Wind power blade flexible pitch control system based on adaptive fuzzy control

Abstract

The invention relates to the technical field of new energy power generation, in particular to a wind power blade flexible variable pitch system based on adaptive fuzzy control, which comprises an acquisition module used for acquiring real-time strain data of a blade waving direction and a shimmy direction through a fiber grating sensor mounted at the root of a wind power blade; the mirroring algorithm module is used for applying mirroring algorithm processing to the real-time strain data to generate a symmetric compensation data set for eliminating unilateral deviation; and the convex hull algorithm module is used for extracting strain envelope feature points representing a load distribution mode through a convex hull algorithm based on the symmetric compensation data set. According to the method, through perception-analysis-decision-execution-anti-interference full-link optimization, the response speed of the wind power blade variable-pitch system is increased, and the blade load impact is effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation technology, and more specifically, to a flexible pitch control system for wind turbine blades based on adaptive fuzzy control. Background Technology

[0002] Wind power generation is crucial in the energy transition, and the pitch control performance of wind turbine blades directly affects the stability, efficiency, and lifespan of the unit. However, current mainstream pitch systems employ rigid mechanisms combined with PID or simple fuzzy control, which can meet the needs of onshore wind farms with stable wind speeds. But as wind power develops towards larger capacities (6MW and above) and deep-sea applications, their limitations become apparent. They may struggle to cope with nonlinear loads under extreme conditions such as typhoons, strong gusts, and low-temperature icing at sea, potentially leading to increased operational risks and decreased efficiency.

[0003] A case study of a 7MW offshore wind power cluster shows that traditional systems have prominent problems in strong gusts during typhoons (wind speed increases from 15m / s to 28m / s within 10 minutes): resistance strain gauges have a slow response (20ms) and sometimes cannot capture sudden strain changes; PID controller parameters are fixed, resulting in lag in pitch adjustment; when blades are icing, the adjustment error may reach ±3.5° due to the lack of adaptation rules; strong electromagnetic interference causes signal distortion, and the final blade flapping amplitude may exceed the standard (±2.8 meters), causing an 8-hour shutdown and potentially resulting in microcracks.

[0004] This reflects the following limitations of traditional systems: insufficient dynamic response of sensors (delay of 15-20ms), poor generalization of control rules (error exceeds ±2.5° when icing), reduced efficiency of actuators (30% decrease in response at low temperatures and 40%-60% increase in load impact), and weak resistance to electromagnetic interference (signal-to-noise ratio ≤20dB). Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a flexible pitch control system for wind turbine blades based on adaptive fuzzy control. Through the full-link optimization of perception-analysis-decision-execution-anti-interference, the response speed of the wind turbine blade pitch control system is improved and the blade load impact is effectively reduced.

[0006] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is as follows: A flexible pitch control system for wind turbine blades based on adaptive fuzzy control includes: The acquisition module is used to acquire real-time strain data of the blade flapping and oscillation directions through fiber optic grating sensors installed at the root of the wind turbine blade; The mirror algorithm module is used to apply the mirror algorithm to real-time strain data to generate a symmetrical compensation dataset that eliminates one-sided deviations. The convex hull algorithm module is used to extract strain envelope feature points that characterize the load distribution pattern based on the symmetric compensation dataset using the convex hull algorithm. The fuzzy control module is used to input the strain envelope feature points into the adaptive fuzzy controller, dynamically reconstruct the membership relationship features and generate pitch angle adjustment commands; at the same time, based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment commands, the fuzzy rule base is expanded in real time in combination with the blade surface condition monitoring results. The matching module is used to input the pitch angle adjustment command into the coordination mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern. The temperature feedback module is used to start phase change material heating and dynamically adjust current parameters to maintain torque output efficiency based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper. The electromagnetic shielding module is used to deploy multi-level electromagnetic shielding layers to suppress electromagnetic interference in the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision and rule base expansion, buffer torque value and temperature feedback adjustment.

[0007] Secondly, a control method for a flexible pitch system of wind turbine blades based on adaptive fuzzy control includes the following steps: Real-time strain data of the blade flapping and oscillation directions are collected by fiber optic grating sensors installed at the root of the wind turbine blades. A mirror algorithm is applied to real-time strain data to generate a symmetrical compensation dataset that eliminates unilateral bias. Based on the symmetric compensation dataset, strain envelope feature points characterizing the load distribution pattern are extracted using the convex hull algorithm; The strain envelope feature points are input into an adaptive fuzzy controller to dynamically reconstruct the membership relationship features and generate pitch angle adjustment commands; at the same time, the fuzzy rule base is expanded in real time based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment commands, combined with the blade surface condition monitoring results. The pitch angle adjustment command is input into the coordination mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern. Based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper, the phase change material heating is initiated and the current parameters are dynamically adjusted to maintain torque output efficiency. In the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision-making and rule base expansion, buffer torque value and temperature feedback adjustment, multi-level electromagnetic shielding layers are deployed to suppress electromagnetic interference.

[0008] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: By using a pre-defined spatial distribution matrix of a fiber optic grating sensor array, combined with time-domain alignment and temperature drift correction algorithms, strain data in the blade flapping / wobbling direction can be acquired in real time and synchronously, reducing data transmission latency and improving the accuracy of resistance strain gauges. A load validity verification mechanism automatically identifies and replaces abnormal data, ensuring the authenticity of the input data. A mirror algorithm eliminates unilateral load deviations through symmetrical compensation, improving the symmetry of strain data. A convex hull algorithm, combined with gradient sensitivity coefficients, locates key load areas, eliminating redundant data and improving feature point extraction efficiency. An adaptive fuzzy controller can dynamically reconstruct membership features, shortening the pitch angle adjustment response time for sudden load mode changes. The fuzzy rule base can be expanded in real time according to special operating conditions such as icing, enhancing the system's adaptability to complex environments. Coordinated control of the magnetorheological damper and servo motor can output a buffer torque matched to the load. Multi-level electromagnetic shielding layers, optimized through targeted spectrum analysis and electromagnetic field simulation, improve the signal-to-noise ratio and reduce signal distortion in strong electromagnetic environments. Dynamic grounding impedance adjustment ensures the integrity of the entire signal chain, reducing the pitch command anomaly rate and guaranteeing the unit's continuous and stable operation in harsh environments such as thunderstorms and salt spray.

[0009] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0010] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. Some specific embodiments of this application will be described in detail below with reference to the accompanying drawings in an exemplary and non-limiting manner. The same reference numerals in the drawings designate the same or similar parts or components. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a schematic diagram of the flexible pitch control system for wind turbine blades based on adaptive fuzzy control, as described in this invention.

[0011] Figure 2 This is a schematic diagram of the control method of the wind turbine blade flexible pitch system based on adaptive fuzzy control according to the present invention. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort should fall within the scope of protection of the present application.

[0013] The following embodiments of this application use a flexible pitch system for wind turbine blades based on adaptive fuzzy control as an example to illustrate the solution of this application in detail. However, this embodiment does not limit the scope of protection of this application.

[0014] like Figure 1 As shown, this invention provides a flexible pitch control system for wind turbine blades based on adaptive fuzzy control, comprising: The acquisition module is used to acquire real-time strain data of the blade flapping and oscillation directions through fiber optic grating sensors installed at the root of the wind turbine blade; The mirror algorithm module is used to apply the mirror algorithm to real-time strain data to generate a symmetrical compensation dataset that eliminates one-sided deviations. The convex hull algorithm module is used to extract strain envelope feature points that characterize the load distribution pattern based on the symmetric compensation dataset using the convex hull algorithm. The fuzzy control module is used to input the strain envelope feature points into the adaptive fuzzy controller, dynamically reconstruct the membership relationship features and generate pitch angle adjustment commands; at the same time, based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment commands, the fuzzy rule base is expanded in real time in combination with the blade surface condition monitoring results. The matching module is used to input the pitch angle adjustment command into the coordination mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern. The temperature feedback module is used to start phase change material heating and dynamically adjust current parameters to maintain torque output efficiency based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper. The electromagnetic shielding module is used to deploy multi-level electromagnetic shielding layers to suppress electromagnetic interference in the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision and rule base expansion, buffer torque value and temperature feedback adjustment.

[0015] In this embodiment of the invention, real-time strain data of the blade root flapping and oscillation directions are collected by fiber optic grating sensors, enabling real-time capture of dynamic load changes of the blade under complex operating conditions. Symmetrical compensation datasets effectively eliminate data distortion caused by unilateral load deviations. Strain envelope feature points can extract key features characterizing the load distribution pattern, simplifying data dimensions while retaining core load information. Dynamically reconstructing membership features and generating pitch angle adjustment commands enables adaptive control of blade load changes, ensuring a high degree of matching between pitch action and actual operating conditions. Obtaining a buffer torque value matching the load distribution pattern allows the synergistic effect of the magnetorheological damper and servo motor to better meet the real-time stress requirements of the blade, effectively buffering the impact of gusts or sudden load changes on the blade. Maintaining torque output efficiency through phase change material heating and dynamic adjustment of current parameters solves the performance degradation problem of the magnetorheological damper caused by temperature changes. Deploying multi-level electromagnetic shielding layers in the closed-loop signal transmission path suppresses interference from the strong electromagnetic environment at sea, ensuring the integrity and accuracy of strain data, control commands, and other signals.

[0016] In the wind turbine blade flexible pitch system based on adaptive fuzzy control described in this embodiment of the invention, real-time strain data of the blade flapping direction and yaw direction are collected by a fiber optic grating sensor installed at the root of the wind turbine blade, including: Step 111: Activate the fiber optic grating sensor array according to the preset spatial distribution matrix, synchronously collect the original light intensity attenuation value, and calculate the initial strain components of each sensor node in the waving direction and the swinging direction. Step 112: Perform time-domain alignment calculation on the initial strain components, compensate for the signal transmission delay of different sensor nodes by sliding time window, and generate a time-synchronized strain data sequence. Step 113: Based on the calibration curve, the amplitude correction calculation is performed on the time-synchronized strain data sequence. The cubic spline interpolation method is used to eliminate the nonlinear error caused by the ambient temperature drift, and the preprocessed strain data is obtained. Step 114: Perform load validity verification calculation on the preprocessed strain data. When the strain ratio in the swing direction to the oscillation direction exceeds the preset safety threshold, trigger the sensor self-test protocol and replace it with historical average data, and finally generate real-time strain data.

[0017] In this embodiment of the invention, activating the sensor array according to a preset spatial distribution matrix and synchronously acquiring data ensures the synchronicity and comprehensiveness of the initial strain component acquisition of each sensor node in the flapping and oscillating directions. This provides complete and consistent basic data for subsequent data processing, ensuring accurate initial perception of the blade strain state. Compensating for signal transmission delay through a sliding time window eliminates time differences caused by transmission path variations between different sensor nodes, keeping the generated strain data sequence synchronized in the time dimension and avoiding analysis errors caused by data asynchrony. Amplitude correction based on the calibration curve and the use of cubic spline interpolation to eliminate nonlinear errors caused by temperature drift effectively reduce the interference of ambient temperature on the strain data. Performing load validity verification calculations and triggering self-checks and data replacement when the strain ratio exceeds a threshold can promptly detect and process abnormal data, ensuring that the final output real-time strain data is authentic and valid. This ensures that the system performs subsequent operations based on accurate data, improving the safety and stability of system operation.

[0018] In this embodiment of the invention, when applied in a specific way, it can be implemented through the following technical solutions, for example: Step 111 above, the "pre-setting" of the spatial distribution matrix, means that according to the structure, stress, and pitch control requirements of the wind turbine blade root, the installation position and arrangement of the fiber Bragg grating sensors are planned in advance to form a layout covering the key monitoring area. The process is as follows: Establish a three-dimensional model of the root and combine it with finite element simulation to determine the strain-sensitive areas (upper and lower edges, left and right edges) in the swinging and oscillating directions, and divide them into monitoring sub-regions; adjust the density according to the strain gradient, set one every 10cm in the area of ​​severe strain (such as the root-hub transition section), and set one every 20-30cm in the area of ​​gentle strain (such as the middle section of the root); place 8-16 in a single root to fully cover the strain in both directions.

[0019] A three-dimensional coordinate system is established with the center of the root section as the origin (Y-axis corresponds to waving and X-axis corresponds to swinging), and the coordinates of each node are marked to ensure that the axis of the sensor grating is consistent with the monitoring direction; 1-2 redundant nodes are set for key monitoring points to form a "main + redundant" structure; ensure that the signal transmission of each node is interference-free and can be triggered synchronously through the same module; verify the layout effect through scaled-down model tests or static loading tests, adjust the node position or density for monitoring blind spots or low signal-to-noise ratio areas, and determine the final scheme.

[0020] Activate the fiber Bragg grating sensor array arranged according to a preset spatial distribution matrix and obtain the original light intensity attenuation value of each sensor node; based on the strain-light intensity attenuation characteristics of the fiber Bragg grating sensor, convert the original light intensity attenuation value into the initial strain components of each node in the waving direction and the swinging direction.

[0021] In step 112 above, the initial strain components output by each sensor node are time-domain aligned based on time; a sliding time window is set, and the delay time of each node signal is determined according to the signal transmission path length and transmission speed of different sensor nodes. The delay signal is compensated by adjusting the window to obtain a time-synchronized strain data sequence.

[0022] In step 113 above, the pre-calibration of the strain-temperature characteristic curve is carried out in the laboratory using a fiber optic grating sensor at the root of the wind turbine blade. This simulates actual temperature changes and strain conditions, establishing the correspondence between temperature and strain measurements. The specific steps are as follows: The sensor was fixed on a specimen made of the same material as the blade root, and the installation method was the same as the actual installation. It was placed in a constant temperature chamber equipped with a reference strain gauge and a temperature sensor. The range covering the actual working temperature was set, and the temperature was adjusted according to the preset step size. Each temperature point was stabilized for a sufficient time to ensure that the specimen and the ambient temperature were balanced. When the specimen was not under external load, the light intensity signal of the sensor at each temperature point was recorded to determine the basic impact of temperature drift on the output. The temperature was fixed, and a known gradient strain was applied to the specimen. The sensor measurement value and the reference value were recorded. The operation was repeated at multiple key temperature points to eliminate the measurement error of the sensor itself.

[0023] By integrating the data and fitting a characteristic curve with temperature on the horizontal axis and strain measurement deviation on the vertical axis, a basis for amplitude correction is provided. The error is verified using temperature points and strain values ​​that were not involved in the fitting to ensure that the error is within the preset range. The sensor needs to be recalibrated after long-term use or changes in operating conditions.

[0024] The amplitude of the time-synchronized strain data sequence is corrected using the pre-calibrated strain-temperature characteristic curve. The cubic spline interpolation method is used to correct the nonlinear error caused by temperature drift according to the changes in ambient temperature, so as to obtain the pre-processed strain data.

[0025] In step 114 above, the strain ratio between the swinging direction and the oscillation direction in the preprocessed strain data is calculated and compared with a preset safety threshold. If the threshold is exceeded, the sensor self-check is initiated, and the historical average data of the sensor node is used to replace the current data. If the threshold is not exceeded, the preprocessed strain data is directly output as real-time strain data.

[0026] In the wind turbine blade flexible pitch system based on adaptive fuzzy control described in this embodiment of the invention, a mirror algorithm is applied to the real-time strain data to generate a symmetrical compensation dataset that eliminates unilateral deviations, including: Step 121: Based on real-time strain data, a sliding time window is used to extract the data according to the blade rotation phase angle, and strain data slices in the flapping direction and oscillation direction within the same cycle are calculated. Step 122: Perform centerline fitting calculation on the strain data slices, and use the least squares method to fit the strain distribution curve of the blade neutral layer to generate a reference symmetry axis. Step 123: Perform mirror symmetry compensation calculation based on the reference symmetry axis. Map the strain values ​​of the windward and leeward sides of the blade equidistantly relative to the reference axis, solve for the difference in compensation amount, and superimpose it onto the original data to obtain the compensated strain data. Step 124: Perform dynamic weight fusion calculation on the compensated strain data, assign weight coefficients according to the blade section position, and generate a symmetrical compensation dataset to eliminate unilateral deviation.

[0027] In this embodiment of the invention, strain data of the same period are extracted using a sliding time window according to the blade rotation phase angle, and strain data slices in the flapping and oscillation directions are separated. This ensures that the processed data comes from the same period of blade motion, avoiding analysis deviations caused by different periods. The least squares method is used to fit the strain distribution curve of the blade neutral layer to generate a reference axis of symmetry, which establishes an objective and accurate symmetry reference standard. This axis can truly reflect the neutral distribution state of the blade strain, providing a reliable benchmark for subsequent mirror symmetry compensation calculations and ensuring the rationality of the compensation direction and amplitude. Based on the reference axis of symmetry, the strain values ​​on the windward and leeward sides of the blade are mapped at equal intervals and the compensation difference is superimposed, which can effectively eliminate the unilateral deviation caused by uneven force on both sides of the blade, making the compensation strain data closer to the symmetrical state in distribution, improving the symmetry and accuracy of the data. According to the weight coefficients assigned to the blade section position, the compensation strain data is dynamically weighted and fused, which can highlight the influence of the strain data of key sections and weaken the interference of secondary section data, so that the generated symmetrical compensation dataset can accurately reflect the overall strain distribution of the blade.

[0028] In this embodiment of the invention, when applied in a specific way, it can be implemented through the following technical solutions, for example: In step 121 above, the start and end times and duration of the sliding time window are set based on the blade rotation phase angle, and the portion of the real-time strain data corresponding to the same rotation cycle is extracted; from the extracted data, the strain data segments of the flapping direction and the oscillation direction are separated to form strain data slices.

[0029] In step 122 above, the least squares method is used to calculate the strain values ​​in the strain data slices; with the theoretical position of the blade neutral layer as a reference, a curve that reflects the strain distribution trend is determined by fitting, and this curve is the reference axis of symmetry.

[0030] In step 123 above, the positions of the strain values ​​on the windward and leeward sides of the blade relative to the reference axis of symmetry are determined. The difference in compensation caused by the deviation of the strain values ​​on both sides from the symmetrical position is calculated by means of equidistant mapping. This difference is added to the original strain data to obtain the compensated strain data and output it.

[0031] In step 124 above, based on the importance of different cross-sectional positions of the blade during the stress process and the range of strain influence, corresponding weighting coefficients are assigned to the post-compensation strain data of each cross-section. The post-compensation strain data of each cross-section are then fused and calculated according to these weighting coefficients to ultimately generate a symmetrical compensation dataset that eliminates unilateral bias.

[0032] In the wind turbine blade flexible pitch system based on adaptive fuzzy control described in this embodiment of the invention, based on the symmetric compensation dataset, strain envelope feature points characterizing the load distribution pattern are extracted using the convex hull algorithm, including: Step 131: Perform strain gradient preprocessing calculation on the symmetric compensation dataset, divide the data into segments according to the blade spanwise position, and calculate the ratio of strain dispersion to mean of each segment as the gradient sensitivity coefficient. Step 132: Perform dynamic convex hull construction calculation based on gradient sensitivity coefficient, select the segment with sensitivity coefficient exceeding the preset threshold as key region, and connect the strain extreme points in the region to form the initial convex hull. Step 133: Based on the initial convex hull, redundant vertices are removed by a geometric curvature analysis algorithm, and inflection points with large curvature changes are retained as candidate feature points. Step 134: Input the candidate feature points into the load pattern matching calculation, compare them with the envelope template of the historical working condition feature library, and select feature points with matching degree reaching the preset standard as strain envelope feature points.

[0033] In this embodiment of the invention, by dividing the area into segments and calculating the gradient sensitivity coefficient, the region of strain change can be accurately located, providing targeted data for subsequent convex hull construction and improving the efficiency of feature extraction. Focusing on key regions to construct the initial convex hull can reduce interference from irrelevant data, enabling the convex hull to reflect the main characteristics of the load distribution, laying the foundation for subsequent feature point extraction. After removing redundant vertices, candidate feature points can accurately reflect the key shape of the convex hull, improving the representativeness of the feature points and facilitating subsequent load pattern matching. By comparing and filtering with historical templates, it can be ensured that the strain envelope feature points conform to the actual load distribution pattern, improving the reliability of the feature points.

[0034] In this embodiment of the invention, when applied in a specific way, it can be implemented through the following technical solutions, for example: In step 131 above, for the symmetric compensation dataset, the data is divided into different segments according to the blade spanwise position; the dispersion and mean of the strain data in each segment are calculated respectively, and the ratio of dispersion to mean is used as the gradient sensitivity coefficient of that segment to reflect the degree of strain change at each spanwise position.

[0035] In step 132 above, based on the obtained gradient sensitivity coefficient, a preset threshold is set; segments with sensitivity coefficients exceeding the threshold are selected and identified as key regions; extreme points such as the maximum and minimum strain points are found within the key regions, and then these extreme points are connected sequentially to form the initial convex hull.

[0036] In step 133 above, based on the initial convex hull, a geometric curvature analysis algorithm is used to analyze each vertex of the convex hull; the curvature magnitude at each vertex is calculated, redundant vertices with small curvature changes and little impact on the overall contour are eliminated, and only those inflection points with obvious curvature changes are retained, and these inflection points are used as candidate feature points.

[0037] In step 134 above, the envelope template in the historical operating condition feature library is a standardized feature set constructed based on actual operating data of wind turbine blades under various typical operating conditions, specifically defined as follows: The core data of the envelope template comes from the measured strain data of the blade under different operating conditions (such as normal wind speed, gusts, extreme wind speed, blade icing, different wind direction angles, etc.). By extracting convex hull features from these historical data, a set of strain envelope feature points under the corresponding operating conditions is obtained, which serves as the basis of the template. Each envelope template must clearly record the key parameters of the corresponding operating condition (such as wind speed range, temperature range, blade operating status, etc.), and also include the spatial coordinates (position along the blade spanwise, strain value) of the strain envelope feature points under that operating condition, the connection relationship between feature points (such as curvature change law), and the distribution density of feature points, forming a complete feature contour description.

[0038] The envelope templates are categorized according to operating conditions (such as "normal operation template", "gust impact template", "low temperature icing template", etc.), and each template is assigned a unique index identifier to facilitate quick retrieval of the matching template based on real-time operating condition parameters. New measured operating condition data are regularly incorporated to verify and correct existing templates. When new operating conditions that are not covered (such as strain modes under special meteorological conditions) appear, corresponding envelope templates are added to ensure the comprehensiveness and timeliness of the feature library.

[0039] The candidate feature points are input into the load pattern matching calculation, and the envelope template in the historical working condition feature library is called. The candidate feature points are compared with the envelope template, the degree of matching between them is calculated, and the feature points that meet the preset standard are selected. These feature points are the final strain envelope feature points.

[0040] In the wind turbine blade flexible pitch system based on adaptive fuzzy control described in this embodiment of the invention, the strain envelope feature points are input into the adaptive fuzzy controller to dynamically reconstruct the membership relationship features and generate pitch angle adjustment commands; simultaneously, based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment commands, the fuzzy rule base is expanded in real time in conjunction with the blade surface condition monitoring results, including: Step 141: Based on the strain envelope feature points, the key dimension parameters of the geometric shape formed by the feature points are obtained through spatial parameterization measurement, and a feature vector representing the load distribution pattern is generated. Step 142: When the feature vector deviates from the reference template by more than a preset threshold, the fuzzy set center position parameter and distribution width parameter are dynamically adjusted according to the degree of deviation to obtain the reconstructed membership relationship features; Step 143: Input the membership relationship features into fuzzy inference calculation, activate the fuzzy rule base to generate the pitch angle adjustment command, and simultaneously generate the confidence parameter characterizing the reliability of the control decision; Step 144: When the confidence parameter is lower than the preset confidence threshold, an abnormal load distribution pattern signal is triggered. Combined with the icing coverage data of the blade surface condition monitoring, a parameter characterizing the severity of icing is calculated. Step 145: Perform fuzzy rule expansion calculation based on the parameters characterizing the severity of icing, add a pitch angle adjustment amplitude enhancement rule that matches the severity of icing, and update the real-time rule library.

[0041] In this embodiment of the invention, abstract feature points are transformed into specific feature vectors, which facilitates subsequent fuzzy control processing and can more accurately reflect the load distribution pattern. This allows the membership features to adapt to changes in the load distribution pattern, improving the adaptability and accuracy of fuzzy control. It also enables the rapid generation of pitch angle adjustment commands that conform to the current load conditions, and the reliability of the decisions can be intuitively understood through confidence parameters, providing a reference for subsequent processing. Furthermore, it allows for the timely detection of abnormal load distribution patterns and the quantification of icing severity based on icing data, providing a basis for rule base expansion. Finally, it enables the fuzzy rule base to adapt to special operating conditions such as blade icing, improving the system's ability to cope with complex environments and enhancing the effectiveness of pitch control.

[0042] In this embodiment of the invention, when applied in a specific way, it can be implemented through the following technical solutions, for example: In step 141 above, for the strain envelope feature points, a spatial parameterization measurement method is used to determine the key dimensional parameters of the geometric figure formed by these feature points, such as the length and width of the figure, the distance between each feature point, etc.; these key dimensional parameters are integrated to form a feature vector that can characterize the load distribution pattern.

[0043] In step 142 above, the feature vector is compared with the benchmark template to calculate the degree of deviation. When the degree of deviation exceeds the preset threshold, the center position parameter and distribution width parameter of the fuzzy set are dynamically adjusted according to the specific deviation to obtain the reconstructed membership features.

[0044] In step 143 above, the rules of the fuzzy rule base are formulated based on the actual needs and historical operating data of wind turbine blade pitch control. They are used to guide the reasoning process from membership relationship features to pitch angle adjustment commands, and are specifically defined as follows: Each rule consists of two parts: preconditions and conclusions. The preconditions are based on key parameters in the membership relationship characteristics (such as the distribution density of strain envelope feature points, geometric size deviation, etc.) and are described by fuzzy linguistic variables (such as "large", "medium", "small", "drastic", "gradual"). The conclusions specify the specific direction (such as increasing, decreasing) and range (such as fine adjustment, moderate adjustment, large adjustment) of the strain pitch angle adjustment. They are formulated based on the load distribution pattern of the blades under different operating conditions, combined with the safety threshold of pitch control (such as the maximum allowable pitch angle, minimum response time) and optimization objectives (such as minimizing load impact and maximizing power generation efficiency). For example, when "large and drastic deviation in strain distribution" is detected, the rule conclusion will point to "rapidly increase the pitch angle to a moderate range".

[0045] Rules are categorized by operating condition type into routine operating condition rules (such as fine-tuning rules under normal wind speed) and special operating condition rules (such as emergency adjustment rules during gusts or icing). Priority levels exist between rules, with special operating condition rules having higher priority than routine rules, ensuring that safety-oriented adjustment instructions are executed first in extreme situations. Multiple rules may correspond to the same input feature; priority is differentiated by setting confidence weights, with weight values ​​determined based on historical control effectiveness assessments. When rule conclusions conflict, a weighted fusion algorithm is used to select the adjustment instruction with the highest overall confidence, avoiding decision-making contradictions. Rule extension interfaces are reserved, allowing for the addition or modification of rules based on new operating conditions (such as new load distribution patterns), and new rules must be compatible with the existing rule system to ensure consistency of reasoning logic.

[0046] The reconstructed membership features are input into the fuzzy inference calculation, which calls and activates the relevant rules in the fuzzy rule base and generates the pitch angle adjustment command based on the rules. At the same time, during the command generation process, the confidence parameter used to characterize the reliability of the control decision is calculated and output synchronously.

[0047] In step 144 above, the confidence parameter of the output is monitored. When the parameter is lower than the preset confidence threshold (the preset confidence threshold is determined based on the rule characteristics of the fuzzy rule base, historical control effects and system safety requirements, and is used to judge the reliability of control decisions), the load distribution mode abnormal signal is triggered. At the same time, the icing coverage data obtained from the blade surface condition monitoring is acquired, and parameters that can characterize the severity of icing are calculated based on the area, thickness and other information of the icing coverage.

[0048] In step 145 above, fuzzy rule expansion calculations are performed based on parameters characterizing the severity of icing. For different degrees of icing severity, matching rules for increasing the pitch angle adjustment range are added, such as rules that increase the adjustment range as the icing severity increases. These new rules are then added to the real-time rule base to complete the rule base update.

[0049] In the wind turbine blade flexible pitch system based on adaptive fuzzy control described in this embodiment of the invention, the pitch angle adjustment command is input to the coordinated mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern, including: Step 151: Perform dynamic command parsing and calculation on the pitch angle adjustment command, decomposing it into pitch angle increment value and pitch angle change rate command. Step 152: Based on the pitch angle increment value, the damping force threshold is dynamically adjusted through the load distribution pattern matching index to generate the target output damping force of the magnetorheological damper. Step 153: Based on the pitch angle change rate command, the motion hysteresis error of the pitch mechanism is eliminated by the proportional-integral algorithm to generate the torque compensation value of the servo motor. Step 154: Based on the target output damping force and torque compensation value, a weighted moving average algorithm is used to fuse the target output damping force and torque compensation value to generate the initial buffer torque value of the cooperative mechanism. Step 155: Perform load matching degree verification calculation on the initial buffer torque value. When the measured torque fluctuation rate exceeds the preset tolerance range, correct the torque value according to the fluctuation amplitude ratio to obtain a buffer torque value that matches the load distribution pattern.

[0050] In this embodiment of the invention, the pitch angle adjustment command is decomposed into specific parameters, providing a clear and explicit control basis for the subsequent coordinated control of the magnetorheological damper and the servo motor, ensuring that the two actions are coordinated and consistent; enabling the damping force output of the magnetorheological damper to adapt to different load distribution patterns, improving the matching degree between the damping effect and the load, and reducing unnecessary energy consumption; effectively compensating for the motion lag of the servo motor, ensuring that the pitch action can be executed accurately at the command rate, improving the timeliness and accuracy of pitch control; integrating the effects of both, so that the initial buffer torque value can comprehensively reflect the synergistic effect of damping and drive, providing a basis for subsequent adjustments; through verification and correction, ensuring that the output buffer torque value is stable and highly matched with the load distribution pattern, effectively buffering load impact and protecting the wind turbine blades.

[0051] In this embodiment of the invention, when applied in a specific way, it can be implemented through the following technical solutions, for example: In step 151 above, after receiving the pitch angle adjustment command, the dynamic command is parsed; the pitch angle adjustment information contained in the command is analyzed and broken down into two parts: one is the specific value that the pitch angle needs to be increased or decreased, i.e., the pitch angle increment value; the other is the rate of change of the pitch angle per unit time, i.e., the pitch angle change rate command.

[0052] In step 152 above, based on the obtained pitch angle increment value, and combined with the matching degree index between the current load distribution mode and the preset mode, when the matching degree is high, the damping force threshold is appropriately relaxed; when the matching degree is low, the damping force threshold is tightened, thereby dynamically adjusting the damping force threshold; based on the adjusted threshold, the target damping force that the magnetorheological damper needs to output is determined.

[0053] In step 153 above, based on the pitch angle change rate command, the motion state of the servo motor during the actual pitch change process is monitored, and the hysteresis error between the servo motor and the rate required by the command is calculated. Using the proportional-integral algorithm, the torque value that needs to be supplemented, i.e. the torque compensation value of the servo motor, is calculated according to the magnitude and duration of the error, in order to eliminate the hysteresis error.

[0054] In step 154 ​​above, the target output damping force of the magnetorheological damper and the torque compensation value of the servo motor are obtained, and corresponding weights are set for the two (such as assigning weights according to the degree of influence of the two on the buffering effect). The target output damping force and the torque compensation value are fused and calculated using a weighted moving average algorithm to obtain the initial buffering torque value of the cooperative mechanism.

[0055] In step 155 above, the initial buffer torque value is actually measured and its volatility is calculated. The volatility is compared with the preset tolerance range. If it exceeds the range, the torque value is corrected proportionally according to the magnitude of the volatility (more correction for large volatility and less correction for small volatility), and finally a buffer torque value that matches the current load distribution pattern is obtained.

[0056] In the wind turbine blade flexible pitch system based on adaptive fuzzy control described in this embodiment of the invention, based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper, phase change material heating is initiated and current parameters are dynamically adjusted to maintain torque output efficiency, including: Step 161: Based on the buffer torque value, calculate the theoretical optimal efficiency value under the current operating condition using the torque-speed characteristic curve pre-stored in the magnetorheological damper. Step 162: Based on the theoretical optimal efficiency value and the measured efficiency value monitored in real time by the magnetorheological damper, when the absolute value of the efficiency deviation exceeds the preset threshold and the real-time temperature is lower than the critical activation temperature of the phase change material, a heating command for the phase change material is triggered. Step 163: Real-time temperature feedback of the magnetorheological damper, establish a temperature-current compensation mapping table, and generate current reference parameters to maintain the target torque output. Step 164: Based on the heating command and current reference parameters, dynamically allocate the weight ratio of phase change material heating power and current compensation intensity to obtain the current parameter adjustment command. Step 165: Monitor the torque output efficiency after the current parameter adjustment command is executed. When the efficiency fluctuation rate exceeds the allowable range, correct the current parameter in reverse according to the efficiency deviation gradient value to maintain the torque output efficiency within the preset working range.

[0057] In this embodiment of the invention, the theoretical optimal efficiency value provides a clear reference standard for judging whether the actual working efficiency of the magnetorheological damper meets the standard, facilitating the timely detection of efficiency anomalies; heating is promptly initiated when the efficiency is low and the temperature conditions are met, preventing the damper's performance from deteriorating due to excessively low temperatures, thus ensuring efficiency maintenance; the current reference parameter provides a temperature-based basis for adjusting the current parameter, ensuring a suitable current foundation at different temperatures, which helps maintain stable torque output; the current parameter adjustment command enables heating and current compensation to work synergistically, ensuring the heating effect while further improving the torque output efficiency maintenance effect through reasonable adjustment of the current parameter; by monitoring and feedback correction of the adjusted efficiency, the torque output efficiency is ensured to remain stable within a reasonable range, guaranteeing the working performance of the magnetorheological damper.

[0058] In this embodiment of the invention, when applied in a specific way, it can be implemented through the following technical solutions, for example: In step 161 above, after obtaining the buffer torque value, the pre-stored torque-speed characteristic curve in the magnetorheological damper is called; based on the current buffer torque value, the corresponding speed information is found on the characteristic curve, and then the theoretical optimal efficiency value that the magnetorheological damper should achieve under the current operating conditions is calculated.

[0059] In step 162 above, the calculated theoretical optimal efficiency value is compared with the measured efficiency value monitored in real time by the magnetorheological damper, and the absolute value of the efficiency deviation between the two is calculated. When the absolute value exceeds the preset threshold, and at the same time the real-time temperature of the magnetorheological damper is lower than the critical activation temperature of the phase change material, a phase change material heating command is issued to start the heating device.

[0060] In step 163 above, real-time temperature feedback data from the magnetorheological damper is continuously received. Based on the influence of current on torque output at different temperatures, a mapping table between temperature and current compensation is constructed. The specific process of constructing the mapping table is as follows: In a laboratory environment, the operating temperature range of the magnetorheological damper is simulated (e.g., -20℃ to 50℃), and the temperature is set according to a preset temperature gradient (e.g., every 5℃ is an interval). At each temperature point, the input current value is adjusted, and the torque output data corresponding to different currents is recorded. At the same time, other operating parameters (e.g., target value of buffer torque, speed, etc.) are kept stable.

[0061] Using the current value required to achieve the target torque output at room temperature (e.g., 25℃) as a benchmark, calculate the additional or reduced current value required to achieve the target torque output at other temperature points, i.e., the current compensation amount; for example, at -5℃, if the benchmark current is 2A, and 3A is actually needed to achieve the target torque, then the current compensation amount is 1A.

[0062] The temperature values ​​are used as indexes, and the corresponding current compensation values ​​are used as mapping values, which are then organized into a table. For intermediate temperature values ​​between temperature gradients, the corresponding current compensation values ​​are determined by interpolation to ensure the continuity and accuracy of the mapping table across the entire temperature range.

[0063] In the actual operating environment, some temperature points are selected for testing. The current compensation amount obtained from the mapping table is applied to the magnetorheological damper to check whether the torque output meets the target value. If there is a deviation, the corresponding data in the mapping table is corrected until the requirements are met.

[0064] In a specific embodiment, taking the magnetorheological damper of this system applied to offshore wind power clusters (single unit capacity of 6MW or more) as an example, part of its temperature and current compensation mapping table is as follows: When the temperature is 25℃ (normal temperature), the current compensation is 0A. At this time, the reference current of 2A can make the damper output the target torque.

[0065] When the temperature drops to -5℃, the viscosity of the magnetorheological fluid increases. In order to maintain the target torque, the current compensation needs to be increased to 1A, that is, the input current is 3A.

[0066] When the temperature rises to 40℃, the performance of the magnetorheological fluid decreases slightly due to the temperature. The current compensation is 0.3A, and the target torque can be achieved with an input current of 2.3A.

[0067] Using this mapping table, the system determines the current reference parameters required to maintain the target torque output after receiving real-time temperature feedback.

[0068] In step 164 above, based on the requirements of the phase change material heating command and the determined current reference parameters, the weight ratio of the phase change material heating power to the current compensation intensity is dynamically allocated, taking into account the heating demand and the importance of current compensation; and based on this ratio, a specific current parameter adjustment command is generated.

[0069] In step 165 above, after the current parameter adjustment command is executed, the torque output efficiency of the magnetorheological damper is continuously monitored and its fluctuation rate is calculated. When the fluctuation rate exceeds the allowable range, the current parameter is adjusted in reverse according to the gradient value of the efficiency deviation (if the efficiency is too low, the current-related parameters are appropriately increased; if the efficiency is too high, the current-related parameters are appropriately decreased) so that the torque output efficiency is kept within the preset working range.

[0070] In the wind turbine blade flexible pitch system based on adaptive fuzzy control described in this embodiment of the invention, a multi-level electromagnetic shielding layer is deployed to suppress electromagnetic interference in the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision-making and rule base expansion, buffer torque value and temperature feedback adjustment, including: Step 171: Perform electromagnetic interference spectrum analysis on the closed-loop signal transmission path, detect the characteristic interference frequency data of the strain data acquisition section and the temperature feedback adjustment section, and generate an interference spectrum feature dataset containing frequency values ​​and amplitudes. Step 172: Based on the interference spectrum feature dataset, assign shielding material type and layer thickness parameters to interference sources in different frequency bands to obtain shielding parameters; Step 173: Based on the shielding parameters, determine the deployment coordinates of the multi-level shielding layer using the spatial topology data of the signal transmission path; Step 174: Based on the deployment location coordinates, verify the shielding effectiveness through electromagnetic field simulation and generate a signal-to-noise ratio improvement command. Step 175: According to the signal-to-noise ratio improvement target instruction, dynamically adjust the grounding impedance parameter of the electromagnetic shielding layer to maintain the integrity of the entire link signal transmission.

[0071] In this embodiment of the invention, the key characteristics of electromagnetic interference in the closed-loop path are grasped, providing clear data support for the subsequent targeted deployment of shielding layers. Shielding parameters ensure that the selection of shielding materials and the setting of layer thickness are adapted to the interference frequency and intensity, ensuring shielding effectiveness while avoiding material waste and over-design. The spatial deployment location of the shielding layer is clearly defined to ensure that the shielding layer can accurately cover key interference areas, maximizing the shielding effect and reducing the impact of electromagnetic interference on signal transmission. The effectiveness of the shielding scheme is confirmed in advance through simulation verification, avoiding the problem of poor shielding effect after actual deployment, reducing trial and error costs, and ensuring that the shielding measures can effectively improve signal quality. By dynamically adjusting the grounding impedance, the shielding layer is ensured to continuously exert good shielding effectiveness, ensuring the signal integrity of the closed-loop signal transmission path and improving the overall anti-interference capability of the system.

[0072] In this embodiment of the invention, when applied in a specific way, it can be implemented through the following technical solutions, for example: Step 171 above involves a comprehensive electromagnetic interference spectrum analysis of the closed-loop signal transmission path, with a focus on monitoring the signals in the strain data acquisition section and the temperature feedback adjustment section. The characteristic interference frequencies in these two key sections are identified through analysis, and the specific values ​​of each interference frequency and the corresponding signal amplitude are recorded. These data are then compiled and summarized to form an interference spectrum feature dataset containing frequency values ​​and amplitude values.

[0073] In step 172 above, based on the interference spectrum feature dataset, interference sources in different frequency bands are classified; for the interference characteristics of each frequency band, the appropriate shielding material type is matched according to the shielding effectiveness of various shielding materials against electromagnetic waves of different frequencies; at the same time, the layer thickness parameters of the corresponding shielding material are determined according to the amplitude of the interference signal, and finally integrated to form complete shielding parameters.

[0074] Step 173 above involves obtaining spatial topology data of the signal transmission path based on the determined shielding parameters, including the direction, layout, and relative positional relationship of each line segment to other devices; combining this spatial information, planning the specific installation locations of the multi-level shielding layers, marking the deployment coordinates of each shielding layer, and ensuring that the shielding layers can effectively cover the signal transmission segments susceptible to interference.

[0075] Step 174 above, the construction of the electromagnetic field simulation model, is based on the deployment coordinates of the multi-level shielding layers and combined with the electromagnetic characteristics of the closed-loop signal transmission path. It is achieved through step-by-step modeling and parameter configuration, as follows: Based on the physical layout of the closed-loop signal transmission path, the simulation software recreates the direction, length, and connection nodes of the signal transmission lines (such as the signal lines of fiber optic grating sensors and the control lines of magnetorheological dampers). At the same time, the spatial morphology of the multi-level shielding layers (such as the coverage area of ​​the shielding layers, the interlayer spacing, and the relative position with the signal path) is accurately drawn according to the deployment location coordinates to ensure that the geometric dimensions are consistent with the actual installation dimensions.

[0076] Local geometric refinement is performed on sections susceptible to interference (such as strain data acquisition section and temperature feedback adjustment section) to clarify the fit between the shielding layer and the signal transmission line (such as whether it is tightly wrapped or whether there are gaps), as well as the splicing method of the shielding layer (such as overlap length and corner treatment) to avoid simulation deviations caused by geometric simplification.

[0077] Based on the shielding parameters, corresponding material properties are assigned to each level of shielding layer, such as the conductivity and permeability of the metal shielding layer, and the dielectric constant of the insulation layer. At the same time, the material parameters of the signal transmission line (such as conductor resistance and insulation layer characteristics) and the electromagnetic properties of the surrounding environment (such as air and equipment housing) are defined to ensure that the material's reflection and absorption characteristics of electromagnetic fields are consistent with reality. Taking the actual installation environment of the closed-loop signal transmission path as a reference, the boundary range of the model is set (such as simulating the limited space inside the offshore wind turbine tower), and the electromagnetic field constraints at the boundary are defined (such as setting the far-field boundary as an "absorbing boundary" to avoid electromagnetic wave reflection interfering with the simulation results).

[0078] Based on the interference spectrum feature dataset, corresponding interference sources are set in the model: the frequency attributes of the interference sources are defined according to the characteristic interference frequencies (such as the high-frequency band of thunderstorm electromagnetic pulses, the specific frequency of equipment electromagnetic radiation); the power intensity is set according to the interference amplitude; the spatial location of the interference sources is determined according to the interference distribution law (such as electromagnetic interference sources near the strain acquisition section, equipment radiation sources near the temperature feedback line); the useful signals in the closed-loop path (such as strain data signals, temperature feedback signals) are simulated, and their frequency, amplitude and transmission characteristics (such as signal attenuation rate) are set as the benchmark signals for evaluating the signal-to-noise ratio.

[0079] Simulate different electromagnetic environment scenarios, such as "strong electromagnetic pulse scenario" (corresponding to thunderstorm weather) and "equipment conventional radiation scenario" (corresponding to electromagnetic interference from wind power equipment itself). For each scenario, configure the working parameters of the interference source according to the preset interference intensity and duration. Select an appropriate solution method based on the frequency characteristics of the interference signal (e.g., use a frequency domain solver for high-frequency interference and a time domain solver for transient pulse interference). Set the simulation accuracy level (e.g., improve the calculation accuracy in the coupling region between the shielding layer and the signal path), and define the solution objectives as "the attenuation of the interference signal by the shielding layer" and "the signal-to-noise ratio of the signal transmission line".

[0080] After running the simulation, the electromagnetic field distribution data within the shielding layer coverage area is extracted, and the signal-to-noise ratio changes before and after shielding are compared. If the simulation results deviate from the theoretical shielding effectiveness by more than the preset range (e.g., the difference between the measured shielding attenuation and the simulation value in a certain frequency band is >5dB), the model parameters are adjusted (e.g., the geometric modeling of the shielding layer corners is refined, and the material conductivity is corrected). The simulation is repeated until the simulation results are stable and reliable, and finally an electromagnetic field simulation model that can accurately reflect the working state of the shielding layer is formed.

[0081] The shielding effectiveness is calculated and evaluated by simulating the working state of the shielding layer under different electromagnetic environments, with a focus on the improvement of the signal-to-noise ratio (SNR). When the simulation results show that the SNR improvement reaches the preset standard, an SNR improvement compliance command is generated.

[0082] In step 175 above, after receiving the signal-to-noise ratio improvement target instruction, the integrity of the entire link signal transmission is monitored in real time; based on the monitoring results, the grounding impedance parameter of the electromagnetic shielding layer is dynamically adjusted. When signal transmission fluctuates, the grounding performance of the shielding layer is optimized by adjusting the impedance, thereby maintaining the stability and integrity of the entire link signal transmission.

[0083] like Figure 2 As shown, a control method for a flexible pitch system of wind turbine blades based on adaptive fuzzy control is disclosed. The control method includes: Real-time strain data of the blade flapping and oscillation directions are collected by fiber optic grating sensors installed at the root of the wind turbine blades. A mirror algorithm is applied to real-time strain data to generate a symmetrical compensation dataset that eliminates unilateral bias. Based on the symmetric compensation dataset, strain envelope feature points characterizing the load distribution pattern are extracted using the convex hull algorithm; The strain envelope feature points are input into an adaptive fuzzy controller to dynamically reconstruct the membership relationship features and generate pitch angle adjustment commands; at the same time, the fuzzy rule base is expanded in real time based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment commands, combined with the blade surface condition monitoring results. The pitch angle adjustment command is input into the coordination mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern. Based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper, the phase change material heating is initiated and the current parameters are dynamically adjusted to maintain torque output efficiency. In the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision-making and rule base expansion, buffer torque value and temperature feedback adjustment, multi-level electromagnetic shielding layers are deployed to suppress electromagnetic interference.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A flexible pitch control system for wind turbine blades based on adaptive fuzzy control, characterized in that, include: The acquisition module is used to acquire real-time strain data of the blade flapping and oscillation directions through fiber optic grating sensors installed at the root of the wind turbine blade; The mirror algorithm module is used to apply the mirror algorithm to real-time strain data to generate a symmetrical compensation dataset that eliminates one-sided deviations. The convex hull algorithm module is used to extract strain envelope feature points that characterize the load distribution pattern based on the symmetric compensation dataset using the convex hull algorithm. The fuzzy control module is used to input the strain envelope feature points into the adaptive fuzzy controller, dynamically reconstruct the membership relationship features and generate pitch angle adjustment commands; at the same time, based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment commands, the fuzzy rule base is expanded in real time in combination with the blade surface condition monitoring results. The matching module is used to input the pitch angle adjustment command into the coordination mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern. The temperature feedback module is used to start phase change material heating and dynamically adjust current parameters to maintain torque output efficiency based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper. The electromagnetic shielding module is used to deploy multi-level electromagnetic shielding layers to suppress electromagnetic interference in the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision and rule base expansion, buffer torque value and temperature feedback adjustment.

2. The flexible pitch system for wind turbine blades based on adaptive fuzzy control according to claim 1, characterized in that, Real-time strain data in the flapping and swaying directions of the wind turbine blades are collected by fiber optic grating sensors installed at the root of the blades, including: The fiber optic grating sensor array is activated according to the preset spatial distribution matrix, and the original light intensity attenuation value is collected synchronously. The initial strain components of each sensor node in the waving direction and the swinging direction are calculated. The initial strain components are time-domain aligned and calculated. The signal transmission delay of different sensor nodes is compensated by a sliding time window to generate a time-synchronized strain data sequence. The amplitude correction calculation is performed on the time-synchronized strain data sequence based on the calibration curve. The cubic spline interpolation method is used to eliminate the nonlinear error caused by the ambient temperature drift, and the preprocessed strain data is obtained. The preprocessed strain data is subjected to load validity verification calculation. When the strain ratio in the swing direction to the oscillation direction exceeds the preset safety threshold, the sensor self-test protocol is triggered and replaced with historical average data, and finally real-time strain data is generated.

3. The wind turbine blade flexible pitch system based on adaptive fuzzy control according to claim 2, characterized in that, A mirror algorithm is applied to real-time strain data to generate a symmetrical compensation dataset that eliminates unilateral bias, including: Based on real-time strain data, a sliding time window is used to extract the data according to the blade rotation phase angle, and strain data slices in the flapping and oscillation directions within the same cycle are calculated. Perform centerline fitting calculations on the strain data slices, and use the least squares method to fit the strain distribution curve of the blade neutral layer to generate a reference symmetry axis; Based on the reference axis of symmetry, a mirror symmetry compensation calculation is performed. The strain values ​​of the windward and leeward sides of the blade are mapped equidistantly relative to the reference axis. The difference in compensation is solved and superimposed on the original data to obtain the compensated strain data. Dynamic weighted fusion calculations are performed on the compensated strain data, and weight coefficients are assigned according to the blade section position to generate a symmetrical compensation dataset that eliminates unilateral deviations.

4. The wind turbine blade flexible pitch system based on adaptive fuzzy control according to claim 3, characterized in that, Based on the aforementioned symmetric compensation dataset, strain envelope feature points characterizing the load distribution pattern are extracted using the convex hull algorithm, including: Strain gradient preprocessing calculation is performed on the symmetric compensation dataset. The data segments are divided according to the blade spanwise position, and the ratio of strain dispersion to mean of each segment is calculated as the gradient sensitivity coefficient. Dynamic convex hull construction calculation is performed based on gradient sensitivity coefficient. The segment with sensitivity coefficient exceeding the preset threshold is selected as the key region, and the strain extreme points in the region are connected to form the initial convex hull. Based on the initial convex hull, redundant vertices are eliminated by a geometric curvature analysis algorithm, and inflection points with large curvature changes are retained as candidate feature points. The candidate feature points are input into the load pattern matching calculation, compared with the envelope template of the historical working condition feature library, and feature points that meet the preset matching degree are selected as strain envelope feature points.

5. The wind turbine blade flexible pitch system based on adaptive fuzzy control according to claim 4, characterized in that, The strain envelope feature points are input into an adaptive fuzzy controller to dynamically reconstruct the membership relationship features and generate a pitch angle adjustment command. Simultaneously, based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment command, and combined with the blade surface condition monitoring results, the fuzzy rule base is expanded in real time, including: Based on strain envelope feature points, key dimensional parameters of the geometric shape formed by the feature points are obtained through spatial parameterization measurement, and feature vectors representing the load distribution pattern are generated. When the deviation of the feature vector from the reference template exceeds a preset threshold, the center position parameter and distribution width parameter of the fuzzy set are dynamically adjusted according to the degree of deviation to obtain the reconstructed membership features. The membership relationship features are input into fuzzy inference calculation, the fuzzy rule base is activated to generate the pitch angle adjustment command, and a confidence parameter characterizing the reliability of the control decision is generated simultaneously. When the confidence parameter is lower than the preset confidence threshold, an abnormal load distribution pattern signal is triggered. Combined with the icing coverage data of the blade surface condition monitoring, a parameter characterizing the severity of icing is calculated. Based on the parameters characterizing the severity of icing, perform fuzzy rule expansion calculations, add rules to enhance the pitch angle adjustment range that match the severity of icing, and update the real-time rule base.

6. The wind turbine blade flexible pitch system based on adaptive fuzzy control according to claim 5, characterized in that, The pitch angle adjustment command is input into the coordination mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern, including: The pitch angle adjustment command is dynamically parsed and calculated, and decomposed into pitch angle increment value and pitch angle change rate command. Based on the pitch angle increment value, the damping force threshold is dynamically adjusted by the load distribution pattern matching index to generate the target output damping force of the magnetorheological damper. Based on the pitch angle change rate command, the motion hysteresis error of the pitch mechanism is eliminated by the proportional-integral algorithm, and the torque compensation value of the servo motor is generated. Based on the target output damping force and torque compensation value, a weighted moving average algorithm is used to fuse the target output damping force and torque compensation value to generate the initial buffer torque value of the cooperative mechanism. The load matching degree verification calculation is performed on the initial buffer torque value. When the measured torque fluctuation rate exceeds the preset tolerance range, the torque value is corrected according to the fluctuation amplitude ratio to obtain a buffer torque value that matches the load distribution pattern.

7. The wind turbine blade flexible pitch system based on adaptive fuzzy control according to claim 6, characterized in that, Based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper, the phase change material heating is initiated and the current parameters are dynamically adjusted to maintain torque output efficiency, including: Based on the buffer torque value, the theoretical optimal efficiency value under the current operating condition is calculated using the torque-speed characteristic curve pre-stored by the magnetorheological damper. Based on the theoretical optimal efficiency value and the measured efficiency value monitored in real time by the magnetorheological damper, when the absolute value of the efficiency deviation exceeds the preset threshold and the real-time temperature is lower than the critical activation temperature of the phase change material, a heating command for the phase change material is triggered. Real-time temperature feedback is provided for the magnetorheological damper, a temperature-current compensation mapping table is established, and current reference parameters for maintaining the target torque output are generated. Based on the heating command and current reference parameters, the weight ratio of the phase change material heating power and the current compensation intensity is dynamically allocated to obtain the current parameter adjustment command. The torque output efficiency is monitored after the current parameter adjustment command is executed. When the efficiency fluctuation rate exceeds the allowable range, the current parameter is corrected in reverse according to the efficiency deviation gradient value to maintain the torque output efficiency within the preset working range.

8. The wind turbine blade flexible pitch system based on adaptive fuzzy control according to claim 7, characterized in that, In the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision-making and rule base expansion, buffer torque value and temperature feedback adjustment, multi-level electromagnetic shielding layers are deployed to suppress electromagnetic interference, including: Electromagnetic interference spectrum analysis is performed on the closed-loop signal transmission path to detect characteristic interference frequency data of the strain data acquisition section and the temperature feedback adjustment section, and to generate an interference spectrum feature dataset containing frequency values ​​and amplitudes. Based on the aforementioned interference spectrum feature dataset, shielding material type and layer thickness parameters are assigned to interference sources in different frequency bands to obtain shielding parameters; Based on the shielding parameters, the deployment coordinates of the multi-level shielding layer are determined by the spatial topology data of the signal transmission path. Based on the deployment location coordinates, the shielding effectiveness is verified through electromagnetic field simulation, and a signal-to-noise ratio improvement target is generated. According to the signal-to-noise ratio improvement target instruction, the grounding impedance parameter of the electromagnetic shielding layer is dynamically adjusted to maintain the integrity of the entire link signal transmission.

9. A control method for a flexible pitch system of wind turbine blades based on adaptive fuzzy control, characterized in that, Applied to the system as described in any one of claims 1 to 8, the method comprises: Real-time strain data of the blade flapping and oscillation directions are collected by fiber optic grating sensors installed at the root of the wind turbine blades. A mirror algorithm is applied to real-time strain data to generate a symmetrical compensation dataset that eliminates unilateral bias. Based on the symmetric compensation dataset, strain envelope feature points characterizing the load distribution pattern are extracted using the convex hull algorithm; The strain envelope feature points are input into an adaptive fuzzy controller to dynamically reconstruct the membership relationship features and generate pitch angle adjustment commands; at the same time, the fuzzy rule base is expanded in real time based on the load distribution pattern anomaly signal triggered when generating the pitch angle adjustment commands, combined with the blade surface condition monitoring results. The pitch angle adjustment command is input into the coordination mechanism of the magnetorheological damper and the servo motor to obtain a buffer torque value that matches the load distribution pattern. Based on the buffer torque value and the real-time temperature feedback of the magnetorheological damper, the phase change material heating is initiated and the current parameters are dynamically adjusted to maintain torque output efficiency. In the closed-loop signal transmission path consisting of strain data acquisition, mirror algorithm processing, convex hull feature extraction, fuzzy control decision-making and rule base expansion, buffer torque value and temperature feedback adjustment, multi-level electromagnetic shielding layers are deployed to suppress electromagnetic interference.

10. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to perform the method as described in any one of claims 1-8.

Images