than the preset wind speed threshold, it is determined that the training data subset is in a high wind speed condition, and the optimization weight of the Glauert critical induction factor threshold is increased to suppress the influence of wake turbulence effect on load prediction.

[0077] (2) When the yaw angle corresponding to the training data subset is higher than the preset yaw angle threshold, it is determined that the training data subset is in a large yaw angle condition. The torque deviation caused by the skewed flow is compensated by correcting the lateral velocity distribution, and the rotation enhancement coefficient is updated based on the correlation between the blade element position and the rotation speed to optimize the boundary layer stability of the yaw airflow due to the rotation effect.

[0078] (3) When the turbulence intensity corresponding to the training data subset is higher than the preset turbulence intensity threshold, the training data subset is determined to be in transient strong turbulence condition. A time delay coefficient is introduced to simulate the time-varying effect of the sudden change in angle of attack on the lift coefficient. The rotation enhancement coefficient is adjusted according to the instantaneous change in turbulence intensity to balance the influence of rotational mixing and airflow disturbance.

[0079] (4) During the parameter adjustment process, based on the sensitivity of the tip loss correction factor to thrust prediction, the tip loss correction factor is optimized first, and then other correction coefficients are adjusted iteratively.

[0080] Specifically, when the wind speed corresponding to a subset of training data exceeds a preset wind speed threshold (e.g., 12 m / s), the subset of training data can be determined to be operating under high wind speed conditions. Under high wind speed conditions, the wake turbulence effect of wind turbines becomes more pronounced, which significantly impacts load prediction. Therefore, increasing the optimization weight of the Glauert critical induction factor threshold makes the model pay more attention to adjusting this parameter, thereby suppressing the impact of wake turbulence on load prediction. By adjusting the Glauert critical induction factor threshold, the model's correction strategy under high load conditions can be changed, improving the accuracy of load prediction.

[0081] Meanwhile, when the yaw angle corresponding to a subset of training data exceeds a preset yaw angle threshold (e.g., 15°), it can be determined that the subset of training data is under a large yaw angle condition. Under large yaw angle conditions, the airflow distribution of the wind turbine will change significantly, leading to torque deviation. To correct this deviation, the torque deviation caused by skewed flow can be compensated by correcting the lateral velocity distribution. Simultaneously, the rotational intensification coefficient is updated based on the correlation between blade element position and rotational speed. Since the rotational speed and stress conditions differ at different blade positions, the rotational intensification coefficient also needs to be adjusted accordingly to optimize the boundary layer stability of the yaw airflow due to the rotational effect, thereby improving the model's prediction accuracy under large yaw angle conditions.

[0082] Similarly, when the turbulence intensity corresponding to a subset of training data exceeds a preset turbulence intensity threshold (e.g., 15%), the subset of training data can be determined to be under transient strong turbulence conditions. Under transient strong turbulence conditions, the angle of attack of the airflow changes abruptly, causing the lift coefficient to change over time. To simulate this time-varying effect, a time lag coefficient can be introduced. Here, the time lag coefficient reflects the time delay in the impact of the abrupt change in angle of attack on the lift coefficient, enabling the model to more accurately predict changes in aerodynamic performance. Simultaneously, the rotational intensification coefficient is adjusted according to the instantaneous change in turbulence intensity to balance the effects of rotational mixing and airflow disturbance. In strong turbulence environments, both rotational mixing and airflow disturbance affect aerodynamic performance. By dynamically adjusting the rotational intensification coefficient, the model can better adapt to this complex condition.

[0083] It should be noted that the above three operating conditions can exist individually or simultaneously, without specific limitations. Furthermore, since the tip loss correction factor has a significant impact on thrust prediction, optimizing it first can improve the accuracy of thrust prediction. Specifically, during parameter adjustment, based on the sensitivity of the tip loss correction factor to thrust prediction, it should be optimized first. Then, other correction coefficients, such as the critical induction factor threshold and rotational intensification coefficient of the Glauret correction, should be iteratively adjusted. Through this optimization of the parameter adjustment sequence, the predictive performance of the model can be gradually improved, enabling the model to provide accurate correction coefficient prediction results under various operating conditions.

[0084] Furthermore, to ensure the accuracy and physical plausibility of the correction coefficient prediction model, the optimization objective of the correction coefficient in this disclosure is defined as minimizing the difference between the BEMT (Blade Element Momentum Theory) correction load and the measured value. This is achieved through a weighted mean square error function, where the thrust and torque error weights are set to 0.6 and 0.4, respectively. These weight settings take into account the importance of thrust and torque in aerodynamic load prediction, allowing the model to more reasonably balance the errors of both during optimization. Simultaneously, to ensure physical plausibility, all correction parameters must meet constraints. For example, the tip loss parameter is limited to the range of 0 to 3, and the Glauert critical threshold is between 0.3 and 0.5. These constraints prevent the model from exhibiting non-physical interpretations, ensuring that the model's prediction results conform to actual physical laws, and improving the model's credibility and reliability.

[0085] In this way, by following the above steps and methods, a well-trained prediction model with corrected coefficients can be constructed, and the accuracy and adaptability of the model can be further improved by the working condition model optimization method, thereby providing reliable support for the performance analysis and optimization of wind turbine units.

[0086] In some other embodiments, to further improve the model's prediction accuracy and generalization ability, after obtaining the trained correction coefficient prediction model, it can be validated, such as using spatiotemporal cross-validation. This method divides the data into training, validation, and test sets in a 7:2:1 ratio to ensure sufficient training, while the validation and test sets are used to evaluate the model's performance and stability. Furthermore, to verify the model's adaptability under different environments and scenarios, independent datasets from different wind turbines can be extracted for validation, further testing the model's generalization ability and avoiding overfitting. In addition, regarding uncertainty analysis, Monte Carlo simulation can be used to quantify the uncertainty of the input data. Specifically, ±5% noise can be added to the model input, and thousands of samples can be performed to simulate the fluctuation range of the correction coefficients under different noise levels. This simulation process can help evaluate the model's robustness to data fluctuations, providing a quantitative basis for uncertainty in practical applications.

[0087] S104, the set of predicted correction coefficients is configured into the standard blade element momentum theory model, and the dynamically corrected aerodynamic load result is determined based on the standard blade element momentum theory model after coefficient configuration, the sensing data and the aerodynamic time series.

[0088] Understandably, while the standard blade element momentum theory model provides a basic framework for aerodynamic load calculation in the aerodynamic performance analysis of wind turbines, its results often deviate from reality due to a series of ideal assumptions based on it, especially under complex and variable operating conditions. This disclosure, by incorporating a set of predicted correction coefficients into the standard blade element momentum theory model, enables targeted correction and optimization of the model, making the aerodynamic load calculation of wind turbines more closely reflect actual conditions and thus yielding more accurate aerodynamic load results (i.e., dynamically corrected aerodynamic load results).

[0089] Here, parameters such as the tip loss correction factor, the critical induction factor threshold of the Glauert correction, and the rotational intensification coefficient in the prediction correction coefficient set can modify the standard blade element momentum theory model from different perspectives. Specifically, the tip loss correction factor compensates for aerodynamic losses at the blade tip caused by airflow around the blade, making the model's assessment of the overall aerodynamic performance of the blade more accurate; the critical induction factor threshold of the Glauert correction adjusts the model's correction strategy under high load conditions, avoiding unreasonable predictions from the theoretical model; and the rotational intensification coefficient considers the aerodynamic performance enhancement effects of centrifugal force, Coriolis force, and other factors during blade rotation, further improving the model's prediction accuracy. After configuring these correction coefficients into the standard blade element momentum theory model, the model can dynamically adjust the calculation parameters according to different operating conditions and environmental conditions, thereby achieving dynamic correction of aerodynamic loads.

[0090] Specifically, the dynamically corrected aerodynamic load results can provide crucial information for subsequent wind turbine performance evaluation, optimization design, and health monitoring. In performance evaluation, accurate aerodynamic load results more accurately reflect the output power, efficiency, and other performance indicators of the wind turbine under different environmental and operating conditions, helping operators understand the actual operating status of the unit. During the optimization design phase, based on accurate aerodynamic load results, the geometry of the blades and the selection of materials can be optimized to improve the aerodynamic performance and power generation efficiency of the wind turbine. In health monitoring, real-time monitoring and analysis of the dynamically corrected aerodynamic load results can promptly identify potential faults in the wind turbine, such as blade fatigue and bearing wear, allowing for appropriate maintenance measures to ensure efficient operation of the wind turbine under different environmental and operating conditions, extend the unit's service life, and reduce operation and maintenance costs.

[0091] For example, to further improve the accuracy and adaptability of the correction coefficient prediction model, this disclosure also proposes a model optimization method, referring to... Figure 5 As shown, the process may include the following steps S501 to S502:

[0092] S501, obtain the real result corresponding to the sensing data with the same time length as the dynamically corrected aerodynamic load result, and construct a model optimization sample set based on the environmental parameter data, the operating status data, the aerodynamic time series and the real result.

[0093] Specifically, the actual results corresponding to the sensing data are obtained for the same time period as the dynamically corrected aerodynamic load results. Here, the actual results are the measured values ​​of the aerodynamic loads. Simultaneously, a model optimization sample set is constructed by combining previously acquired environmental parameter data, operating status data, and aerodynamic time series. By correlating and integrating these data with the actual results, the constructed model optimization sample set can more comprehensively reflect the actual operating conditions of wind turbines under different operating conditions, providing strong data support for the optimization training of the correction coefficient prediction model.

[0094] S502, the modified coefficient prediction model is optimized and trained based on the model optimization sample set and the standard leaf element momentum theory model, so as to use the optimized and trained modified coefficient prediction model to predict the new set of modified coefficients.

[0095] Specifically, the correction coefficient prediction model is optimized and trained based on the model optimization sample set and the standard blade element momentum theory model. During the optimization training process, data from the model optimization sample set is input into the correction coefficient prediction model, and the model outputs a set of predicted correction coefficients. These predicted correction coefficients are then configured into the standard blade element momentum theory model to obtain the predicted aerodynamic load results. The predicted aerodynamic load results are compared with the actual results in the model optimization sample set, and the error between the two is calculated using a preset loss function.

[0096] Understandably, based on the magnitude of the error, optimization algorithms (such as gradient descent and Adam optimization algorithms) are used to adjust the parameters of the correction coefficient prediction model, gradually bringing the model's prediction results closer to the actual results. After multiple iterations of training, until the model's prediction accuracy meets the preset requirements, the optimized correction coefficient prediction model is obtained. Using this optimized model, predictions of new correction coefficient sets can be achieved, further improving the accuracy and reliability of wind turbine aerodynamic load prediction.

[0097] In some possible implementations, to meet real-time requirements, an edge computing and cloud-based collaborative architecture can be adopted. The edge computing layer deploys a lightweight correction coefficient prediction model, which takes real results from a past period (e.g., ten minutes) and corresponding environmental parameter data and aerodynamic time series as input. This lightweight model can quickly predict the correction coefficient for the next moment. In this way, real-time prediction can provide timely adjustments to the wind turbine's control strategy, improving the turbine's adaptability to environmental changes.

[0098] Here, the cloud platform aggregates edge data every 24 hours. This edge data contains a wealth of real-world operational information and prediction results. The cloud platform uses this rich data to retrain the full-parameter model. The full-parameter model has a more complex structure and stronger learning capabilities, enabling it to further uncover potential patterns in the data. After training, the optimized model parameters can be wirelessly updated and synchronized to local devices, forming a collaborative architecture of "real-time edge response - deep cloud optimization." This architecture ensures the system's real-time performance while fully utilizing the powerful computing capabilities and abundant data resources of the cloud to continuously improve the performance of the correction coefficient prediction model.

[0099] In some other embodiments, during the engineering deployment phase, the optimized correction coefficients are stored as a three-dimensional lookup table, categorized by wind speed, yaw angle, and turbulence intensity. Parameter groups are independently configured for each 1 m / s wind speed interval. This categorized storage method allows for a more refined description of the correction coefficients under different operating conditions, improving the model's adaptability to various conditions. The dynamic coefficient table is integrated into the wind turbine main control system via the Modbus / TCP protocol, achieving seamless integration with the wind turbine control system and providing real-time guidance for pitch and yaw control strategies.

[0100] For example, under sudden wind speed conditions, after the edge model detects a sudden increase in wind speed, it can automatically lower the Glauert critical threshold and increase the correction intensity. This adjustment can reduce thrust prediction errors, trigger pitch protection earlier, and prevent wind turbines from being damaged by excessive thrust. During continuous yaw, the lateral inflow correction model can compress torque prediction errors, enabling wind turbines to adjust pitch and yaw angles more accurately, improving power generation efficiency and achieving efficient and stable operation of wind turbines.

[0101] The dynamic correction method, apparatus, medium, and equipment for wind turbine aerodynamic load provided in this disclosure embodiment obtain a set of correction coefficients by utilizing sensor data and a trained correction coefficient prediction model, and realize the dynamic correction of the wind turbine aerodynamic load based on the set of correction coefficients and the standard blade element momentum theory model, so as to continuously optimize the performance of the unit in complex environments and different operating conditions, and improve the operating safety and power generation efficiency of the wind turbine.

[0102] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0103] Based on the same inventive concept, this disclosure also provides a dynamic correction device for the aerodynamic load of a wind turbine corresponding to the dynamic correction method for the aerodynamic load of a wind turbine. Since the principle of the device in this disclosure for solving the problem is similar to the dynamic correction method for the aerodynamic load of a wind turbine described above, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0104] Reference Figure 6 The diagram shown is a schematic of a dynamic correction device 600 for aerodynamic loads of a wind turbine provided in an embodiment of this disclosure. The device includes:

[0105] The data acquisition module 601 is used to acquire sensing data about the target wind turbine based on sensors pre-installed on the target wind turbine; wherein, the sensing data includes structural load data, environmental parameter data and operating status data;

[0106] The sequence generation module 602 is used to perform inverse calculations on the aerodynamic thrust and torque of each blade element using the blade finite element model and the natural frequency of the target wind turbine, combined with the structural load data, to generate an aerodynamic time series.

[0107] The coefficient prediction module 603 is used to input the aerodynamic time series, the environmental parameter data and the operating status data into the trained correction coefficient prediction model, and predict the correction coefficient set based on the trained correction coefficient prediction model to obtain the predicted correction coefficient set.

[0108] The result correction module 604 is used to configure the set of prediction correction coefficients to the standard blade element momentum theory model, and determine the dynamically corrected aerodynamic load result based on the standard blade element momentum theory model after coefficient configuration, the sensing data and the aerodynamic time series.

[0109] In some possible embodiments, the structural load data includes blade root bending moment and tower vibration acceleration; the environmental parameter data includes wind speed and turbulence intensity; and the operating status data includes yaw angle.

[0110] In some possible embodiments, the sequence generation module 602 is specifically used for:

[0111] A finite element model of the blades of the target wind turbine is established, and modal analysis is performed on the finite element model of the blades to obtain the natural frequencies of each blade of the target wind turbine.

[0112] Based on the natural frequency of each blade, the bending moment at the blade root, and the vibration acceleration of the tower, the aerodynamic thrust and torque of each blade element are calculated in reverse to generate the aerodynamic time series.

[0113] In some possible embodiments, refer to Figure 7 As shown, the device further includes:

[0114] Model building module 605 is used to build a standard leaf element momentum theory model;

[0115] The data acquisition module 606 is used to acquire the correction coefficient prediction model to be trained and the training dataset; wherein, the training dataset includes multiple training data subsets, each training data subset includes multiple time series data, and each time series data includes aerodynamic training data, environmental parameter training data and operating status training data at the corresponding time.

[0116] The model training module 607 is used to train the correction coefficient prediction model to be trained based on the training dataset and the standard leaf element momentum theory model, so as to obtain the trained correction coefficient prediction model.

[0117] In some possible embodiments, the model training module 607 is specifically used for:

[0118] For each subset of training data, a set of training prediction correction coefficients corresponding to the training data subset is determined based on the prediction model of the correction coefficients to be trained and the training data subset; the standard leaf element momentum theory model is configured based on the set of training prediction correction coefficients; and the training aerodynamic load result corresponding to the training data subset is determined based on the configured standard leaf element momentum theory model and the training data subset.

[0119] The loss value between the training aerodynamic load result corresponding to the training data subset and the sample label corresponding to the training data subset is determined based on a preset loss function, and the model parameters of the correction coefficient prediction model to be trained are adjusted based on the loss value.

[0120] Repeat the above steps until the training results meet the preset requirements, and obtain the trained correction coefficient prediction model.

[0121] In some possible embodiments, the model training module 607 is specifically used for:

[0122] When the wind speed corresponding to the training data subset is higher than the preset wind speed threshold, the training data subset is determined to be in a high wind speed condition, and the optimization weight of the Glauert critical induction factor threshold is increased to suppress the influence of wake turbulence effect on load prediction.

[0123] When the yaw angle corresponding to the training data subset is higher than the preset yaw angle threshold, the training data subset is determined to be in a large yaw angle condition. The torque deviation caused by the skewed flow is compensated by correcting the lateral velocity distribution, and the rotation enhancement coefficient is updated based on the correlation between blade element position and rotational speed to optimize the boundary layer stability of the yaw airflow due to the rotation effect.

[0124] When the turbulence intensity corresponding to the training data subset is higher than the preset turbulence intensity threshold, the training data subset is determined to be in transient strong turbulence condition. A time delay coefficient is introduced to simulate the time-varying effect of sudden angle of attack on the lift coefficient. The rotation enhancement coefficient is adjusted according to the instantaneous change of turbulence intensity to balance the influence of rotational mixing and airflow disturbance.

[0125] During the parameter adjustment process, based on the sensitivity of the tip loss correction factor to thrust prediction, the tip loss correction factor is optimized first, and then other correction coefficients are adjusted iteratively.

[0126] In some possible embodiments, the result correction module 604 is further configured to:

[0127] Obtain the real results corresponding to the sensing data with the same time length as the dynamically corrected aerodynamic load results, and construct a model optimization sample set based on the environmental parameter data, the operating status data, the aerodynamic time series, and the real results;

[0128] The modified coefficient prediction model is optimized and trained based on the model optimization sample set and the standard leaf element momentum theory model, so as to use the optimized and trained modified coefficient prediction model to predict the new set of modified coefficients.

[0129] Based on the same technical concept, this disclosure also provides a computer device. (See also...) Figure 8 The diagram shows the structure of a computer device 800 provided in this embodiment of the present disclosure, including a processor 801, a memory 802, and a bus 803. The memory 802 stores execution instructions and includes a main memory 8021 and an external memory 8022. The main memory 8021, also called internal memory, is used to temporarily store computational data in the processor 801 and data exchanged with external memory 8022 such as a hard disk. The processor 801 exchanges data with the external memory 8022 through the main memory 8021.

[0130] In this embodiment, the memory 802 is specifically used to store application code that executes the solution of this application, and its execution is controlled by the processor 801. That is, when the computer device 800 is running, the processor 801 communicates with the memory 802 through the bus 803, so that the processor 801 executes the application code stored in the memory 802, and then executes the method described in any of the foregoing embodiments.

[0131] The memory 802 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0132] Processor 801 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0133] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the computer device 800. In other embodiments of this application, the computer device 800 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0134] This disclosure also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the steps of the dynamic correction method for aerodynamic loads of a wind turbine generator described in the above-described method embodiments. The storage medium can be either volatile or non-volatile computer-readable storage.

[0135] This disclosure also provides a computer program product carrying program code. The program code includes instructions that can be used to execute the steps of the dynamic correction method for aerodynamic loads of wind turbine generators described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0136] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0137] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this disclosure, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0138] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0139] In addition, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0140] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A method of dynamic correction of aerodynamic loads of a wind turbine, characterized in that, The method comprises the following steps: collecting sensing data about the target wind turbine based on sensors pre-installed on the target wind turbine; wherein the sensing data comprises structural load data, environmental parameter data and operating state data; using a blade finite element model and the natural frequency of the blade of the target wind turbine, and combining the structural load data, to inversely calculate the aerodynamic thrust and torque of each blade element and generate an aerodynamic force time series; inputting the aerodynamic force time series, the environmental parameter data and the operating state data into a trained correction coefficient prediction model, predicting a correction coefficient set based on the trained correction coefficient prediction model, and obtaining a predicted correction coefficient set; configuring the predicted correction coefficient set to a standard blade element momentum theory model, and determining a dynamically corrected aerodynamic load result based on the standard blade element momentum theory model after the coefficient configuration, the sensing data and the aerodynamic force time series; wherein the correction coefficient prediction model is trained through the following steps: Step 1: Construct a standard blade element momentum theory model; Step 2: Obtain a correction coefficient prediction model to be trained and a training data set; wherein the training data set comprises a plurality of training data subsets, each training data subset comprises a plurality of time series data, and each time series data comprises aerodynamic training data, environmental parameter training data and operating state training data at a corresponding time; Step 3: For each training data subset, determine a training prediction correction coefficient set corresponding to the training data subset based on the correction coefficient prediction model to be trained and the training data subset, configure the standard blade element momentum theory model based on the training prediction correction coefficient set, and determine a training aerodynamic load result corresponding to the training data subset based on the standard blade element momentum theory model after the configuration and the training data subset; Step 4: Determine the loss value between the training aerodynamic load result corresponding to the training data subset and the sample label corresponding to the training data subset based on a preset loss function, and adjust the model parameters of the correction coefficient prediction model to be trained based on the loss value; Step 5: Repeat steps 2 to 4 until the training result meets the preset requirements, and obtain the trained correction coefficient prediction model.

2. The method of claim 1, wherein, The structural load data comprises blade root bending moment and tower vibration acceleration; the environmental parameter data comprises wind speed and turbulence intensity; and the operating state data comprises yaw angle.

3. The method of claim 2, wherein, The inverse calculation of the aerodynamic thrust and torque of each blade element based on the blade finite element model and the natural frequency of the blade of the target wind turbine, in combination with the structural load data, comprises: establishing a blade finite element model of the target wind turbine, and performing modal analysis on the blade finite element model to obtain the natural frequency of each blade of the target wind turbine; based on the natural frequency of each blade, the blade root bending moment and the tower vibration acceleration, inversely calculating the aerodynamic thrust and torque of each blade element to generate the aerodynamic force time series.

4. The method of claim 2, wherein, The model parameter of the to-be-trained correction coefficient prediction model is adjusted based on the loss value, and the adjustment includes: When the wind speed corresponding to the training data subset is higher than a preset wind speed threshold, it is determined that the training data subset is in a high wind speed working condition, the optimization weight of the Glauert critical induction factor threshold is increased to suppress the influence of the wake turbulence effect on the load prediction; When the yaw angle corresponding to the training data subset is higher than a preset yaw angle threshold, it is determined that the training data subset is in a large yaw angle working condition, the torque deviation caused by the skewed flow is compensated by correcting the transverse velocity distribution, and the rotational enhancement coefficient is updated based on the correlation between the blade element position and the rotational speed to optimize the boundary layer stability of the yaw flow under the influence of the rotational effect; When the turbulence intensity corresponding to the training data subset is higher than a preset turbulence intensity threshold, it is determined that the training data subset is in a transient strong turbulence working condition, a time lag coefficient is introduced to simulate the time-varying effect of the angle of attack on the lift coefficient, and the rotational enhancement coefficient is adjusted according to the instantaneous change of the turbulence intensity to balance the influence of the rotational mixing effect and the air flow disturbance; During the parameter adjustment process, based on the sensitivity of the tip loss correction factor to the thrust prediction, the tip loss correction factor is preferentially optimized, and then other correction factors are iteratively adjusted.

5. The method of claim 1, wherein, After determining the dynamically corrected aerodynamic load result based on the standard blade element momentum theory model configured with the correction coefficients, the sensor data, and the aerodynamic force time sequence, the method further includes: Obtaining real results corresponding to the sensor data and having the same time length as the dynamically corrected aerodynamic load result, and constructing a model optimization sample set based on the environmental parameter data, the operating state data, the aerodynamic force time sequence, and the real results; Optimizing and training the correction coefficient prediction model based on the model optimization sample set and the standard blade element momentum theory model, so as to realize the prediction of a new correction coefficient set by using the optimized and trained correction coefficient prediction model.

6. A dynamic correction device for aerodynamic loads of a wind turbine, characterized by It includes: A data acquisition module is configured to acquire sensor data about a target wind turbine based on sensors pre-installed on the target wind turbine; wherein the sensor data includes structural load data, environmental parameter data, and operating state data; A sequence generation module is configured to inversely calculate the aerodynamic thrust and torque of each blade element by using a blade finite element model and the natural frequency of the blade of the target wind turbine, in combination with the structural load data, to generate an aerodynamic force time sequence; A coefficient prediction module is configured to input the aerodynamic force time sequence, the environmental parameter data, and the operating state data into a trained correction coefficient prediction model, predict a correction coefficient set based on the trained correction coefficient prediction model, and obtain a predicted correction coefficient set; A result correction module is configured to configure the predicted correction coefficient set to a standard blade element momentum theory model, and determine a dynamically corrected aerodynamic load result based on the standard blade element momentum theory model configured with the correction coefficients, the sensor data, and the aerodynamic force time sequence; The device further includes: A model construction module is configured to perform step 1: constructing a standard blade element momentum theory model; The data acquisition module is configured to perform step 2: acquire a correction coefficient prediction model to be trained and a training data set; wherein the training data set comprises a plurality of training data subsets, each training data subset comprises a plurality of time series data, and each time series data comprises aerodynamic training data, environmental parameter training data and operating state training data at a corresponding moment; The model training module is configured to perform step 3: for each training data subset, determine a set of training prediction correction coefficients corresponding to the training data subset based on the correction coefficient prediction model to be trained and the training data subset; configure the standard blade element momentum theory model based on the set of training prediction correction coefficients; and determine a training aerodynamic load result corresponding to the training data subset based on the configured standard blade element momentum theory model and the training data subset; Step 4: determine a loss value between the training aerodynamic load result corresponding to the training data subset and a sample label corresponding to the training data subset based on a preset loss function, and adjust model parameters of the correction coefficient prediction model to be trained based on the loss value; Step 5: repeat steps 2 to 4 until the training result meets a preset requirement, and obtain the trained correction coefficient prediction model.

7. A storage medium having stored thereon a computer program, characterized in that The computer program, when executed by a processor, implements the method of any one of claims 1 to 5.

8. A computer device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor, when executing the computer program, implements the method of any one of claims 1 to 5.

Images