Control method, system, device and medium for fan active mass damper

Abstract

The application discloses a control method, system, device and method for a wind turbine active mass damper, wherein the method comprises a floating wind turbine, the floating wind turbine comprising a column, a plurality of pontoons and a support structure connecting the column and the pontoons, and the inside of the pontoon is provided with the active mass damper along the radial direction thereof; the method comprises the following steps: collecting environmental information, state information of the floating wind turbine and operation information of each active mass damper to obtain a current state vector; inputting the current state vector and a historical state vector into a fuzzy short-time prediction module to obtain a predicted state vector of a future time window and an overall prediction confidence, so as to obtain a local prediction confidence corresponding to each active mass damper; and calculating a control force of the active mass damper based on the current state vector, the predicted state vector and the local prediction confidence. Through the fusion control of the fuzzy prediction and the state feedback, the application can adaptively adjust the control gain under different sea conditions, and improve the platform stability.

Description

Technical Field

[0001] This invention belongs to the field of offshore wind power generation technology, and in particular relates to a control method, system, equipment and medium for active mass dampers of wind turbines. Background Technology

[0002] Active mass dampers (AMD), often referred to as active tuned mass dampers (ATMD), are well-established structural control devices. An AMD typically consists of a controllable mass element, a drive mechanism (e.g., a linear motor), a spring element, and a damping element. By actively applying force to the mass, it generates a reaction force to mitigate structural vibrations, offering enhanced damping performance compared to purely passive tuned mass dampers (TMD).

[0003] The diameter and height of the pontoons of a floating offshore wind turbine determine the rigid body motions of the platform, such as pitch, roll, and heave. Therefore, deploying active mass dampers inside the pontoons of a floating offshore wind turbine allows for more effective countermeasures against these basic platform motions, thereby reducing the vibrations of the tower and wind turbine caused by the platform motion. However, traditional control methods have limitations when used to control the active mass dampers of floating wind turbines. For example, PID controllers perform poorly when applied to complex dynamic systems with large time delays, rapid parameter changes, significant uncertainties, and severe nonlinearities. LQR-based control requires the system dynamics to be accurately represented by state-space equations. H-infinity control mainly focuses on minimizing the maximum gain at a specific frequency, which may lead to a lack of robustness over a very large spectral range of disturbances. The accuracy of traditional control methods depends on accurate modeling of environmental disturbances and structural dynamics. In unpredictable marine environments, when facing the dynamic control problem of floating wind turbines with high dimensions, high frequencies, continuous and complex dynamics, they cannot effectively and adaptively manage the complex, multimodal and stochastic dynamic responses of floating wind turbines in highly variable marine environments, resulting in poor control performance. Summary of the Invention

[0004] The purpose of this invention is to provide a control method, system, device, and medium for active mass dampers of wind turbines. By integrating fuzzy prediction and state feedback control, the control gain can be adaptively adjusted under different sea conditions, thereby improving platform stability.

[0005] This invention is achieved through the following technical solution: A control method for an active mass damper for a wind turbine includes a floating wind turbine, the floating wind turbine including a column, a plurality of floats spaced apart along the circumference of the column, and a support structure connecting the column and the floats, wherein an active mass damper is provided inside the floats along their radial direction. The method includes the following steps: The system collects environmental information, the status information of the floating wind turbine, and the operational information of each active mass damper. It then preprocesses the collected environmental information, status information, and operational information to obtain the current state vector. The environmental information includes wind speed data and wave data. The status information includes IMU data, stress data, torque data, and tilt angle data at multiple different locations of the floating wind turbine. The operational information includes the position data of the mass blocks in the active mass dampers and the force data acting on the mass blocks. The state vector within a preset historical window length is obtained as the historical state vector. The current state vector and the historical state vector are input into the fuzzy short-time prediction module. The fuzzy short-time prediction module outputs the predicted state vector of the future time window and the overall prediction confidence based on the pre-built fuzzy rule base. Based on the overall prediction confidence, the local prediction confidence corresponding to each active mass damper is obtained. Based on the current state vector, the predicted state vector, and the local prediction confidence level of each active mass damper, calculate the control force of the active mass damper; in, (1); (2); (3); In the formula, For the first The control force of an active mass damper For the first Feedforward control force of an active mass damper The first state vector is calculated using LQR control. The state feedback control force of an active mass damper For the first The fusion weights of the active mass dampers For saturation mapping to , , , These are the weighting coefficients. For the first Local response of the platform at the location of an active mass damper For the first The maximum root mean square value of the local response of the platform at the location of each active mass damper within the historical operating window. For the first The ratio of the remaining stroke of the mass block in an active mass damper To predict the feedforward gain matrix; The future number of the output of the fuzzy short-time prediction module The platform predicts the state vector.

[0006] Furthermore, the first value calculated by LQR control The steps for controlling the force of an active mass damper include: Based on the floating wind turbine and active mass damper, a discrete state-space model is established: (4); In the formula, The input vector includes state information and position information. for time, This is the dynamic vector, which includes the force data of each active mass damper. This is a perturbation vector, which includes environmental information. and Let be the coupling matrix. Here is the perturbation matrix; A recursive least squares algorithm is used to adaptively identify discrete state-space models. Construct the cost function: (5); In the formula, Let cost function be The state weight matrix is... To control the weight matrix; Gain is obtained based on discrete state-space model and cost function. K And calculate the control force: (6); In the formula, The control vector is composed of the control forces of each active mass damper.

[0007] Furthermore, after the step of adaptively identifying the discrete state-space model using the recursive least squares algorithm, the method further includes: For each parameter included in the identification result, determine whether the parameter exceeds the preset range. If so, apply saturation processing to the parameter and record an alarm.

[0008] Furthermore, the step of calculating the control force of the active mass damper includes: For each active mass damper, determine whether the active mass damper meets the triggering condition. If so, the active mass damper is recorded as a retreat damper. The triggering condition is that the displacement of the mass block reaches the first preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper exceeds a preset safety threshold. Remove the retreat damper from control and designate the active mass dampers other than the retreat damper as healthy dampers; The control forces of the retreat damper and the health damper are calculated according to formulas (1), (2) and (3), and the control forces that cannot be executed after each retreat damper exits control and the original control forces of each health damper are obtained. Calculate the compensating control force of the health damper according to formula (7): (7); In the formula, For the first The compensating control force of a health damper For the first The original control force of a healthy damper To transfer from the retraction damper to the first The contribution factor of each health damper This refers to the control force that cannot be executed after the damper exits control.

[0009] Furthermore, after the step of calculating the compensating control force of the health damper according to formula (7), the method further includes: Determine whether the retreat damper meets the recovery condition. If so, gradually restore the control force of the retreat damper from 0 to the control force calculated by formula (1), formula (2) and formula (3). The recovery condition is that the displacement of the mass block is less than the second preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper is below the preset safety threshold. Once the control force of the retreating damper is restored to the control force calculated by formulas (1), (2) and (3), the retreating damper is deactivated.

[0010] Furthermore, the step of obtaining the local prediction confidence of each active mass damper based on the overall prediction confidence includes: Based on the overall prediction confidence level, the local prediction confidence level for each active mass damper is calculated using the following formula: (8); (9); in, The overall prediction confidence level output by the fuzzy short-time prediction module. For the first The correlation coefficient of the platform response at the location of each active mass damper. For the first The current actual platform response at the installation location of each active mass damper. The first fuzzy short-time prediction module obtained Response of an active mass damper at its installation location. This is the local response error scale parameter.

[0011] Furthermore, the method further includes the following steps: The operation of each active mass damper is controlled by the control force of each active mass damper, and the operating information of each active mass damper is collected. Based on the operational information, the deviation between the force data acting on the mass block and the target value of the control force is calculated, and it is determined whether the deviation exceeds the preset deviation threshold. If so, the control force of the active mass damper is modified.

[0012] The present invention also provides a control system for an active mass damper for a wind turbine, including a floating wind turbine, the floating wind turbine including a column, a plurality of floats spaced apart along the circumference of the column, and a support structure connecting the column and the floats, wherein an active mass damper is provided inside the floats along its radial direction. The system also includes: The acquisition module is used to acquire environmental information, the status information of the floating wind turbine, and the operating information of each active mass damper. It also preprocesses the acquired environmental information, status information, and operating information to obtain the current state vector. The environmental information includes wind speed data and wave data. The status information includes IMU data, stress data, torque data, and tilt angle data at multiple different locations of the floating wind turbine. The operating information includes the position data of the mass blocks in the active mass dampers and the force data acting on the mass blocks. The acquisition module is used to acquire the state vector within a preset historical window length as the historical state vector. The current state vector and the historical state vector are input into the fuzzy short-time prediction module. The fuzzy short-time prediction module outputs the predicted state vector of the future time window and the overall prediction confidence based on the pre-built fuzzy rule base, and obtains the local prediction confidence corresponding to each active mass damper based on the overall prediction confidence. The calculation module is used to calculate the control force of the active mass damper based on the current state vector, the predicted state vector, and the local prediction confidence of each active mass damper. in, (1); (2); (3); In the formula, For the first The control force of an active mass damper For the first Feedforward control force of an active mass damper The first state vector is calculated using LQR control. The state feedback control force of an active mass damper For the first The fusion weights of the active mass dampers For saturation mapping to , , , These are the weighting coefficients. For the first Local response of the platform at the location of an active mass damper For the first The maximum root mean square value of the local response of the platform at the location of each active mass damper within the historical operating window. For the first The ratio of the remaining stroke of the mass block in an active mass damper To predict the feedforward gain matrix; The future number of the output of the fuzzy short-time prediction module The platform predicts the state vector.

[0013] The present invention also provides an electronic device, the electronic device comprising: processor; Memory is used to store executable computer programs; Among them, the steps of implementing the control method for the active mass damper of the wind turbine when the processor executes the computer program.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a control method for an active mass damper of a wind turbine.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The active mass damper is arranged radially along the pontoon and can directly generate the reverse inertial force coupled with the typical attitude mode of the platform. It has an efficient suppression effect on the platform pitch, roll and deck acceleration, and significantly reduces the amplitude of the platform attitude response. Furthermore, the active mass damper is installed inside the pontoon and does not affect the shape of the pontoon and the overall buoyancy layout. The guide rail structure is simple, the maintenance space is sufficient, and it has good compatibility with existing semi-submersible platforms. It can be directly deployed through later modification or new platform construction. (2) The control method has adaptive capability and can adapt to different sea state changes. The present invention adopts an adaptive control strategy that integrates fuzzy prediction and state feedback. It can automatically adjust the control parameters according to the real-time wind and wave amplitude, platform motion RMS index and mass block response to ensure that the control performance can be maintained at the best under light wind, medium wave and strong wind and wave conditions. (3) It has a fast response and is suitable for engineering applications. Fuzzy prediction is used for short-term trend judgment, and LQR state feedback provides stable control compensation. The fused control law has a small calculation volume and strong real-time performance. It is compatible with the controller hardware of actual offshore wind power platforms and realizes engineering-level deployment. (4) Low energy consumption and high operational reliability under all working conditions; the control algorithm can automatically optimize the action force under the premise of meeting the attitude suppression target, so that the actuator can work in the low energy consumption area, extend the life of the servo system, and at the same time ensure that the AMD displacement does not exceed the limit, thus improving the long-term operational reliability. Attached Figure Description

[0016] Figure 1 This is a flowchart of the control method for an active mass damper of a wind turbine according to the present invention; Figure 2 A schematic diagram of a floating wind turbine used in the invention of a control method for an active mass damper of a wind turbine; Figure 3 This is a schematic diagram of the control system of the present invention for an active mass damper of a wind turbine; Figure 4 This is a hardware structure diagram of the electronic device of the present invention.

[0017] In the diagram, 1-column, 2-buoy, 3-support structure, 4-active mass damper. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0022] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed when in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0023] Please see Figures 1 to 3 , Figure 1 This is a flowchart illustrating the steps of the control method for an active mass damper in a wind turbine according to the present invention. Figure 2 This is a schematic diagram of a floating wind turbine used in a control method for an active mass damper in a wind turbine. The invention provides a control method for an active mass damper in a wind turbine, comprising a floating wind turbine. The floating wind turbine includes a column 1, multiple floats 2 spaced circumferentially along the column 1, and a support structure 3 connecting the column 1 and the floats 2. An active mass damper 4 is radially arranged inside each float 1. Each active mass damper 4 includes: a mass block, a high-density material block accounting for 1%-2% of the platform's total displacement; a guide rail mounted on the bottom plate inside the float to restrict the mass block's radial linear movement; and a driver for driving the mass block to move along the guide rail, the driver being an electro-hydraulic servo system or a high-thrust linear motor.

[0024] The control method for an active mass damper of a wind turbine according to the present invention includes the following steps: S1. Collect environmental information, the status information of the floating wind turbine, and the operating information of each active mass damper. Preprocess the collected environmental information, status information, and operating information to obtain the current state vector. The environmental information includes wind speed data and wave data. The status information includes IMU data, stress data, torque data, and tilt angle data at multiple different locations of the floating wind turbine. The operating information includes the position data of the mass blocks in the active mass dampers and the force data acting on the mass blocks. S2. Obtain the state vector within the preset historical window length as the historical state vector, input the current state vector and the historical state vector into the fuzzy short-time prediction module, the fuzzy short-time prediction module outputs the predicted state vector of the future time window and the overall prediction confidence based on the pre-built fuzzy rule base, and obtains the local prediction confidence corresponding to each active mass damper based on the overall prediction confidence. S3. Calculate the control force of the active mass damper based on the current state vector, the predicted state vector, and the confidence level. in, (1); (2); (3); In the formula, For the first The control force of an active mass damper For the first Feedforward control force of an active mass damper The first state vector is calculated using LQR control. The state feedback control force of an active mass damper For the first The fusion weights of the active mass dampers For saturation mapping to , , , These are the weighting coefficients. For the first Local response of the platform at the location of an active mass damper For the first The maximum root mean square value of the platform's local response at each active mass damper location within the historical operating window is used to normalize the current local response. This is because the coupling relationship between the float locations of different active mass dampers and the attitude modes of the floating wind turbines is different, resulting in differences in the response amplitude at each location. Therefore, normalization using the local historical maximum response better reflects the vibration state of each active mass damper at its current location. For the first The ratio of the remaining stroke of the mass block in an active mass damper The details of predicting the feedforward gain matrix will not be elaborated here. The future number of the output of the fuzzy short-time prediction module The platform predicts the state vector.

[0025] In step S1 above, IMUs (accelerometers and gyroscopes), tilt sensors, strain gauges, and torque sensors are respectively installed at the column, the top of the three pontoons, the tower top, and the tower base. High-precision displacement and force sensors are installed at the guide rails of the active mass damper inside each pontoon. Environmental sensors (ultrasonic anemometers, wave radars, or wave height sensors) are installed on the pontoons at or near the tower top. The sampling frequency is synchronized with the IMUs; a clock synchronization mechanism can be used to ensure consistent data timestamps and a unified sampling frequency. Collection, suggestions (Based on control bandwidth settings). The tower top refers to the very top of the tower (or the mast above the central column), the highest point; the tower base refers to the location where the tower connects to the top platform of the central column, the root of the tower. The IMU provides the platform's triaxial acceleration and angular velocity, the tilt sensor provides static attitude angle information, the strain gauge provides structural deformation and stress distribution, and the torque sensor provides the actual bending moment and shear force between the tower base and the pontoons. High-precision linear displacement sensors (encoders / laser rulers) are used at each pontoon guide rail to measure the real-time position of the mass block relative to the guide rail, and force sensors (load cells or hydraulic pressure sensors) are arranged between the mass block and the drive mechanism to directly measure the applied force. The velocity signal is estimated by filtering the position signal differentially or using a Kalman filter based on a mass-damped dynamic model. If a motor drive is used, the drive current can also be collected in parallel as thrust redundancy information; if a hydraulic drive is used, the thrust is calculated using the piston pressure difference and effective area.

[0026] Before entering the control loop, all acquired raw sensor data undergoes a unified preprocessing procedure. This includes removing discrete high-frequency noise from the raw data using a second-order Butterworth low-pass filter, synchronizing the timestamps and sampling frequencies of different sensors, transforming the IMU signal from the body coordinate system to the platform coordinate system and performing gravity compensation, suppressing noise and drift using low-pass or Kalman filtering, estimating the platform attitude, tower top acceleration, angular velocity, and displacement and velocity of the mass blocks of the active mass damper using differential / integral and fusion algorithms, constructing external disturbance estimates by combining anemometer and wave radar data, and eliminating sudden noise through amplitude limiting and anomaly detection. The preprocessed attitude data, inertial data, and environmental excitations are normalized to a unified scale and then concatenated into the current state vector, providing a stable and reliable input for the control strategy.

[0027] In addition, preprocessing includes threshold detection and interpolation repair for sensor abrupt changes: if a sensor experiences multiple consecutive sampling anomalies (N_bad), N_bad=5 can be set to trigger redundant sensor switching or enter fault-tolerant mode. Specifically, triggering redundant sensor switching employs the following three redundancy strategies: hardware redundancy, where dual or triple sets of sensors are deployed at some critical locations (such as platform attitude measurement). It can be first determined whether there is a replacement sensor for the abnormal sensor; if so, the sensor is switched; otherwise, analytical redundancy or functional redundancy is entered. Analytical redundancy relies on the physical correlation model between multiple sensors to calculate replacement values; functional redundancy utilizes coordinate transformations of data from sensors with similar functions at other locations for mutual backup. Fault-tolerant mode is a tiered, progressively decreasing emergency operation mechanism, operating according to the following steps: Step 1, Fault Detection: Confirm the fault through signal integrity check, physical range check, and multi-sensor consistency check; Step 2, Tiered Switching: Level 1: Only the faulty sensor is switched (replaced with a backup or estimated value), and the system operates normally; Level 2: Functional degradation, reducing reliance on the sensor data, resulting in decreased control performance but without system crash; Level 3: Enter safety mode (e.g., lock the mass block of the active mass damper to the neutral position) to avoid equipment damage caused by erroneous commands; Step 3, Control Law Reconstruction: Switch signal sources, reallocate control weights among multiple AMDs, and reduce control gain; Step 4, Recovery Mechanism: After the fault is cleared, smoothly switch back to normal mode; if automatic recovery fails, an alarm will be triggered to notify the operations and maintenance department.

[0028] In step S2 above, the state vector within a preset historical window length can be obtained first as the historical state vector, and the preset historical window length is... Can be set to The current state vector and historical state vectors are combined in time series to form the input matrix for fuzzy inference. The fuzzy short-term prediction module maps the above input matrix to the membership degree of fuzzy sets through membership functions (trigonometric or Gaussian), such as classifying the wind speed rise rate into fuzzy concepts like "high, medium, and low". Based on a pre-built fuzzy rule base, the activation intensity of each rule is calculated. Finally, Mamdani-type inference is used to convert the results of all rules (fuzzy sets) into specific numerical prediction results through a confidence-based weighted average method, that is, to obtain the predicted state vector for the future time window, and at the same time output the overall prediction confidence γ, reflecting the reliability of this prediction. The Adaptive Fuzzy Short-term Prediction Module (AFPM) is a modular fuzzy prediction unit for short-term time series prediction. Its core relies on fuzzy mathematics and adaptive rule optimization, and it is specifically designed to handle the problems of random disturbances in short-term data and fuzzy uncertainty of influencing factors.

[0029] Among them, the rule form of the fuzzy rule base adopts Structure, for example: (rate of wind speed increase) ) (Current pitch angle) ) (Pitch will increase in the future). The fuzzy rule base is pre-defined and its design is based on expert experience, system dynamics characteristics and control objectives. The construction of the fuzzy rule base needs to cover the platform's main control modes (pitch, roll), key outputs (tower top acceleration) and actuator constraints (mass block travel, motion limit). The number of rules, membership function parameters and rule weights are determined through engineering experience and simulation iteration.

[0030] Then, based on the consistency between the local platform response and the predicted state at each active mass damper installation location, the overall prediction confidence γ is mapped to the local prediction confidence corresponding to each active mass damper. This is used to calculate the fusion weights for different active mass dampers. Since different active mass dampers are positioned at different pontoon locations, their local platform responses differ. Therefore, it is necessary to further map the global prediction confidence to the local prediction confidence at the corresponding location of each active mass damper. The specific calculations are as follows: (8); (9); in, The overall prediction confidence level output by the fuzzy short-time prediction module. For the first The platform response correlation coefficient at the location of each active mass damper is used to describe the degree of consistency between the local response at that location and the overall predicted state. For the first The current actual platform response at the installation location of each active mass damper. The first fuzzy short-time prediction module obtained Response of an active mass damper at its installation location. This is the local response error scale parameter.

[0031] In step S3 above, the future time window prediction state vector output by the fuzzy short-time prediction module is... Substituting the pre-designed feedforward control law, i.e., formula (3), the feedforward control force of each active mass damper is calculated. The prediction feedforward gain matrix in formula (3) Based on the coupled dynamic model of the active mass damper and the floating platform, the optimal control algorithm is used to solve offline to establish the mapping relationship between the predicted state vector and the control force of the active mass damper. The calculation of the feedforward gain matrix based on the state-space model is a conventional calculation process in existing optimal control techniques, such as Model Predictive Control (MPC) and Linear Quadratic Regulation (LQR), and will not be elaborated further here. Then, based on the current state vector, the LQR control is used to calculate the... State feedback control force of an active mass damper feedforward control force and state feedback control force According to fusion weight By performing a weighted combination, we obtain the first... The final control command for each active mass damper mainly consists of the instantaneous action force of the active mass damper and necessary auxiliary target quantities (target acceleration or velocity of the mass block). This directly drives the mass block to move along the guide rail, generating inertial forces to compensate for the platform's attitude. The control command is transmitted to the actuator of each active mass damper, causing the mass block to generate corresponding acceleration and inertial forces along the guide rail. The fusion weight is the core hub connecting fuzzy prediction and LQR feedback. It dynamically adjusts the control commands of each active mass damper. and The proportion of [a certain percentage] is used to achieve multi-objective adaptive balance. When the value approaches 1, fuzzy prediction feedforward is the main method for rapid suppression. When the confidence level approaches 0, LQR feedback becomes the primary method, ensuring robustness and reliability. Furthermore, when the fuzzy prediction confidence level... High, local response When it is big, Increase the size of the feedforward prediction to enhance its rapid suppression capability; when the mass block has remaining travel time... When the confidence level is small or low. Automatic reduction allows control to degenerate into a more robust LQR feedback.

[0032] Furthermore, in step S3, the first value calculated by LQR control is... The steps for controlling the force of an active mass damper include: S31. Based on the floating wind turbine and active mass damper, establish a discrete state-space model: (4); In the formula, The input vector includes state information and position information. for time, This is the dynamic vector, which includes the force data of each active mass damper. This is a perturbation vector, which includes environmental information. and Let be the coupling matrix. Here is the perturbation matrix; S32. Adaptive identification of discrete state-space models is performed using the recursive least squares algorithm. S33. Constructing the cost function: (5); In the formula, Let cost function be The state weight matrix is... To control the weight matrix; S34. Obtaining Gain Based on Discrete State Space Model and Cost Function K And calculate the control force: (6); In the formula, The control vector is composed of the control forces of each active mass damper.

[0033] In step S31 above, a discrete state-space model (sampling period) is established, taking the floating wind turbine and each active mass damper as a whole. ), initial coupling matrix and Based on The offline dynamics modeling or empirical parameters are obtained; to reduce model complexity, the coupling matrix can be approximated by symmetry and the main coupling terms can be retained (such as pitch-mass coupling, roll-mass coupling).

[0034] In step S32 above, the recursive least squares (RLS) method is used to update the parameters of the discrete state-space model online, enabling the model to adapt to changes in actual operating conditions. The parameter update formula is as follows: ,further, Covariance matrix update ,in, For the first The model parameter vector to be identified at each time step is derived from... and The elements in the are obtained; The parameter estimation error covariance matrix; This is the recursive gain matrix; The regression vector; the regression vector includes , and , The actual measured output vector (or "measurement vector") of the system is used to drive the correction term for online parameter identification in RLS, and a forgetting factor is introduced during the recursive least squares identification process. By reducing the weight of historical data, the model's adaptability to changes in construction process parameters is improved. The forgetting factor ranges from 0.98 to 0.995. The closer the parameter estimate is to 1, the more stable it is; when... When the size decreases, the model becomes more sensitive to newly acquired data. The input to RLS primarily consists of estimates derived from the current state. Control input and environmental disturbances The regression vector is formed by , , The column vector formed by extracting and combining the selected parameters to be identified. The system outputs measured by the sensors (such as status information and mass block displacement) and the updated column vector of parameters to be identified are also included. This is used to correct the corresponding coupling matrix elements in the discrete state-space model, from the updated column vector of parameters to be identified. Parsing , , The updated values ​​at the corresponding positions are filled into the discrete state-space model for use in the LQR gain calculation and fuzzy prediction of the next control cycle, thereby ensuring that subsequent LQR control and fuzzy prediction are always based on a high-fidelity dynamic model that matches the current environment.

[0035] Furthermore, after step S32, the following steps are also included: S32a. For each parameter included in the identification result, determine whether the parameter exceeds the preset range. If so, apply saturation processing to the parameter and record an alarm.

[0036] In step S22a above, the online identification result should be subject to physical constraints. Generally speaking, the parameter vector to be estimated... The parameters include mass, stiffness, damping, added mass coefficient, restoring force and stiffness coupling term, time delay parameter, friction coefficient, modal damping ratio, etc. Applying physical constraints to these parameters aims to prevent the identification results from diverging to non-physical values ​​due to noise or numerical problems, ensuring the model still possesses reasonable extrapolation capabilities under extreme conditions, and preventing system instability caused by control commands due to model distortion. Therefore, each parameter has a preset range, which represents the upper and lower physical limits of the parameter. When the identified parameter vector... When a parameter exceeds the preset physical upper and lower limits, the system adopts saturation to constrain and correct the parameter. Specifically, this includes: 1. clamping the parameter to the boundary value through hard truncation to ensure that the model parameter is always within the physical feasible region; 2. correcting the covariance matrix (such as resetting the corresponding elements or using a projection algorithm) to avoid oversensitivity or divergence in subsequent identification due to parameter truncation.

[0037] In steps S33 and S34 above, the state weight matrix Designed to emphasize the platform's overall attitude energy. For example, the weights of the platform's pitch, roll, and angular velocities are set much higher than the weights of individual mass displacements. This allows LQR to prioritize control allocation schemes that most effectively reduce the overall platform vibration during optimization, rather than simply suppressing the local motion of individual buoys. Control weight matrix This not only includes penalty terms (diagonal elements) for the energy consumption of each active mass damper, but also allows for the introduction of off-diagonal elements to constrain the mutual cancellation behavior among the three sets of active mass dampers. For example, if two sets of active mass dampers simultaneously output large forces in opposite directions, although they may not contribute much to the platform's attitude, they will waste energy and increase structural fatigue. Through appropriate settings... The off-diagonal elements are positive, and LQR will penalize such inefficient internal friction and encourage cooperative output. The feedback gain matrix is ​​then obtained by solving the algebraic Riccati equation. K ,Depend on Calculate the control vector for the active mass dampers. The control vector includes the control forces of all active mass dampers. The control force of each active mass damper can be obtained from the control vector. In this embodiment, there are three sets of active mass dampers, namely... .in K The dimension is 3×n ( n (as the dimension of the state vector), a manifestation of coupling: KEach row (corresponding to one AMD) includes not only the feedback gain of the AMD's own mass block state (displacement, velocity), but also the feedback gain terms of the platform attitude state and the mass block states of the other two active mass dampers.

[0038] Furthermore, to prevent the active mass damper from exceeding the actuator's capacity due to excessive control output, the following step is added after step S34: S35. The control force calculated through LQR control. Perform saturation constraint processing: (10).

[0039] In step S35 above, firstly, step S34... Calculate the theoretical control force, and then through By limiting it to the maximum permissible output range of the active mass damper, the final control force calculated by LQR control is obtained. Ultimately, feedback control force after amplitude limiting is adopted. Participation in fusion control, i.e. As Used to substitute into formula (1) for fusion calculation.

[0040] Furthermore, the steps for calculating the control force of the active mass damper include: S35. For each active mass damper, determine whether the active mass damper meets the triggering condition. If so, record the active mass damper as a retreat damper. The triggering condition is that the displacement of the mass block reaches the first preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper exceeds the preset safety threshold. S36. Remove the retreat damper from control and designate the active mass damper other than the retreat damper as the healthy damper. S37. Calculate the control forces of the retreat damper and the health damper according to formula (1), formula (2) and formula (3) to obtain the control force that cannot be executed after each retreat damper exits control and the original control force of each health damper. S38. Calculate the compensating control force of the health damper according to formula (7): (7); In the formula, For the first The compensating control force of a health damper For the first The original control force of a healthy damper To transfer from the retraction damper to the first The contribution factor of each health damper This refers to the control force that cannot be executed after the damper exits control.

[0041] In steps S36 to S38 above, if a set of active mass dampers faces physical limit constraints, the control output of that channel will be stopped, and the control task will be transferred to the other active mass dampers to share the burden, so as to protect the equipment safety and maintain the basic stability of the platform. To determine whether an active mass damper is facing physical limits, it can be determined that the displacement of the mass block of the active mass damper reaches the first preset percentage of the maximum allowable stroke, which is 90%, or the temperature rise of the active mass damper exceeds the preset safety threshold. When any of the above conditions are met, it is considered that the active mass damper has reached physical limits, and the active mass damper is marked as a retreat damper. The retreat damper is then removed from control, that is, it is no longer controlled according to the calculated control force. After the retreat damper is removed from control, the control force that was originally planned to be output but cannot be executed needs to be borne by other healthy dampers. Therefore, for retreat dampers and healthy dampers, the control force that cannot be executed after each retreat damper is removed from control is first calculated by formula (1), formula (2) and formula (3). and the original control force of the health damper Then, the compensation control force after the healthy damper shares the control force that cannot be executed after the retreat damper exits is calculated by formula (7). Where, if there is only one retreat damper, the control force that cannot be executed after the retreat damper exits control is calculated. If there are multiple retreat dampers, then according to the direction of action of the control force of each retreat damper, the control quantities that each retreat damper cannot execute are combined to obtain the total control quantity to be compensated. When the control forces of each active mass damper have been mapped to a unified control mode, it can be expressed as: In the formula, It is the set of all current retreat dampers. Sharing coefficient The following factors should be considered: the coupling relationship between the spatial position of each active mass damper on its pontoon and the platform's attitude modes; the remaining stroke margin of the healthy damper itself; and the current temperature rise of the healthy damper. First, a comprehensive evaluation index should be constructed based on the attitude mode coupling contribution, remaining stroke margin, and temperature rise safety margin. Then, the fault control force sharing coefficient is calculated using a normalization method: (11); In the formula, Taking into account the coupling degree between the active mass damper and the platform attitude mode, the remaining travel margin, and the temperature rise safety margin, the control force is adaptively distributed according to the remaining control capability of each healthy damper.

[0042] Furthermore, after calculating the compensating control force of the health damper according to formula (7), the following steps are also included: S39. Determine whether the retreat damper meets the recovery condition. If so, gradually restore the control force of the retreat damper from 0 to the control force calculated by formula (1), formula (2) and formula (3). The recovery condition is that the displacement of the mass block is less than the second preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper is below the preset safety threshold. S310. After the control force of the damper is restored to the control force calculated by formula (1), formula (2) and formula (3), the mark of the damper is cancelled.

[0043] In steps S39 to S310 above, if the displacement of the mass block of the damper is less than the second preset percentage of the maximum allowable stroke, and the second preset percentage is 80%, it means that the damper has returned to the safe area. If the temperature rise of the active mass damper is below the preset safe threshold, it means that the temperature rise of the damper is also within the safe threshold. At this time, the control force of the damper will be gradually restored from 0 to the control force calculated by formula (1), formula (2) and formula (3). The control force of the damper will be gradually restored to prevent impact. When the control force of the damper is restored to the control force calculated by formula (1), formula (2) and formula (3), it means that the damper is a healthy damper. Therefore, the mark of the damper is canceled, and the active mass damper is controlled again. All dampers calculate the control force according to formula (1) to formula (3).

[0044] Furthermore, after calculating the compensating control force of the health damper according to formula (7), the following steps are also included: S311. Determine whether the duration of the retreat damper exceeds the preset duration. If so, trigger an alarm notification.

[0045] In step S311 above, if the damper fails to return to normal operation for an extended period, an alarm will be triggered to notify maintenance personnel to manually intervene and perform repairs.

[0046] Furthermore, the method also includes the following steps: S4. Control the operation of each active mass damper according to the control force of each active mass damper, and collect the operating information of each active mass damper. S5. Based on the operating information, calculate the deviation between the force data acting on the mass block and the target value of the control force, and determine whether the deviation exceeds the preset deviation threshold. S6. If so, then the control force of the active mass damper is modified.

[0047] In steps S4 to S6 above, the control command output in step S3 includes the target value of the control force, from which the target value of the control force of the mass block of the active mass damper can be obtained. Final control quantity. Force / current commands are mapped to electro-hydraulic servo or linear motors via actuator distributors, while also considering speed limits. With displacement limit The position, velocity, and actual output force of the mass block are transmitted back in real time; if the deviation between the actual force and the target value of the desired control force exceeds a threshold... If the deviation is caused by nonlinear factors (such as friction or saturation), a feedback correction is triggered and fault diagnosis information is recorded. Feedback correction refers to the system activating a low-level closed-loop compensation mechanism. Based on the original LQR and fuzzy prediction output control commands, a fast correction amount based on force deviation is superimposed, allowing the actual output force to return to the expected value as quickly as possible. Specifically: 1. Proportional-Integral (PI) Correction: The force deviation is input into an independent PI controller, which outputs a correction amount, which is then superimposed on the original command and sent to the actuator. This method can quickly eliminate steady-state errors. 2. State Observer Compensation: If the system is equipped with a Kalman filter or state observer, the force deviation is used to correct the state estimation of the mass block's velocity or acceleration, thereby indirectly adjusting the control command for the next cycle, making the model prediction more closely match the actual situation. 3. Command Remapping: When the deviation is caused by nonlinear factors (such as friction or saturation), the feedforward model of the driver is corrected through table lookup or adaptive algorithms, for example, adjusting the zero bias of the electro-hydraulic servo valve or the current-force mapping coefficient of the linear motor. This correction process runs independently within a millisecond-level control cycle, while simultaneously recording deviation exceeding events for fault diagnosis. If the correction amount is consistently too large or frequently triggered, it indicates that there may be sensor failure, actuator performance degradation, or model mismatch. The system will reduce the weight of the channel or switch to a backup strategy according to the fault tolerance logic in step S1.

[0048] Furthermore, the method also includes the following steps: S6. Real-time calculation of performance metrics, including attitude and acceleration. Tower foundation stress statistics At least one of the cumulative energy consumption and cumulative fatigue indicators; S7. Determine whether the performance indicators deviate from the target. If so, perform predictive control parameter self-tuning. The predictive control parameters include adjusting the parameters of the fuzzy short-time prediction module, the predictive fusion weights, and the parameters of the online identification model, so that the predictive control part can adapt to the dynamic characteristics of the platform.

[0049] In steps S6 and S7 above, based on The performance index is calculated using the counting method. If long-term operating data shows that the performance index deviates from the target (e.g., attitude), then... If the expected decrease is not achieved, then self-tuning of the predictive control parameters will be performed: adjusting the parameters of the fuzzy short-time prediction module, such as the fuzzy membership function parameters and rule weights; recalibrating. Or improve the energy consumption / stability trade-off through small-scale optimization adjustments; update The forgetting factor is adjusted to adapt to new sea state characteristics. The LQR feedback control gain matrix is ​​not involved. Real-time adjustment of LQR feedback control gain matrix The parameters remain unchanged, serving only as a basic control element in the stable feedback control loop. Specifically, adjusting the fuzzy membership function parameters and rule weights employs a two-stage self-tuning strategy: online small-step learning and periodic batch recalibration. In the online stage, the sliding window cost function J is used as the objective. The parameter gradient is estimated through finite difference, and the membership function center, width, and rule consequent coefficients are updated with small steps. The update uses gradient descent with driving forces, constrained by amplitude and protected by rollback. In the batch stage, data from the most recent few hours is uploaded to the offline simulation platform. ANFIS+RLS is used to uniformly re-evaluate the membership functions and consequent coefficients. Finally, the controller is deployed after successful simulation verification. To ensure safety, all online updates limit the magnitude of single changes and monitor performance for several windows after the update. If performance deteriorates, automatic rollback occurs.

[0050] Please see Figure 3 , Figure 3 This is a schematic diagram of the control system for an active mass damper of a wind turbine according to the present invention. Corresponding to the aforementioned embodiment of the control method for an active mass damper of a wind turbine according to the present invention, the present invention also provides a control system for an active mass damper of a wind turbine, including a floating wind turbine, the floating wind turbine including a column, a plurality of floats spaced apart along the circumference of the column, and a support structure connecting the column and the floats, wherein an active mass damper is provided radially inside the floats; The control system for the active mass damper of a wind turbine of the present invention further includes: The acquisition module 1 is used to acquire environmental information, the status information of the floating wind turbine, and the operating information of each active mass damper. It also preprocesses the acquired environmental information, status information, and operating information to obtain the current state vector. The environmental information includes wind speed data and wave data. The status information includes IMU data, stress data, torque data, and tilt angle data at multiple different locations of the floating wind turbine. The operating information includes the position data of the mass blocks in the active mass dampers and the force data acting on the mass blocks. The acquisition module 2 is used to acquire the state vector within a preset historical window length as the historical state vector. The current state vector and the historical state vector are input into the fuzzy short-time prediction module. The fuzzy short-time prediction module outputs the predicted state vector of the future time window and the overall prediction confidence based on the pre-built fuzzy rule base, and obtains the local prediction confidence corresponding to each active mass damper based on the overall prediction confidence. Calculation module 3 is used to calculate the control force of the active mass damper based on the current state vector, the predicted state vector, and the local prediction confidence of each active mass damper. in, (1); (2); (3); In the formula, For the first The control force of an active mass damper For the first Feedforward control force of an active mass damper The first state vector is calculated using LQR control. The state feedback control force of an active mass damper For the first The fusion weights of the active mass dampers For saturation mapping to , , , These are the weighting coefficients. For the first Local response of the platform at the location of an active mass damper For the first The maximum root mean square value of the local response of the platform at the location of each active mass damper within the historical operating window. For the first The ratio of the remaining stroke of the mass block in an active mass damper To predict the feedforward gain matrix; The future number of the output of the fuzzy short-time prediction module The platform predicts the state vector.

[0051] Furthermore, the data acquisition module 1 includes: A submodule is established to create a discrete state-space model based on a floating wind turbine and an active mass damper: (4); In the formula, The input vector includes state information and position information. for time, This is the dynamic vector, which includes the force data of each active mass damper. This is a perturbation vector, which includes environmental information. and Let be the coupling matrix. Here is the perturbation matrix; The identification submodule is used to adaptively identify discrete state-space models using a recursive least squares algorithm. Construct a submodule for building the cost function: (5); In the formula, Let cost function be The state weight matrix is... To control the weight matrix; The first computational submodule is used to obtain the gain based on the discrete state-space model and cost function. K And calculate the control force: (6); In the formula, The control vector is composed of the control forces of each active mass damper.

[0052] Furthermore, the acquisition module 1 also includes: The processing submodule is used to process the control force calculated through LQR control. Perform saturation constraint processing: (10).

[0053] Furthermore, the acquisition module 1 also includes: The first judgment submodule is used to determine whether each parameter in the identification result exceeds the preset range. If so, the parameter is saturated and an alarm is recorded.

[0054] Furthermore, the computing module 3 includes: The second judgment submodule is used to determine whether each active mass damper meets the triggering condition. If so, the active mass damper is recorded as a retreat damper. The triggering condition is that the displacement of the mass block reaches the first preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper exceeds a preset safety threshold. The marking module is used to remove the retreat damper from control and to mark the active mass dampers other than the retreat damper as healthy dampers. The second calculation submodule is used to calculate the control forces of the retreat damper and the health damper according to formula (1), formula (2) and formula (3), and obtain the control force that cannot be executed after the retreat damper exits control and the original control force of the health damper. The third calculation submodule is used to calculate the compensating control force of the health damper according to formula (7): (7); In the formula, For the first The compensating control force of a health damper For the first The original control force of a healthy damper To transfer from the retraction damper to the first The contribution factor of each health damper This refers to the control force that cannot be executed after the damper exits control.

[0055] Furthermore, the computing module 3 also includes: The third judgment submodule is used to determine whether the retreat damper meets the recovery condition. If so, the control force of the retreat damper is gradually restored from 0 to the control force calculated by formula (1), formula (2) and formula (3). The recovery condition is that the displacement of the mass block is less than the second preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper is lower than the preset safety threshold. Cancel submodule, used to cancel the mark of the damper after the control force of the damper is restored to the control force calculated by formula (1), formula (2) and formula (3).

[0056] Furthermore, module 2 includes: The fourth calculation submodule is used to calculate the local prediction confidence level for each active mass damper based on the overall prediction confidence level, according to the following formula: (8); (9); in, The overall prediction confidence level output by the fuzzy short-time prediction module. For the first The correlation coefficient of the platform response at the location of each active mass damper For the first The current actual platform response at the installation location of each active mass damper. The first fuzzy short-time prediction module obtained Response of an active mass damper at its installation location. This is the local response error scale parameter.

[0057] Furthermore, the system also includes: The operation module is used to control the action of each active mass damper according to the control force of each active mass damper, and to collect the operation information of each active mass damper. The first judgment module is used to calculate the deviation between the force data of the mass block and the target value of the control force based on the running information, and to determine whether the deviation exceeds the preset deviation threshold. The correction module is used to correct the control force of the active mass damper if the judgment module determines that it is correct.

[0058] Furthermore, the system also includes: The metrics module is used to calculate performance metrics in real time, including attitude and acceleration. Tower foundation stress statistics At least one of the cumulative energy consumption and cumulative fatigue indicators; The second judgment module is used to determine whether the performance indicators deviate from the target. If so, it performs predictive control parameter self-tuning. The predictive control parameters include adjusting the parameters of the fuzzy short-time prediction module, the predictive fusion weights, and the parameters of the online identification model, so that the predictive control part can adapt to the dynamic characteristics of the platform.

[0059] The implementation process of the functions and roles of each module and submodule in the above system is detailed in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0060] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units.

[0061] Corresponding to the aforementioned embodiments of the control method for an active mass damper of a wind turbine, the present invention also provides an electronic device, which may include: a processor; a memory for storing an executable computer program; wherein, when the processor executes the computer program, it implements the control method for an active mass damper of a wind turbine in any of the aforementioned method embodiments.

[0062] The embodiments of the control and system for active mass dampers of wind turbines provided in this invention can all be applied to electronic devices. Taking software implementation as an example, as a logical device, it is formed by the processor of the electronic device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 4 As shown, except Figure 4In addition to the processor, memory, network interface, and non-volatile memory shown, the electronic device may also include other hardware, such as a camera module; or, depending on the actual function of the electronic device, it may also include other hardware, which will not be elaborated further.

[0063] Corresponding to the aforementioned embodiments of the control method for an active mass damper of a wind turbine, this embodiment of the invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the control method for an active mass damper of a wind turbine in any of the aforementioned method embodiments.

[0064] Embodiments of the present invention may take the form of a computer program product implemented on one or more storage media containing program code (including but not limited to disk storage, CD-ROM, optical storage, etc.). The computer-readable storage medium may include: permanent or non-permanent removable or non-removable media. The information storage function of the computer-readable storage medium can be implemented by any feasible method or technology. The information may be computer-readable instructions, data structures, program models, or other data.

[0065] In addition, computer-readable storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or other non-transfer media that can be used to store information accessible by a computing device.

[0066] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A control method for an active mass damper in a wind turbine, characterized in that, The invention includes a floating wind turbine, which comprises a column, a plurality of pontoons spaced apart along the circumference of the column, and a support structure connecting the column and the pontoons. An active mass damper is provided inside the pontoons along their radial direction. The method includes the following steps: The system collects environmental information, the status information of the floating wind turbine, and the operational information of each active mass damper. It then preprocesses the collected environmental information, status information, and operational information to obtain the current state vector. The environmental information includes wind speed data and wave data. The status information includes IMU data, stress data, torque data, and tilt angle data at multiple different locations of the floating wind turbine. The operational information includes the position data of the mass blocks in the active mass dampers and the force data acting on the mass blocks. The state vector within a preset historical window length is obtained as the historical state vector. The current state vector and the historical state vector are input into the fuzzy short-time prediction module. The fuzzy short-time prediction module outputs the predicted state vector of the future time window and the overall prediction confidence based on the pre-built fuzzy rule base. Based on the overall prediction confidence, the local prediction confidence corresponding to each active mass damper is obtained. Based on the current state vector, the predicted state vector, and the local prediction confidence level of each active mass damper, calculate the control force of the active mass damper; in, (1); （2）； （3）； In the formula, For the first The control force of an active mass damper For the first Feedforward control force of an active mass damper The first state vector is calculated using LQR control. The state feedback control force of an active mass damper For the first The fusion weights of the active mass dampers For saturation mapping to , , , These are the weighting coefficients. For the first Local response of the platform at the location of an active mass damper For the first The maximum root mean square value of the local response of the platform at the location of each active mass damper within the historical operating window. For the first The ratio of the remaining stroke of the mass block in an active mass damper To predict the feedforward gain matrix; The future number of the output of the fuzzy short-time prediction module The platform predicts the state vector.

2. The control method for an active mass damper of a wind turbine according to claim 1, characterized in that, The first calculated by LQR control The steps for controlling the force of an active mass damper include: Based on the floating wind turbine and active mass damper, a discrete state-space model is established: （4）； In the formula, The input vector includes state information and position information. for time, This is the dynamic vector, which includes the force data of each active mass damper. This is a perturbation vector, which includes environmental information. and Let be the coupling matrix. Here is the perturbation matrix; A recursive least squares algorithm is used to adaptively identify discrete state-space models. Construct the cost function: （5）； In the formula, Let cost function be The state weight matrix is... To control the weight matrix; Gain is obtained based on discrete state-space model and cost function. K And calculate the control force: （6）； In the formula, The control vector is composed of the control forces of each active mass damper.

3. The control method for an active mass damper of a wind turbine according to claim 2, characterized in that, Following the step of adaptively identifying the discrete state-space model using the recursive least squares algorithm, the method further includes: For each parameter included in the identification result, determine whether the parameter exceeds the preset range. If so, apply saturation processing to the parameter and record an alarm.

4. The control method for an active mass damper of a wind turbine according to claim 1, characterized in that, The steps for calculating the control force of the active mass damper include: For each active mass damper, determine whether the active mass damper meets the triggering condition. If so, the active mass damper is recorded as a retreat damper. The triggering condition is that the displacement of the mass block reaches the first preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper exceeds a preset safety threshold. Remove the retreat damper from control and designate the active mass dampers other than the retreat damper as healthy dampers; The control forces of the retreat damper and the health damper are calculated according to formulas (1), (2) and (3), and the control forces that cannot be executed after each retreat damper exits control and the original control forces of each health damper are obtained. Calculate the compensating control force of the health damper according to formula (7): （7）； In the formula, For the first The compensating control force of a health damper For the first The original control force of a healthy damper To transfer from the retraction damper to the first The contribution factor of each health damper This refers to the control force that cannot be executed after the damper exits control.

5. The control method for an active mass damper of a wind turbine according to claim 4, characterized in that, After the step of calculating the compensating control force of the health damper according to formula (7), the method further includes: Determine whether the retreat damper meets the recovery condition. If so, gradually restore the control force of the retreat damper from 0 to the control force calculated by formula (1), formula (2) and formula (3). The recovery condition is that the displacement of the mass block is less than the second preset percentage of the maximum allowable stroke, or the temperature rise of the active mass damper is below the preset safety threshold. Once the control force of the retreating damper is restored to the control force calculated by formulas (1), (2) and (3), the retreating damper is deactivated.

6. The control method for an active mass damper of a wind turbine according to claim 1, characterized in that, The step of obtaining the local prediction confidence of each active mass damper based on the overall prediction confidence includes: Based on the overall prediction confidence level, the local prediction confidence level for each active mass damper is calculated using the following formula: （8）； （9）； in, The overall prediction confidence level output by the fuzzy short-time prediction module. For the first The correlation coefficient of the platform response at the location of each active mass damper. For the first The current actual platform response at the installation location of each active mass damper. The first fuzzy short-time prediction module obtained Response of an active mass damper at its installation location. This is the local response error scale parameter.

7. The control method for an active mass damper of a wind turbine according to claim 1, characterized in that, The method further includes the following steps: The operation of each active mass damper is controlled by the control force of each active mass damper, and the operating information of each active mass damper is collected. Based on the operational information, the deviation between the force data acting on the mass block and the target value of the control force is calculated, and it is determined whether the deviation exceeds the preset deviation threshold. If so, the control force of the active mass damper is modified.

8. A control system for an active mass damper of a wind turbine, characterized in that, The invention includes a floating wind turbine, which comprises a column, a plurality of pontoons spaced apart along the circumference of the column, and a support structure connecting the column and the pontoons. An active mass damper is provided inside the pontoons along their radial direction. The system also includes: The acquisition module is used to acquire environmental information, the status information of the floating wind turbine, and the operating information of each active mass damper. It also preprocesses the acquired environmental information, status information, and operating information to obtain the current state vector. The environmental information includes wind speed data and wave data. The status information includes IMU data, stress data, torque data, and tilt angle data at multiple different locations of the floating wind turbine. The operating information includes the position data of the mass blocks in the active mass dampers and the force data acting on the mass blocks. The acquisition module is used to acquire the state vector within a preset historical window length as the historical state vector. The current state vector and the historical state vector are input into the fuzzy short-time prediction module. The fuzzy short-time prediction module outputs the predicted state vector of the future time window and the overall prediction confidence based on the pre-built fuzzy rule base, and obtains the local prediction confidence corresponding to each active mass damper based on the overall prediction confidence. The calculation module is used to calculate the control force of the active mass damper based on the current state vector, the predicted state vector, and the local prediction confidence of each active mass damper. in, (1); （2）； （3）； In the formula, For the first The control force of an active mass damper For the first Feedforward control force of an active mass damper The first state vector is calculated using LQR control. The state feedback control force of an active mass damper For the first The fusion weights of the active mass dampers For saturation mapping to , , , These are the weighting coefficients. For the first Local response of the platform at the location of an active mass damper For the first The maximum root mean square value of the local response of the platform at the location of each active mass damper within the historical operating window. For the first The ratio of the remaining stroke of the mass block in an active mass damper To predict the feedforward gain matrix; The future number of the output of the fuzzy short-time prediction module The platform predicts the state vector.

9. An electronic device, characterized in that, include: processor; Memory is used to store executable computer programs; Wherein, when the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images