Multi-source collaborative super-long wind power blade flexible variable pitch control method and multi-source collaborative super-long wind power blade flexible variable pitch control system

Abstract

The invention relates to the technical field of data processing, and discloses a multi-source collaborative super-long wind power blade flexible variable pitch control method and system, and the method comprises the steps: constructing a response frequency band distribution matrix, effectively recognizing a frequency band overlapping interval caused by the distribution coupling of a modal frequency sequence and node coordinates in a flexible variable pitch blade, and obtaining a flexible variable pitch control result. The problem that in the prior art, control path frequency decoupling cannot be achieved under the condition of flexible response frequency band overlapping is solved, further, on the basis of an initial paddle angle path sequence, nonlinear projection analysis processing is introduced, a mapping drift vector is generated, and the path frequency decoupling effect is improved. The method is used for depicting a nonlinear projection drift trend generated by target propeller angle response in a variable-pitch control path, the problem of dynamic error accumulation caused by nonlinear projection drift of the variable-pitch path is solved, and through the two key steps, it is ensured that in an ultra-long flexible blade system, the nonlinear projection drift of the variable-pitch path is reduced. Cooperative execution of frequency domain mapping constraint and path offset correction can be achieved through variable pitch control.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a multi-source coordinated ultra-long wind turbine blade flexible pitch control method and system. Background Art

[0002] As the core equipment of the wind energy conversion system, wind turbines usually adjust the pitch angle of the wind rotor blades in real time through pitch control strategies, so as to achieve a dynamic balance between the wind energy capture capability and the unit load. In the prior art, common pitch control methods include constant speed pitch control, optimal power point tracking pitch control and regional gain adjustment control. The basic idea is to build a desired pitch angle target based on measurement signals such as wind speed, output power or rotor speed, and implement closed-loop regulation through a controller. In most industrial units, the pitch system uses an electro-hydraulic servo drive to quickly execute control commands to cope with wind disturbances and changes in unit load, ensure stable operation of the wind turbine near the rated wind speed, and prevent overspeed and loss of control.

[0003] However, as the power level of wind turbines continues to increase, the size of blades has increased significantly, and the ultra-long flexible blade system has gradually become the mainstream configuration. In such systems, the structural flexibility of the blades is significantly enhanced, resulting in a significant decrease in their low-order modal frequencies, overlapping frequency bands, and a significant increase in the difficulty of decoupling control between multiple vibration modes. At the same time, due to the nonlinear coupling relationship between flexible deformation and wind pressure load, the execution response of the target blade angle in the pitch control path shows obvious nonlinear projection drift characteristics, resulting in the accumulation of dynamic errors between the control target and the actual blade angle. Under the above background, the existing linear control strategy is difficult to ensure the timing consistency of the pitch adjustment process, which may lead to problems such as control instability, enhanced unit vibration and increased blade fatigue. Therefore, how to ensure the timing stability of the pitch control output under the conditions of overlapping flexible response bands of ultra-long blades and nonlinear projection drift in the pitch control path has become a key technical problem that needs to be solved in the current operation control of ultra-large megawatt wind turbines.

[0004] In view of this, the present invention proposes a multi-source coordinated ultra-long wind turbine blade flexible pitch control method and system to solve the above-mentioned problem. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a multi-source coordinated ultra-long wind turbine blade flexible pitch control method and system.

[0006] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, a multi-source coordinated ultra-long wind turbine blade flexible pitch control method is provided, comprising: Obtain blade structure parameters and environmental input data, where the blade structure parameters include a modal frequency sequence and a node coordinate distribution, and the environmental input data includes a wind speed sequence and a load response sequence; Perform a frequency band mapping process on the node coordinate distribution according to the modal frequency sequence to obtain an initial frequency band set, and perform an overlapping region extraction process on the frequency axis according to the initial frequency band set to obtain a response frequency band distribution matrix, where the response frequency band distribution matrix is used to characterize the overlapping interval of the flexible response mode in the frequency dimension; Perform a pitch target calculation process on the wind speed sequence and the load response sequence to obtain an initial pitch angle path sequence, where the initial pitch angle path sequence is used to represent the input state of the pitch control path; Perform a non-linear projection analysis process on the initial pitch angle path sequence according to the response frequency band distribution matrix to obtain a mapping drift vector, where the mapping drift vector is used to represent the offset trend of the control target under flexible response; Perform a trajectory adjustment process according to the mapping drift vector and the response frequency band distribution matrix to obtain a pitch angle output sequence.

[0007] In some embodiments, the method for performing a frequency band mapping process on the node coordinate distribution according to the modal frequency sequence to obtain an initial frequency band set includes: According to each modal frequency in the modal frequency sequence, retrieve the set of mode shape function values of the corresponding order from the preset simulation database, align the modal amplitudes in each set of mode shape function values with the node numbers in the node coordinate distribution, and generate a modal response amplitude matrix; According to each column in the modal response amplitude matrix, calculate the arithmetic mean of all node response amplitudes in each column, perform a calculation of the ratio of the response amplitude to the arithmetic mean for each node's response amplitude to obtain a response ratio. When the response ratio is greater than the preset ratio, form a frequency-node pair with the node number and the current modal frequency, and summarize all frequency-node pairs to form a frequency-node correspondence set; Extract the list of all node numbers corresponding to each modal frequency from the frequency-node correspondence set, sort the node numbers in each node number list in ascending order according to the axial position in the node coordinate distribution, and generate a frequency-node sorting result; Select two adjacent modal frequencies in sequence from the frequency-node sorting result, extract the node number sequences corresponding to the two modal frequencies, and count the number of repeated nodes in the two node number sequences. When the proportion of the number of repeated nodes in the total number of nodes in the two sequences is greater than the preset node proportion, mark the frequency-node correspondence set as having modal overlap, and summarize all frequency-node correspondence sets with modal overlap to generate an initial frequency band set.

[0008] In some embodiments, the method for constructing the preset simulation database includes: Perform mesh generation processing on the preset blade structure design model, divide the blade axially from the root to the tip into multiple structural units, and set node numbers and initial geometric parameters for each structural unit; Perform material parameter assignment processing on each structural unit, and assign elastic modulus, density and damping ratio to it according to the blade design input to form a finite element model with complete parameters; Perform modal simulation processing on the finite element model that has completed parameter assignment, and obtain the modal frequencies and corresponding mode shape function values of the finite element model under the set boundary conditions; Convert each modal frequency and corresponding mode shape function value into a structured data entry, and bind the number identifier of the finite element model and the boundary condition parameters to each structured data entry; Establish an index table for all structured data entries according to the modal numbers to form a simulation database that can be quickly retrieved according to the modal numbers.

[0009] In some embodiments, the method for obtaining the modal frequencies and corresponding mode shape function values of the finite element model under the set boundary conditions includes: ; In the formula, represents the mode shape function value of the -th order modal frequency, represents the original stiffness matrix of the finite element model, represents the perturbation adjustment factor, represents the linear combination of local perturbation stiffness matrices composed of spatial position weighting functions, represents the position at the -th perturbation region of the -th spatial position weighting function, represents the perturbation stiffness matrix corresponding to the -th perturbation sub-segment, represents the inverse matrix operation of the combined stiffness matrix, represents the square value of the -th order modal frequency,

[0010] In some embodiments, the method for performing overlapping region extraction processing on the frequency axis according to the initial frequency band set to obtain the response frequency band distribution matrix includes: According to each frequency-node corresponding set in the initial frequency band set, extract the frequency value and the node number list corresponding to the frequency, arrange the frequency values in ascending order and form an ordered structure with the node number list to generate a frequency node sequence; For each pair of adjacent structures in the frequency node sequence, respectively extract the set of node numbers corresponding to the previous frequency and the set of node numbers corresponding to the next frequency, perform a set intersection operation to obtain a set of duplicate nodes; Calculate the ratio of the number of nodes in the set of duplicate nodes to the average length of the previous and next node number sets to obtain a duplicate node ratio. When the duplicate node ratio is greater than a preset node ratio, calibrate the current frequency as having an overlapping relationship; According to all frequency pairs with an overlapping relationship, merge the frequency values therein by continuous sections, organize all frequency value sets that satisfy the overlapping relationship into several continuous frequency sections, and generate a set of frequency overlapping sections; According to each frequency section in the set of frequency overlapping sections, count all node numbers that fall into this frequency section, use the frequency section as the row index and the node number as the column index to construct a frequency node boolean matrix, and use the frequency node boolean matrix as the response frequency band distribution matrix.

[0011] In some embodiments, the method for performing pitch target calculation processing on the wind speed sequence and the load response sequence to obtain the initial pitch angle path sequence includes: According to each wind speed measurement value in the wind speed sequence, extract the time index corresponding to the wind speed measurement value, and extract the response amplitude at the same position as this time index in the load response sequence to construct a wind-load joint sample sequence; Perform time window partitioning processing on the wind-load joint sample sequence, set a fixed window length and a sliding step size, and divide the wind-load joint sample sequence into multiple period sample segments in a sliding window manner to generate a set of wind-load window sequences; According to each window segment in the set of wind-load window sequences, calculate the wind speed change gradient and the load mean deviation of all sample points in this window, combine the wind speed change gradient and the load mean deviation into a two-dimensional adjustment factor sequence, and construct an adjustment factor matrix; According to each row in the adjustment factor matrix, calculate the corresponding target pitch angle value according to the weighted summation formula of the wind speed change gradient and the load mean deviation, and construct an initial pitch angle value sequence.

[0012] In some embodiments, the method for performing non-linear projection analysis processing on the initial pitch angle path sequence according to the response frequency band distribution matrix to obtain the mapping drift vector includes: According to each frequency section in the response frequency band distribution matrix, extract the set of node number indices corresponding to the frequency section, and combine the node numbers corresponding to each time point in the initial pitch angle path sequence to generate a frequency index list; Number each node in the frequency index list, extract the blade angle values of the corresponding nodes in the initial blade angle path sequence at all time points, and construct a node blade angle time series matrix, where the row index of the node blade angle time series matrix is the node number and the column index is the time point number; According to each row in the node blade angle time series matrix, calculate the time derivative of the blade angle time series change curve to obtain a node response rate matrix, and classify the node response rate matrix by frequency bands to generate a frequency band response set; For each frequency band in the frequency band response set, extract the response rate sequences of all nodes in this band, calculate its average response rate in the time axis direction, calculate the difference residual based on the average response rate and the corresponding time series value in the initial blade angle path sequence, and construct a residual response matrix according to the difference residual; According to each row in the residual response matrix, calculate its cumulative offset in the time direction, and arrange the cumulative offset in a vector structure according to the node number to obtain a mapping drift vector.

[0013] In some embodiments, the method for calculating the time derivative of the blade angle time series change curve to obtain a node response rate matrix according to each row in the node blade angle time series matrix includes: Extract the blade angle change sequence for each row in the node blade angle time series matrix, set a fixed time interval Δt, construct a sliding window structure in time order, form a sample pair for every two adjacent time points, and generate a sliding window sample sequence; According to each sample pair in the sliding window sample sequence, calculate the blade angle difference between two adjacent time points, and divide the difference by the time interval Δt to obtain the blade angle change rate of each node in the time interval Δt, and generate a local rate set; Reorder all the rate values in the local rate set according to the time index and fill them into positions of the same length as the original blade angle change sequence to construct a preliminary response rate matrix; Perform edge point compensation processing on the preliminary response rate matrix and fill in the missing points by interpolation to construct a node response rate matrix.

[0014] In some embodiments, the method for calculating the blade angle difference between two adjacent time points according to each sample pair in the sliding window sample sequence includes: ; In the formula, represents the blade angle difference corresponding to the th sample pair, and respectively represent the original blade angle values at the th and the +1th time points, β represents a non-linear amplification factor, represents a disturbance response scale factor, represents the central time point in the sliding window sample sequence, represents the exponential decay coefficient, which is used to perform weight decay according to the distance of the sample from the central time point in terms of proximity or distance, is the arctangent function, constitutes a dynamic memory mechanism, is the natural constant.

[0015] In a second aspect, a multi-source collaborative flexible pitch control system for ultra-long wind turbine blades is provided, which is used to implement the above-mentioned multi-source collaborative flexible pitch control method for ultra-long wind turbine blades, and includes: Data acquisition module: used to acquire blade structure parameters and environmental input data, where the blade structure parameters include modal frequency sequences and node coordinate distributions, and the environmental input data includes wind speed sequences and load response sequences; First processing module: used to perform frequency band mapping processing on the node coordinate distribution according to the modal frequency sequence to obtain an initial set of frequency bands, and perform overlapping region extraction processing on the frequency axis according to the initial set of frequency bands to obtain a response frequency band distribution matrix, where the response frequency band distribution matrix is used to characterize the overlapping interval of the flexible response mode in the frequency dimension; Second processing module: used to perform pitch target calculation processing on the wind speed sequence and the load response sequence to obtain an initial pitch angle path sequence, where the initial pitch angle path sequence is used to represent the input state of the pitch control path; Third processing module: used to perform non-linear projection analysis processing on the initial pitch angle path sequence according to the response frequency band distribution matrix to obtain a mapping drift vector, where the mapping drift vector is used to represent the offset trend of the control target under flexible response; Adjustment module: used to perform trajectory adjustment processing according to the mapping drift vector and the response frequency band distribution matrix to obtain a pitch angle output sequence.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By constructing a response frequency band distribution matrix, the present invention effectively identifies the frequency band overlapping intervals caused by the coupling of modal frequency sequences and node coordinate distributions in flexible pitch blades, solving the problem in the prior art that it is impossible to achieve decoupling of control path frequencies under the condition of flexible response frequency band overlap. Further, on the basis of the initial pitch angle path sequence, the present invention introduces non-linear projection analysis processing to generate a mapping drift vector, which is used to characterize the non-linear projection drift trend of the target pitch angle response in the pitch control path, solving the problem of dynamic error accumulation caused by non-linear projection drift in the pitch path. Through the above two key steps, it is ensured that in an ultra-long flexible blade system, pitch control can achieve the coordinated execution of frequency domain mapping constraints and path offset correction, thereby improving the timing stability and response accuracy of the pitch angle output sequence and avoiding risks such as control instability and fatigue aggravation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic flowchart of a multi-source collaborative flexible pitch control method for ultra-long wind turbine blades in the present invention; Figure 2 It is a schematic structural diagram of a multi-source collaborative flexible pitch control system for ultra-long wind turbine blades in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the following further describes the present invention in detail with reference to specific embodiments and the accompanying drawings. In the following detailed description, many specific details are set forth to provide a thorough understanding of the described exemplary embodiments. However, it is obvious to those skilled in the art that some or all of these specific details may not be required to practice the described embodiments. In other exemplary embodiments, well-known structures are not described in detail to avoid unnecessarily obscuring the concepts of the present disclosure. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. At the same time, without conflict, the various aspects described in the embodiments can be combined arbitrarily.

[0019] Embodiment 1 Please refer to Figure 1 As shown, this embodiment discloses and provides a multi-source collaborative flexible pitch control method for ultra-long wind turbine blades, including: S10: Obtain blade structure parameters and environmental input data, where the blade structure parameters include modal frequency sequences and node coordinate distributions, and the environmental input data includes wind speed sequences and load response sequences; In this embodiment, the modal frequency sequence refers to the set of natural frequencies extracted based on the free vibration responses excited by a wind turbine blade at different positions during the structural response test or simulation of the wind turbine blade, and is used to describe the low-order frequency characteristics of a flexible structure under multi-modal coupling. Taking an ultra-long wind turbine blade with a length of 107 meters as an example, its first three modal frequencies under typical boundary conditions are 0.32 Hz, 0.91 Hz, and 1.64 Hz respectively. This modal frequency sequence is used for subsequent frequency-domain mapping of control targets and can reveal the distribution characteristics of the frequency band overlapping region, and is the basic input for characterizing flexible responses in the frequency dimension.

[0020] The node coordinate distribution refers to the set of spatial coordinates of a group of discrete measurement points set along the axial direction of the blade from the root to the tip during the blade structure modeling or modal test, and is usually expressed as the radial position (unit: meter) of each node and the corresponding cross-section coordinates (unit: millimeter). For example, a total of 45 nodes are arranged in the finite element modeling of a certain type of blade, and the coordinates corresponding to node number 20 are ( ), and this node is located at the position of the main beam in the middle of the blade and is one of the main response positions of the second mode. The node coordinate distribution is not only used for modal identification and frequency mapping, but also participates in the non-linear projection calculation in the pitch path optimization.

[0021] Similarly, the wind speed sequence refers to the set of time-continuous wind speed measurement values collected by the nacelle anemometer, the root sensor of the blade, or the remote wind measurement radar during the operation of the wind turbine, and usually forms a time series structure by sampling at a fixed time interval, reflecting the dynamic change trend of the wind field disturbance in the time dimension. For example, in a certain ultra-long wind turbine blade system, the wind speed at the leading edge of the blade is recorded at a frequency of once per second, and the following wind speed sequence is obtained by continuously recording for 600 seconds: , where , , , showing the short-term fluctuation characteristics of a typical wind field. As a dynamic input signal, the wind speed sequence directly affects the generation process of the pitch angle target and jointly determines the adjustment window and non-linear error distribution of the pitch path with the flexible response mode.

[0022] The load response sequence is a time series set composed of the response amplitudes of the structure under load collected by devices such as strain gauges, accelerometers, and force sensors during the operation of the wind turbine blade. Usually, the local or overall load state of the blade is recorded at a fixed sampling period, reflecting the dynamic mechanical response characteristics of the blade under wind field disturbance. Taking a 107-meter blade as an example, the strain gauge arranged at the 25th cross-section records its bending moment response at a sampling rate of 200 Hz, and the following sequence consisting of 6000 data points is obtained during a certain operation period: , where , this load response sequence is not only used to calculate the pitch target, but also participates in the extraction of non-linear drift characteristics and the timing correction process of the control path together with the modal frequency band mapping matrix.

[0023] It should be added that in this embodiment, to avoid the data misalignment problem caused by directly aligning the wind speed sequence (1Hz) and the load sequence (200Hz) in time index, preferably, two synchronization methods can be used. One is to perform downsampling on the load response sequence to reduce its sampling rate from 200Hz to 1Hz; the other is to perform interpolation on the wind speed sequence to increase its sampling rate to 200Hz and align it with the original load sampling points, ultimately ensuring that there is a one-to-one corresponding sampling pair under the same time index.

[0024] S20: Perform frequency band mapping processing on the node coordinate distribution according to the modal frequency sequence to obtain an initial frequency band set, and perform overlapping region extraction processing on the frequency axis according to the initial frequency band set to obtain a response frequency band distribution matrix, where the response frequency band distribution matrix is used to characterize the overlapping interval of the flexible response mode in the frequency dimension; The method for performing frequency band mapping processing on the node coordinate distribution according to the modal frequency sequence to obtain an initial frequency band set includes: According to each modal frequency in the modal frequency sequence, retrieve the set of mode shape function values corresponding to the order of this modal frequency from the preset simulation database, align the modal amplitudes in each set of mode shape function values with the node numbers in the corresponding node coordinate distribution to generate a modal response amplitude matrix; According to each column in the modal response amplitude matrix, calculate the arithmetic mean of the response amplitudes of all nodes in each column, perform the calculation of the ratio of the response amplitude to the arithmetic mean for each node's response amplitude to obtain a response ratio. When the response ratio is greater than the preset ratio, form a frequency-node pair by combining the node number and the current modal frequency, and summarize all frequency-node pairs to form a frequency-node correspondence set; Extract the list of all node numbers corresponding to each modal frequency from the frequency-node correspondence set, sort the node numbers in each node number list in ascending order according to the axial position in the node coordinate distribution to generate a frequency-node sorting result; Select two adjacent modal frequencies in sequence from the frequency-node sorting result, extract the node number sequences corresponding to the two modal frequencies, count the number of repeated nodes in the two node number sequences. When the proportion of the number of repeated nodes in the total number of nodes in the two sequences is greater than the preset node proportion, mark the frequency-node correspondence set as having modal overlap, and summarize all frequency-node correspondence sets with modal overlap to generate an initial frequency band set.

[0025] In this embodiment, according to each modal frequency in the modal frequency sequence, the set of mode shape function values corresponding to the order of this frequency is retrieved from the preset simulation database, and the modal response amplitude matrix is generated by aligning the modal response amplitudes with the node numbers in the node coordinate distribution. It should be noted that the row index of the modal response amplitude matrix is the node number, and the column index is the modal frequency. Taking the 1st and 2nd frequencies in the modal frequency sequence as 0.91 Hz and 1.64 Hz respectively, the corresponding sets of modal response amplitudes are 0.32, 0.47, 0.55, 0.92, 0.20 and 0.12, 0.60, 0.68, 0.91, 0.25, and the node numbers are 1, 2, 3, 4, 5. Then the following response amplitude matrix can be constructed: ; With the above structure (row: node number, column: modal frequency), subsequent normalization and screening processing can be performed on the response intensity of each node at each modal frequency.

[0026] It can be understood that to screen out the nodes with significant responses at a certain modal frequency, it is necessary to perform a normalized ratio calculation on the response amplitudes of each node in the column corresponding to this modal frequency. The method adopted is the ratio of each response amplitude to the arithmetic mean of all response values in the current column. For example, for the frequency 0.91 Hz, the corresponding node response amplitudes are 0.32, 0.47, 0.55, 0.92, 0.20, and their average value is 0.492. After calculating the ratio, the node ratio vector is obtained: ; Furthermore, a preset ratio threshold is set to 1.0, and the node ratios greater than this threshold are screened. The corresponding node numbers screened out are 3 and 4, indicating that their response amplitudes are higher than the average level, forming the frequency-node pairs (0.91 Hz, 3) and (0.91 Hz, 4). By analogy, all modal frequencies can be traversed to construct a frequency-node correspondence set, providing basic data for subsequent frequency band mapping and the generation of the initial frequency band set.

[0027] The construction method of the preset simulation database includes: Perform mesh generation processing on the preset blade structure design model, divide the blade axially from the root to the tip into multiple structural units, and set node numbers and initial geometric parameters for each structural unit; Perform material parameter assignment processing on each structural unit, and assign elastic modulus, density, and damping ratio to it according to the blade design input to form a finite element model with complete parameters; Perform modal simulation processing on the finite element model with completed parameter assignment to obtain the modal frequencies and corresponding mode shape function values of the finite element model under the set boundary conditions; Convert the modal frequencies and corresponding eigenfunction values of each order into structured data entries, and bind the number identifier of the finite element model and the boundary condition parameters to each structured data entry; Establish an index table for all structured data entries according to the modal number to form a simulation database that can be quickly retrieved by the modal number.

[0028] In this embodiment, to construct a preset simulation database for frequency mapping and modal analysis, it is first necessary to complete the structural division of the finite element model based on the actual blade structure in the design stage. Specifically, by performing mesh division processing on the preset blade structure design model, the blade can be axially discretized into multiple structural units from the root to the tip, and a unique node number and initial geometric parameters are calibrated for each structural unit. The geometric parameters include the coordinate position of the node in three-dimensional space and the type of element it is connected to. For example, in a 107-meter blade model of type 1, the node numbers are numbered from 1 to 45 one by one, and the geometric parameters of the node numbered 20 are ( ), indicating that it is located at the position of the main beam in the middle of the blade.

[0029] It can be understood that after the structural division is completed, to ensure that the finite element model can truly reflect the physical characteristics of the structure, it is necessary to perform material parameter assignment processing on each structural unit. This processing will assign material properties such as elastic modulus E, density rho, damping ratio zeta, etc. to each unit according to the blade design input data to form a finite element model with complete physical parameters. For example, the assignment of the structural unit numbered 1 is as follows: E = 70 GPa, rho = 1800 kg / m³, zeta = 0.01, thus constituting the basic data set of the blade simulation system.

[0030] Furthermore, to extract the inherent response characteristics of the structure under different excitation modes, perform modal simulation processing on the finite element model that has completed the material parameter assignment. Under the given boundary conditions (such as fixed root and free tip), use commercial simulation software (such as ANSYS, Abaqus, etc.) to calculate the modal frequencies and corresponding eigenfunction values of the model at different orders. For example, the first-order modal frequency can be 0.32 Hz, and the eigenfunction is a vector of displacement amplitudes corresponding to a set of nodes. For example: [0.01, 0.03, 0.05, 0.08, 0.11], indicating the response amplitudes of nodes 1 to 5 in this mode.

[0031] It can be understood that for the convenience of unified management and retrieval, the modal frequency and eigenfunction values need to be standardized into structured data entries and record the model source and boundary parameter information. For example, convert the above first-order mode into the entry format as follows: Modal number: 1; Modal frequency: 0.32 Hz; Eigenfunction values: [0.01, 0.03, 0.05, 0.08, 0.11]; Boundary conditions: fixed at the root and free at the tip; Finite element model number: Model_A; Finally, to make the above structured data searchable, an index table is established for all data entries according to the modal number, and a preset simulation database that can be quickly searched is constructed.

[0032] The method for obtaining the modal frequencies and corresponding eigenfunction values of each order of the finite element model under the set boundary conditions includes: ; In the formula, represents the eigenfunction value of the th-order modal frequency, represents the original stiffness matrix of the finite element model, represents the perturbation adjustment factor, represents the linear combination of local perturbation stiffness matrices composed of spatial position weighting functions, represents the position at the th perturbation region of the spatial position weighting function, represents the th perturbation stiffness matrix corresponding to the perturbation sub-segment, represents the inverse matrix operation of the combined stiffness matrix, represents the th-order modal frequency squared value, represents the mass matrix of the model.

[0033] The spatial position weighting function is as follows: ; In the formula, is the spatial position weighting function, is the perturbation center position, is the influence radius, is the exponential function with base e, where e is the natural constant.

[0034] In this embodiment, by introducing a linear combination of a perturbation adjustment factor and a spatial position weighting function into the stiffness matrix term of the traditional modal eigenfunction calculation formula, a structural improvement is essentially made to the commonly used inherent formula in the prior art. The existing modal solution methods usually assume that the structural stiffness is globally uniformly fixed. When dealing with problems such as structural boundary deformation, local stiffness degradation, or complex material distribution, there are problems such as large frequency drift and eigenfunction mismatch, lacking sensitivity and adaptability. And this method is based on the original stiffness matrix Based on this, a perturbation stiffness matrix composed of multiple perturbation regions is constructed and its corresponding spatial position weighting function to form a linear combination of local perturbation stiffness matrices, which is then multiplied by a perturbation adjustment factor , forming a spatially adjustable and region-sensitive perturbation stiffness combination structure, thereby modifying and dynamically enhancing the overall stiffness response characteristics. This improvement not only maintains the numerical stability of the main structure solution method, but also realizes the dynamic perception and local enhancement of the modal response region by adjusting the parameters of the perturbation adjustment factor and the spatial position weighting function , thus improving the accuracy and adaptability of the modal frequency and vibration mode in non-ideal structure scenarios. Therefore, while maintaining the existing solution logic framework, this formula realizes the deep coupling of local enhancement of the stiffness term and structural perturbation modeling. The technical solution has clear formula structure traceability, adjustable enhancement mechanism, and mathematical closed-loop under theoretical support, which is a substantial improvement to the existing modal function calculation method.

[0035] The method for performing overlapping region extraction processing on the frequency axis according to the initial frequency band set to obtain the response frequency band distribution matrix includes:[[]] According to each frequency-node corresponding set in the initial frequency band set, extract the frequency value and the node number list corresponding to this frequency, arrange the frequency values in ascending order and form an ordered structure with the node number list to generate a frequency node sequence; According to each pair of adjacent structures in the frequency node sequence, respectively extract the node number set corresponding to the previous frequency and the node number set corresponding to the next frequency, and perform a set intersection operation to obtain a repeated node set; Calculate the ratio of the number of nodes in the repeated node set to the average length of the front and back two node number sets to obtain the repeated node ratio. When the repeated node ratio is greater than the preset node ratio, calibrate the current frequency as having an overlapping relationship; According to all frequency pairs with overlapping relationships, merge the frequency values in them by continuous sections, organize all frequency value sets that satisfy the overlapping relationship into several continuous frequency sections, and generate a frequency overlapping section set; According to each frequency section in the frequency overlapping section set, count all the node numbers that fall into this frequency section, use the frequency section as the row index and the node number as the column index to construct a frequency node boolean matrix, and use the frequency node boolean matrix as the response frequency band distribution matrix.

[0036] In this embodiment, by structuring and organizing each frequency value in the initial frequency band set and its corresponding node number set, the system first constructs a frequency node sequence to establish a stable mapping relationship between the frequency value and the response space. Subsequently, the system sequentially performs an intersection operation on the node sets corresponding to any two adjacent frequency values, extracts the repeated node set, and calculates the proportion of repeated nodes based on this. If this proportion exceeds the preset threshold, it is determined that there is a spatial overlap relationship of modal responses for this frequency pair. This processing logic effectively avoids the fuzzy judgment problem caused by only using the frequency difference as the overlap criterion in the traditional method, ensuring that the identification of frequency response overlap is closer to the structural physical characteristics and has stronger interpretability and engineering applicability.

[0037] On this basis, all the frequency pairs determined to have an overlap relationship are aggregated into several continuous frequency segments, and a frequency node Boolean matrix is constructed according to the node number set corresponding to each frequency segment. This Boolean matrix uses the frequency segment as the row index and the node number as the column index, and represents the state of whether the node participates in the response of this frequency segment in a binary form, providing key inputs for subsequent execution of modal direction adjustment, frequency window control, and pitch response adjustment. Generally speaking, this step improves the ambiguity problem of the existing frequency response discrimination method through the overlap determination logic based on the node repetition ratio, and the constructed response frequency band distribution matrix has good structural resolvability, providing an executable and adjustable frequency domain structure expression form for the flexible pitch control strategy.

[0038] S30: Perform pitch target calculation processing on the wind speed sequence and the load response sequence to obtain an initial pitch angle path sequence, where the initial pitch angle path sequence is used to represent the input state of the pitch control path; The method for performing pitch target calculation processing on the wind speed sequence and the load response sequence to obtain an initial pitch angle path sequence includes: According to each wind speed measurement value in the wind speed sequence, extract the time index corresponding to the wind speed measurement value, and extract the response amplitude at the same position as the time index in the load response sequence to construct a wind-load joint sample sequence; Perform time window partitioning processing on the wind-load joint sample sequence, set a fixed window length and a sliding step size, and divide the wind-load joint sample sequence into multiple period sample paragraphs in a sliding window manner to generate a wind-load window sequence set; According to each window paragraph in the wind-load window sequence set, calculate the wind speed change gradient and the load mean deviation of all sample points in this window, and combine the wind speed change gradient and the load mean deviation into a two-dimensional adjustment factor sequence to construct an adjustment factor matrix; According to each row in the adjustment factor matrix, calculate the corresponding target pitch angle value according to the weighted summation formula of the wind speed change gradient and the load mean deviation to construct an initial pitch angle value sequence.

[0039] In this embodiment, in order to effectively combine the wind speed sequence and the load response sequence, first, according to each wind speed measurement value in the wind speed sequence, the time index of the measurement value in the time series is extracted, and the response amplitude consistent with the time index is retrieved in the load response sequence to construct a wind-load combined sample sequence. The processing principle of this step is to achieve the spatial response linkage pairing between the wind speed and the load based on time synchronization, so as to construct a sample basis with dynamic coupling characteristics and provide data support for subsequent gradient calculation and factor construction. The purpose of this step is to avoid the matching deviation caused by asynchronous sampling or response dislocation and ensure the physical response consistency between samples.

[0040] For example, assume that in the wind speed sequence, the wind speed values at the 1st to 5th moments are respectively , and the corresponding time indices are t1 to t5. In the load response sequence, the load response amplitudes at the t1 to t5 moments are respectively , then the wind-load combined sample sequence can be constructed as: .

[0041] This structure is used for subsequent sliding window division and gradient-deviation index extraction to ensure that the analysis window has the characteristics of wind-load collaborative response.

[0042] It can be understood that in order to capture the change trend and adjustment law of the wind-load response in different time periods, it is necessary to perform time window division processing on the wind-load combined sample sequence. The division method uses a fixed window length and a sliding step length for overlapping division to enhance the sample continuity and detail sensitivity. The principle is to generate sample paragraphs in multiple time periods through the sliding window method and maintain a certain time overlap to facilitate capturing the boundary change process. The purpose of this step is to divide the global sequence into local response units and enhance the time resolution and sample richness of the adjustment factor calculation.

[0043] For example, dividing the above 5 groups of wind-load samples with a window length of 3 and a sliding step length of 1, three window paragraphs can be obtained: Window 1: ; Window 2: ; Window 3: ; These windows form a wind-load window sequence set and provide sectional sample input for the construction of the adjustment factor matrix.

[0044] Further, to quantify the response characteristics within each window paragraph, it is necessary to calculate the wind speed change gradient and the deviation of the load mean for each window segment separately. The wind speed change gradient can be calculated by the difference between the maximum and minimum wind speeds in the window, reflecting the disturbance amplitude of the wind field; the load mean deviation represents the concentration degree or fluctuation trend of the load response within the window. After combining the two indicators, a two-dimensional adjustment factor sequence is formed, and its matrix form is used to construct the factor input space. The purpose of this step is to characterize the wind load co-response state with two variables, so as to realize the design of a differential pitch angle adjustment strategy in the subsequent process.

[0045] For example, the wind speed gradient in window 1 is 11.8 - 10.2 = 1.6 m / s, and the load mean deviation is (400 - 350) / 2 = 25 N. Then the adjustment factor for window 1 is (1.6, 25). By analogy, the adjustment factor matrix can be constructed as follows:

[0046] It can be understood that in this embodiment, for the sake of clarity, the adjustment factor matrix is presented in tabular form. Based on the constructed adjustment factor matrix, by performing weighted summation processing on each row, the target pitch angle value corresponding to each window can be calculated. This weighted formula considers the relative influence weights of the wind speed change gradient and the load deviation, thereby reflecting the dynamic sensitivity of the system to disturbance changes and structural responses. The design principle of the formula is to uniformly incorporate the physical change rate and the load response intensity into the pitch angle adjustment mechanism to achieve the energy-saving steady-state adjustment goal under flexible pitch control. This step dynamically generates the initial pitch angle path sequence in a structure data-driven manner, improving the system's adaptability and adjustment accuracy to the changing environment.

[0047] S40: Perform non-linear projection analysis processing on the initial pitch angle path sequence according to the response frequency band distribution matrix to obtain a mapping drift vector, where the mapping drift vector is used to represent the offset trend of the control target under flexible response; The method of performing non-linear projection analysis processing on the initial pitch angle path sequence according to the response frequency band distribution matrix to obtain a mapping drift vector includes: According to each frequency section in the response frequency band distribution matrix, extract the set of node number indexes corresponding to the frequency section, and combine the node numbers corresponding to each time point in the initial pitch angle path sequence to generate a frequency index list; For each node number in the frequency index list, extract the pitch angle values of the corresponding node in the initial pitch angle path sequence at all time points, and construct a node pitch angle time series matrix, where the row index of the node pitch angle time series matrix is the node number and the column index is the time point number; For each row in the node pitch angle time series matrix, calculate the time derivative of the pitch angle time series change curve to obtain the node response rate matrix, and classify the node response rate matrix by frequency range to generate a frequency band response set; For each frequency range in the frequency band response set, extract the response rate sequences of all nodes in this range, calculate their average response rate in the time axis direction, calculate the difference residual based on the average response rate and the corresponding time series value in the initial pitch angle path sequence, and construct a residual response matrix according to the difference residual; For each row in the residual response matrix, calculate its cumulative offset in the time direction, and arrange the cumulative offset in a vector structure according to the node number to obtain a mapping drift vector.

[0048] In this embodiment, first, for each frequency range in the response frequency band distribution matrix, extract the corresponding node number index set, and combine the node number information corresponding to each time point in the initial pitch angle path sequence to generate a frequency index list. This frequency index list is used to clarify the spatial node range covered in each frequency range and is a prerequisite structure for realizing the coupling of pitch angle response distribution and frequency structure. For each node number in the frequency index list, extract the pitch angle values of this node at all time points from the initial pitch angle path sequence, and construct a node pitch angle time series matrix with the node number as the row index and the time point number as the column index. This matrix can completely express the pitch angle time evolution trajectory at the node level and is the basic data structure for performing dynamic response modeling.

[0049] Subsequently, to capture the change trend of the node response rate, it is necessary to calculate the derivative value of the pitch angle with respect to time for each row in the node pitch angle time series matrix, so as to obtain the node response rate matrix, and classify the response rate matrix according to the frequency range defined by the response frequency band distribution matrix to generate a frequency band response set. This processing step realizes the dynamic mapping between the pitch angle change rate and the frequency range and provides a structured input for subsequent residual calculation. For each frequency range in the frequency band response set, extract the response rate sequences of all nodes in this range, calculate their average response rate along the time axis direction, and perform a difference calculation with the pitch angle value at the corresponding time point in the initial pitch angle path sequence to construct a difference residual set, and further generate a residual response matrix according to the difference residual set, so as to realize the quantitative characterization of the dynamic offset between the pitch angle path and the frequency structure.

[0050] For example, if the frequency range corresponds to the node number set , and the pitch angle values of these three nodes in the initial pitch angle path sequence at time points to are respectively: Node 1: [1.2, 1.4, 1.6, 1.7, 1.9]; Node 3: [1.1, 1.3, 1.5, 1.6, 1.8]; Node 5: [1.0, 1.2, 1.4, 1.5, 1.7]; Then the calculated response rate is: Node 1: [0.2, 0.2, 0.1, 0.2]; Node 3: [0.2, 0.2, 0.1, 0.2]; Node 5: [0.2, 0.2, 0.1, 0.2]; Taking the average value of the rates of the three nodes is: [0.2, 0.2, 0.1, 0.2]; The average difference in paddle angles in this frequency band from the paddle angle path is [0.1, 0.1, 0.1, 0.1], obtaining a row of the residual response matrix: [0.1, 0.1, 0.1, 0.1]; Finally, perform a cumulative summation operation on each row in the residual response matrix along the time direction. After sorting the cumulative offsets by node number, organize them into a one-dimensional vector structure, which is the mapping drift vector. This vector is used to characterize the overall response offset trend of the initial paddle angle path under the frequency structure mapping, thereby assisting subsequent path correction and adjustment control.

[0051] The method for calculating the node response rate matrix by calculating the time derivative of the paddle angle time series curve according to each row in the node paddle angle time series matrix includes: Extract the paddle angle change sequence for each row in the node paddle angle time series matrix, and set a fixed time interval Δt. Construct a sliding window structure in chronological order, form a sample pair for every two adjacent time points, and generate a sliding window sample sequence; According to each sample pair in the sliding window sample sequence, calculate the paddle angle difference between two adjacent time points, and divide this difference by the time interval Δt to obtain the paddle angle change rate of each node in the time interval Δt, generating a local rate set; Re - sort all the rate values in the local rate set according to the time index and fill them into positions of the same length as the original paddle angle change sequence to construct a preliminary response rate matrix; Perform edge - point compensation processing on the preliminary response rate matrix and fill in the missing points by interpolation to construct the node response rate matrix.

[0052] In this embodiment, first, the pitch angle change sequence is extracted from each row of the node pitch angle time series matrix. A fixed time interval Δt is set on the time axis, and in the form of a sliding window, every two adjacent time points are used as a sample pair to form a sliding window sample sequence. This process can achieve local modeling of the pitch angle change trend. Each group of sample pairs in the sliding window sample sequence can be used to calculate the local change rate of the pitch angle. Specifically, the pitch angle difference between two adjacent time points is divided by Δt to obtain the instantaneous change rate within each time period, and a local rate set is formed. This set initially characterizes the response speed of the pitch angle changing with time.

[0053] Subsequently, the rate values in the local rate set are re - sorted according to their corresponding time indices and filled into a position structure with the same length as the original pitch angle change sequence to obtain a preliminary response rate matrix. The main function of this step is to restore the integrity of the time axis and ensure the time alignment basis for subsequent matrix analysis.

[0054] It should be noted that there are gaps in the undefined rate values at the head and tail time points in the preliminary response rate matrix. To address this problem, this step performs edge point compensation processing on this matrix, that is, interpolation is used to fill the missing points. The interpolation method can be linear interpolation or neighboring point extension, etc. The core is to fill the edge gaps through the smooth transition of the front and back rate values, ensuring that the finally constructed node response rate matrix has continuity and differentiability on the time axis, thereby improving the expression accuracy and stability of the response trend in subsequent dynamic modeling.

[0055] The method for calculating the pitch angle difference between two adjacent time points according to each group of sample pairs in the sliding window sample sequence includes: ; In the formula, represents the pitch angle difference corresponding to the th sample pair, and respectively represent the original pitch angle values at the th and the +1th time points, β represents the non - linear amplification factor, represents the perturbation response scale factor, represents the central time point in the sliding window sample sequence, represents the exponential decay coefficient, which is used to perform weighted decay according to the distance of the sample from the central time point , is the arctangent function, constitutes a dynamic memory mechanism, is the natural constant.

[0056] In this embodiment, to ensure that the calculation of the dynamic response characteristics based on the pitch angle difference has sufficient physical interpretability and response adjustment capabilities, relevant parameters need to be clearly set: the nonlinear amplification factor β is usually set to 0.5 to enhance the influence weight of significant disturbances; the disturbance response scale factor is set to 0.1 to prevent numerical explosion caused by too small a denominator in the derivative term; the exponential decay coefficient is used to control the weight decay degree of sample points in the time window relative to the central time point, and the typical value is set to 0.01; represents the index value of the central time point in the sliding window sample sequence and serves as the reference anchor point of the dynamic memory function. Based on the traditional linear difference logic, the calculation method of the pitch angle difference introduces a nonlinear amplification mechanism, a time decay factor, and a central position perception mechanism, constituting an improved structural difference model with local enhancement ability and dynamic adjustment ability, which can more accurately describe the non-uniform disturbance characteristics in the flexible pitch path.

[0057] Compared with the difference formula of "the current time point minus the previous time point" in the traditional technology, in this embodiment, by introducing a nonlinear function, the absolute value of the pitch angle change is amplified within a specific interval, significantly improving the response sensitivity to small disturbances and effectively avoiding the problem that small signals are ignored or misjudged as noise in the conventional method. Further, this method also introduces a time distance decay factor in the difference calculation, so that the pitch angle differences farther from the current center point automatically obtain lower weights, thereby enhancing the ability to extract local disturbance characteristics and reducing the interference of strong changes at the far end on the current difference estimation result, reflecting the dynamic memory control mechanism based on time distance.

[0058] S50: Perform trajectory adjustment processing according to the mapping drift vector and the response frequency band distribution matrix to obtain the pitch angle output sequence.

[0059] In this embodiment, the method of performing trajectory adjustment processing according to the mapping drift vector and the response frequency band distribution matrix to obtain the pitch angle output sequence can be to classify and group each node offset value in the mapping drift vector according to its corresponding frequency section index in the response frequency band distribution matrix, and calculate the average offset amount of all node offset values within each frequency section to form a frequency band average offset sequence; then, according to each item in the frequency band average offset sequence, extract the pitch angle value sequence corresponding to the corresponding time period in the initial pitch angle path sequence, and perform unified offset correction processing on all values in this sequence to obtain a set of frequency band adjusted pitch angle sequences; finally, splice all the sets of frequency band adjusted pitch angle sequences in the order of their corresponding time periods into a complete pitch angle output sequence, and perform linear interpolation smoothing operations at the splicing boundaries to ensure the continuity and smoothness of the pitch angle transition, thereby completing the trajectory correction and dynamic adjustment process of the initial pitch angle path.

[0060] In this embodiment, to achieve fine adjustment of the blade angle trajectory deviation caused by flexible response, a trajectory adjustment method based on the mapping drift vector and the response frequency band distribution matrix is proposed. This method constructs a dynamic correction mechanism for the blade angle output sequence through frequency band classification and intra-segment aggregation. It can achieve flexible adjustment of the local control rhythm under the influence of multi-frequency modes while maintaining the continuity of the overall control path. It can be understood that first, the node offset values in the mapping drift vector are classified and summarized according to the frequency sections to which the nodes belong in the response frequency band distribution matrix, so that the offset responses within each frequency section can form a unified expression. Then, by calculating the average value of all node offset values within this section, a frequency band average offset sequence is constructed, thereby mapping and compressing the structural response to the frequency dimension to achieve a low-dimensional abstract expression.

[0061] Furthermore, to map the frequency band average offset back to the time dimension of the initial blade angle path sequence, the system performs unified offset correction processing on the blade angle values of each time period in the initial blade angle path sequence according to the correspondence between the frequency band and the time period, and generates a set of frequency band adjusted blade angle sequences. This processing method completes the quantitative calibration of the local response influence area without changing the overall trend of the control path, effectively eliminating the adjustment lag and target deviation problems caused by flexible response. Finally, the system stitches together all the sets of frequency band adjusted blade angle sequences in the order of their appearance on the time axis, and inserts linear transition segments at the boundaries of each stitched segment to construct a continuous and smooth complete blade angle output sequence. This step ensures that the corrected pitch control path is physically feasible at the structural level and maintains the stability and realizability of system control at the adjustment level.

[0062] Embodiment 2 Please refer to Figure 2 As shown, based on the same inventive concept, this embodiment discloses and provides a multi-source collaborative flexible pitch control system for ultra-long wind turbine blades. For the details not described in this embodiment, please refer to the relevant parts in Embodiment 1. The system includes: Data acquisition module: used to acquire blade structure parameters and environmental input data, where the blade structure parameters include the modal frequency sequence and the node coordinate distribution, and the environmental input data includes the wind speed sequence and the load response sequence; First processing module: used to perform frequency band mapping processing on the node coordinate distribution according to the modal frequency sequence to obtain an initial frequency band set, and perform overlapping area extraction processing on the frequency axis according to the initial frequency band set to obtain a response frequency band distribution matrix, where the response frequency band distribution matrix is used to characterize the overlapping intervals of the flexible response mode in the frequency dimension; Second processing module: used to perform pitch target calculation processing on the wind speed sequence and the load response sequence to obtain an initial blade angle path sequence, where the initial blade angle path sequence is used to represent the input state of the pitch control path; The third processing module: used to perform non-linear projection analysis processing on the initial blade angle path sequence according to the response frequency band distribution matrix to obtain a mapping drift vector, where the mapping drift vector is used to represent the offset trend of the control target under flexible response; The adjustment module: used to perform trajectory adjustment processing according to the mapping drift vector and the response frequency band distribution matrix to obtain a blade angle output sequence.

[0063] The detailed description set forth above in connection with the accompanying drawings describes the examples and does not represent all examples that can be implemented or fall within the scope of the claims. The terms "example" and "exemplary" as used in this specification mean "serving as an example, instance, or illustration" and do not mean "superior to or better than other examples".

[0064] The phrase "one embodiment" or "an embodiment" recited throughout this specification means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Therefore, the use of these phrases may refer to more than just one embodiment. In addition, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0065] It should also be noted that these embodiments may be described as processes depicted as flowcharts, structure diagrams, or block diagrams. Although a flowchart may describe the operations as sequential processes, many of these operations can be performed in parallel or concurrently. Additionally, the order of these operations may be rearranged.

Claims

1. A flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration, characterized in that include: Obtain blade structural parameters and environmental input data, wherein the blade structural parameters include modal frequency sequence and node coordinate distribution, and the environmental input data include wind speed sequence and load response sequence; Perform frequency band mapping processing on the node coordinate distribution according to the modal frequency sequence to obtain an initial set of frequency bands, perform overlapping area extraction processing on the frequency axis according to the initial set of frequency bands to obtain a response frequency band distribution matrix, wherein the response frequency band distribution matrix is used to characterize the overlapping interval of the flexible response mode in the frequency dimension; Performing pitch target calculation processing on the wind speed sequence and the load response sequence to obtain an initial blade angle path sequence, wherein the initial blade angle path sequence is used to represent the input state of the pitch control path; A nonlinear projection analysis is performed on the initial blade angle path sequence according to the response frequency band distribution matrix to obtain a mapping drift vector, wherein the mapping drift vector is used to represent the deviation trend of the control target under the flexible response; The trajectory adjustment processing is performed according to the mapped drift vector and the response frequency band distribution matrix to obtain the blade angle output sequence.

2. The flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 1, characterized in that, The method of performing frequency band mapping processing on the node coordinate distribution according to the modal frequency sequence to obtain the initial set of frequency bands includes: According to each modal frequency in the modal frequency sequence, a set of vibration mode function values of the order corresponding to the modal frequency is retrieved from a preset simulation database, and the vibration mode amplitude in each vibration mode function value set is aligned with the node number in the corresponding node coordinate distribution to generate a modal response amplitude matrix; According to each column in the modal response amplitude matrix, the arithmetic mean of the response amplitudes of all nodes in each column is calculated. For the response amplitude of each node, the ratio of the response amplitude to the arithmetic mean is calculated to obtain the response ratio. When the response ratio is greater than the preset ratio, the node number and the current modal frequency are combined into a frequency-node pair, and all frequency-node pairs are aggregated to form a frequency-node corresponding set. Extract all node number lists corresponding to each modal frequency from the frequency-node correspondence set, sort the node numbers in each node number list from small to large according to the axial position in the node coordinate distribution, and generate a frequency-node sorting result; Two adjacent modal frequencies are selected in order from the frequency-node sorting results, the node number sequences corresponding to the two modal frequencies are extracted, and the number of repeated nodes in the two node number sequences is counted. When the ratio of the number of repeated nodes to the total number of nodes in the two sequences is greater than the preset node ratio, the frequency-node corresponding set is marked as having modal overlap, and all frequency-node corresponding sets with modal overlap are aggregated to generate an initial set of frequency bands.

3. A flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 2, characterized in that The method for constructing the preset simulation database includes: Performing meshing processing on the preset blade structure design model, dividing the blade into multiple structural units along the axial direction from the root to the tip, and setting the node number and initial geometric parameters for each structural unit; Perform material parameter assignment processing on each structural unit, assign elastic modulus, density and damping ratio to it according to the blade design input, and form a finite element model with complete parameters; Performing modal simulation processing on the finite element model with completed parameter assignment to obtain the modal frequencies of each order and the corresponding vibration shape function values of the finite element model under the set boundary conditions; Convert the modal frequencies and corresponding eigenfunction values of each order into structured data entries, and bind the number identifier of the finite element model and the boundary condition parameters to each structured data entry; Establish an index table for all structured data entries according to the modal number to form a simulation database that can be quickly retrieved by modal number.

4. A flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 3, characterized in that, The methods for obtaining the modal frequencies of each order and the corresponding eigenfunction values of the finite element model under the set boundary conditions include: ; In the formula, represents the modal shape function value of the -th order modal frequency, represents the original stiffness matrix of the finite element model, represents the perturbation adjustment factor, represents the linear combination of local perturbation stiffness matrices composed of spatial position weighting functions, represents the position at which the -th spatial position weighting function of the -th perturbation region, represents the perturbation stiffness matrix corresponding to the -th perturbation sub-segment, represents the inverse matrix operation of the combined stiffness matrix, represents the square value of the -th order modal frequency, represents the mass matrix of the model.

5. A flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 4, characterized in that The methods for performing overlapping region extraction processing on the frequency axis according to the initial frequency band set to obtain the response frequency band distribution matrix include: According to each frequency-node corresponding set in the initial frequency band set, extract the frequency value corresponding to this frequency and the node number list, arrange the frequency values in ascending order and form an ordered structure with the node number list to generate a frequency node sequence; According to each pair of adjacent structures in the frequency node sequence, respectively extract the node number set corresponding to the previous frequency and the node number set corresponding to the next frequency, and perform a set intersection operation to obtain a repeated node set; Calculate the ratio of the number of nodes in the repeated node set to the average length of the previous and next node number sets to obtain the repeated node ratio. When the repeated node ratio is greater than the preset node ratio, calibrate the current frequency as having an overlapping relationship; According to all frequency pairs with overlapping relationships, perform a merging process on the frequency values therein by continuous sections, organize all frequency value sets that satisfy the overlapping relationship into several continuous frequency sections, and generate a frequency overlapping section set; According to each frequency section in the frequency overlapping section set, count all the node numbers falling into this frequency section, use the frequency section as the row index and the node number as the column index to construct a frequency node boolean matrix, and use the frequency node boolean matrix as the response frequency band distribution matrix.

6. The flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 3, characterized in that The methods for performing pitch target calculation processing on the wind speed sequence and the load response sequence to obtain the initial pitch angle path sequence include: According to each wind speed measurement value in the wind speed sequence, extract the time index corresponding to the wind speed measurement value, and extract the response amplitude at the same position as this time index in the load response sequence to construct a wind load joint sample sequence; Perform a time window division process on the wind load joint sample sequence, set a fixed window length and a sliding step, and divide the wind load joint sample sequence into multiple time period sample paragraphs in a sliding window manner to generate a wind load window sequence set; According to each window paragraph in the wind load window sequence set, calculate the wind speed change gradient and the load mean deviation of all sample points in this window, and combine the wind speed change gradient and the load mean deviation into a two-dimensional adjustment factor sequence to construct an adjustment factor matrix; According to each row in the adjustment factor matrix, calculate the corresponding target pitch angle value according to the weighted summation formula of the wind speed change gradient and the load mean deviation to construct an initial pitch angle value sequence.

7. A flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 6, characterized in that, The methods for performing non-linear projection analysis processing on the initial pitch angle path sequence according to the response frequency band distribution matrix to obtain the mapping drift vector include: According to each frequency section in the response frequency band distribution matrix, extract the node number index set corresponding to the frequency section, and combine the node numbers corresponding to each time point in the initial pitch angle path sequence to generate a frequency index list; Number each node in the frequency index list, extract the pitch angle values of the corresponding nodes in the initial pitch angle path sequence at all time points, and construct a node pitch angle time series matrix, where the row index of the node pitch angle time series matrix is the node number and the column index is the time point number; According to each row in the node pitch angle time series matrix, calculate the time derivative of the pitch angle time series change curve to obtain a node response rate matrix, and classify the node response rate matrix by frequency bands to generate a frequency band response set; For each frequency band in the frequency band response set, extract the response rate sequences of all nodes in this band, calculate their average response rate in the time axis direction, calculate the difference residual based on the average response rate and the corresponding time series value in the initial pitch angle path sequence, and construct a residual response matrix according to the difference residual; According to each row in the residual response matrix, calculate its cumulative offset in the time direction, and arrange the cumulative offset in a vector structure according to the node number to obtain a mapping drift vector.

8. A flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 7, characterized in that, The method of calculating the node response rate matrix by calculating the time derivative of the pitch angle time series change curve according to each row in the node pitch angle time series matrix includes: Extract the pitch angle change sequence for each row in the node pitch angle time series matrix, set a fixed time interval Δt, construct a sliding window structure in time order, and form a sample pair for every two adjacent time points to generate a sliding window sample sequence; According to each sample pair in the sliding window sample sequence, calculate the pitch angle difference between two adjacent time points, and divide the difference by the time interval Δt to obtain the pitch angle change rate of each node in the time interval Δt, and generate a local rate set; Reorder all the rate values in the local rate set according to the time index and fill them in the positions with the same length as the original pitch angle change sequence to construct a preliminary response rate matrix; Perform edge point compensation processing on the preliminary response rate matrix and fill in the missing points by interpolation to construct a node response rate matrix.

9. A flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to claim 8, characterized in that, The method of calculating the pitch angle difference between two adjacent time points according to each sample pair in the sliding window sample sequence includes: ； In the formula, represents the paddle angle difference corresponding to the th sample pair, and respectively represent the original paddle angle values at the th and the +1th time points. β represents the non-linear amplification factor, represents the disturbance response scale factor, represents the central time point in the sliding window sample sequence, represents the exponential decay coefficient, which is used to perform weight decay according to the distance of the sample from the central time point , is the arctangent function, constitutes the dynamic memory mechanism, is the natural constant.

10. A flexible pitch control system for ultra-long wind turbine blades with multi-source collaboration, which is used to implement a flexible pitch control method for ultra-long wind turbine blades with multi-source collaboration according to any one of claims 1-9, characterized in that, including: Data acquisition module: used to acquire blade structure parameters and environmental input data, where the blade structure parameters include a modal frequency sequence and a node coordinate distribution, and the environmental input data includes a wind speed sequence and a load response sequence; The first processing module: used to perform frequency band mapping processing on the node coordinate distribution according to the modal frequency sequence to obtain an initial frequency band set, and perform overlapping area extraction processing on the frequency axis according to the initial frequency band set to obtain a response frequency band distribution matrix, where the response frequency band distribution matrix is used to characterize the overlapping interval of the flexible response mode in the frequency dimension; The second processing module: used to perform pitch change target calculation processing on the wind speed sequence and the load response sequence to obtain an initial pitch angle path sequence, where the initial pitch angle path sequence is used to represent the input state of the pitch change control path; The third processing module: used to perform non-linear projection analysis processing on the initial pitch angle path sequence according to the response frequency band distribution matrix to obtain a mapping drift vector, where the mapping drift vector is used to represent the offset trend of the control target under flexible response; Adjustment module: used to perform trajectory adjustment processing according to the mapping drift vector and the response frequency band distribution matrix to obtain the paddle angle output sequence.

Images