Digital twinning-based wind turbine generator full life cycle management system

Abstract

The invention relates to the technical field of digital twinning, in particular to a digital twinning-based wind turbine generator full life cycle management system, which comprises a deformation parameter resolving module, a skeleton driving reconstruction module, a load damage calculation module, a health difference evaluation module and a balance control decision module. According to the method, the instantaneous curvature and the torsion rate are calculated by collecting cross section wavelength drift data, the dynamic bending posture of the flexible part in the space is reconstructed in real time, continuous tracking of stress circulation and a closed lag loop is achieved by combining pneumatic simulation and the residual peak-valley sequence stack technology, the accumulated damage degree of the part in the whole service cycle is accurately quantified, and the reliability of the part is improved. By analyzing the discrete deviation of individual damage relative to a group mean value, an asymmetric load distribution mechanism based on health state difference is established, the fatigue extension trend of high-damage parts is effectively relieved, premature failure caused by uneven load distribution is avoided, and life balance and collaborative life extension of key parts of a unit in the whole life cycle are achieved.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a wind turbine full lifecycle management system based on digital twins. Background Technology

[0002] The field of digital twin technology involves the mapping and construction of physical entities in virtual space, dynamic data-driven state evolution representation, continuous simulation analysis of cross-cycle operation behavior, and model-based operation management support. Its core aspects include the digital representation of physical unit structural parameters, the acquisition and mapping of real-time operational data, the bidirectional correlation between the virtual model and the actual unit, and monitoring, diagnosis, and management capabilities throughout the entire lifecycle. This technology generally supports unified management of equipment during the design, manufacturing, operation, and maintenance stages by establishing a dynamically updated virtual twin. In contrast, the traditional wind turbine lifecycle management system refers to a system covering the entire lifecycle of a wind turbine, from installation and commissioning to decommissioning. The traditional operation and management system typically relies on information such as the unit's operation logs, periodic inspection reports, manual measurement records, and status quantities output by the unit controller. It infers the fatigue state of unit components through manual methods or fixed thresholds, arranges unit maintenance plans through preset maintenance cycles, analyzes the unit's operational health status based on single-point monitoring values ​​such as wind speed, engine speed, and yaw angle, judges the structural state changes of key unit components such as gearboxes, main shafts, hubs, and blades, and identifies unit operational anomalies by comparing historical operating curves. Traditional methods usually rely on manual processing and static rules when processing multi-source data, lacking the ability to dynamically express complex changes in unit operating conditions.

[0003] Traditional unit management systems rely on manual inspection records and static threshold judgments, ignoring the nonlinear cumulative characteristics of material fatigue under dynamic wind loads. Relying on single-point monitoring data such as rotational speed or wind speed cannot accurately reconstruct the real-time spatial bending and twisting attitude of large flexible components in complex flow fields. The use of fixed-cycle maintenance plans leads to a disconnect between the actual component health and the timing of maintenance. Relying solely on single-dimensional operational data makes it difficult to capture the hidden microstructural damage evolution process. Static rule-based analysis models lack fine-grained differentiation of individual damage accumulation differences among multi-blade components. Generalized control strategies do not consider the differences in the service history of individual components, causing damaged components to continuously bear excessive loads and accelerate aging, resulting in a short-board effect in the overall life of the unit and reducing the operational reliability throughout its entire life cycle. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a wind turbine full lifecycle management system based on digital twins.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a wind turbine full lifecycle management system based on digital twins includes: The deformation parameter calculation module collects the center wavelength drift data of the underside and belly regions of the blade cross section, calculates the difference results of the center wavelength drift data of the same cross section, normalizes them by combining the vertical distance, calculates the instantaneous curvature vector and torsional scalar of the corresponding cross section, and establishes the geometric morphology parameters. The skeleton-driven reconstruction module extracts the instantaneous curvature vector based on the geometric morphology parameters, constructs the neutral axis skeleton curve, calculates the rigid body rotation and translation matrix by combining the skeleton node indexes bound to the virtual mesh vertices, performs linear transformation on the original local coordinates of the mesh vertices, and outputs the real-time position coordinates. The load damage calculation module combines the real-time location coordinates and wind speed simulation to generate a stress change sequence, injects residual peak and valley sequence stack data, identifies closed stress hysteresis loops to accumulate damage variables, overwrites the residual peak and valley sequence stack with unclosed data, and outputs the total accumulated damage variable. The health difference assessment module calculates the arithmetic mean of the total cumulative damage variables, analyzes the deviation between the damage degree of each leaf and the average result, and performs inverse linear mapping calculation to establish the load setpoint bias of the corrected control target. The balanced control decision module superimposes the load setpoint offset onto the rated root bending moment parameter to establish a dynamic load target, determines the expected load level to adapt to the current health status difference, calculates the difference result with the virtual load, constructs pitch command to adjust the stress state, and outputs the unit operating status parameters.

[0006] As a further aspect of the present invention, the geometric parameters include the cross-sectional bending curvature component and the cross-sectional torsional angle component; the real-time position coordinates are specifically a spatial coordinate triplet of discrete points on the blade surface; the total cumulative damage variable includes the count statistics of closed stress cycles and the percentage of material fatigue life consumed; the load setpoint offset includes the weighting coefficient of the load distribution of each blade and the dead zone limit parameter of the pitch controller; and the unit operating status parameters include the action rate command value of the pitch motor of each blade and the dynamic adjustment set value of the unit's active power.

[0007] As a further aspect of the present invention, the deformation parameter calculation module includes: The drift data acquisition submodule uses a fiber optic grating sensor to collect center wavelength drift data for each monitoring point by performing spectral scanning and signal demodulation on the underside and belly of the leaf. The data differential processing submodule performs differential calculations on the center wavelength drift data at the same cross-sectional position, and normalizes the differential results by combining the vertical distance between the monitoring point on the back of the leaf and the monitoring point on the belly of the leaf to generate a normalized strain difference value for the cross section. The parameter calculation and analysis submodule acquires real-time ambient temperature data based on the normalized strain difference value of the cross section, calculates the cross section bending characteristic index, performs vector synthesis by combining the direction vector of the cross section coordinate system, solves the instantaneous curvature vector and torsional scalar of the corresponding cross section, evaluates the local bending degree and torsional state of the blade, and generates geometric morphology parameters.

[0008] As a further aspect of the present invention, the skeleton-driven reconstruction module includes: The skeleton curve construction submodule analyzes the instantaneous curvature vector distributed along the blade span based on geometric morphology parameters, performs spatial cumulative integration along the blade span, derives the translation and rotation of the center of each discrete section relative to the blade root, analyzes the overall bending attitude of the blade, connects the center points of each section, constructs a spatial geometric path, and generates the neutral axis skeleton curve. The transformation matrix calculation submodule, for the neutral axis skeleton curve, combines the skeleton node index bound to the virtual mesh vertex to extract the initial reference coordinates of the skeleton node in the undeformed state, calculates the rotation angle and displacement distance of each skeleton node in the current deformed state relative to the initial state, and constructs a rigid body rotation and translation matrix containing a rotation submatrix and a translation vector. The mesh reconstruction mapping submodule, based on the rigid body rotation and translation matrix, obtains the original local coordinate data of the virtual mesh vertices on the blade surface, performs a linear transformation on the original local coordinates of the mesh vertices, maps and transforms the mesh vertices from the local reference system to the global observation coordinate system, calculates the three-dimensional spatial position of each vertex at the current moment, and generates real-time position coordinates.

[0009] As a further aspect of the present invention, the load damage calculation module includes: The stress simulation analysis submodule extracts the real-time position coordinates, calls the collected environmental wind speed data, analyzes the stress state at the blade root through aeroelastic simulation, and generates time series data of root stress change. The loop counting and identification submodule reads the residual peak and valley sequence stack and injects it into the starting position of the stress change time series for the root stress change time series. Through peak and valley extraction and loop pairing operation, it identifies closed stress hysteresis loops and separates unclosed residual peak and valley data to generate hysteresis loops and residual feature sets. The damage accumulation assessment submodule calculates the periodic damage increment value based on the hysteresis loop and residual feature set, accumulates the damage variable, extracts the unclosed residual peak and valley data and overwrites it to the residual peak and valley sequence stack, and outputs the total cumulative damage variable for each blade.

[0010] As a further aspect of the present invention, the health difference assessment module includes: The damage discrete analysis submodule calculates the arithmetic mean of the total cumulative damage variables of multiple blades, analyzes the discrete deviation of the damage degree and the average result of each blade, quantifies the damage distribution difference parameters of the blades during their life cycle, and generates the damage discrete deviation degree. The load allocation and adjustment submodule, based on the damage discrete deviation degree, performs an inverse linear mapping operation according to the discrete deviation degree corresponding to each blade, quantitatively evaluates the relative remaining life difference of each blade component, reallocates the preset load share of the blade, and generates a load bearing ratio factor. The constraint configuration construction submodule calculates the load setpoint offset required to correct the independent control target of each blade based on the load bearing ratio factor, establishes pitch control boundary conditions adapted to the health state of each blade, constructs an asymmetric execution framework including the angle of attack limit and torque distribution rules for each blade, and establishes asymmetric control constraint configuration.

[0011] As a further aspect of the present invention, the equilibrium control decision module includes: The target dynamic setting submodule, by superimposing the load setting point offset to the rated root bending moment setting parameter, differentiates the theoretical control target of each blade, calculates the expected load level to adapt to the current health state difference, and generates a dynamic load target value. The instruction generation decision submodule acquires and calculates the current real-time virtual load value at the blade root for the dynamic load target value. It combines differential operation to calculate the numerical deviation between the real-time force state and the expected target. Based on the polarity and amplitude of the deviation, it determines the required angle adjustment range and maps it to the action step size and rate limit of the matching pitch actuator to generate a pitch angle adjustment instruction set. The state control feedback submodule drives the asymmetric pitch action of each blade pitch motor according to the pitch angle adjustment instruction set, adjusts the aerodynamic force distribution on the blade surface, and collects the unit power output, rotor speed and tower vibration data after the pitch adjustment response in real time to generate unit operating status parameters.

[0012] As a further aspect of the present invention, the process of superimposing the load setting point offset to the rated root bending moment setting parameter specifically comprises: Real-time acquisition of hub height wind speed data of the unit's operating environment, querying the standard reference bending moment corresponding to the hub height wind speed data in the preset unit control strategy database, and marking the standard reference bending moment as the rated root bending moment setting parameter; The ultimate failure load value in the material properties of the blade component is retrieved, the remaining bearing margin between the ultimate failure load value and the rated root bending moment setting parameter is calculated, and the remaining bearing margin is multiplied by the preset safety margin utilization coefficient to define the maximum allowable offset adjustment range threshold. An amplitude out-of-bounds detection is performed on the load setpoint offset to determine whether the absolute value of the load setpoint offset is greater than the maximum offset adjustment amplitude threshold. If the absolute value of the load setpoint offset is greater than the maximum offset adjustment range threshold, then the positive and negative directions of the load setpoint offset are maintained, and the absolute value of the load setpoint offset is truncated to the maximum offset adjustment range threshold to generate an effective injected offset that is safely limited. A linear superposition operation is performed to accumulate the effective injected bias amount to the rated root bending moment setting parameter, generating the dynamic load target value used to guide subsequent pitch control.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, instantaneous curvature and torsional rate are calculated by collecting cross-sectional wavelength drift data, and the dynamic bending attitude of flexible components in space is reconstructed in real time. Combined with aerodynamic simulation and residual peak-valley sequence stacking technology, continuous tracking of stress cycles and closed hysteresis loops is achieved, and the cumulative damage degree of components throughout the entire service life is accurately quantified. By analyzing the discrete deviation of individual damage relative to the group mean, an asymmetric load distribution mechanism based on health status differences is established, which effectively alleviates the fatigue propagation trend of high-damage components, avoids premature failure caused by uneven load distribution, and achieves life balance and collaborative life extension of key components of the unit throughout the entire life cycle. Attached Figure Description

[0014] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a flowchart of the deformation parameter calculation module of the present invention; Figure 3 This is a flowchart of the skeleton-driven reconstruction module of the present invention; Figure 4 This is a flowchart of the load damage calculation module of the present invention; Figure 5 This is a flowchart of the health difference assessment module of the present invention; Figure 6 This is a flowchart of the equilibrium control decision module of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0017] Please see Figure 1 A digital twin-based wind turbine lifecycle management system includes: The deformation parameter calculation module collects the center wavelength drift data of the underside and belly regions of the blade cross section, calculates the difference results of the center wavelength drift data of the same cross section, normalizes them by combining the vertical distance, calculates the instantaneous curvature vector and torsional scalar of the corresponding cross section, and establishes the geometric morphology parameters. The skeleton-driven reconstruction module extracts instantaneous curvature vectors based on geometric morphology parameters, constructs neutral axis skeleton curves, calculates rigid body rotation and translation matrices by combining the skeleton node indices bound to virtual mesh vertices, performs linear transformations on the original local coordinates of mesh vertices, and outputs real-time position coordinates. The load damage calculation module generates a stress change sequence by combining real-time location coordinates and wind speed simulation, injects residual peak and valley sequence stack data, identifies closed stress hysteresis loops to accumulate damage variables, overwrites the residual peak and valley sequence stack with unclosed data, and outputs the total accumulated damage variable. The health difference assessment module calculates the arithmetic mean of the total cumulative damage variable, analyzes the deviation between the damage degree of each leaf and the average result, and performs inverse linear mapping calculation to establish the load setpoint bias of the corrected control target. The balanced control decision module superimposes the load setpoint offset onto the rated root bending moment parameter to establish a dynamic load target, determines the expected load level to adapt to the current health status difference, calculates the difference result with the virtual load, constructs pitch command to adjust the stress state, and outputs the unit operating status parameters.

[0018] The geometric parameters include the cross-sectional bending curvature component and the cross-sectional torsional angle component. The real-time position coordinates are specifically the spatial coordinate triplet of discrete points on the blade surface. The total cumulative damage variables include the count statistics of closed stress cycles and the percentage of material fatigue life consumed. The asymmetric control constraint configuration includes the weighting coefficient of the load distribution of each blade and the dead zone limit parameter of the pitch controller. The unit operating status parameters include the action rate command value of the pitch motor of each blade and the dynamic adjustment setting value of the unit's active power.

[0019] Please see Figure 2 The deformation parameter calculation module includes: The drift data acquisition submodule uses a fiber optic grating sensor to collect center wavelength drift data for each monitoring point by performing spectral scanning and signal demodulation on the underside and belly of the leaf. Using a fiber Bragg grating sensor network pre-laid along the blade span, a high-frequency broadband light source emits a continuous spectral signal with a center wavelength of 1550 nm into the fiber. When the light signal passes through the grating region, the grating reflects light of a specific wavelength that satisfies the Bragg condition. At this time, the signal demodulator receives the reflected light signal in real time at a sampling frequency of 100 Hz and performs peak-finding calculation on the center peak position of the reflected spectrum. In actual operation, the system synchronously scans paired monitoring points on the blade's back (tension side) and front (compression side). For example, for the monitoring section numbered S-01 at the blade root, the system reads the values ​​of the back sensor FBG-Back-01 at the current... The reflection center wavelength at the previous moment was 1550.52nm, and the reflection center wavelength of the leaf underside sensor FBG-Belly-01 was 1549.48nm. At the same time, the reference wavelengths of these two sensors in the initial unstressed state stored in the system were retrieved, both of which were 1550.00nm. By performing numerical subtraction between the current measured wavelength and the initial reference wavelength, the positive drift amount of the leaf underside region was obtained as +0.52nm and the negative drift amount of the leaf underside region was obtained as -0.48nm, respectively. In this way, the spectral signal changes caused by micro-deformation of each section of the leaf during dynamic operation were accurately captured, and these original wavelength drift data containing structural response characteristics were transmitted to the subsequent processing unit.

[0020] The data differential processing submodule performs differential calculations on the center wavelength drift data at the same cross-sectional location, and combines the vertical distance between the monitoring points on the back of the leaf and the monitoring points on the belly of the leaf to normalize the differential results and generate the cross-sectional normalized strain difference value. The system retrieves paired wavelength drift data from the same cross-sectional location, namely the positive drift on the blade back and the negative drift on the blade belly, and performs differential calculations to eliminate the influence of axial tensile strain and part of the common-mode temperature strain. Specifically, the system reads pre-calibrated and stored blade cross-sectional geometric parameters to obtain the straight-line distance along the chordal vertical plane between the monitoring points on the blade back and the blade belly at that cross-section. For example, for cross-section S-01, its calibrated vertical distance parameter is 2.8 meters. The system first calculates the algebraic difference of the wavelength drift, i.e. Then, the difference is divided by the corresponding vertical distance for spatial geometric normalization. This is achieved through calculation. This converts changes in the optical signal level into a physical quantity reflecting the degree of pure bending deformation of the cross-section. In this process, a sensor sensitivity coefficient (e.g., 1.2 pm / ) is introduced. The wavelength difference is converted into a micro-strain difference value, and finally the cross-sectional normalized strain difference value at the current moment is generated. This value intuitively eliminates the influence of the cross-sectional thickness, and provides a standard input with a unified dimension for the subsequent lateral comparison of the bending degree of cross-sections with different thicknesses.

[0021] The parameter calculation and analysis submodule, based on the normalized strain difference value of the cross section, obtains real-time ambient temperature data and uses the following formula: ; The bending characteristic index of the cross section is calculated, and vector synthesis is performed by combining the direction vector of the cross section coordinate system. The instantaneous curvature vector and torsional scalar of the corresponding cross section are solved to evaluate the local bending degree and torsional state of the blade and generate geometric morphological parameters. in, The index representing the bending characteristic of the cross section. The normalized strain difference value of the representative cross section is obtained by collecting the center wavelength drift data of the blade back and belly and performing differential and normalization processing. The temperature-strain coupling weight is represented by a value obtained through prior calibration of the strain response of the grating material under varying temperature conditions. This represents the normalized ambient temperature compensation value, calculated by collecting real-time ambient temperature data and comparing it with a preset reference temperature. This represents the normalized cross-sectional geometry factor, calculated by measuring the effective bending diameter of the blade cross-section and comparing it with a reference diameter. The matrix strain transfer correction factor is obtained by analyzing the shear hysteresis effect of the adhesive layer between the grating sensor and the blade matrix. This represents the grating strain sensitivity coefficient, which can be obtained by querying the sensor's factory parameters. Based on the normalized strain difference of the cross section, the system first synchronously acquires the real-time ambient temperature data (e.g., 25°C) using thermocouples or fiber optic temperature sensors arranged on the blade surface, and then calculates the temperature change by calling a preset reference temperature (e.g., 20°C), followed by applying the formula... To perform the calculation, in this formula, The first part is the normalized strain difference of the cross section. (set as) That is, 200 Perform squaring operations and simultaneously calculate the temperature-strain coupling weights. (set as) ) and normalized ambient temperature compensation value (Right now The absolute value of the product of ) is used in this part of the calculation, which aims to eliminate the spurious strain interference caused by thermal expansion and contraction under large temperature difference environments through nonlinear modulus synthesis. The denominator term includes... As a normalized cross-sectional geometry factor (set to 1.0), combined with the matrix strain transfer correction factor (Set to 0.05, representing 5% shear hysteresis loss), the calculation results are geometrically dimensionless and compensated for transmission efficiency; the latter half of the formula... Then, the grating strain sensitivity coefficient is introduced. (Set as 1.2) Maintaining the linear characteristics under small deformations, the specific calculation is as follows: First, calculate the numerator of the nonlinear term. ; The denominator is The first part of the results is ; The second part of the linear terms is: ; The final result is obtained by superposition. ; in, The (section bending characteristic index) is a comprehensive state parameter that has undergone multi-physics decoupling and dimensionless processing. Specifically, after eliminating spurious strain components caused by environmental temperature fluctuations and transmission losses due to sensor adhesive layers, it normalizes the degree of pure geometric bending of the blade section relative to its initial undeformed state. Its specific effect is that it provides a standard input source for subsequent blade skeleton reconstruction, unaffected by differences in cross-sectional dimensions and environmental interference. This significantly improves the morphological inversion accuracy of digital twin models in complex and variable wind field environments, effectively avoids false damage identification caused by thermal expansion and contraction or sensor drift, and thus ensures the accuracy of load assessment in the entire life cycle management. By combining the direction vector of the cross-sectional coordinate system (determined by the blade twist angle, such as 15°), the instantaneous curvature vector (with a modulus of 15°) of the corresponding cross-section is finally calculated through vector synthesis. ) and torsional scalar ( This allows for the establishment of geometric morphological parameters that contain precise deformation information for each section of the blade.

[0022] Please see Figure 3 The skeleton-driven reconstruction module includes: The skeleton curve construction submodule analyzes the instantaneous curvature vector distributed along the blade span based on geometric morphology parameters, performs spatial cumulative integration along the blade span, derives the translation and rotation of the center of each discrete section relative to the blade root, analyzes the overall bending attitude of the blade, connects the center points of each section, constructs a spatial geometric path, and generates the neutral axis skeleton curve. Based on geometric morphological parameters, the instantaneous curvature vector distributed along the blade span is analyzed, with the blade root as the coordinate origin. Discretize the blade along the spanwise direction (Z-axis) as follows: There are infinitesimal segments (e.g., 100 segments), and each segment has a length of . (For example, 0.8 meters), for the first For each cross section, extract the instantaneous curvature vector calculated in the preceding steps. (For example and torsion rate Using the recursive integral method of infinitesimal elements, the increment of the tangential rotation angle of the center of the current cross section relative to the center of the previous cross section is calculated. and displacement increment Through cumulative operation (in (The attitude matrix of the previous node) is used to derive the absolute spatial coordinates of the center of each discrete section in the leaf root coordinate system step by step (for example, the center coordinates of section 50 are...). The 100 discrete cross-sectional center points (Roll, Pitch, Yaw) are then smoothly connected using a cubic spline interpolation algorithm to construct a continuous three-dimensional spatial curve, namely the neutral axis skeleton curve. This curve fully describes the overall flapping and oscillating bending attitude of the blade under the current wind load.

[0023] The transformation matrix calculation submodule, for the neutral axis skeleton curve, combines the skeleton node index bound to the virtual mesh vertex to extract the initial reference coordinates of the skeleton node in the undeformed state, calculates the rotation angle and displacement distance of each skeleton node in the current deformed state relative to the initial state, and constructs a rigid body rotation and translation matrix containing rotation submatrices and translation vectors. For the neutral axis skeleton curve, the system calls a pre-built virtual mesh model, in which each mesh vertex is bound to the index of its nearest neighbor skeleton node (e.g., vertex). Bind skeleton nodes First, extract the skeleton nodes. Initial reference coordinates in the undeformed state (e.g.) ) and initial attitude quaternion Next, obtain the real-time coordinates of the skeleton node on the skeleton curve under the current deformation state (e.g., ) and real-time attitude quaternions quaternion division operation Calculate the rigid body rotation of the node relative to the initial state, and then convert the quaternion. Convert to Rotation submatrix Simultaneously calculate the displacement vector (Right now ), ultimately rotating the submatrix With translation vector Combine and construct a control area suitable for this skeleton node. The rigid body rotation and translation matrix is ​​used to subsequently describe the rigid body transformation relationship of the mesh in this region as the skeleton moves.

[0024] The mesh reconstruction mapping submodule, based on the rigid body rotation and translation matrix, obtains the original local coordinate data of the virtual mesh vertices on the blade surface, performs a linear transformation on the original local coordinates of the mesh vertices, maps and transforms the mesh vertices from the local reference system to the global observation coordinate system, solves the three-dimensional spatial position of each vertex at the current moment, and generates real-time position coordinates. Based on the rigid body rotation and translation matrix, the system traverses each virtual mesh vertex on the blade surface to obtain the original coordinate data of the vertex in the local reference frame (e.g., vertex coordinates). The local coordinates are (representing the local offset relative to the skeleton node), according to the linear algebraic transformation rules, the original local coordinates are regarded as column vectors. Multiply by the corresponding rigid body rotation and translation matrix on the left. That is, to perform operations ;in This represents the global absolute coordinate vector of the skeleton node at the current moment. If the rotation matrix... This corresponds to a 5-degree rotation around the X-axis, and the current absolute coordinates of the skeleton node are... (i.e., from the initial reference coordinates) Superimposed displacement The result is that the calculation process involves rotating and transforming the local coordinates and then superimposing the absolute position vector of the node, ultimately calculating the current three-dimensional spatial position of the vertex in the global observation coordinate system (e.g., By repeating this process for all tens of thousands of mesh vertices, a deformable mesh model with the same height as the physical blade can be generated in real time in digital space, and its real-time position coordinates can be output.

[0025] Please see Figure 4The load damage calculation module includes: The stress simulation analysis submodule extracts real-time position coordinates, calls the collected environmental wind speed data, analyzes the stress state at the blade root through aeroelastic simulation, and generates time series data of root stress change. The real-time position coordinates are extracted to reconstruct the aerodynamic shape of the blade at the current moment. Simultaneously, environmental wind speed data collected by a lidar or anemometer mounted on the top of the nacelle (e.g., wind speed at hub height 12 m / s, turbulence intensity 10%) is retrieved. Using blade element momentum theory (BEM) or computational fluid dynamics (CFD) solvers, the wind speed vector field is coupled with the real-time position, angle of attack, and relative inflow velocity of each blade section to simulate the aerodynamic forces (lift and drag) distributed along the blade spanwise. The bending moment load at the blade root is then calculated by integration. For example, the current moment... The root flapping moment is 6500 kNm. The system continuously executes this simulation process with a time step of 0.1 seconds to generate time series data of root stress change within a time window (such as 10 minutes). This data sequence (such as [50, 52, 48, 60, ...] MPa) accurately reflects the stress fluctuation of the blade root material under dynamic wind field.

[0026] The loop counting and identification submodule reads the residual peak and valley sequence stack and injects it into the starting position of the stress change time series for the root stress change time series. Through peak and valley extraction and loop pairing operation, it identifies closed stress hysteresis loops and separates unclosed residual peak and valley data, generating hysteresis loops and residual feature sets. For the time series data of root stress changes, the system first performs peak and valley value extraction processing, retaining only the local maxima and minima of the waveform and filtering out non-reversed intermediate data points to form a simplified load spectrum (e.g., 10->80->20->70->30). Then, it reads the residual peak and valley sequence stack left over from the previous calculation cycle (e.g., [100,10]), splices the newly generated peak and valley sequence to the end of the stack, and scans it using the RainflowCounting rule to determine whether there is a closed condition that satisfies the condition that "the amplitude of the next cycle contains the amplitude of the previous cycle". For example, if a closed small cycle is identified that rises from 20MPa to 70MPa and then returns to 20MPa, the system extracts the closed loop for damage calculation and retains the unclosed divergent data (e.g., the last half wave) in the stack as the starting condition for the next cycle. Finally, it generates a set of hysteresis loops containing multiple closed loops (with defined mean and amplitude) and an unclosed residual feature set.

[0027] The damage accumulation assessment submodule, based on hysteresis loops and residual feature sets, uses the following formula: ; Calculate the incremental value of periodic damage, accumulate the damage variable, extract the unclosed residual peak and valley data and overwrite it to the residual peak and valley sequence stack, and output the total cumulative damage variable for each blade. in, Represents the incremental value of periodic damage. The counting index representing the closed stress hysteresis loop is determined by the cyclic pairing operation. The total number of identified closed stress hysteresis loops is obtained through a cycle counting process. Representing the The normalized stress amplitude of each closed loop is obtained by dividing the analytically obtained physical stress amplitude by the material yield strength. The average stress correction factor is obtained through material fatigue testing. Representing the The normalized average stress of a closed loop is obtained by dividing the analytically obtained physical average stress by the material's yield strength. The fatigue damage index is obtained by referring to the slope of the SN curve of the blade material. This represents the dimensionless fatigue strength baseline, obtained by consulting material handbooks to find the standard fatigue strength constant and then performing normalization conversion. The weight representing the impact of fluctuations is set through statistical analysis of the impact of historical wind conditions on damage. The wind speed fluctuation coefficient is obtained by calculating the variance of environmental wind speed data and then normalizing it. Based on the recognition results shown in Table 1, using the hysteresis loop and residual feature set, the system employs the formula... Calculations are performed, in which, (Periodic damage increment) refers to the sum of the proportions of fatigue life consumed by all identified closed stress hysteresis loops within the current calculation period (corresponding to the length of the input stress change time series). Specifically, by directly coupling environmental turbulence characteristics (wind speed fluctuations) into Miner's linear accumulation theory, an "environmentally sensitive" damage assessment mechanism is achieved. This mechanism automatically improves the sensitivity of damage assessment under severe wind conditions (high fluctuations), thus capturing the latent damage to the blade root from dynamic wind loads more conservatively and accurately than the traditional static SN curve method, effectively preventing the risk of structural fracture due to underestimating the impact of high-frequency gusts. In the formula... For closed loop indexes, To determine the total number (e.g., 2 in Table 1), for the data in the first row of Table 1, first calculate the normalized stress amplitude. (Assuming the material yield strength is 600 MPa), normalized mean stress Introducing a mean stress correction factor (Set as 0.3), then the base of the numerator is Fatigue damage index Take the slope of the SN curve of the material (set to 4.0), and the molecule value is... The denominator is the dimensionless fatigue strength base. (Set to 1000), fluctuation affects weight (Set to 0.5), wind speed fluctuation coefficient (Obtained by normalizing the wind speed variance, set to 0.2), then the denominator is... The damage increment per cycle is The system applies to all The damage increments of each cycle are accumulated, and the residual stack is updated. Finally, the total cumulative damage variable of the blade in the current time period is output (e.g., ).

[0028] Table 1. Data for Identifying Closed Stress Hysteresis Loops See Table 1, which lists the specific parameters of two typical closed stress hysteresis loops identified by the rainflow counting method during this calculation period, serving as the input for the formula calculation.

[0029] Please see Figure 5 The health difference assessment module includes: The damage discrete analysis submodule calculates the arithmetic mean of the total cumulative damage variables of multiple blades, analyzes the discrete deviation of the damage degree and the average result of each blade, quantifies the damage distribution difference parameters of the blades during their life cycle, and generates the damage discrete deviation degree. The system acquires the total cumulative damage variables of each of the three blades (BladeA, BladeB, BladeC) of the wind turbine in real time. Assume the cumulative damage values ​​of the three blades at the current moment are as follows: , , The system first performs an arithmetic mean calculation, which is the baseline damage mean at the computer group level. Then, the degree of dispersion deviation of each leaf relative to the mean is calculated, i.e. Calculations show that the deviation of blade A is: (Minor damage), the deviation of blade B is... (Severe damage), the deviation of blade C is Through this quantification process, a significant "health deficiency" in leaf B was clearly characterized, generating a damage dispersion bias that includes these three percentage values.

[0030] The load allocation and adjustment submodule, based on the damage discrete deviation degree, performs an inverse linear mapping operation according to the discrete deviation degree corresponding to each blade, quantitatively evaluates the relative remaining life difference of each blade component, reallocates the preset load share of the blade, and generates a load bearing ratio factor. Based on the damage dispersion deviation, the system sets a principle of equalizing the health status of the entire unit, meaning that components with significant damage should bear less load, while components with less damage can bear more load. An inverse linear mapping operation is performed based on the dispersion deviation degree corresponding to each blade, and the mapping function is set as follows: ,in To adjust the sensitivity coefficient (set to 0.5), substitute the preceding data to calculate: the load-bearing ratio factor for blade A is... The factor for leaf B is The factor for leaf C is The results indicate that, in order to balance lifespan, blades A and C will bear 1.083 times the rated load in subsequent operations, while the more severely damaged blade B will only bear 0.833 times the rated load, thereby generating a load-bearing ratio factor to guide subsequent control.

[0031] The constraint configuration construction submodule calculates the load setpoint offset required to correct the independent control target of each blade based on the load bearing ratio factor, establishes pitch control boundary conditions adapted to the health state of each blade, constructs an asymmetric execution framework that includes angle of attack limits and torque distribution rules for each blade, and establishes asymmetric control constraint configuration. Based on the load bearing ratio factor, the system, in conjunction with the rated control target of the main control system (e.g., rated root bending moment of 8000 kNm), calculates the load setpoint offset required to correct the independent control target of each blade. For blade A, the target correction is as follows: kNm, bias is kNm, for blade B, the target is corrected to kNm, bias is Based on these offset values, the system further establishes pitch control boundary conditions, such as setting a smaller maximum angle of attack limit for blade B (from 90 degrees to 88 degrees) to prevent overload, and constructs an asymmetric execution framework that allows the three blades to maintain different pitch angles at the same wind speed. Finally, an asymmetric control constraint configuration containing their respective offset targets and action limits is established.

[0032] Please see Figure 6 The equilibrium control decision module includes: The target dynamic setting submodule reconstructs the theoretical control target of each blade by superimposing the load setpoint offset to the rated root bending moment setting parameter, calculates the expected load level to adapt to the current health status difference, and generates dynamic load target value. The system first collects real-time hub height wind speed data (e.g., 10.5 m / s) of the unit's operating environment, and then queries the pre-set unit control strategy database (e.g., Cp-Lambda curve table) for the standard reference bending moment corresponding to that wind speed. For example, it finds that the rated root bending moment setting parameter at 10.5 m / s is 7500 kNm. Next, it retrieves the ultimate failure load value (e.g., 15000 kNm) from the material properties of the blade components and calculates the remaining load margin. kN / m, combined with the preset engineering safety factor (set to 0.2, i.e., retaining a 20% margin), the maximum allowable offset adjustment threshold is defined as follows: The system performs boundary checks on the load setpoint offsets calculated in the preceding steps (e.g., +664kNm for blade A, -1336kNm for blade B). If the calculated offset at a certain moment is -1800kNm, and its absolute value of 1800 is greater than the threshold of 1500, then it is truncated and limited to -1500kNm, generating a safe-limited effective injected offset. Finally, a linear superposition operation is performed to accumulate the effective offset to the rated root bending moment setting parameter. For example, the dynamic target of blade B becomes... kNm, thereby generating differentiated dynamic load target values ​​for each blade.

[0033] The instruction generation decision submodule acquires and calculates the current real-time virtual load value at the blade root for the dynamic load target value. It combines differential operation to calculate the numerical deviation between the real-time force state and the expected target. Based on the polarity and amplitude of the deviation, it determines the required angle adjustment range and maps it to the action step size and rate limit of the matching pitch actuator to generate a pitch angle adjustment instruction set. For a target dynamic load value (e.g., 6164 kNm for blade B), the system obtains the current real-time virtual load value (e.g., 6500 kNm as measured) using strain gauges or fiber Bragg grating sensors installed at the blade root, and performs differential calculations to determine the deviation. The value of kNm indicates that the current load is too large and feathering (increasing the pitch angle) is required to unload it. The system, based on the deviation amplitude of +336 kNm and the proportional gain coefficient in the PID control algorithm, determines the load. (For example The theoretically required angle adjustment range was calculated as follows: This amplitude is then mapped to the physical limits of the pitch actuator to check if it exceeds the maximum pitch rate per single cycle (e.g., The maximum change within a 0.1s period If the output exceeds the limit, the output will be restricted to a certain value. Finally, control commands containing pitch direction and step size are constructed to generate a pitch angle adjustment command set for blade B.

[0034] The state control feedback submodule drives the asymmetric pitch action of each blade pitch motor according to the pitch angle adjustment instruction set, adjusts the aerodynamic force distribution on the blade surface, and collects the unit power output, rotor speed and tower vibration data after the pitch adjustment response in real time to generate unit operating status parameters. According to the pitch angle adjustment command set, the system sends asymmetric control signals to the independent pitch drivers of the three blades. This drives blade A to maintain its original angle or slightly reduce the pitch angle to increase the force, while driving blade B to rapidly increase the pitch angle (e.g., from 12 degrees to 12.5 degrees) to reduce the angle of attack and lift. Through this asymmetric pitch action, the aerodynamic force distribution of the impeller plane is physically changed to match the health status of each blade. At the same time, the system collects the unit's operating data in real time after the pitch adjustment response, including generator power output (e.g., fluctuating from 2.1MW to 2.0MW), rotor speed (e.g., 14.5rpm), and vibration acceleration data at the top of the tower. This confirms that the vibration level does not exceed the safety threshold (e.g., 0.1g), thereby verifying the effectiveness of the control strategy and recording this data to generate unit operating status parameters that include health correction features.

[0035] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A wind turbine full life cycle management system based on digital twinning, characterized in that, The system comprises: a deformation parameter solving module, which collects center wavelength drift data of the suction side and the pressure side in a blade section, calculates a difference result of the center wavelength drift data, normalizes the difference result in combination with a vertical distance, solves an instantaneous curvature vector and a torsion rate scalar of the corresponding section, and establishes geometric morphological parameters; a skeleton drive reconstruction module, which extracts the instantaneous curvature vector based on the geometric morphological parameters, constructs a neutral axis skeleton curve, calculates a rigid body rotation and translation matrix in combination with a skeleton node index bound to a virtual grid vertex, performs linear transformation on original local coordinates of the grid vertex, and outputs real-time position coordinates; a load damage calculation module, which generates a stress change sequence in combination with the real-time position coordinates and a wind speed simulation, injects residual peak-valley sequence stack data, identifies a closed stress hysteresis loop cumulative damage variable, overwrites the residual peak-valley sequence stack with unclosed data, and outputs a total cumulative damage variable; a health difference evaluation module, which calculates an arithmetic mean result of the total cumulative damage variable, analyzes a deviation value of the damage degree of each blade and the average result, and performs inverse linear mapping operation to establish a load setpoint bias of a correction control target.

2. The digital-twin-based wind turbine full life cycle management system according to claim 1, characterized in that, The geometric morphological parameters comprise a section bending curvature component and a section torsion angle component, the real-time position coordinates are specifically spatial coordinate triplets of blade surface discrete points, the total cumulative damage variable comprises a count statistical value of a closed stress cycle and a material fatigue life consumption percentage, and the load setpoint bias comprises a weight coefficient of blade load distribution and a dead zone limit parameter of a variable pitch controller.

3. The digital-twin-based wind turbine full life cycle management system of claim 1, wherein, The deformation parameter solving module comprises: a drift data collection submodule, which collects center wavelength drift data of each monitoring point by performing spectrum scanning and signal demodulation on the suction side area and the pressure side area by using a fiber grating sensor; a data difference processing submodule, which performs difference operation on the center wavelength drift data at the same section position, normalizes the difference result in combination with a vertical distance between the suction side monitoring point and the pressure side monitoring point, and generates a section normalized strain difference value; a parameter solving and analyzing submodule, which obtains real-time environmental temperature data based on the section normalized strain difference value, calculates a section bending characteristic index by using a formula:, performs vector synthesis in combination with a section coordinate system direction vector, solves an instantaneous curvature vector and a torsion rate scalar of the corresponding section, evaluates a local bending degree and a torsion state of the blade, and generates geometric morphological parameters; ； The skeleton drive reconstruction module comprises: wherein, represents a cross-section bending characteristic index, represents a cross-section normalized strain difference, represents a temperature-strain coupling weight, represents a normalized ambient temperature compensation value, represents a normalized cross-section geometry factor, represents a matrix strain transfer correction factor, represents a grating strain sensitivity coefficient.

4. The digital twin-based wind turbine full lifecycle management system of claim 3, wherein, a skeleton curve construction submodule, which analyzes the instantaneous curvature vector distributed along the blade span, performs spatial cumulative integral operation along the blade span, deduces a translation amount and a rotation amount of each discrete section center relative to a blade root, analyzes a whole bending posture of the blade, connects center points of each section, constructs a spatial geometric path, and generates a neutral axis skeleton curve. ​ The transformation matrix calculation submodule extracts initial reference coordinates of the skeleton nodes in an undeformed state, calculates rotation angles and displacement distances of each skeleton node relative to the initial state in a current deformed state, and constructs a rigid rotation and translation matrix including a rotation submatrix and a translation vector, for the neutral-axis skeleton curve and in combination with the skeleton node indexes bound to the virtual grid vertices. The grid reconstruction mapping submodule obtains original local coordinate data of the virtual grid vertices on the blade surface based on the rigid rotation and translation matrix, performs linear transformation on the original local coordinates of the grid vertices, maps and converts the grid vertices from a local reference system to a global observation coordinate system, solves three-dimensional space positions of each vertex at the current time, and generates real-time position coordinates.

5. The digital twin-based wind turbine full lifecycle management system of claim 4, wherein, The load damage calculation module includes: The stress simulation analysis submodule extracts the real-time position coordinates, calls the collected environmental wind speed data, analyzes the stress state of the blade root through aeroelastic simulation, and generates root stress change time series data; The cycle count identification submodule reads residual peak and valley sequence stacks and injects the starting position of the stress change time series for the root stress change time series data, identifies closed stress hysteresis loops and separates unclosed residual peak and valley data through peak and valley extraction and cycle pairing operations, generates a hysteresis loop and a residual feature set, and outputs the total cumulative damage variable of each blade. The damage accumulation evaluation submodule calculates the periodic damage increment value based on the hysteresis loop and the residual feature set, accumulates the damage variable, extracts the unclosed residual peak and valley data and overwrites it to the residual peak and valley sequence stack, and outputs the total cumulative damage variable of each blade.

6. The digital twin-based wind turbine full lifecycle management system of claim 5, wherein, The health difference evaluation module includes: The damage dispersion analysis submodule calculates the arithmetic mean of the total cumulative damage variables of multiple blades, analyzes the dispersion deviation value of the damage degree of each blade and the average result, quantifies the damage distribution difference parameter of the blades in the life cycle, and generates a damage dispersion deviation degree; The load distribution adjustment submodule performs reverse linear mapping operations according to the dispersion deviation degree of each blade, quantifies and evaluates the relative residual life difference of each blade component, redistributes the preset load share of the blades, and generates a load bearing proportion factor; The constraint configuration construction submodule calculates the load setpoint bias required to modify each blade independent control target based on the load bearing proportion factor, establishes a variable pitch control boundary condition that adapts to the health status of each blade, constructs an asymmetric execution framework including the attack angle limit and torque distribution rule of each blade, and establishes an asymmetric control constraint configuration.

7. The digital twin-based wind turbine full lifecycle management system of claim 1, wherein, The system further includes: The equalization control decision module adds the load setpoint bias to the rated root moment parameter to establish a dynamic load target, judges the expected load level that adapts to the current health status difference, calculates the difference result with the virtual load, constructs a variable pitch instruction adjustment stress state, and outputs a unit operation state parameter; The unit operation state parameter includes the action rate instruction value of each blade variable pitch motor and the dynamic adjustment set value of the unit active power.

8. The digital twin-based wind turbine full lifecycle management system of claim 7, wherein, The equalization control decision module includes: The target dynamic setting sub-module differentiates and reconstructs the theoretical control target of each blade by superimposing the load set point bias on the rated root bending moment setting parameter, calculates the expected load level adapted to the difference in current health state, and generates a dynamic load target value; The instruction generation decision sub-module obtains and calculates the real-time virtual load value of the blade root in the current state in response to the dynamic load target value, calculates the numerical deviation between the real-time stress state and the expected target through differential operation, determines the required angle adjustment amplitude according to the deviation polarity and amplitude, maps it to the action step and rate limit of the matching variable pitch actuator, and generates a variable pitch angle adjustment instruction set; The state regulation feedback sub-module drives the asymmetric variable pitch action of each blade variable pitch motor according to the variable pitch angle adjustment instruction set, adjusts the aerodynamic stress distribution state of the blade surface, and collects the unit power output, rotor speed and tower vibration data after the variable pitch adjustment response in real time, and generates unit operation state parameters.

9. The digital twin-based wind turbine full lifecycle management system of claim 8, wherein, The process of superimposing the load set point bias on the rated root bending moment setting parameter is specifically: Real-time acquisition of hub height wind speed data of the unit operating environment, and query of the standard reference bending moment corresponding to the hub height wind speed data in the pre-set unit control strategy database, the standard reference bending moment is marked as the rated root bending moment setting parameter; The limit damage load value in the blade component material attribute is called, the residual bearing margin between the limit damage load value and the rated root bending moment setting parameter is calculated, and the residual bearing margin is multiplied by the preset safety margin utilization coefficient to define the maximum bias adjustment amplitude threshold allowed to be applied; The amplitude out-of-range detection is performed on the load set point bias, and it is judged whether the absolute value of the load set point bias is greater than the maximum bias adjustment amplitude threshold; If the absolute value of the load set point bias is greater than the maximum bias adjustment amplitude threshold, the positive and negative action directions of the load set point bias are maintained, the absolute value of the load set point bias is truncated to the maximum bias adjustment amplitude threshold, and an effective injection bias after safety limiting is generated; The linear superposition operation is performed, the effective injection bias is added to the rated root bending moment setting parameter, and the dynamic load target value for guiding subsequent variable pitch control is generated.

Images