Tower wall vision measurement-based full-blade load monitoring method for wind turbine generator

Abstract

The invention discloses a wind turbine generator full-blade load monitoring method based on tower wall vision measurement, and the method comprises the steps: S1, arranging an observation platform, and installing a camera; s2, arranging a blade mark group; s3, synchronizing image acquisition and time; s4, camera calibration and distortion correction; s5, feature detection and tracking; s6, performing three-dimensional reconstruction; s7, centrifugal force and gravity compensation calculation; s8, blade net displacement calculation; and S9, carrying out aerodynamic force load inversion and result output. According to the invention, monitoring of the load in the edge direction of the whole blade of the wind turbine generator can be effectively realized, and reliable technical support is provided for safe operation, optimal design and economical efficiency improvement of the blade of the large-scale wind turbine generator.

Description

Technical Field

[0001] This invention relates to the technical field of wind power generation, and in particular to a method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall. Background Technology

[0002] As the core component for capturing wind energy, the load distribution and deformation state of the blades across their entire length directly determine the operational safety and power generation efficiency of the turbine. However, existing blade load monitoring technologies have significant limitations. The widely adopted root fiber optic load monitoring technology, which uses fiber optic sensors deployed at the blade root flange or hub connection, can only acquire the oscillation load at the blade root. The core drawback of this technology is that the monitoring range is limited to the root region where the blade connects to the main shaft (usually within 0-5m of the blade root), failing to reflect load changes in critical areas such as the blade middle and tip. The localized load peaks generated by airflow separation and turbulent impact in the blade middle to tip section are often the main causes of blade fatigue damage. Relying solely on root load data cannot comprehensively assess the overall stress state of the blade, easily leading to the omission of early damage in the blade middle and tip regions. Some wind turbines supplement load monitoring by attaching strain gauges to the blade cavity, but their coverage is limited by the blade cavity structure and installation process, typically only deployable within 30-40m of the blade length. On the one hand, the inner cavity of the blade gradually narrows from the blade root to the blade tip, and the narrow space at the blade tip cannot accommodate the installation of strain gauges and data transmission cables; on the other hand, the strain gauges need to be connected to the data acquisition module through wires, but the wires are prone to breakage due to centrifugal force during the blade rotation, causing the monitoring points in the blade tip area to fail frequently.

[0003] In summary, the coverage limitations of existing technologies result in a "monitoring blind spot" on the blades—that is, the area from 40m to the blade tip is completely unmonitored by load, lacking deformation and load data for this area. This makes it impossible to accurately establish an overall aeroelastic model of the blade, resulting in a lack of data support for the unit's pitch control strategy and blade design optimization. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for monitoring the load of the entire blade of a wind turbine based on visual measurement of the tower wall. This method enables the monitoring of the edge load of the entire blade of the wind turbine, providing reliable technical support for the safe operation, optimized design and economic improvement of large wind turbine blades.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall, comprising the following steps:

[0007] S1. Observation Platform Layout and Camera Installation

[0008] Several observation platforms are arranged at different heights on the wind turbine tower wall. Each observation platform is equipped with two cameras, and there is a fixed baseline length between the two cameras.

[0009] S2, arrangement of blade marker groups

[0010] Dot clusters are arranged on several cross sections along the blade length, each dot cluster consisting of at least 4 high-contrast markers;

[0011] S3, Image Acquisition and Time Synchronization

[0012] The multiple cameras in step S1 are triggered to simultaneously acquire images of the leaf marker points, thereby obtaining a multi-view, time-aligned leaf image sequence;

[0013] S4. Camera Calibration and Distortion Correction

[0014] After completing the camera's intrinsic and extrinsic parameter calibrations, distortion correction is performed on the blade images, and then the corrected image frames are obtained.

[0015] S5, Feature Detection and Tracking

[0016] The Harris corner detection algorithm and QR code recognition method are used to extract the initial pixel coordinates of the marker points in the corrected image frame. Then, the improved Lucas-Kanade optical flow method is used to process the initial pixel coordinates of the marker points by combining multi-resolution pyramid iteration, sub-pixel level accuracy optimization, topological constraints and Kalman filtering smoothing to obtain the sub-pixel level trajectory coordinates of each marker point in the time series.

[0017] S6, 3D Reconstruction

[0018] Based on the epipolar constraint principle of binocular cameras, the three-dimensional position sequence of the leaf marker group in the world coordinate system is obtained according to the fixed baseline length and the sub-pixel-level trajectory coordinates of each marker point in the time series.

[0019] S7. Centrifugal Force and Gravity Compensation Calculation

[0020] The axial tension and gravity load distributed along the arc length of the blade are calculated based on the three-dimensional position sequence of the blade marker group in the world coordinate system and the preset blade design parameters.

[0021] S8, Calculation of blade net displacement

[0022] The original three-dimensional displacement is obtained by the three-dimensional position sequence of the blade marker group in the world coordinate system, and then the compensated and corrected net blade displacement is calculated based on the original three-dimensional displacement, blade installation deviation and preset blade compensation displacement.

[0023] S9. Aerodynamic Load Inversion and Result Output

[0024] The total distributed load relationship is established based on the gravity load distributed along the blade arc length, the external load of the blade, and the aerodynamic load to be inverted. Based on the preset blade design parameters, the axial tension distributed along the blade arc length, the net displacement of the blade, and the total distributed load, a bending equation model with axial force is established. The aerodynamic load inversion calculation is performed based on the total distributed load relationship and the bending equation model to obtain the aerodynamic load distribution of the blade over the entire length range, thereby realizing the monitoring of the load in the edge direction of the entire blade of the wind turbine.

[0025] Furthermore, in step S1,

[0026] The specific number of observation platforms is determined based on the unit size and field of view obstruction. Two high-speed industrial cameras are installed on each observation platform to form a stereo vision measurement unit with a fixed baseline length B, where B is the straight-line distance between the optical centers of the two cameras. Based on the principle of stereo vision, the depth measurement of the target point is achieved by utilizing the imaging difference between the two cameras, given the fixed baseline length of the cameras. The installation position of the cameras is obtained by measuring the external parameters using a total station.

[0027] Furthermore, in step S2,

[0028] The markers are made of weather-resistant reflective material, and the geometric topology of each lattice cluster is kept unique. The spacing between adjacent lattice clusters is 5% to 15% of the blade length to ensure that the camera can accurately identify and match them.

[0029] Furthermore, in step S3,

[0030] Multi-camera synchronous acquisition is achieved through hardware trigger signals, with time synchronization accuracy better than 1ms; all image frames are appended with a unified timestamp and transmitted to the data processing unit in real time via gigabit Ethernet; a double buffering mechanism is adopted to avoid frame loss and ensure that all cameras acquire leaf marker images at the same time.

[0031] Furthermore, in step S4,

[0032] The focal length f, principal point coordinates (cx, cy), and distortion coefficients (k1, k2, p1, p2, k3) of the camera lens were obtained using the Zhang Zhengyou calibration method, and a temperature-focal length change lookup table was established to achieve self-correction during operation.

[0033] With the center of the tower base as the origin of the world coordinate system O w +Z w The axis is vertically upward, +X w The axis points directly in front of the cabin in its normal position, and the camera's rotation matrix R and translation vector T are calculated by combining the camera installation parameters measured by the total station and the observation platform installation parameters;

[0034] The Brown-Conrady distortion model is used to perform geometric correction on the leaf image, converting the original pixel coordinates of the marked points into ideal projected coordinates.

[0035] Furthermore, step S5 includes,

[0036] S5.1 Feature Point Selection

[0037] The Harris corner detection algorithm is used to extract the initial pixel coordinates (x, y) of the marker points in the corrected image frame, and combined with the QR code recognition method for auxiliary confirmation;

[0038] S5.2, Basic Model of Optical Flow

[0039] Based on the initial pixel coordinates (x, y) of the marker points, the improved Lucas-Kanade optical flow method is used to track the motion trajectory of the marker points in the time series.

[0040] Assuming that the grayscale intensity of two adjacent frames remains consistent, the basic constraint equation is:

[0041] I(x,y,t)=I(x+u,y+v,t+Δt)

[0042] Where I(x,y,t) is the gray value of pixel (x,y) at time t, and (u,v) is the displacement of pixel (x,y) within the time interval Δt.

[0043] S5.3, Multi-resolution Pyramid Iterative Improvement

[0044] An image pyramid model is established based on the basic constraint equations. Coarse optical flow (u0, v0) is calculated at the low-resolution layer and iteratively corrected layer by layer to the high-resolution layer until the original resolution layer is reached, thus obtaining the initial value of the optical flow vector after multi-layer iterative correction.

[0045] S5.4, sub-pixel level precision optimization

[0046] Within a local window region, a quadratic interpolation method is used to fit the optical flow calculation results to obtain high-precision sub-pixel level marker point displacement;

[0047] S5.5 Marker Cluster Topological Constraints

[0048] Within the same cluster of marker points, geometric topological constraints are applied to the optical flow solution results. If the solution result of a certain point deviates from the range of the topological constraints, it is automatically removed, resulting in a set of topologically consistent marker point trajectories.

[0049] S5.6 Dynamic Filtering and Smoothing Processing

[0050] The optical flow results of the time series were smoothed using a standard Kalman filter to obtain the smoothed optical flow trajectory.

[0051] Furthermore, in step S6,

[0052] Based on the epipolar constraint principle of binocular cameras, the 3D coordinates of the marker points are recovered through parallax calculation. The calculation formula is as follows:

[0053]

[0054] Where Z is the depth, which is the vertical distance from the target point to the camera's imaging plane, f is the focal length of the camera lens, B is the fixed baseline length, and d is the parallax, which is the difference in horizontal coordinates of the same object point in the left and right camera images.

[0055] The three-dimensional position (X) of the marker point in the camera coordinate system is obtained through back projection calculation. c ,Y c Z c Then, the coordinates are transformed to the base world coordinate system (X) using a coordinate transformation matrix. w ,Y w Z w The RANSAC algorithm is used to remove outliers from the reconstruction results.

[0056] Furthermore, step S7 includes,

[0057] S7.1, Coordinate and Parameter Preparation:

[0058] Based on the three-dimensional position sequence of the blade marker group in the world coordinate system, the blade arc length coordinates s∈[0,L] are established, where s=0 is the blade root and s=L is the blade tip. Then, combined with the blade arc length coordinates and preset blade design parameters, the blade mass distribution along the arc length μ(s), blade angular velocity Ω, radial distance r(s) from the blade section to the rotation axis, and the unit vector in the local edge direction of the blade are determined. Parameter set;

[0059] S7.2 Calculation of centrifugal force axial tension:

[0060] The centrifugal force generated by the blade rotation is manifested as an axially distributed force along the arc length. The axially distributed force f is calculated based on the blade's mass distribution along the arc length μ(s), the blade's angular velocity Ω, and the radial distance r(s) from the blade cross-section to the axis of rotation. c (s),

[0061] f c (s)=μ(s)Ω 2 r(s)

[0062] Next, the axial tension at section s of the blade is calculated based on the axial distributed force, thus obtaining the numerical distribution data of the axial tension N(s) along the arc length s of the blade.

[0063]

[0064] S7.3, Based on the mass distribution μ(s) of the blade along the arc length and the unit vector in the local edge direction of the blade. Calculate the gravity load q distributed along the arc length s of the blade. g The numerical distribution data of (s),

[0065]

[0066] Where g is the gravitational acceleration, and in the world coordinate system, the gravity vector g = (0, 0, -g).

[0067] Furthermore, in step S8,

[0068] The original three-dimensional displacement ω is obtained from the three-dimensional position sequence of the blade marker group in the world coordinate system. meas (s), and then based on the original three-dimensional displacement ω meas (s), blade installation deviation ω install_bias (s) and preset blade compensation displacement ω other_known (s) Calculate the compensated and corrected blade net displacement ω net (s),

[0069] ω net (s)=ω meas (s)-ω install_bias (s)-ω other_known (s).

[0070] Furthermore, in step S9,

[0071] Based on the gravity load q distributed along the arc length s of the blade g (s), blade external load q ext (s) and the aerodynamic load q of the blade to be inverted along the arc length s. aero (s) Establish the total distributed load relationship.

[0072] q total (s)=q aero (s)+q g (s)+q ext (s)

[0073] Where, q total (s) represents the total distributed load, q ext (s) represents the external loads on the blades, namely the set of lateral loads caused by the wind vane, sensors, and installation of the wind turbine.

[0074] In the presence of axial tension, the transverse bending of the blade satisfies the Euler-Bernoulli beam equation with axial force. Therefore, a bending equation model with axial force is established.

[0075]

[0076] Where EI(s) is the blade's flexural stiffness along the edge direction, N(s) is the axial tension distributed along the blade's arc length s, and ω net (s) represents the net displacement of the blade, q total (s) represents the total distributed load;

[0077] Based on the total distributed load relationship and the bending equation model, the aerodynamic load inversion formula is obtained by inverting the aerodynamic distribution.

[0078]

[0079] Where, ω” net (s) represents the net blade displacement ω net The differential result of (s);

[0080] The aerodynamic load distribution of the blade over its entire length is calculated using the aerodynamic distribution inversion formula, thereby enabling the monitoring of the load in the edge direction of the entire wind turbine blade.

[0081] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0082] 1. This invention upgrades blade deformation monitoring from a qualitative description to a quantifiable calculation system. By establishing standardized coordinate calculations, load estimation parameters become traceable. Simultaneously, a centrifugal force-added displacement formula is introduced to directly generate a dynamic database, thereby reducing solution adaptation costs by at least 30%. Furthermore, a new high-speed camera system covers the blade from 40m to the blade tip blind zone, filling the gap in load data.

[0083] 2. Based on this invention, a complete aeroelastic model can be established to support full-domain pitch control, thereby greatly improving power generation efficiency; a rapid load database can be established to promote lightweight blade design and shorten the iteration cycle.

[0084] 3. This invention adopts reproducible engineering standards, upgrading blade monitoring from "local inference" to "full-domain measurement", providing reliable technical support for the safe operation, optimized design and economic improvement of large wind turbine blades, and laying the foundation for industrialization. Attached Figure Description

[0085] Figure 1 This is a flowchart of the method of the present invention.

[0086] Figure 2 This is a schematic diagram of the present invention, which arranges lattice clusters on several cross sections along the length of the blade. Detailed Implementation

[0087] 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. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0088] like Figure 1 As shown, this embodiment provides a method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall, including the following steps:

[0089] S1. Observation Platform Layout and Camera Installation

[0090] Several observation platforms are arranged at different heights on the wind turbine tower wall. Each observation platform is equipped with two cameras, with a fixed baseline length between the two cameras. The specific number of observation platforms is determined based on the turbine size and field of view obstruction. Two high-speed industrial cameras are installed on each observation platform, forming a stereo vision measurement unit with a fixed baseline length B, where B is the straight-line distance between the optical centers of the two cameras. Based on the principle of stereo vision, under the condition of knowing the fixed baseline length of the cameras, the depth measurement of the target point is achieved by utilizing the imaging difference between the two cameras. The installation position of the cameras is obtained by measuring the external parameters using a total station, and high frame rate cameras are used to ensure clear capture under high-speed blade movement.

[0091] S2, arrangement of blade marker groups

[0092] like Figure 2 As shown, dot matrix clusters are arranged on several cross-sections along the blade's length. Each cluster consists of at least four high-contrast markers. The markers are made of weather-resistant reflective material, and the geometric topology of each cluster is unique. The spacing between adjacent clusters is 5%–15% of the blade length to ensure accurate camera identification and matching. Active luminescent markers or reflective rings are added at some locations to ensure visibility at night and in low-light conditions.

[0093] S3, Image Acquisition and Time Synchronization

[0094] In step S1, multiple cameras are triggered to simultaneously acquire images of the leaf marker points, thereby obtaining a multi-view, time-aligned leaf image sequence. The simultaneous acquisition of multiple cameras is achieved through hardware trigger signals, with a time synchronization accuracy better than 1ms. All image frames are appended with a unified timestamp and transmitted to the data processing unit in real time via gigabit Ethernet. A double buffering mechanism is used to avoid frame loss and ensure that all cameras acquire leaf marker images at the same time.

[0095] S4. Camera Calibration and Distortion Correction

[0096] After completing the camera's intrinsic and extrinsic parameter calibrations, the blade images are then subjected to distortion correction to obtain the corrected image frames. The Zhang Zhengyou calibration method is used to determine the camera lens's focal length f, principal point coordinates (cx, cy), and distortion coefficients (k1, k2, p1, p2, k3), and a temperature-focal length change lookup table is established to achieve self-correction during operation.

[0097] With the center of the tower base as the origin of the world coordinate system O w +Z w The axis is vertically upward, +X w The axis points directly in front of the cabin in its normal position, and the camera's rotation matrix R and translation vector T are calculated by combining the camera installation parameters measured by the total station and the observation platform installation parameters;

[0098] The Brown-Conrady distortion model was used to perform geometric correction on the acquired leaf images, converting the original pixel coordinates of the marked points into ideal projected coordinates.

[0099] S5, Feature Detection and Tracking

[0100] The Harris corner detection algorithm and QR code recognition method are used to extract the initial pixel coordinates of the marker points in the corrected image frame. Then, the initial pixel coordinates of the marker points are processed by an improved Lucas-Kanade optical flow method combined with multi-resolution pyramid iteration, sub-pixel level accuracy optimization, topological constraints, and Kalman filtering smoothing to obtain the sub-pixel level trajectory coordinates of each marker point in the time series. The specific steps are as follows:

[0101] S5.1 Feature Point Selection

[0102] The Harris corner detection algorithm is used to extract the initial pixel coordinates (x, y) of the marker points in the corrected image frame, and combined with the QR code recognition method for auxiliary confirmation;

[0103] S5.2, Basic Model of Optical Flow

[0104] Based on the initial pixel coordinates (x, y) of the marker points, the improved Lucas-Kanade optical flow method is used to track the motion trajectory of the marker points in the time series.

[0105] Assuming that the grayscale intensity of two adjacent frames remains consistent, the basic constraint equation is:

[0106] I(x,y,t)=I(x+u,y+v,t+Δt)

[0107] Where I(x,y,t) is the gray value of pixel (x,y) at time t, and (u,v) is the displacement of pixel (x,y) within the time interval Δt.

[0108] S5.3, Multi-resolution Pyramid Iterative Improvement

[0109] An image pyramid model is established based on the basic constraint equations. Coarse optical flow (u0, v0) is calculated at the low-resolution layer and iteratively corrected layer by layer to the high-resolution layer until the original resolution layer is reached, thus obtaining the initial value of the optical flow vector after multi-layer iterative correction.

[0110] S5.4, sub-pixel level precision optimization

[0111] Within a local window region, a quadratic interpolation method is used to fit the optical flow calculation results to obtain high-precision sub-pixel level marker point displacement;

[0112] S 5.5 Marker Cluster Topological Constraints

[0113] Within the same cluster of marker points, geometric topological constraints are applied to the optical flow solution results. If the solution result of a certain point deviates from the range of the topological constraints, it is automatically removed, resulting in a set of topologically consistent marker point trajectories.

[0114] S 5.6 Dynamic Filtering and Smoothing Processing

[0115] The optical flow results of the time series were smoothed using a standard Kalman filter to obtain the smoothed optical flow trajectory.

[0116] S6, 3D Reconstruction

[0117] Based on the epipolar constraint principle of a stereo camera, the 3D position sequence of the leaf marker group in the world coordinate system is obtained according to the fixed baseline length and the sub-pixel-level trajectory coordinates of each marker point in the time series. Specifically, based on the epipolar constraint principle of the stereo camera, 3D coordinate recovery is achieved through parallax calculation. The depth calculation formula is as follows:

[0118]

[0119] Where Z is the depth, which is the vertical distance from the target point to the camera's imaging plane, f is the focal length of the camera lens, B is the fixed baseline length, and d is the parallax, which is the difference in horizontal coordinates of the same object point in the left and right camera images.

[0120] The three-dimensional position (X) of the marker point in the camera coordinate system is obtained through back projection calculation. c ,Y c Z c Then, the coordinates are transformed to the base world coordinate system (X) using a coordinate transformation matrix. w ,Y w Z w The RANSAC algorithm is used to remove outliers from the reconstruction results.

[0121] S7. Centrifugal Force and Gravity Compensation Calculation

[0122] Calculate the axial tension and gravity load distributed along the arc length of the blade based on the three-dimensional position sequence of the blade marker group in the world coordinate system and the preset blade design parameters; including the steps,

[0123] S7.1, Coordinate and Parameter Preparation:

[0124] Based on the three-dimensional position sequence of the blade marker group in the world coordinate system, the blade arc length coordinates s∈[0,L] are established, where s=0 is the blade root and s=L is the blade tip. Then, combined with the blade arc length coordinates and preset blade design parameters, the blade mass distribution along the arc length μ(s), blade angular velocity Ω, radial distance r(s) from the blade section to the rotation axis, and the unit vector in the local edge direction of the blade are determined. Parameter set;

[0125] S7.2 Calculation of centrifugal force axial tension:

[0126] The centrifugal force generated by the blade rotation is manifested as an axially distributed force along the arc length. The axially distributed force f is calculated based on the blade's mass distribution along the arc length μ(s), the blade's angular velocity Ω, and the radial distance r(s) from the blade cross-section to the axis of rotation. c (s),

[0127] f c (s)=μ(s)Ω 2 r(s)

[0128] Next, the axial tension at section s of the blade is calculated based on the axial distributed force, thus obtaining the numerical distribution data of the axial tension N(s) along the arc length s of the blade.

[0129]

[0130] S7.3, Based on the mass distribution μ(s) of the blade along the arc length and the unit vector in the local edge direction of the blade. Calculate the gravity load q distributed along the arc length s of the blade. g The numerical distribution data of (s),

[0131]

[0132] Where g is the gravitational acceleration, and in the world coordinate system, the gravity vector g = (0, 0, -g).

[0133] S8, Calculation of blade net displacement

[0134] The original three-dimensional displacement is obtained based on the three-dimensional position sequence of the blade marker group in the world coordinate system. Then, the compensated net displacement of the blade is calculated based on the original three-dimensional displacement, blade installation deviation, and preset blade compensation displacement. Among them, the original three-dimensional displacement ω is obtained based on the three-dimensional position sequence of the blade marker group in the world coordinate system. meas (s), and then based on the original three-dimensional displacement ω meas (s), blade installation deviation ω install_bias (s) and preset blade compensation displacement ω other_known (s) Calculate the compensated and corrected net blade displacement ω caused solely by aerodynamic loads. net (s),

[0135] ω net (s)=ω meas (s)-ω install_bias (s)-ω other_known (s).

[0136] S9. Aerodynamic Load Inversion and Result Output

[0137] Based on the gravity load distributed along the blade's arc length, the external load on the blade, and the aerodynamic load to be inverted, a total distributed load relationship is established. A bending equation model containing axial force is established based on preset blade design parameters, the axial tension distributed along the blade's arc length, the blade's net displacement, and the total distributed load. Aerodynamic load inversion calculations are performed based on the total distributed load relationship and the bending equation model to obtain the aerodynamic load distribution along the entire blade length, thereby enabling monitoring of the load in the edge direction of the entire wind turbine blade. Specifically, the gravity load q distributed along the blade's arc length s is... g (s), blade external load q ext (s) and the aerodynamic load q of the blade to be inverted along the arc length s. aero (s) Establish the total distributed load relationship.

[0138] q total (s)=q aero (s)+q g (s)+q ext (s)

[0139] Where, q total (s) represents the total distributed load, q ext (s) represents the external loads on the blades, namely the set of lateral loads caused by the wind vane, sensors, and installation of the wind turbine.

[0140] In the presence of axial tension, the transverse bending of the blade satisfies the Euler-Bernoulli beam equation with axial force. Therefore, a bending equation model with axial force is established.

[0141]

[0142] Where EI(s) is the blade's flexural stiffness along the edge direction, N(s) is the axial tension distributed along the blade's arc length s, and ω net (s) represents the net displacement of the blade, q total (s) represents the total distributed load;

[0143] Based on the total distributed load relationship and the bending equation model, the aerodynamic load inversion formula is obtained by inverting the aerodynamic distribution.

[0144]

[0145] Where, ω” net (s) represents the net blade displacement ω net The differential result of (s);

[0146] The aerodynamic load distribution of the blade over its entire length is calculated using the aerodynamic distribution inversion formula, thereby enabling the monitoring of the load in the edge direction of the entire wind turbine blade.

[0147] In step S9, the net displacement ω of the blade is... net (s) Spline fitting or regularization smoothing is required before differentiation to avoid noise amplification; Boundary conditions should satisfy: root clamping ω(0)=0, ω'(0)=0, tip free ω””L)=0, ω”’(L)=0; Axial tension N(s) distributed along the blade along the arc length s is calculated by numerical integration, and attention should be paid to the consistency of parameter units; The aerodynamic distribution obtained by inversion needs to be back-substituted into the beam equation for verification to check the consistency with the measured displacement.

[0148] This invention provides a verified and reliable aerodynamic distribution dataset, and through back-substitution verification, ensures the numerical stability and computational accuracy of the inversion process, providing accurate data support for blade condition assessment, control strategy optimization, and design improvement.

[0149] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall, characterized in that, Including steps, S1. Observation Platform Layout and Camera Installation Several observation platforms are arranged at different heights on the wind turbine tower wall. Each observation platform is equipped with two cameras, and there is a fixed baseline length between the two cameras. S2, Blade Marker Group Arrangement Dot clusters are arranged on several cross sections along the blade length, each dot cluster consisting of at least 4 high-contrast markers; S3, Image Acquisition and Time Synchronization The multiple cameras in step S1 are triggered to simultaneously acquire images of the leaf marker points, thereby obtaining a multi-view, time-aligned leaf image sequence; S4. Camera Calibration and Distortion Correction After completing the camera's intrinsic and extrinsic parameter calibrations, distortion correction is performed on the blade images, and then the corrected image frames are obtained. S5, Feature Detection and Tracking The Harris corner detection algorithm and QR code recognition method are used to extract the initial pixel coordinates of the marker points in the corrected image frame. Then, the improved Lucas-Kanade optical flow method is used to process the initial pixel coordinates of the marker points by combining multi-resolution pyramid iteration, sub-pixel level accuracy optimization, topological constraints and Kalman filtering smoothing to obtain the sub-pixel level trajectory coordinates of each marker point in the time series. S6, 3D Reconstruction Based on the epipolar constraint principle of binocular cameras, the three-dimensional position sequence of the leaf marker group in the world coordinate system is obtained according to the fixed baseline length and the sub-pixel-level trajectory coordinates of each marker point in the time series. S7. Centrifugal Force and Gravity Compensation Calculation The axial tension and gravity load distributed along the arc length of the blade are calculated based on the three-dimensional position sequence of the blade marker group in the world coordinate system and the preset blade design parameters. S8, Calculation of blade net displacement The original three-dimensional displacement is obtained by the three-dimensional position sequence of the blade marker group in the world coordinate system, and then the compensated and corrected net blade displacement is calculated based on the original three-dimensional displacement, blade installation deviation and preset blade compensation displacement. S9. Aerodynamic Load Inversion and Result Output The total distributed load relationship is established based on the gravity load distributed along the blade arc length, the external load of the blade, and the aerodynamic load to be inverted. Based on the preset blade design parameters, the axial tension distributed along the blade arc length, the net displacement of the blade, and the total distributed load, a bending equation model with axial force is established. The aerodynamic load inversion calculation is performed based on the total distributed load relationship and the bending equation model to obtain the aerodynamic load distribution of the blade over the entire length range, thereby realizing the monitoring of the load in the edge direction of the entire blade of the wind turbine.

2. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall as described in claim 1, characterized in that, In step S1, The specific number of observation platforms is determined based on the unit size and field of view obstruction. Two high-speed industrial cameras are installed on each observation platform to form a stereo vision measurement unit with a fixed baseline length B, where B is the straight-line distance between the optical centers of the two cameras. Based on the principle of stereo vision, the depth measurement of the target point is achieved by utilizing the imaging difference between the two cameras, given the fixed baseline length of the cameras. The installation position of the cameras is obtained by measuring the external parameters using a total station.

3. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall as described in claim 1, characterized in that, In step S2, The markers are made of weather-resistant reflective material, and the geometric topology of each lattice cluster is kept unique. The spacing between adjacent lattice clusters is 5% to 15% of the blade length to ensure that the camera can accurately identify and match them.

4. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall as described in claim 1, characterized in that, In step S3, Multi-camera synchronous acquisition is achieved through hardware trigger signals, with time synchronization accuracy better than 1ms; all image frames are appended with a unified timestamp and transmitted to the data processing unit in real time via gigabit Ethernet. A double buffering mechanism is used to avoid frame drops and ensure that all cameras acquire leaf marker images at the same time.

5. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall as described in claim 1, characterized in that, In step S4, The focal length f, principal point coordinates (cx, cy), and distortion coefficients (k1, k2, p1, p2, k3) of the camera lens were obtained using the Zhang Zhengyou calibration method, and a temperature-focal length change lookup table was established to achieve self-correction during operation. With the center of the tower base as the origin of the world coordinate system O w +Z w The axis is vertically upward, +X w The axis points directly in front of the cabin in its normal position, and the camera's rotation matrix R and translation vector T are calculated by combining the camera installation parameters measured by the total station and the observation platform installation parameters; The Brown-Conrady distortion model is used to perform geometric correction on the leaf image, converting the original pixel coordinates of the marked points into ideal projected coordinates.

6. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall according to claim 1, characterized in that, Step S5 includes, S5.1 Feature Point Selection The Harris corner detection algorithm is used to extract the initial pixel coordinates (x, y) of the marker points in the corrected image frame, and combined with the QR code recognition method for auxiliary confirmation; S5.2, Basic Model of Optical Flow Based on the initial pixel coordinates (x, y) of the marker points, the improved Lucas-Kanade optical flow method is used to track the motion trajectory of the marker points in the time series. Assuming that the grayscale intensity of two adjacent frames remains consistent, the basic constraint equation is: I(x,y,t)=I(x+u,y+v,t+Δt) Where I(x,y,t) is the gray value of pixel (x,y) at time t, and (u,v) is the displacement of pixel (x,y) within the time interval Δt. S5.3, Multi-resolution Pyramid Iterative Improvement An image pyramid model is established based on the basic constraint equations. Coarse optical flow (u0, v0) is calculated at the low-resolution layer and iteratively corrected layer by layer to the high-resolution layer until the original resolution layer is reached, thus obtaining the initial value of the optical flow vector after multi-layer iterative correction. S5.4, sub-pixel level precision optimization Within a local window region, a quadratic interpolation method is used to fit the optical flow calculation results to obtain high-precision sub-pixel level marker point displacement; S5.5 Marker Cluster Topological Constraints Within the same cluster of marker points, geometric topological constraints are applied to the optical flow solution results. If the solution result of a certain point deviates from the range of the topological constraints, it is automatically removed, resulting in a set of topologically consistent marker point trajectories. S5.6 Dynamic Filtering and Smoothing Processing The optical flow results of the time series were smoothed using a standard Kalman filter to obtain the smoothed optical flow trajectory.

7. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall according to claim 1, characterized in that, In step S6, Based on the epipolar constraint principle of binocular cameras, the 3D coordinates of the marker points are recovered through parallax calculation. The calculation formula is as follows: Where Z is the depth, which is the vertical distance from the target point to the camera's imaging plane, f is the focal length of the camera lens, B is the fixed baseline length, and d is the parallax, which is the difference in horizontal coordinates of the same object point in the left and right camera images. The three-dimensional position (X) of the marker point in the camera coordinate system is obtained through back projection calculation. c ,Y c Z c Then, the coordinates are transformed to the base world coordinate system (X) using a coordinate transformation matrix. w ,Y w Z w The RANSAC algorithm is used to remove outliers from the reconstruction results.

8. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall according to claim 1, characterized in that, Step S7 includes... S7.1, Coordinate and Parameter Preparation: Based on the three-dimensional position sequence of the blade marker group in the world coordinate system, the blade arc length coordinates s∈[0,L] are established, where s=0 is the blade root and s=L is the blade tip. Then, combined with the blade arc length coordinates and preset blade design parameters, the blade mass distribution along the arc length μ(s), blade angular velocity Ω, radial distance r(s) from the blade section to the rotation axis, and the unit vector in the local edge direction of the blade are determined. Parameter set; S7.2 Calculation of centrifugal force axial tension: The centrifugal force generated by the blade rotation is manifested as an axially distributed force along the arc length. The axially distributed force f is calculated based on the blade's mass distribution along the arc length μ(s), the blade's angular velocity Ω, and the radial distance r(s) from the blade cross-section to the axis of rotation. c (s), f c (s)=μ(s)Ω 2 r(s) Next, the axial tension at section s of the blade is calculated based on the axial distributed force, thus obtaining the numerical distribution data of the axial tension N(s) along the arc length s of the blade. S7.3, Based on the mass distribution μ(s) of the blade along the arc length and the unit vector in the local edge direction of the blade. Calculate the gravity load q distributed along the arc length s of the blade. g The numerical distribution data of (s), Where g is the gravitational acceleration, and in the world coordinate system, the gravity vector g = (0, 0, -g).

9. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall as described in claim 1, characterized in that, In step S8, The original three-dimensional displacement ω is obtained from the three-dimensional position sequence of the blade marker group in the world coordinate system. meas (s), and then based on the original three-dimensional displacement ω meas (s), blade installation deviation ω install_bias (s) and preset blade compensation displacement ω other_known (s) Calculate the compensated and corrected blade net displacement ω net (s), oh net (s)=ω meas (s)-ω install_bias (s)-ω other_known (s)。 10. The method for monitoring the load on the entire blade of a wind turbine based on visual measurement of the tower wall according to claim 1, characterized in that, In step S9, Based on the gravity load q distributed along the arc length s of the blade g (s), blade external load q ext (s) and the aerodynamic load q of the blade distributed along the arc length s to be inverted. aero (s) Establish the total distributed load relationship. q total (s)=q aero (s)+q g (s)+q ext (s) Where, q total (s) represents the total distributed load, q ext (s) represents the external loads on the blades, namely the set of lateral loads caused by the wind vane, sensors, and installation of the wind turbine. In the presence of axial tension, the transverse bending of the blade satisfies the Euler-Bernoulli beam equation with axial force. Therefore, a bending equation model with axial force is established. Where EI(s) is the blade's flexural stiffness in the edge direction, N(s) is the axial tension distributed along the blade's arc length s, and ω net (s) represents the net displacement of the blade, q total (s) represents the total distributed load; Based on the total distributed load relationship and the bending equation model, the aerodynamic load inversion formula is obtained by inverting the aerodynamic distribution. Where, ω” net (s) represents the net blade displacement ω net The differential result of (s); The aerodynamic load distribution of the blade over its entire length is calculated using the aerodynamic distribution inversion formula, thereby enabling the monitoring of the load in the edge direction of the entire wind turbine blade.

Images