A fan blade instability monitoring control method and system

Abstract

This invention relates to the field of wind power generation technology and provides a method and system for monitoring and controlling wind turbine blade instability. The method includes: constructing a one-dimensional ANCF dynamic model of the wind turbine blade, calculating aerodynamic loads, and combining this model with structural dynamic equations to form a structure-aerodynamic coupled dynamic model; performing multi-condition dynamic simulation calculations on the wind turbine blade; extracting dynamic response characteristics by incorporating static load test conditions and measured response results; constructing a parameterized database of blade conditions and responses; establishing a blade instability evaluation system and training a neural network to obtain a surrogate model; and using the above-ground operation to predict the dynamic response characteristics of the blade, and judging the stability state of the blade based on the instability evaluation system and generating control commands, thereby making it possible to perform feedforward prediction and proactive intervention before blade instability occurs.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation technology, and in particular relates to a method and system for monitoring and controlling wind turbine blade instability. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] As wind turbine blades become larger and more flexible, they are subjected to strong coupling effects of aerodynamic, inertial, and elastic forces in complex wind fields, making them highly susceptible to large deformations and aeroelastic instability. To ensure the safe and efficient operation of wind turbines, the control system needs to make rapid judgments and adjustments to the dynamic response of the blades based on real-time wind field conditions (wind speed, wind direction) and operating parameters (pitch angle, rotational speed). However, existing technologies have certain limitations in achieving "rapid prediction" and "active control": First, high-precision finite element simulation models are computationally time-consuming and cannot be directly integrated into the control system for real-time decision-making; second, existing wind turbine control and prediction methods mostly rely on sensor feedback with hysteresis or datasets built based on historical wind turbine operating data. Sensor feedback has an inherent time delay, causing control adjustments to often occur after the blade response; while data-driven methods require long-term operation to accumulate samples, making it difficult to provide effective prediction and adjustment basis in the early stages of wind turbine operation or during periods of rapid changes in operating conditions, thus limiting their real-time and forward-looking control capabilities under complex wind field conditions.

[0004] More critically, existing methods have not yet established a parameterized database of "condition-response" results that can cover all operating conditions and support rapid prediction. Limited by factors such as the high cost of high-precision simulation calculations and insufficient offline result organization, it is difficult to pre-build a complete response result database before wind turbine commissioning or during operation. This makes it impossible for the control system to quickly match and obtain the corresponding blade dynamic response evolution trend within the extremely short timescale before instability occurs when encountering sudden wind loads, thus hindering timely feedforward compensation and optimization adjustments to pitch angle, heading angle, and rotational speed. This limitation weakens the ability to proactively predict large blade deformation and torsional behavior, restricting the stability monitoring and safe operation of large flexible blades under extreme and variable operating conditions. Summary of the Invention

[0005] To address the technical problems mentioned above, this invention provides a method and system for monitoring and controlling wind turbine blade instability. By establishing a high-fidelity structure-aerodynamic coupling dynamic model and constructing a parameterized database of blade operating conditions and responses under multiple operating conditions, a neural network surrogate model is introduced to characterize the nonlinear mapping relationship between blade operating conditions and dynamic responses. This surrogate model is then used to rapidly predict and evaluate the blade's stability state, effectively reducing computational complexity and latency. This makes it possible to perform feedforward prediction and proactive intervention before blade instability occurs, which is of great significance for the safe operation of wind turbines.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a method for monitoring and controlling wind turbine blade instability, comprising: A one-dimensional beam dynamic model of a wind turbine blade is constructed. A local coordinate system is established at the centroid of the airfoil section of the wind turbine blade. The chord length vector is determined by the positions of the leading and trailing edge points. Combined with the two-dimensional relative wind speed of the wind turbine blade section, aerodynamic loads are calculated and integrated with the structural dynamic equations to form a structure-aerodynamic coupled dynamic model. Based on the structure-aerodynamic coupled dynamic model, the operating parameters are changed to perform multi-condition dynamic simulation calculations on the wind turbine blade. By introducing the static load test conditions and actual measured response results of the blade, the simulation calculation results are corrected, and dynamic response characteristic quantities are extracted from the simulation calculation results. Using operating condition parameters and dynamic response features obtained from simulation, a parameterized database of blade operating condition-response is constructed, and a neural network is trained to obtain a neural network surrogate model of blade operating condition-response relationship. During blade instability monitoring, real-time operating data of the wind turbine blades are used as input. The dynamic response characteristics are predicted through the blade condition-response relationship neural network surrogate model to determine the stability state of the blades and generate control commands.

[0008] Furthermore, the step of determining the stability state of the blade and generating control commands includes: determining whether the predicted dynamic response characteristic quantities satisfy the blade instability control evaluation system; if not, using the operating condition parameters as known conditions, and using the pitch angle, speed, and heading angle as adjustable control variables, and minimizing the change in the control variables to solve in reverse the solution to obtain the change in pitch angle, speed, and heading angle.

[0009] Furthermore, the blade instability control evaluation system is as follows: the predicted geometric response does not exceed the safety threshold or the predicted equivalent response characteristic does not show a sudden change compared with the sensor monitoring value.

[0010] Furthermore, the operating parameters include wind speed, wind direction, pitch angle, speed, and heading angle.

[0011] Furthermore, the dynamic response characteristic quantities include geometric response quantities and equivalent response characteristic quantities; The geometric response quantities include tip flexure deformation, tip torsion angle, mid-blade flexure deformation, and mid-blade torsion angle; The equivalent response characteristics include the equivalent bending moment response of the blade root section along the flapping and oscillation directions, the axial strain and bending strain responses at the blade root and key sections, the acceleration response at the blade root or the mid-node of the blade, and the root mean square value of the acceleration response.

[0012] Furthermore, the aerodynamic load is: In the formula, intermediate parameters The unit vector perpendicular to the normal vector of the two-dimensional cross-section frame. , and Let represent the slope vectors of the global position coordinates in the y and z directions, respectively. and The upward resistance coefficient of the ANCF unit. for antisymmetric matrix, It is a 3-order identity matrix. air density, Let be the chord length vector on the airfoil section. It is the relative velocity after the three-dimensional relative velocity is projected onto the two-dimensional cross section.

[0013] Furthermore, the relative velocity projected onto the two-dimensional cross-section after the three-dimensional relative velocity is further defined as: Among them, the three-dimensional frame field composed of gradient vectors , , and These represent the slope vectors of the global position coordinates in the x, y, and z directions, respectively, and are diagonal matrices. The relative velocity experienced by the blade element in actual three-dimensional space In three-dimensional space, the blade element is subjected to an ideal relative velocity. , For the incoming wind speed, The velocity of the leading edge of the blade element during blade operation is the induced velocity of the blade element. , and These are the axial and tangential induction factors, respectively. and These are the projected velocities in the axial and tangential directions, respectively. and These are the axial and tangential projection matrices of the blade, respectively. This represents the velocity of the leading edge of the airfoil of the blade micro-element during blade operation.

[0014] A second aspect of the present invention provides a wind turbine blade instability monitoring and control system, comprising: The modeling and simulation module is configured to: construct a one-dimensional beam dynamic model of the wind turbine blade; establish a local coordinate system at the centroid of the airfoil section of the wind turbine blade; determine the chord length vector by the positions of the leading and trailing edge points; calculate the aerodynamic loads by combining the two-dimensional relative wind speed of the wind turbine blade section; and combine the aerodynamic loads with the structural dynamic equations to form a structure-aerodynamic coupled dynamic model. Based on the structure-aerodynamic coupled dynamic model, the operating parameters are changed to perform multi-condition dynamic simulation calculations on the wind turbine blade; the simulation calculation results are corrected by introducing the blade static load test conditions and actual measured response results; and the dynamic response characteristic quantities are extracted from the simulation calculation results. The proxy model construction module is configured to: construct a blade condition-response parameterized database using operating condition parameters and dynamic response features obtained from simulation, train a neural network, and obtain a blade condition-response relationship neural network proxy model. The safety monitoring and control module is configured to: during blade instability monitoring operation, take the real-time operating condition data of the wind turbine blade as input, and predict the dynamic response characteristic quantity through the blade condition-response relationship neural network proxy model to determine the stability state of the blade and generate control commands.

[0015] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the wind turbine blade instability monitoring and control method described above.

[0016] A fourth aspect of the present invention provides a computer device including a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor executes the program to implement the steps of the wind turbine blade instability monitoring and control method described above.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention establishes a high-fidelity structure-aerodynamic coupling dynamic model and constructs a parameterized database of blade operating conditions and responses under multiple operating conditions. Based on this, a neural network surrogate model is introduced to characterize the nonlinear mapping relationship between blade operating conditions and dynamic responses. This surrogate model is then used to quickly predict and evaluate the blade stability state, effectively reducing computational complexity and latency. This makes it possible to perform feedforward prediction and active intervention before blade instability occurs, which is of great significance for the safe operation of wind turbines.

[0018] Based on the structural model, this invention introduces a quasi-steady-state aerodynamic model to achieve aerodynamic-structural coupling. By calculating the chord length vector of the blade section and the local two-dimensional relative airflow velocity, the aerodynamic load of each section is obtained and mapped to the absolute nodal coordinate system. This is then combined with the structural dynamic equations to form the nonlinear dynamic control equations for blade structure-aerodynamic coupling, which can accurately characterize the dynamic response characteristics of the blade under different wind speeds, wind directions, and blade pitch angles.

[0019] This invention combines the results of multi-condition dynamic simulation with the measured results of static load testing to form a blade condition-response parameterized dataset. Based on this dataset, a blade instability evaluation system is established and a blade condition-response surrogate model is trained and generated.

[0020] This invention creates an instability control scheme that, without considering power generation efficiency and prioritizing safety, can provide early warning information before the risk of blade instability becomes significant. By outputting control commands such as pitch angle, heading angle, and speed, it achieves closed-loop adaptive regulation, thereby effectively suppressing the occurrence of blade instability. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a schematic diagram illustrating the principle of a wind turbine blade instability monitoring and control method according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the structure-operating condition-response relationship neural network surrogate model and instability evaluation in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the monitoring and adjustment reverse neural network proxy model of Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the airfoil cross-section of the blade according to Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the structure of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0024] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] Example 1 This embodiment provides a method for monitoring and controlling wind turbine blade instability.

[0026] To address the issue that large wind turbine blades are prone to large deflection and torsion under complex operating conditions, leading to instability risks, and that traditional real-time analysis and control methods based on high-fidelity finite element models are computationally expensive and difficult to apply online, this embodiment provides a wind turbine blade instability monitoring and control method. This method uses a neural network surrogate model for inverse mapping to monitor and control blade instability, enabling rapid prediction of dynamic response and deformation threshold constraint control of blades under varying operating conditions.

[0027] like Figure 1 As shown in this embodiment, a method for monitoring and controlling wind turbine blade instability is proposed. First, numerical simulations under multiple operating conditions are conducted based on the blade structure-aerodynamic coupling dynamic model, and the results of blade static load test experiments are integrated to construct a parameterized database of "operating condition-response". On this basis, a blade instability evaluation system is formed. Subsequently, a neural network surrogate model of the operating condition-response relationship is established using the database as training data. The operating condition parameters and response data synchronously collected by the prototype sensors are compared and verified against the output of the surrogate model. When the verification results meet the accuracy requirements, the model can be run on the machine, using the operating conditions collected by real-time sensors as input to the surrogate model to quickly predict the key blade responses and perform instability evaluation according to the instability evaluation system. Finally, the evaluation results determine whether control adjustments are needed: if the stability threshold is not met, the reverse calculation capability of the surrogate model is used to solve the optimal operating condition adjustment strategy that meets the blade deformation threshold constraint, obtaining the changes in controllable parameters such as pitch angle, speed, and heading angle; if the stability threshold is met, the current control state is maintained and continuous cyclic monitoring and evaluation are performed.

[0028] This embodiment provides a method for monitoring and controlling wind turbine blade instability, including: First, a structural dynamics model of the blade is established in the global coordinate system using one-dimensional beam elements and the Absolute Nodal Coordinates (ANCF) method. A local frame is established at the centroid of the blade airfoil section to extract the spatial position information of the leading and trailing edges. The chord length vector and section attitude are used to characterize the geometric nonlinear characteristics of the blade under large deformation conditions. The blade is discretized into several ANCF elements along its span. Based on the absolute position and gradient degrees of freedom of the element nodes, the system mass matrix is ​​constructed, and a curvature-like vector is used to describe the nonlinear elastic deformation behavior of the blade, thus forming a structural dynamics expression suitable for large deflection and large torsional conditions.

[0029] Based on the structural model, a quasi-steady-state aerodynamic model is introduced to achieve aerodynamic-structural coupling. By calculating the chord length vector of the blade cross-section and the local two-dimensional relative airflow velocity, the aerodynamic loads at each cross-section are obtained. These loads, along with gravity and aerodynamic torque, are substituted into the generalized external force calculation formula to obtain the generalized external force. Then, the generalized elastic force is subtracted to combine these forces into the generalized force Q in the dynamic control equations, forming the nonlinear dynamic control equations for the blade structure-aerodynamic coupling. This model can accurately characterize the dynamic response characteristics of the blade under different wind speeds, wind directions, and blade pitch angles.

[0030] Based on the aforementioned coupled dynamics model, rapid numerical simulations are performed for various operating conditions. By systematically changing input parameters such as wind speed, wind direction, and blade pitch angle, relevant dynamic response characteristics are obtained, including blade flexural deformation, torsional deformation, equivalent bending moment response of the blade root section along the flapping and oscillation directions, axial strain and bending strain response at the blade root and key sections, acceleration response at the blade root node, and root mean square value of the acceleration response. Simultaneously, the blade static load test conditions and their measured response results are introduced, using the strain and deformation at various points on the blade measured under ultimate loading in the static load test conditions as the safe threshold for blade failure. Then, operating condition parameters such as wind speed, wind direction, pitch angle, rotational speed and heading angle are used as operating condition dimensions, and geometric deformation response and strain, bending moment and acceleration characteristics consistent with sensor output are used as response dimensions. The simulation results are parametrically processed to construct a parametric dynamic response database of "operating condition-response" covering multiple wind speed, multiple wind direction and multiple pitch angle combinations. This database is used to characterize the structural response law and stability evolution characteristics of the blade under different operating conditions, and a blade instability evaluation system is established based on this database.

[0031] Based on this, a neural network surrogate model is constructed using the aforementioned parameterized dynamic response database. Wind speed, wind direction, blade pitch angle, and rotational speed are used as input parameters. Output parameters include the deformation response and torsional response at key blade points, the equivalent bending moment response of the blade root section along the flapping and oscillation directions, the axial strain and bending strain responses at the blade root and key sections, the acceleration response at the blade root node, and their root mean square values. The neural network is used to extract features and train high-dimensional nonlinear mapping relationships, establishing a condition-response surrogate model capable of rapid prediction, thus replacing the traditional computationally intensive numerical simulation process.

[0032] To ensure the reliability and physical consistency of the established blade operating condition-dynamic response surrogate model in engineering applications, and to verify its predictive ability for the aforementioned key dynamic response quantities, prototype measured data were introduced to verify the consistency of the surrogate model after its construction. Specifically, during the prototype or experimental phase, typical response quantities corresponding to the model output were selected as verification objects. The corresponding operating condition parameters were input into the surrogate model to obtain the predicted results of the key blade responses. Simultaneously, the actual dynamic response of the prototype under the same operating conditions was measured. By comparing the consistency between the surrogate model's predicted results and the prototype's measured results in terms of time-domain evolution characteristics and statistical significance, evaluation indicators such as relative error and root mean square error were calculated. This assessed the ability of the operating condition-response mapping relationship established by the surrogate model to characterize the real dynamic behavior of the blade, providing a reliable basis for subsequent blade response prediction and instability identification based on operating condition inputs.

[0033] The blade instability evaluation system, under the control condition of prioritizing operational safety and temporarily disregarding power generation efficiency, inputs real-time operating data collected by sensors into a pre-established surrogate model during actual wind turbine operation to rapidly predict and feedforward assess the dynamic response and stability of the blade. When the prediction results indicate that the blade flexure deformation or torsion angle is close to 80% of the preset safety threshold, or when the predicted equivalent bending moment response of the blade root section along the flapping and oscillation directions, the axial strain and bending strain response at the blade root and key sections, the acceleration response at the blade root node, and the root mean square value of the acceleration response show a sudden change compared to the current sensor monitoring values, the system determines that the blade has a potential instability risk and triggers the corresponding safety monitoring and control module. Based on the predicted instability development trend, the control module reversely utilizes a proxy model to comprehensively generate pitch angle adjustment commands, heading angle compensation commands, and speed optimization control commands. This guides the adjustment of the operating conditions to a safe range where the corresponding blade deformation level is approximately 50% of the safety threshold, or where the equivalent bending moment response, axial strain and bending strain response, acceleration response, and root mean square value of acceleration response at the blade root section remain stable or gradually decrease by 20%. This forms a closed-loop control mechanism that proactively intervenes in the wind turbine's operating status before instability occurs, effectively suppressing the risk of blade instability and ensuring the safe and stable operation of the wind turbine.

[0034] This embodiment provides a wind turbine blade instability monitoring and control method, which mainly involves constructing a wind turbine blade structure-aerodynamic coupling dynamic model based on the absolute nodal coordinate method, and combining a parameterized database and a neural network surrogate model to achieve blade instability monitoring, early warning, and adaptive control. The method includes the following steps: Step 1: Obtain known dynamic characteristic data of wind turbine blades under multiple operating conditions.

[0035] The dynamic characteristic data includes structural data, operating condition data, and corresponding blade dynamic response data.

[0036] Step 101: Construct a one-dimensional beam dynamic model of the wind turbine blade based on the Absolute Nodal Coordinates (ANCF) method, and construct the element mass matrix and the generalized elastic force based on the curvature vector based on the absolute nodal coordinates.

[0037] Specifically, the wind turbine blades are divided into several ANCF elements along the spanwise direction. Each ANCF element node uses absolute node coordinates to describe its position and orientation degrees of freedom, thereby achieving an accurate description of the blade's large deformation and large rotation behavior.

[0038] Step 102, as follows Figure 4 As shown, a local coordinate system (r) is established at the centroid of the airfoil section of the blade. y ,r z The chord length vector is determined by the positions of the leading and trailing edge points to characterize the change in cross-sectional attitude. Based on the two-dimensional relative wind speed and chord length vector of the blade cross-section, the quasi-steady-state aerodynamic force is calculated, and the aerodynamic load and the structural dynamic equation are combined in a unified manner to form the nonlinear dynamic control equation of structure-aerodynamic coupling.

[0039] (1) Based on the coordinates of the leading and trailing edges on the airfoil section frame, calculate the chord length vector on the airfoil section: ; in, , , , The coordinates of the leading and trailing edges of the airfoil section in the section coordinate system.

[0040] (2) Determine the formulas for calculating the relative velocity (i.e., the two-dimensional relative wind speed of the blade section) and the angle of attack on the two-dimensional cross section of the airfoil: ; ; ; ; ; ; ; In the formula, For a blade element in three-dimensional space to be subjected to an ideal relative velocity, The incoming wind speed at infinity. The velocity of the leading edge of the airfoil element during blade operation. The induced velocity experienced by the blade unit. and These are the projected velocities in the axial and tangential directions, respectively. and These are the axial and tangential induction factors, respectively. and These are the axial and tangential projection matrices of the blade, respectively. This refers to the position of the leading edge of the airfoil of the blade micro-element during blade operation. This represents the relative velocity experienced by the blade element in actual three-dimensional space. It is the relative velocity projected onto a two-dimensional cross-section from the three-dimensional relative velocity, where G is the three-dimensional frame field composed of gradient vectors. , and These represent the slope vectors of the global position coordinates in the x, y, and z directions, respectively. E is a diagonal matrix used to construct the two-dimensional projection matrix of the relative velocity with respect to the frame G. The angle of attack of the blade airfoil section, It is the unit vector perpendicular to the normal vector of the two-dimensional cross-section frame.

[0041] Determine the angle of attack Then, the corresponding lift coefficient can be found based on the airfoil element. and drag coefficient ,set up and With an initial value of 0, we combine momentum theory and leaf element theory, and consider the Prandtl loss correction coefficient and the Glauert correction method to obtain the iterative equation for the induction factor. Then, we solve for the induction factor: ; ; In the formula, intermediate parameters , for antisymmetric matrix, It is a 3-order identity matrix. The chord length vector on the airfoil section B represents the number of blades, and r represents the distance from the blade unit to the center of the hub. This is the critical value of the inducing factor, which is [value missing] here. F is the tip-hub loss factor.

[0042] In this embodiment, the chord length vector and cross-sectional attitude are used to characterize the geometric nonlinear characteristics of the blade under large deformation conditions.

[0043] (3) Substitute the induction factor obtained from the iterative solution in the previous step into the relative velocity solution formula to obtain the relative velocity. Then, according to the aerodynamic calculation formula, the quasi-steady-state aerodynamic forces on the ANCF element can be obtained as follows: ; In the formula, This refers to air density.

[0044] Current aerodynamic load models can meet accuracy requirements under normal uniform incoming wind speeds, but they do not consider the induced terms of actual wind loads and related corrections. This embodiment takes into account the actual wind load effects, making the aerodynamic load modeling more accurate.

[0045] (4) The blade is discretized into several ANCF elements along the spanwise direction. The system mass matrix is ​​constructed based on the absolute position and gradient degrees of freedom of the element nodes. The nonlinear elastic deformation behavior of the blade is described by a curvature-like vector, thus forming a structural dynamic expression suitable for large deflection and large torsion conditions.

[0046] The specific form of the structural dynamics equation is: ; In the formula, It is the mass matrix composed of all ANCF elements. It is the generalized force matrix composed of all ANCF elements. These are constraint equations. yes about Jacobian matrix, It is a Lagrange multiplier. The velocity vector of all ANCF element nodes of the blade. Let t represent the acceleration vector of all ANCF element nodes of the blade, where t represents time. This represents the position vector of all ANCF element nodes of the blade.

[0047] (A) Construct the mass matrix based on the absolute position and gradient degrees of freedom of the unit nodes.

[0048] The specific formula for calculating the mass matrix of each ANCF element of the blade is as follows: , It is a 3rd order identity matrix. The mass matrix of the nth element is represented as: ; in, , , , Let x be the linear density function of the beam element along the arc length coordinate x of the centroid line, y be the moment of inertia of the beam element about the y-axis, z be the moment of inertia of the beam element about the z-axis, and z be the product of mass inertia of the beam element about the y and z axes, respectively. , , , , , Represents shape functions, , This indicates the position of the tail node of the nth unit. This indicates the position of the first node of the nth unit.

[0049] (B) The generalized force matrix of each ANCF element of the blade is divided into a generalized external force matrix and a generalized elastic force matrix: The specific formula for calculating the generalized external force matrix is ​​as follows: ; in, and These represent the distributed force and distributed moment of the beam element along the arc length coordinate x of the centroid line, respectively. Here, the distributed force is a combination of gravity and quasi-steady-state aerodynamic forces. The sum, intermediate parameters ; Representing shape functions The first derivative with respect to the arc length coordinate x; Representing shape functions The first derivative with respect to the arc length coordinate x.

[0050] The specific formula for calculating the generalized elastic force matrix is ​​as follows: ; Among them, the stiffness terms related to bending and torsion of the beam element under the initial condition. , Let c be a constant coefficient tensor derived from curvature-like vectors related to bending and torsion on beam elements, where c is the viscoelastic damping coefficient. , , For generalized coordinates, the subscript range is... The range is consistent. , For the first derivative of the generalized coordinates with respect to time, , The generalized position coordinates are used as a reference configuration.

[0051] (5) Introduce the aerodynamic load into the generalized external force in the absolute nodal coordinate system and combine it with the structural dynamic equation to form the nonlinear dynamic control equation of structure-aerodynamic coupling (i.e., the structure-aerodynamic coupling dynamic model), which can accurately characterize the dynamic response characteristics of the blade under different wind speeds, wind directions and blade pitch angles.

[0052] Step 103: Based on the structure-aerodynamic coupling dynamic model, systematically change the operating condition parameters, and perform multi-condition dynamic simulation calculations on the blade based on the Generalized-α method to obtain the multi-condition dynamic simulation results.

[0053] The operating parameters include, but are not limited to, wind speed, wind direction, pitch angle, speed, and heading angle.

[0054] By performing numerical integration calculations under the above multi-condition combination, the dynamic response of the blade under wind load is obtained.

[0055] Key dynamic response features of the blades are extracted from the simulation results and used as the basis for subsequent surrogate model training and online prediction.

[0056] Among them, the dynamic response characteristics include not only the geometric response characteristics that describe the overall deformation state of the blade, but also the equivalent response characteristics that can establish a physical consistency with the sensor monitoring values ​​during the actual operation of the wind turbine.

[0057] Among them, the geometric response quantities include, but are not limited to, tip flexure deformation, tip torsion angle, mid-blade flexure deformation, and mid-blade torsion angle.

[0058] Based on the ANCF dynamic model, the mechanical state quantities at the blade root and key sections are calculated simultaneously during the simulation process, and the equivalent response characteristic quantities consistent with the sensor output are further extracted.

[0059] The equivalent response characteristics include, but are not limited to, the equivalent bending moment response of the blade root section along the flapping and oscillation directions, the axial strain and bending strain response at the blade root and key sections, the acceleration response at the blade root or the mid-node of the blade, and the root mean square value (RMS) of the acceleration response.

[0060] The aforementioned equivalent response characteristics can directly correspond to the monitoring data collected by fiber Bragg grating (FBG) strain sensors, acceleration sensors, and speed sensors during actual wind turbine operation, thereby achieving a consistent description between numerical simulation results and engineering measurable quantities.

[0061] Step 104: Conduct a full-size static load test on the actual manufactured wind turbine blades.

[0062] Based on the blade design load conditions and test specifications, loading points are set at different positions along the blade span, and static loads of different magnitudes and directions are applied through loading devices to simulate the equivalent aerodynamic loads experienced by the blade under typical operating conditions.

[0063] Step 105: During the static load test, collect the structural response data of the blade under various loading conditions in real time.

[0064] The measured structural response data includes the flexural deformation at the blade tip and loading point, the torsional deformation angle at the key cross section of the blade, and the strain values ​​at the blade root and the pressure and negative pressure surfaces along the blade span direction, as well as the leading and trailing edges.

[0065] Step 106: During the static load test, record the static load test condition information corresponding to the structural response data, including the location of the loading point, the magnitude of the loading force, the loading direction, and the loading path.

[0066] Step 2: Construct a parameterized database of blade operating conditions and responses.

[0067] By introducing the blade static load test conditions and their measured response results, the dynamic simulation results under multiple operating conditions are supplemented, and a parameterized database of blade operating conditions and responses that can reflect the true characteristics is constructed.

[0068] Step 201: After the simulation results and measured data are fused, the blade dynamic response data from multiple sources are uniformly parameterized and organized.

[0069] Specifically, operating parameters such as wind speed, wind direction, pitch angle, rotational speed, and heading angle are used as operating condition dimensions, while geometric deformation response and strain, bending moment, and acceleration characteristics consistent with sensor outputs are used as response dimensions. The blade responses under different operating conditions are stored one-to-one. At the same time, the strain and deformation at various points on the blade measured under extreme loading during static load testing are used as the safety threshold for blade failure.

[0070] Step 202: Through the above steps, the multi-condition response data and the measured data from the static load test are stored in a unified manner to form a blade condition-response parameterized database covering multiple wind speeds, wind directions, pitch angles, and speed combinations. While reflecting the blade's operating characteristics, the database also incorporates deformation data from the measured ultimate load test as a supplement. Furthermore, a blade instability evaluation system is established based on this database, providing a reliable data foundation for the subsequent training of the condition-response relationship neural network surrogate model, feedforward prediction, and solution of the inverse control strategy.

[0071] Step 3: Based on the dynamic characteristic database (i.e., the blade operating condition-response parameterized database), construct a neural network surrogate model of the blade operating condition-response relationship.

[0072] Based on the blade structure-aerodynamic coupling dynamic characteristic database established in step 2, a neural network surrogate model of blade operating condition-response relationship is constructed using neural network methods to quickly characterize the nonlinear mapping relationship between blade operating conditions and dynamic response.

[0073] In this embodiment, based on the multi-condition dynamic characteristic database of blades, the blade operating condition data is used as input, and high-dimensional nonlinear features are extracted through a neural network to output key dynamic response data of the blades, clarifying the mapping relationship between blade operating conditions and dynamic response, thereby obtaining a neural network surrogate model based on the operating condition-response relationship.

[0074] Specifically, due to the combined effects of various operating conditions such as wind speed, wind direction, rotational speed, pitch angle, and heading angle, the blade exhibits significant high-dimensional, strongly nonlinear characteristics in its tip flexure deformation, torsional response, root bending moment-strain, and acceleration response. Furthermore, these nonlinear relationships dynamically evolve with changing operating conditions during blade operation. To meet the requirements for online monitoring and rapid assessment of blade operating status, a response prediction model capable of both ensuring prediction accuracy and rapid computation is needed. Considering that the nonlinear activation function introduced in the neural network unit can effectively characterize complex nonlinear mapping relationships, and that the neural network surrogate model has high computational efficiency during the inference phase and is suitable for real-time applications, this embodiment selects a neural network algorithm to establish a surrogate model of the blade operating condition-response relationship for subsequent real-time monitoring and stability assessment of the blade's dynamic response.

[0075] In this embodiment, the input layer variables and output layer variables of the blade operating condition-response relationship neural network surrogate model are as follows: Figure 2 As shown, the input layer variables include blade operating condition parameters, and the output layer variables include: flexural and torsional deformation at the blade tip and midsection, equivalent bending moment response of the blade root section along the flapping and oscillation directions, axial strain and bending strain response at the blade root and key sections, acceleration response at the blade root node, and root mean square value of the acceleration response.

[0076] Specifically, based on the established dynamic characteristic database of blade operating condition-response relationship, 200,000 sets of data were randomly generated as training samples for the neural network, and 20,000 sets of data were generated as validation samples. The training and validation sets consist of high-fidelity numerical simulation results and some measured data, in order to improve the generalization ability and prediction accuracy of the blade operating condition-response relationship neural network surrogate model.

[0077] The blade operating condition-response relationship neural network surrogate model consists of a linear weighting function and a nonlinear activation function. The linear weighting function is used to perform a linear transformation on the input variables, while the nonlinear activation function is used to enhance the network's ability to fit complex nonlinear relationships.

[0078] In this embodiment, for the regression prediction task, the ReLU function (rectified linear function) is selected as the activation function of the neural network.

[0079] Neural network units are interconnected according to a pre-defined structure to form a deep feedforward neural network. For example... Figure 2As shown, the forward operating condition-response neural network has two hidden layers: hidden layer 1 contains 32 computational nodes, and hidden layer 2 contains 16 computational nodes. The input layer receives operating condition parameters such as wind speed, wind direction, rotor speed, pitch angle, and heading angle. The output layer outputs the flexural deformation and torsional angle responses at the blade tip and midpoint, the equivalent bending moment responses of the blade root section along the flapping and flaring directions, the axial strain and bending strain responses at the blade root and key sections, the acceleration response at the blade root node, and the root mean square value of the acceleration response.

[0080] In this embodiment, the normalized mean square error is used as the evaluation index for the prediction accuracy of the neural network surrogate model of the blade operating condition-response relationship. Its calculation method is as follows: ; in, The actual response value is obtained from dynamic simulation or actual measurement. is the predicted value of the neural network surrogate model, and N is the number of samples in the training set.

[0081] To avoid overfitting of the neural network surrogate model for blade operating condition-response relationship, the learning rate is set to a preset constant, the optimizer is selected as the Adam (Adaptive Moment Estimation) optimization algorithm, and the number of training rounds is set to 1000. The convergence of the mean square error during the iteration process is used to determine whether the neural network training is complete.

[0082] To ensure the reliability and physical consistency of the established blade operating condition-dynamic response surrogate model in engineering applications, prototype measured data are introduced to verify the consistency of the surrogate model after its construction. Specifically, during the prototype or testing phase, the corresponding operating condition parameters are input into the surrogate model, and the response results predicted by the surrogate model are extracted. By comparing the consistency between the predicted values ​​of the surrogate model and the measured values ​​of the prototype in the time domain and statistically, evaluation indicators such as relative error and root mean square error are calculated. This verifies whether the operating condition-response mapping relationship established by the surrogate model can accurately reflect the dynamic response characteristics of the blade under real operating conditions.

[0083] Step 4: Use a neural network surrogate model to predict the blade dynamic response based on the real-time operating data sensed by the sensor.

[0084] Using the working condition-response relationship neural network surrogate model constructed, trained, and verified in step 3, the system performs online prediction of the operating condition data collected in real time by the sensor to obtain the dynamic response state of the blade under the current working condition.

[0085] Specifically, sensors deployed in the nacelle and blades collect real-time operating parameters such as wind speed, wind direction, rotational speed, pitch angle, and heading angle as current status information. These real-time operating parameters are then input into a trained operating condition-response relationship neural network surrogate model to quickly predict the blade's tip and mid-section flexural deformation and torsional response, as well as the equivalent bending moment response of the blade root section along the flapping and oscillation directions, the axial strain and bending strain response at the blade root and key sections, the acceleration response at the blade root node, and the root mean square value of the acceleration response. These predictions are then compared with preset safety thresholds to achieve feedforward assessment of the blade's operating status and early warning of instability risks.

[0086] Step 5: Based on the feedforward prediction results of the working condition-response relationship neural network surrogate model, determine the operation control adjustment strategy that meets the blade deformation threshold constraint.

[0087] Specifically, control commands are generated based on the blade instability evaluation system and the predicted response results from the surrogate model.

[0088] The aforementioned blade instability control evaluation system refers to the following: when the blade flexure deformation or torsion angle predicted by the working condition-response relationship neural network surrogate model in step 4 approaches the preset safety threshold at subsequent times or under adjacent working conditions, or when the equivalent bending moment response of the blade root section along the flapping and oscillation directions, the axial strain and bending strain response at the blade root and key sections, the acceleration response at the blade root node, and the root mean square value of the acceleration response predicted by the surrogate model show a sudden change compared with the current sensor monitoring values, the reverse control solution process based on the surrogate model is triggered; otherwise, the control adjustment command is not triggered.

[0089] Specifically, the aforementioned blade instability control evaluation system refers to the following: under the condition that power generation efficiency is not considered and safety is the key factor, the strain and deformation of the blade measured at various points under the ultimate loading in the static load test condition are taken as the safe deformation threshold for blade failure. The constraint is that the geometric response quantity (blade flexure deformation and torsion angle) predicted by the surrogate model does not exceed 80% of the safe deformation threshold, or the predicted equivalent response characteristic quantity (equivalent bending moment response of the blade root section along the flapping and oscillation directions, axial strain and bending strain response at the blade root and key sections, acceleration response at the blade root node, and root mean square value of acceleration response) does not show a sudden change compared with the current sensor monitoring value. Within the operating condition parameter space, a set of similar safe operating conditions that meet the constraints are determined by performing reverse search or optimization calculation based on the trained operating condition-response relationship neural network surrogate model.

[0090] Specifically, a mutation refers to a rapid increase or decrease in data within a short period of time, and the data collection or prediction step size is defined as... Based on the sensor time and sensor data collection Predicted data with proxy models The rate of change deviation characterizes abrupt changes: ; In the formula, The change over a time step is expressed in the denominator. To prevent extremely small quantities with denominators, This represents the rate of change threshold, i.e., the maximum relative deviation between the predicted equivalent response characteristic and the current sensor monitoring value within the step size does not exceed [a certain value]. This indicates that no mutation has occurred.

[0091] In the reverse engineering process, the reverse control module uses external operating parameters such as wind speed and direction as known conditions, and pitch angle, speed, and heading angle as adjustable control variables. By minimizing the changes in these control variables, it solves in reverse to obtain the corresponding changes in pitch angle, speed, and heading angle. Figure 3 As shown.

[0092] The changes in the aforementioned pitch angle, rotational speed, and heading angle are output as control commands to the wind turbine control system. The unit's operating status is adjusted in real time, and adjusted to the condition where the obtained deformation value is 50% of the safety threshold, or the equivalent bending moment response, axial strain and bending strain response, acceleration response, and root mean square value of acceleration response of the blade root section remain stable or gradually decrease by 20%. This ensures that the deflection and torsional deformation of the blade are always controlled within the safety threshold range during future operation, thereby achieving active suppression and adaptive control of blade instability risk.

[0093] This embodiment provides a wind turbine blade instability monitoring and control method. By constructing a high-precision parameterized database and a neural network proxy model through wind turbine blade dynamic modeling and simulation, it can monitor and control the blade instability risk, improve monitoring efficiency, and prevent instability risks.

[0094] During operation, the dynamic response characteristics of wind turbine blades are highly coupled with their structural characteristics and operating parameters (including wind speed, wind direction, pitch angle, and rotational speed). Blade instability typically manifests as significant changes in flexural deformation, torsional deformation, and root bending moment-strain or vibration characteristics. When the structural stiffness characteristics, aerodynamic load distribution, or operating state of the blade change under the same operating conditions, the corresponding dynamic response characteristics will show significant differences. Therefore, the wind turbine blade instability monitoring and control method provided in this embodiment, based on blade operating information and dynamic response characteristics, can effectively monitor the blade's stability state and instability risk.

[0095] This embodiment provides a wind turbine blade instability monitoring and control method. Given that existing wind turbine generators are generally equipped with wind measurement systems to obtain operating condition information such as wind speed and direction, and are equipped with sensors such as fiber Bragg gratings (FBG), acceleration, strain, or pitch angle to collect macroscopic dynamic response data of the blades, this method fully utilizes existing wind turbine operating data. A high-fidelity structure-aerodynamic coupling dynamic model based on the absolute nodal coordinate method is established to construct a parameterized database of blade "operating condition-response" under multiple operating conditions. Based on this, a neural network surrogate model is introduced to characterize the nonlinear mapping relationship between blade operating conditions and dynamic response, and this surrogate model is used to quickly predict and evaluate the blade stability state.

[0096] This embodiment provides a wind turbine blade instability monitoring and control method. It utilizes a neural network surrogate model to rapidly predict the dynamic response and stability state of the blade. After model training, the computational efficiency of the online monitoring and prediction process is significantly improved, meeting the engineering application requirements for real-time or near-real-time instability monitoring and control. Compared to traditional instability analysis methods based on high-fidelity numerical simulation, this embodiment effectively reduces computational complexity and latency, making feedforward prediction and proactive intervention possible before blade instability occurs, which is of great significance for the safe operation of wind turbines.

[0097] Existing blade instability monitoring methods often rely on complex, high-precision measurement devices or offline analysis techniques, which suffer from high implementation costs, insufficient real-time performance, and limited engineering applicability. This embodiment provides a wind turbine blade instability monitoring and control method that utilizes existing wind turbine operating condition measurement and response sensors to achieve indirect instability monitoring and control based on easily acquired signals. This reduces reliance on dedicated, high-cost measurement equipment, simplifies sensor placement, minimizes interference with the blade structure, and reduces maintenance costs, making it suitable for widespread engineering application.

[0098] This embodiment provides a wind turbine blade instability monitoring and control method. It establishes an instability control evaluation system, providing early warning information before significant blade instability risks become apparent, even when power generation efficiency is not a primary concern and safety is paramount. By outputting control commands such as pitch angle, heading angle, and speed, it achieves closed-loop adaptive regulation, effectively suppressing blade instability. Based on accurate and continuous stability status assessment results, it enables intelligent operation and maintenance and monitoring-based blade control oriented towards operational status, avoiding over-control or response lag, improving the safety and reliability of wind turbine operation, and helping to extend the service life of the blades and the entire system.

[0099] Example 2 This embodiment provides a wind turbine blade instability monitoring and control system, including: The modeling and simulation module is configured to: construct a one-dimensional beam dynamic model of the wind turbine blade; establish a local coordinate system at the centroid of the airfoil section of the wind turbine blade; determine the chord length vector by the positions of the leading and trailing edge points; calculate the aerodynamic loads by combining the two-dimensional relative wind speed of the wind turbine blade section; and combine the aerodynamic loads with the structural dynamic equations to form a structure-aerodynamic coupled dynamic model. Based on the structure-aerodynamic coupled dynamic model, the operating parameters are changed to perform multi-condition dynamic simulation calculations on the wind turbine blade; the simulation calculation results are corrected by introducing the blade static load test conditions and actual measured response results; and the dynamic response characteristic quantities are extracted from the simulation calculation results. The proxy model construction module is configured to: construct a blade condition-response parameterized database using operating condition parameters and dynamic response features obtained from simulation, train a neural network, and obtain a blade condition-response relationship neural network proxy model. The safety monitoring and control module is configured to: during blade instability monitoring operation, take the real-time operating condition data of the wind turbine blade as input, and predict the dynamic response characteristic quantity through the blade condition-response relationship neural network proxy model to determine the stability state of the blade and generate control commands.

[0100] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0101] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the wind turbine blade instability monitoring and control method described in Embodiment 1 above.

[0102] Example 4 This embodiment provides a computer device, such as... Figure 5 As shown, the system includes a computer-readable storage medium 1003, a processor 1001, a communication interface 1002, and a computer program stored on the computer-readable storage medium 1003 and executable on the processor 1001. The processor 1001, communication interface 1002, and computer-readable storage medium 1003 can be connected via a bus or other means. The communication interface 1002 is used to receive and send data. When the processor 1001 executes the program, it implements the steps of the wind turbine blade instability monitoring and control method described in Embodiment 1 above.

[0103] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring and controlling wind turbine blade instability, characterized in that, include: A one-dimensional beam dynamic model of a wind turbine blade is constructed. A local coordinate system is established at the centroid of the airfoil section of the wind turbine blade. The chord length vector is determined by the positions of the leading and trailing edge points. Combined with the two-dimensional relative wind speed of the wind turbine blade section, aerodynamic loads are calculated and integrated with the structural dynamic equations to form a structure-aerodynamic coupled dynamic model. Based on the structure-aerodynamic coupled dynamic model, the operating parameters are changed to perform multi-condition dynamic simulation calculations on the wind turbine blade. By introducing the static load test conditions and actual measured response results of the blade, the simulation calculation results are corrected, and dynamic response characteristic quantities are extracted from the simulation calculation results. Using operating condition parameters and dynamic response features obtained from simulation, a parameterized database of blade operating condition-response is constructed, and a neural network is trained to obtain a neural network surrogate model of blade operating condition-response relationship. During blade instability monitoring, real-time operating data of the wind turbine blades are used as input. The dynamic response characteristics are predicted through the blade condition-response relationship neural network surrogate model to determine the stability state of the blades and generate control commands.

2. The wind turbine blade instability monitoring and control method as described in claim 1, characterized in that, The step of determining the stability state of the blade and generating control commands includes: determining whether the predicted dynamic response characteristics satisfy the blade instability control evaluation system; if not, using the operating condition parameters as known conditions, and using the pitch angle, speed, and heading angle as adjustable control variables, and minimizing the change in the control variables to solve in reverse the solution to obtain the change in pitch angle, speed, and heading angle.

3. The wind turbine blade instability monitoring and control method as described in claim 2, characterized in that, The blade instability control evaluation system is as follows: the predicted geometric response does not exceed the safety threshold or the predicted equivalent response characteristic does not show a sudden change compared with the sensor monitoring value.

4. The wind turbine blade instability monitoring and control method as described in claim 1, characterized in that, The operating parameters include wind speed, wind direction, pitch angle, speed, and heading angle.

5. The wind turbine blade instability monitoring and control method as described in claim 1, characterized in that, The dynamic response characteristics include geometric response characteristics and equivalent response characteristics. The geometric response quantities include tip flexure deformation, tip torsion angle, mid-blade flexure deformation, and mid-blade torsion angle; The equivalent response characteristics include the equivalent bending moment response of the blade root section along the flapping and oscillation directions, the axial strain and bending strain responses at the blade root and key sections, the acceleration response at the blade root or the mid-node of the blade, and the root mean square value of the acceleration response.

6. The wind turbine blade instability monitoring and control method as described in claim 1, characterized in that, The aerodynamic load is: In the formula, intermediate parameters The unit vector perpendicular to the normal vector of the two-dimensional cross-section frame. , and Let represent the slope vectors of the global position coordinates in the y and z directions, respectively. and The upward resistance coefficient of the ANCF unit. for antisymmetric matrix, It is a 3-order identity matrix. air density, Let be the chord length vector on the airfoil section. It is the relative velocity after the three-dimensional relative velocity is projected onto the two-dimensional cross section.

7. The wind turbine blade instability monitoring and control method as described in claim 1, characterized in that, The relative velocity after the three-dimensional relative velocity is projected onto the two-dimensional cross section is: Among them, the three-dimensional frame field composed of gradient vectors , , and These represent the slope vectors of the global position coordinates in the x, y, and z directions, respectively, and are diagonal matrices. The relative velocity experienced by the blade element in actual three-dimensional space In three-dimensional space, the blade element is subjected to an ideal relative velocity. , For the incoming wind speed, The velocity of the leading edge of the blade element during blade operation is the induced velocity of the blade element. , and These are the axial and tangential induction factors, respectively. and These are the projected velocities in the axial and tangential directions, respectively. and These are the axial and tangential projection matrices of the blade, respectively. This represents the velocity of the leading edge of the airfoil of the blade micro-element during blade operation.

8. A wind turbine blade instability monitoring and control system, characterized in that, include: The modeling and simulation module is configured to: construct a one-dimensional beam dynamic model of the wind turbine blade; establish a local coordinate system at the centroid of the airfoil section of the wind turbine blade; determine the chord length vector by the positions of the leading and trailing edge points; calculate the aerodynamic loads by combining the two-dimensional relative wind speed of the wind turbine blade section; and combine the aerodynamic loads with the structural dynamic equations to form a structure-aerodynamic coupled dynamic model. Based on the structure-aerodynamic coupled dynamic model, the operating parameters are changed to perform multi-condition dynamic simulation calculations on the wind turbine blade; the simulation calculation results are corrected by introducing the blade static load test conditions and actual measured response results; and the dynamic response characteristic quantities are extracted from the simulation calculation results. The proxy model construction module is configured to: construct a blade condition-response parameterized database using operating condition parameters and dynamic response features obtained from simulation, train a neural network, and obtain a blade condition-response relationship neural network proxy model. The safety monitoring and control module is configured to: during blade instability monitoring operation, take the real-time operating condition data of the wind turbine blade as input, and predict the dynamic response characteristic quantity through the blade condition-response relationship neural network proxy model to determine the stability state of the blade and generate control commands.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the wind turbine blade instability monitoring and control method as described in any one of claims 1-7.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the wind turbine blade instability monitoring and control method as described in any one of claims 1-7.

Images