Method for controlling an offshore wind energy system

Abstract

The invention discloses a method for controlling an offshore wind energy system, relating to the technical field of offshore wind energy control. Said offshore wind energy system comprises a wind turbine, a converter, vibration and strain sensors, and a controller, and aims to solve the technical problem of effectively inhibiting key structural loads while ensuring power output.The key points of its technical solution are as follows: obtaining a forward wind speed forecast sequence via a nacelle lidar, while simultaneously collecting real-time tower and blade load data; constructing a parameterized state-space model and using the traceless Kalman filter algorithm to perform online identification and state estimation of aerodynamic parameters and unmodeled dynamics; in each control period, with power tracking and load inhibition as multi-objectives, solving the finite-horizon optimal control problem based on the updated model and the forecast wind speed sequence; and finally, analyzing and executing the optimal incremental pitch angle control and the optimal incremental generator torque control. This method is primarily used to improve the operating life and stability of offshore wind turbines.

Description

1 Method for controlling an offshore wind energy system Technical Field The invention relates to the technical field of wind energy control. More specifically, the present invention relates to a method for controlling an offshore wind energy system. 5 Prior Art Offshore wind energy is an important form of renewable energy use, but its operating environment is complex and hostile, and the wind turbine is subjected to the combined action of irregular aerodynamic loads and long-term wave loads. This leads to a propensity of the key structural components of the turbine (such as the tower and blades) to suffer fatigue damage, which in turn affects the operating life and reliability of the entire system. Therefore,How to effectively suppress structural loads while pursuing high energy generation efficiency is a persistent technical problem in the field of offshore wind energy system control. The control strategy for traditional wind turbines relies mainly on feedback control: it consists of measuring, via sensors, real-time states such as the rotation speed of the wind turbine wheel and the output power, then adjusting the offset angle and the generator torque to achieve maximum power point tracking (MPPT) or a stable power output. However, this control method has an inherent delay: when wind speed changes abruptly, strong turbulence and wind shear already act on the wind turbine and cause irreversible load shocks25 before the control system begins to react, thus limiting the control's effectiveness. This frequently exposes the assembly structure to shocks from hostile weather conditions such as wind gusts,This is not conducive to stable long-term operation. To improve control performance, the anticipatory action control (AAC) technology was introduced, the core of which lies in using wind speed forecast information to make advance control decisions. Initial studies 2025 / 5809 BE2025 / 5809 2 attempted to use anemometers to measure wind speed inside the turbine as an anticipatory action signal. However, this measurement point is located behind the wind turbine, and the wind speed obtained is perturbed by the turbine itself, not being the actual incident wind acting primarily on the entire surface of the turbine. Consequently, the forecast information lacks accuracy and comprehensiveness, making it difficult to accurately predict future aerodynamic loads. Furthermore, establishing an accurate dynamic model of the wind farm is the basis for designing advanced control algorithms such as high-performance model prediction control (MPC). However, the offshore wind farm is a complex multi-body coupling system.whose model comprises several subsystems such as aerodynamic elasticity, the transmission chain, the generator, and the control system. Among these, the aerodynamic parameters (such as the aerodynamic torque and the thrust coefficient) evolve significantly as a function of the offset angle, the ratio of the blade tip speed, as well as the surface roughness and freezing conditions of the blades, exhibiting strong temporally variable characteristics. If fixed aerodynamic parameters are adopted in the control model, this will lead to a mismatch between the model and the real system; particularly in complex turbulent wind conditions, this mismatch will severely affect the performance of the model-based control algorithm, and could even cause control destabilization. Although offline system identification allows obtaining the model parameters, it cannot adapt to the temporally variable characteristics of the parameters during operation. Consequently,How to perform online and real-time identification of key aerodynamic parameters and unmodeled dynamics constitutes a major challenge. Finally, the controller design itself is a multi-objective optimization problem, which requires achieving a balance between monitoring the electrical power setpoint and suppressing structural loads. These two objectives are mutually contradictory in many cases; for example, reducing loads by rapid blade pitch can cause power fluctuations. At the same time, the control actions must strictly comply with the 2025 / 5809 BE2025 / 5809 3 physical constraints of the actuator, such as the limits of the rate of change of the offset angle and the rate of change of the torque. Under conditions where there is the anticipated effect of wind speed, model uncertainty, and multiple constraints, the real-time resolution of this type of complex optimal control problem imposes extremely high demands on the computational capacity of the controller. In the past,Due to the limitations of perception technologies, online identification algorithms, and real-time computing hardware, it was difficult to implement in practical engineering a control process capable of efficiently using wind forecast information, adaptively updating the model, and efficiently solving the multi-objective optimization problem with constraints simultaneously. Description of the Invention: One object of the present invention is to solve at least the problems mentioned above and to offer at least the advantages that will be explained later. Another object of the present invention is to provide a control method for an offshore wind energy system, to find solutions to the technical shortcomings of traditional control methods for offshore wind farms, which rely mainly on feedback control and exhibit a response delay.cannot effectively cope with the shocks of violent weather conditions such as wind gusts, causing serious fatigue damage to key structural components. To achieve these objectives and other advantages of the present invention, the invention provides a method for controlling an offshore wind energy system. Said offshore wind energy system comprises a wind turbine, a converter, vibration and stress sensors deployed at key structural points of the wind turbine, and a controller integrated within the wind turbine. The control method comprises the following steps: Step 1: Scan the wind field in front of the wind turbine using a lidar to obtain a forecast wind speed sequence; simultaneously,Collect real-time load measurement values ​​of the front-to-back bending moment of the tower and the bending moment at the base of the blades via vibration and stress sensors. 2025 / 5809 BE2025 / 5809 4 Step 2: Establish a parameterized state-space model including the aeroelastic dynamics and the dynamics of the generator's transmission chain; adopt a joint state and parameter estimation algorithm based on a traceless Kalman filter, take the forecast wind velocity sequence and the load measurement values ​​as observation inputs, perform the online and real-time identification and updating of the time-varying aerodynamic parameters and the unmodeled dynamics in the parameterized state-space model, and simultaneously estimate the state variables of the entire system. Step 3: In each control period, with the optimization objectives of minimizing power tracking error and minimizing structural loads, based on the updated parameterized state-space model,take the forecast wind speed sequence as input of known anticipated action perturbation and the state variables of the whole system estimated in Step 2 as initial state; solve the optimal control problem at finite horizon subject to the constraint conditions of the actuator such as the rate of change of the offset angle, the rate of change of the rotational speed of the wind turbine and the rate of change of the generator torque, and output the optimal control sequence in the future control horizon. Step 4: Analyze the first element of the optimal control sequence output in Step 3 into an optimal incremental control of the offset angle and an optimal incremental control of the generator torque for the current control period; superimpose these two incremental control commands respectively on the corresponding commands of the previous period, in order to generate and execute the actual control command of the offset angle and the actual control command of the generator torque for the current period. Preferably, in the process of controlling an offshore wind energy system,The process of obtaining the forecast wind velocity sequence using the ulidar scan data in Step 1 includes the following steps: Step 11: Reconstruct the four-dimensional spatio-temporal wind field in a predefined range in front of the wind turbine wheel based on the raw ulidar scan data, said four-dimensional spatio-temporal wind field including three-dimensional spatial coordinates and one-dimensional temporal coordinate. Step 12: In each control period, predict the state trajectory of the wind turbine in the future forecast horizon according to the system state variables estimated in Step 2 of the previous period and the optimal control sequence solved in Step 3 of the previous period, said state trajectory including at least the drift angle trajectory and the rotation angle trajectory of the wind turbine. Step 13: Establish a wind field-wind turbine coupling model.perform a coordinate transformation on the four-dimensional spatio-temporal wind field obtained in Step 11 according to the state trajectory predicted in Step 12, 10 and apply it dynamically in a rotating coordinate system moving with the future wind wheel, in order to generate a dynamic forecast incident wind field sequence acting on each axial element of each blade, indexed by the rotation angle of the wind wheel. Preferably, in the process of controlling an offshore wind energy system, the 15 process of obtaining the forecast wind speed sequence using the ulidar sweep data in Step 1 includes, in addition: based on the error between the actual rotation angle of the wind turbine measured in real time and the trajectory of the rotation angle of the wind turbine predicted in Step 12, performing an online correction of the coordinate transformation parameters in the 20 wind field-wind turbine coupling model,the specific steps being as follows: Step a: Measure in real time the actual current rotation angle of the wind turbine wheel using an encoder installed on the main shaft of the wind turbine; Step b: Compare the actual rotation angle with the trajectory value of the rotation angle of the wind turbine wheel at the current time predicted in Step 12, and generate a 25 error signal of the rotation angle of the wind turbine wheel; Step c: Introduce the error signal of the rotation angle of the wind turbine wheel into a predefined proportional-integral (PI) controller, said proportional-integral (PI) controller delivering an azimuth offset correction quantity used to compensate for the dynamic mapping process in the wind field-wind turbine coupling model; Step d: Use the azimuth offset correction quantity to adjust in time 2025 / 5809 BE2025 / 5809 6 real rotation matrix parameters in the coordinate transformation, and generate a corrected dynamic forecast incident wind field sequence acting on each axial element of each blade. Preferably,in the process of controlling an offshore wind energy system, the resolution of the optimal finite horizon control problem in Step 3 specifically includes the following steps: Step 31: On the basis of the updated parameterized state space model, transform the optimization objectives, which are minimizing the power tracking error and minimizing structural loads, into an objective function, and take into account the constraint conditions of the actuator to construct an optimal continuous-time control problem; Step 32: Adopt the orthogonal interpolation point method to discretize the state variables and control variables in the future forecast horizon over a series of interpolation points, and transform the continuous-time optimal control problem into a nonlinear programming problem; Step 33: Use a real-time iteration algorithm to solve the nonlinear programming problem; in the process of solving each control period,Take the optimal control sequence and the state trajectory solved in the previous control period as initial guess values ​​to present the solution, and perform the iterative calculation to accelerate convergence; Step 34: After reaching the predefined solution accuracy or the maximum number of iterations, output the optimal control sequence of the current control period. Preferably, in the control process of an offshore wind energy system, the joint state and parameter estimation algorithm in Step 2 adjusts in real time the process noise decovariance matrix and the measurement noise decovariance matrix of the traceless Kalman filter by introducing an adaptive mechanism, based on the innovation sequence or the residual sequence. Preferably, in the process of controlling an offshore wind energy system, before Step 4,The process further comprises the following steps: filtering respectively the optimal incremental control of the offset angle and the optimal incremental control of the generator torque outputs in Step 3 by a first-order inertial element, whose time constant is obtained by 2025 / 5809 BE2025 / 5809 7 identification of the dynamic response characteristics of the blade offset mechanism and the converter. Preferably, in the process of controlling an offshore wind energy system, the fitting of the process noise covariance matrix of the joint state and parameter estimation algorithm based on the traceless Kalman filter in Step 2 is carried out by adaptive scaling according to the turbulence intensity value of the forecast wind speed sequence. Preferably, in the process of controlling an offshore wind energy system,The length of the resolution horizon for the optimal control problem at a finite horizon in Step 3 is dynamically adjusted according to the dominant frequency of the wind shear and the turning effect contained in the forecast wind speed sequence. Preferably, in the process of controlling an offshore wind energy system, the time-varying aerodynamic parameters identified online and in real time in Step 2 include at least the equivalent time constant of the dynamic stall model and the relaxation time constant of the induction factor of the incident dynamic flow model. Preferably, in the process of controlling an offshore wind energy system, in Step 4, before generating the actual control commands for the current period,An independent consistency check and a balanced distribution of 20, the optimal incremental control of the offset angle are performed according to the measured offset angles of the three blades. The present invention includes at least the following beneficial effects. 1. The present invention achieves an early perception of future disturbances by predicting the wind field via a lidar, providing a basis for anticipatory action control. Combined with joint state and parameter estimation based on the traceless Kalman filter, it can track variations in the system's dynamic characteristics online and maintain model accuracy. Finally, via optimal finite-horizon control, it integrates anticipatory action information, the updated model, and multi-objective optimization. Subject to actuator constraints, it can significantly reduce fatigue loads on key structural components while maintaining good power tracking performance.1. This effectively extends the lifespan of the turbine, improving its operational stability and economic benefits. 2. The present invention, by reconstructing the four-dimensional spatio-temporal wind field and predicting the state trajectory of the turbine, precisely maps the wind field information in a fixed reference frame onto each element of each blade of the future rotating wind turbine. This means that the predicted wind speed sequence is no longer a sequence from a single spatial point, but an incident wind field directly usable for calculating aerodynamic loads, corresponding precisely to the rotating position of the wind turbine. It significantly improves the accuracy and usability of the anticipated action information, providing a solid basis for precise controls such as high-performance independent timing. 3. The present invention, by introducing a closed-loop correction mechanism based on the real-time measurement of the wind turbine's angle of rotation,can effectively compensate for coordinate transformation errors due to model prediction deviations and external disturbances. It thus guarantees high-precision synchronization between the predicted wind field sequence and the actual position of the wind turbine wheel, significantly improving the reliability and robustness of the control by anticipating action, avoiding the potential negative control effects of misalignment of mapping and ensuring stable operation of control performance. 4. The present invention adopts the orthogonal interpolation point method to transform the continuous-time optimal control problem into a nonlinear programming problem. Combined with the real-time iteration algorithm, which uses the solution from the previous control period as the initial guess value for the iterative computation, it significantly accelerates the convergence speed of the solution process. It efficiently resolves the conflict between the large computational quantity of the complex optimization problem and the short control period.ensuring that the optimal control problem can obtain a high-precision solution within the time range allowed by the engineering applications and realizing the high-frequency, real-time operation of the control algorithm. 5. The present invention adjusts the noise covariance matrix of the 2025 / 5809 BE2025 / 5809 9 Kalman filter in real time via an adaptive mechanism, allowing the filter to dynamically adjust its degree of confidence in the model prediction and measurement information according to the actual operating state of the system. When the dynamic variations of the system are intense or the measurements present anomalies, it can automatically adjust the filtering lean, significantly strengthening the robustness and estimation accuracy of the state and parameter estimation algorithm in the face of uncertainties and disturbances. 6. The present invention filters the optimal incremental control of theoretical control based on the dynamic response characteristics of the actuator,making the variations of the commands finally sent to the actuator smooth and physically feasible. This avoids placing excessive demands on the actuator, reduces delay and overshoot during execution, and ensures that the control commands are executed accurately and stably, thus improving the overall performance and elasticity of the control loop. 7. The present invention associates the covariance of process noise with the intensity of turbulence in the forecast wind field, achieving an adaptive adaptation between the filter parameters and the intensity of external wind disturbances: in conditions of strong turbulence, it automatically increases the covariance of process noise, indicating an increase in model uncertainty and causing the filter to place more trust in the measured values, thus allowing the system variations to be followed more quickly; conversely, in stable wind conditions, it reduces the covariance of process noise,further optimizing filter performance under different wind conditions and improving estimation accuracy. 8. The present invention dynamically adjusts the length of the prediction horizon 25 according to the dominant frequency of wind shear and the turn effect, ensuring that the controller's prediction range can cover one or more complete cycles of these periodic disturbances. This allows the optimized controller to "see" and plan in advance the response to disturbances over the entire cycle, thus more effectively inhibiting vibrations and loads 30 caused by these periodic sources and improving the control's capability and efficiency. 9. The present invention identifies online the key aerodynamic process time constants such as dynamic stall and incident dynamic flow, instead of adopting fixed values. This allows the model to more faithfully reflect the actual response rate of aerodynamics under different operating conditions.significantly improving the accuracy of aerodynamic load calculations, then the accuracy of state estimation and model prediction-based control performance, especially in the face of complex nonlinear phenomena such as dynamic stall. 10. The present invention achieves independent verification and balanced distribution of the optimal offset angle increment calculated in a unified manner, based on the actual positions of each blade. This compensates for asymmetries due to differences between actuators and wind shear, ensuring that all three blades ultimately receive commands that allow for better load balancing, further reducing asymmetric fatigue loads on key components such as the hub, and optimizing the overall performance of the independent offset control. 15. Other advantages, objectives, and features of the present invention will become apparent in part from the following description.and in part will be understood by a person skilled in the art through research and practice of the present invention. Description of embodiments With reference to the drawings below, the present invention will be described in more detail, 20 so that technical personnel skilled in the art can implement it with reference to the text of the description. It should be understood that terms such as “have”, “include” and “include” used herein do not exclude the presence or addition of one or more other elements or combinations thereof. 25 It should be specified that, for the experimental methods described in the following embodiments, unless otherwise indicated, they are all conventional methods; and for the reagents and materials mentioned, unless otherwise indicated, they can be obtained from commercial sources. In one embodiment of the present invention, a method for controlling an offshore wind energy system is provided. Said offshore wind energy system comprises a wind turbine, a converter,vibration and stress sensors deployed at key structural points of the wind turbine assembly, as well as an integrated controller within the wind turbine assembly. The control process includes the following steps: Step 1: Scan the wind field in front of the wind turbine wheel using a lidar to obtain a forecast wind speed sequence; simultaneously, collect in real time 5 real-time load measurement values ​​of the front-to-back bending moment of the tower and the bending moment at the base of the blades via the vibration and stress sensors. Step 2: Establish a parameterized state-space model including aeroelastic dynamics and generator transmission chain dynamics; adopt a joint state and parameter estimation algorithm based on a traceless Kalman filter; take the forecast wind velocity sequence and load measurement values ​​as observation inputs; perform online and real-time identification and updating of the time-varying aerodynamic parameters and the unmodeled dynamics in the parameterized state-space model.and simultaneously estimate the state variables of the entire system. Step 3: In each control period, with the optimization objectives of minimizing the power tracking error and minimizing structural loads, based on the updated parameterized state-space model, take the 20 forecast wind speed sequence as the input of the known anticipated action perturbation and the state variables of the entire system estimated in Step 2 as the initial state; solve the optimal finite-horizon control problem subject to the actuator constraint conditions such as the rate of change of the offset angle, the rate of change of the wind turbine rotational speed and the rate of change of the generator torque,and output the optimal control sequence in the future control horizon. Step 4: Analyze the first element of the optimal control sequence output in Step 3 into an optimal incremental control of the offset angle and an optimal incremental control of the generator torque for the current control period; superimpose these two incremental control commands respectively on the corresponding control commands of the previous period, in order to generate and execute the actual control control command of the offset angle and the actual control command of the generator torque for the current period. Traditional control methods for offshore wind farms rely on a feedback mechanism, exhibit a delayed response, and struggle to cope with severe weather conditions such as wind gusts. This results in key structural components like the tower and blades bearing irreversible fatigue loads over the long term, while also suffering from a lack of power output stability, which affects the lifespan and reliability of the farm. Furthermore,Traditional processes lack the ability to accurately anticipate wind speed and update the system model in real time, thus preventing coordinated optimization control between power and loads under changing wind conditions. In the present embodiment, the specific technical process is as follows: Step 1: Perception and acquisition of anticipated wind speed and load information: Use a lidar installed in front of the wind turbine nacelle to perform continuous wind field sweep measurements over a certain range in front of the wind turbine. Based on the raw lidar sweep data, generate a four-dimensional spatio-temporal forecast wind speed sequence for the next tens of seconds using a wind field reconstruction algorithm. At the same time,Use vibration and stress sensors previously deployed at key structural points of the tower and blade base to collect, in real time and synchronously, the bending moment signals in the front-to-back direction of the tower and the bending moment signals at the base of each blade, which serve as measurement values ​​reflecting the structural loads. Step 2: Adaptive online update of the system model and full state estimation: Establish a parameterized state-space model merging the aeroelastic dynamics of the wind turbine and the dynamics of the generator transmission chain as the base model. Adopt a joint state-space and parameter estimation algorithm based on a traceless Kalman filter, taking the predicted wind speed sequence and the load measurement values ​​acquired in Step 1. BE2025 / 5809 13 as observation inputs, and continuously perform online and real-time identification and updating of key temporally variable aerodynamic parameters and unmodeled dynamics in the model. By means of the algorithm,During the model parameter update, simultaneously estimate the state variables of the entire system, including the rotational speed of the wind turbine, the torque of the transmission chain, the front-to-back bending moment of the tower, the deflection moment of the blades, etc., in order to provide the controller with complete information on the current state of the system. Step 3: Multi-objective optimization solution with constraints based on the anticipated action and the updated model: In each control period, take minimizing the error tracking of generation power and minimizing structural loads as dual optimization objectives. Based on the last parameterized state-space model updated in Step 2,Consider the predicted wind speed sequence obtained in Step 1 as the input for the anticipated action perturbation15, and the state variables of the entire system estimated in Step 2 as the initial state of the optimization problem. Construct and solve a finite-horizon optimal control problem subject to strict physical constraint conditions of the actuator such as the variation of the offset angle.The rate of change of the rotational speed of the wind turbine and the rate of change of the torque of the generator. Calculate the optimal control sequence of the offset angle and the torque of the generator for a future control period using a numerical optimization algorithm. Step 4: Analysis and Execution of Optimal Control Commands: Analyze the first element of the optimal control sequence resolved in Step 325 (i.e., the amount of control to be executed immediately) into an optimal incremental command for the offset angle and an optimal incremental command for the generator torque for the current control period. Superimpose these two incremental commands, respectively, on the corresponding commands from the previous period to generate the final actual control commands for the offset angle and the generator torque for the current period. Then, send these two commands, respectively, to the blade offset execution system and the converter for execution.realizing thus simultaneously the 2025 / 5809 BE2025 / 5809 14 precise control of generator power and effective inhibition of key structural loads. In the present embodiment, the wind field in front of the wind turbine is first scanned using a lidar installed at the front of the crane to obtain a forecast wind speed sequence, while vibration and stress sensors are used to collect real-time load data for the rotor and blades. Next, a parameterized state-space model, including aerodynamic elasticity and transmission chain dynamics, is established. The traceless Kalman filter algorithm is adopted, which combines the forecast wind speed and measured loads to perform online identification and state estimation of the aerodynamic parameters and the dynamics not modeled in the model. In each control period, with the objectives of minimizing power tracking error and minimizing structural loads,The optimal finite-horizon control problem is solved based on the updated model and the forecast wind speed, subject to the physical constraints of the actuator, in order to obtain the optimal incremental control of the offset angle and the optimal incremental control of the generator torque. Finally, these incremental commands are superimposed on the corresponding commands from the previous period to generate the actual control commands, which are then transmitted and executed. Compared to the closest prior technique, traditional methods generally rely solely on feedback control or simple anticipatory action, lacking the ability to accurately perceive the future wind field and adaptively update the model. Consequently, their response is slow in gust or turbulent wind conditions, and their efficiency is limited. Control is limited. The present embodiment performs wind field forecasting via lidar, combined with online model identification and multi-objective optimization control.significantly improving the system's prospecting and adaptability. It can simultaneously guarantee power output and structural safety under complex wind conditions. 30 The technical solution provided by the present embodiment improves the control system's ability to predict future wind disturbances, enhances the model's accuracy under different operating conditions, and achieves a balanced optimization between power and loads. It reduces fatigue damage to key structures, extends the unit's lifespan, improves operational resilience and overall economic benefits, and exhibits strong engineering applicability and robustness. It is suitable for the intelligent control of all types of offshore wind farms. 5 In another embodiment of the present invention,The process of obtaining the forecast wind velocity sequence using lidar scan data in Step 1 includes the following steps: Step 11: Reconstruct the four-dimensional spatio-temporal wind field in a predefined range in front of the wind turbine wheel based on the raw lidar scan data, said four-dimensional spatio-temporal wind field including three-dimensional spatial coordinates and a one-dimensional temporal coordinate; Step 12: In each control period, predict the state trajectory of the wind turbine in the future forecast horizon according to the system state variables estimated in Step 2 of the previous period and the optimal control sequence solved in Step 3 of the previous period, said state trajectory including at least the drift angle trajectory and the rotation angle trajectory of the wind turbine; Step 13: Establish a wind field-wind turbine coupling model,to perform a 20 coordinate transformation on the four-dimensional spatio-temporal wind field obtained in Step 11 according to the state trajectory predicted in Step 12, and to apply it dynamically in a rotating coordinate system moving with the future wind wheel, in order to generate a dynamic forecast incident wind field sequence acting on each axial element of each blade, 25 indexed by the rotation angle of the wind wheel. In traditional methods, the wind turbine lidar measures a fixed wind velocity sequence at a single point or spatial plane, while the wind turbine is in continuous rotation and each blade experiences a different actual incident wind velocity depending on its azimuth angle. Directly using the wind velocity sequence from a fixed point for control does not accurately reflect the specific impacts of three-dimensional wind disturbances such as wind shear and the turning effect on each blade.resulting in a significant spatial discrepancy between anticipated action information and the actual aerodynamic loads of the wind turbine, and limiting the performance improvement of precise control strategies such as independent calibration. Given the above technical problem, the present embodiment 5 begins by processing raw LiDAR scan data to reconstruct a four-dimensional spatiotemporal wind field within a predefined three-dimensional range in front of the wind turbine and for a future period. In each control period, based on the system state estimated in the previous period and the optimal control sequence solved in the previous period, the state trajectory of the wind turbine in the future forecast horizon is predicted, in particular the drift angle trajectory and the rotation angle trajectory of the wind turbine. Then,A wind-wheel coupling model is established: the four-dimensional spatio-temporal wind field reconstructed in the fixed coordinate system undergoes a coordinate transformation according to the predicted future rotation state of the wind turbine and is dynamically mapped into a rotating coordinate system moving with the future wind turbine. Through this process, a dynamic sequence of the predicted incident wind field indexed by the rotation angle of the wind turbine is generated. This sequence precisely describes the wind velocity acting on each axial element of each blade each time the wind turbine rotates at a specific angle during the future period. Compared to the closest previous technique, traditional methods generally simply use the wind speed measured by lidar at a point ahead or along a linear planar surface as an anticipated action signal, neglecting the variations in spatial deposition due to the rotation of the wind turbine and the inhomogeneity of the three-dimensional distribution of the wind field. The present implementation,by reconstructing the four-dimensional wind field and creating a dynamic coordinate map based on the predicted movement trajectory of the wind turbine, transforms the fixed wind field information into a sequence of 30 incident wind fields corresponding precisely to the rotating pale elements, thus resolving the central problem of the spatial mapping disagreement. The technical solution provided by the present embodiment considerably improves the spatial accuracy and usability of the anticipated wind speed information, transforming the "wind from a single point" forecast information into "wind from a single blade." It provides direct and precise input conditions for high-performance independent calibration, thus providing a solid basis for more effectively inhibiting periodic load fluctuations caused by the inhomogeneity of spatial wind distributions and improving the accuracy and efficiency of control. In another embodiment of the present invention,The process of obtaining the forecast wind speed sequence using lidar scan data in Step 1 includes further: based on the error between the real-time measured rotation angle of the wind turbine wheel and the predicted rotation angle trajectory of the wind turbine wheel in Step 12, performing an online correction of the coordinate transformation parameters in the wind field-wind turbine coupling model, the specific steps being as follows: Step a: Real-time measurement of the actual current rotation angle of the wind turbine wheel at 15 using an encoder installed on the main shaft of the wind turbine; Step b: Compare the actual rotation angle with the trajectory value of the wind turbine's rotation angle at the current time predicted in Step 12, and generate an error signal for the wind turbine's rotation angle; Step c: Input the error signal for the wind turbine's rotation angle into a predefined proportional-integral (PI) controller.said proportional-integral (PI) controller delivering an azimuth offset correction quantity used to compensate for the dynamic mapping process in the wind field-wind turbine coupling model; Stepd: Use the azimuth offset correction quantity to adjust in real time 25 parameters of the rotation matrix in the coordinate transformation, and generate a corrected dynamic forecast incident wind field sequence acting on each axial element of each blade. In the process of mapping the predicted wind field in the fixed coordinate system to the rotating coordinate system using the predicted trajectory of the wind turbine's rotation angle, errors may appear between the predicted trajectory and the actual rotation state of the wind turbine due to factors such as model prediction deviations,External disturbances or variations in system parameters. These errors lead to a desynchronization in spatial orientation between the predicted incident wind field sequence after mapping and the input wind field that the wind turbine blades will actually encounter, forming a systemic alignment error. If this misaligned pre-action information is used directly for control, it can not only effectively inhibit loads, but can also generate erroneous control actions, worsen load fluctuations, and reduce the reliability and practicality of pre-action control. The present embodiment measures in real time the actual rotation angle of the main shaft of the wind turbine assembly via a high-precision encoder, thus obtaining the true orientation of the wind turbine blade. In each control period, this angle The actual measured rotation is compared in real time to the trajectory value of the rotation angle of the wind turbine at the current moment, predicted from information15 of the previous period,generating an error signal for the rotation angle of the wind turbine representing the prediction error. This error signal is fed into a predefined proportional-integral (PI) controller, which delivers a corresponding azimuth offset correction quantity. This correction quantity is used to compensate for and adjust in real time the parameters of the rotation matrix on which the coordinate transformation in the wind field-wind turbine coupling model is based, thus correcting the dynamic mapping process online. Finally, the corrected parameters are used to generate a dynamic forecast incident wind field sequence acting on each axial element of each blade, which maintains high-precision synchronization with the true rotation orientation of the wind turbine. Compared to the closest prior technique, traditional methods generally perform the mapping transformation from a fixed coordinate system to a rotating coordinate system in an open loop.relying entirely on the prediction accuracy of the model and lacking the capacity for real-time verification and error correction of the accuracy of the mapping result. Once a prediction deviation occurs, the quality of all anticipated action information decreases without being detectable. The present embodiment 2025 / 5809 BE2025 / 5809 19 introduces a closed-loop correction mechanism based on direct physical measurements, creatively combining the anticipated action pathway with a fast feedback correction loop. It can automatically and continuously eliminate mapping errors due to imperfections in the prediction model and external disturbances.5 The technical solution provided by the present embodiment significantly improves the synchronization accuracy and long-term reliability between the Wind speed information, anticipated action, and the true spatial orientation of the wind turbine wheel. It effectively avoids the risk of reduced control performance due to anticipated action, or even negative effects.caused by the inaccuracy of the model prediction, it strengthens the robustness of the entire control system and ensures that the anticipatory action control strategy based on forecast information can stably develop its expected effectiveness in the real operating environment. In another embodiment of the present invention, solving the optimal finite-horizon control problem in Step 3 specifically comprises the following steps: Step 31: Based on the updated parameterized state-space model, transform the optimization objectives, which are minimizing power tracking error and minimizing structural loads, into an objective function, and take into account the actuator constraint conditions to construct a continuous-time optimal control problem; Step 32: Adopt the orthogonal interpolation point method to discretize the state variables and control variables in the future forecast horizon over a series of interpolation points,and transform the optimal control problem in 25 continuous time into a nonlinear programming problem; Step 33: Use a real-time iteration algorithm to solve the nonlinear programming problem; in the process of solving each control period, take the optimal control sequence and the state path solved in the previous control period as initial guess values ​​for the 30-present solution, and perform the iterative calculation in order to accelerate convergence; Step 34: After reaching the predefined solution accuracy or the maximum number of iterations, output the optimal control sequence of the current control period. Controlling offshore wind farms is a complex continuous-time dynamic optimization problem with multiple objectives and constraints. Directly solving this type of optimal control problem in the continuous-time domain results in excessively high computational complexity.This makes it difficult to obtain results within the limited control period required by engineering practice. Traditional discretization methods can suffer from insufficient precision or numerical instability, while conventional optimization algorithms have a slow resolution speed and cannot meet the high-frequency refresh requirement of real-time control. This prevents many advanced control theories from being applied in engineering practice, limiting further improvement of control performance. The present embodiment is specifically as follows: first, based on the most recently updated parameterized state-space model, the multi-objective optimization problem 15, consisting of minimizing power tracking error and minimizing structural loads, is formally constructed as a continuous-time optimal control problem, taking into account various physical constraints of the actuator. Then, the orthogonal interpolation point method,A numerical computation technique is adopted to discretize the continuous state variables and the control variables in the future forecast horizon over a series of selected interpolation points, thus transforming the initial continuous-time optimal control problem into a structured and easier-to-treat nonlinear programming problem. To efficiently solve this nonlinear programming problem, a real-time iteration algorithm is used. When initializing the solution for each control period, this algorithm takes the optimal control sequence and the state trajectory solved in the previous control period as the initial iteration guess values ​​for the presented solution. By relying on the continuity of the system state and the smoothness of the control commands, it significantly reduces the number of iterations required and accelerates the convergence process. 2025 / 5809 BE2025 / 5809 21 Finally, after meeting the predefined resolution accuracy requirements or reaching the limit of a maximum number of iterations,The optimal control sequence for the current control period is output. Compared to the nearest prior technique, traditional methods may adopt simple discretization methods leading to a loss of precision, or use conventional optimization algorithms to solve the problem for each period from zero, resulting in a long computation time and making it difficult to guarantee the reality of the control time. The present implementation combines the orthogonal interpolation point method (a high-precision numerical conversion method) and the real-time iteration algorithm exploiting information from historical solutions, efficiently resolving the conflict between computational precision, numerical flexibility, and the speed of solving the complex optimization problem. The present embodiment achieves a high-precision and high-efficiency numerical solution to the complex continuous-time optimal control problem.15 ensuring that a multi-objective optimization algorithm with constraints can obtain a reliable solution within the time window strictly limited by the engineering applications. This makes real-time operation of the high-frequency advanced model prediction control possible on the real wind turbine controller, providing a central computational basis for improving control performance. 20 In another embodiment of the present invention, the joint state-parameter estimation algorithm in Step 2 adjusts in real time the process noise covariance matrix and the measurement noise covariance matrix of the traceless Kalman filter by introducing an adaptive mechanism, based on the innovation sequence or the residual sequence. 25 When using the traceless Kalman filter for joint state-parameter estimation,Its performance depends strongly on the parameterization of the process noise decovariance matrix and the measurement noise decovariance matrix. Traditional filters generally adopt a fixed noise decovariance matrix, but the operating environment of offshore wind farms is complex: their system dynamic characteristics and sensor measurement environment undergo significant changes under the effect of elements such as turbulence intensity, meteorological conditions and mechanical wear. The statistical assumptions of fixed noise cannot adapt to these temporally variable characteristics. This leads to the filter exhibiting an estimation lag or divergence during abrupt changes in operating conditions, or placing too much trust in erroneous data when measurements are disturbed. This severely compromises the accuracy and robustness of the state and parameter estimation.and consequently influences the control performance based on this model.10 The present embodiment is specifically as follows: during the operation of the joint state and parameter estimation algorithm based on the traceless Kalman filter, the innovation sequence or the filter residual sequence is continuously monitored and calculated (the innovation sequence reflects the difference between the actual measurement and the value predicted by the filter). An adaptive mechanism15 is introduced, capable of analyzing online the statistical characteristics of the innovation or residual sequence, such as the degree of difference between the actual covariance and the theoretical covariance. Based on this analysis result, the mechanism dynamically and in real time adjusts predefined values ​​of the process noise covariance matrix and the noise covariance matrix20 demeure dans le filtre de Kalman sans trace. When the innovation or residual sequence indicates an increase in the gap between the model prediction and the actual measurement,The adaptive mechanism consequently increases the covariance of the process noise, indicating greater confidence in the measured values; conversely, when the measurements present abnormal noise, it increases the 25 covariance of the measurement noise, indicating greater confidence in the prediction of the model, thus achieving online optimization of the filter gain. Compared to the closest prior technique, traditional methods generally rely on engineering experience to perform offline tuning and define a set of fixed and compromised noise covariance parameters.30 This approach cannot cope with changes in the system's dynamic characteristics during operation, and filter performance decreases significantly outside of design conditions. By introducing an adaptive mechanism based on innovation or residual,The present embodiment transforms the filter of a static fixed-parameter estimator into a dynamic estimator capable of adjusting itself according to the actual operating state of the system. The technical solution provided by the present embodiment significantly improves the adaptability and robustness of the joint state and parameter estimation algorithm in the face of dynamic changes in the system, model uncertainties, and measurement disturbances. It guarantees the provision of constant, stable, and accurate state estimation and parameter identification results across the entire range of operating conditions, providing a reliable and trustworthy model basis for subsequent high-performance model prediction control, and fundamentally enhancing the adaptability and resilience of the entire control system. In another embodiment of the present invention, prior to Step 4,The process further includes the following steps: filtering respectively the optimal incremental control of the offset angle and the optimal incremental control15 of the generator torque outputs in Step 3 by a first-order inertial element, whose time constant is obtained by identifying the dynamic response characteristics of the blade offset mechanism and the converter. The optimal incremental theoretical control commands obtained by solving the optimization algorithm are generally calculated instantaneously based on an ideal system model, and their variations can be intense or discontinuous. However, the actual actuators of the wind turbine (such as the blade shifting mechanism and the converter) all exhibit intrinsic mechanical inertia, response delay, and speed of action limitations. By sending the theoretical commands directly to the actuators, excessive demands are imposed on them, resulting in imperfect control tracking and execution, which can cause delay.an overshoot or even mechanical wear of the actuators. This causes the actual control effect to deviate from the expected objective of the optimization algorithm, affecting the overall performance and stability of the control loop.30 In the present embodiment, after obtaining the optimal incremental control of the offset angle and the optimal incremental control of the generator torque in Step 3, they are not transmitted directly to the actuators, house 2025 / 5809 BE2025 / 5809 24 these two incremental control paths are subjected to smoothing by an independent first-order inertia filtering element respectively. The key parameters of this first-order inertia element (i.e., its time constant) are not fixed arbitrarily by experience,but obtained through specific system identification experiments on the dynamic response characteristics of the actual blade shifting mechanism and converter system. This identification process aims to accurately reflect the dynamic delay characteristics of these actuators between command reception and actual action. After filtering, the theoretical incremental commands, which could exhibit rapid variations, are transformed into smooth-variing control signals, physically easier for the actuators to track and execute. Compared to the closest prior technique, traditional methods may neglect the dynamic characteristics of the actuators and directly transmit the controller output commands, or adopt only fixed and conservative filtering parameters15 to trade response speed for elasticity. The present embodiment, however, precisely regulates the filtering parameters by identifying the specific dynamic characteristics of the physical actuators.performing a customized filtering adapted to the characteristics of the equipment.20 The present embodiment effectively avoids the problem of decreased control performance due to the mismatch between theoretical commands and the dynamic capabilities of the actuators, making the issued control commands smoother and more feasible, reducing the actuator load and mechanical stresses. It guarantees that the control intention obtained by optimization calculation25 can be transformed with precision and stability into actual actions, thus improving the operational reliability and the system's overall control loop resilience. In another embodiment of the present invention, for the joint state and parameter estimation algorithm based on the traceless Kalman filter30 in Step 2,the adjustment of the process noise decovariance amateur is achieved through adaptive scaling according to the turbulence intensity value of the forecast wind speed sequence. 2025 / 5809 BE2025 / 5809 25 When using the traceless Kalman filter for state and parameter estimation, the value of the process noise decovariance matrix is ​​crucial: it reflects the measure of the model uncertainty. Traditional methods generally define this matrix as a fixed value. However, the intensity of turbulence in the offshore wind field varies violently: strong turbulence signifies rapid and unpredictable changes in wind speed, leading to a significant increase in model uncertainty; weak turbulence indicates stable wind conditions and a more reliable model. A fixed process noise decovariance matrix cannot adapt to this temporally variable characteristic of external wind disturbances. in the case of strong turbulence, it leads the filter to have too much confidence in an imprecise model prediction,generating an estimation delay; in the case of low turbulence, it may over-rely on measurement noise, affecting the smoothness and accuracy of the estimation, and ultimately limiting the overall performance of the state and parameter estimator under different wind conditions.15 In the present embodiment, during filter operation, the forecast wind velocity sequence acquired by the lidare is analyzed in real time, and a key index characterizing the volatility of the wind field is calculated: the turbulence intensity. This turbulence intensity value is used to directly guide the online adjustment of the process noise covariance matrix.20 More precisely, a scaling mechanism is designed, with the turbulence intensity as an adaptive factor. When the calculated value of intensity turbulence is high (indicating current conditions of strong turbulence, violent changes in the wind field and an increase in model prediction uncertainty),The elements of the 25-process noise covariance matrix are automatically increased proportionally. This adjustment gives the filter greater algorithmic flexibility, making it more inclined to trust new measured data and thus allowing it to follow the system's dynamic changes more quickly. Conversely, when the turbulence intensity value is low (indicating stable wind conditions and high model reliability), the process noise covariance is reduced accordingly, causing the filter to place more trust in the model's prediction, in order to smooth out measurement noise and improve estimation accuracy. 2025 / 5809 BE2025 / 5809 26 Compared to the closest prior technique,The statistical characteristics of the filter's process noise in traditional methods are static and disconnected from the current external environment. The present embodiment creatively associates the central filter parameter (process noise covariance) with a physical quantity directly characterizing the properties of the main external disturbance source (turbulence intensity), establishing an active correspondence between the filter's intrinsic regulation mechanism and the dynamic changes in the external environment. The technical solution provided by the present embodiment includes a joint state and parameter estimation algorithm with a self-tuning capacity adapted to the intensity of wind disturbances. It significantly improves the adaptability and accuracy of the state and parameter estimation under different wind conditions, providing the overall control system with more reliable and timely system state information.and thus strengthens the robustness and control performance of the controller in the face of complex and changing wind conditions. 15 In another embodiment of the present invention, the length of the resolution horizon of the optimal control problem at a finite horizon in Step 3 is dynamically adjusted according to the dominant frequency of wind shear and the turning effect contained in the forecast wind speed sequence. In model prediction control, the length of the prediction horizon is a 20 key design parameter. Traditional control strategies generally define it as a fixed value. However, in the offshore wind field, wind shear and the turning effect, which exert the greatest influence on the structural loads of the wind turbine, are disturbances with marked periodic characteristics,whose dominant frequency changes dynamically with the variation of the wind turbine's rotational speed and the average wind speed. A fixed prediction horizon cannot accommodate these changing periodic disturbances: if the prediction horizon is too short, it cannot cover a complete disturbance cycle, and the controller cannot "perceive" the entire disturbance or perform effective anticipatory compensation; if the prediction horizon is too long, it unnecessarily increases computational complexity and can reduce control performance due to the accumulation of model prediction errors. This conflict makes it difficult for the fixed-horizon control strategy to achieve optimal suppression of periodic loads under all operating conditions. In the present embodiment, during the operation of the controller, the forecast wind speed sequence acquired by the lidar is continuously analyzed.5. The dominant frequency of the periodic fluctuations in wind speed caused by wind shear and the turning effect is identified in real time. Based on the dominant frequency value identified in real time, the length of the solution horizon for the optimal finite-horizon control problem is dynamically adjusted in Step 3. The specific adjustment strategy consists of ensuring that the length of the prediction horizon adaptively covers one or more complete cycles corresponding to this dominant frequency. For example, when the dominant frequency is high, the corresponding cycle is short, and the prediction horizon is automatically reduced to one or two cycle lengths; when the dominant frequency is low, the cycle is long, and the prediction horizon is extended accordingly to guarantee that the The low-frequency perturbation process is still covered. In this way, the prediction range is dynamically linked to the periodic characteristics of the main perturbations. Compared to the closest previous technique,Traditional methods20 adopt a fixed prediction horizon, the design of which is based on a compromise consideration for typical operating conditions. They cannot adapt to the dynamic changes in disturbance cycle characteristics during wind turbine operation and constitute a passive, compromise solution. The present invention transforms25 the prediction horizon of a fixed control parameter into a dynamic variable closely associated with the physical characteristics of the main current disturbances, implementing an actively adaptable prediction strategy. The present embodiment allows the model prediction controller to constantly associate its anticipated vision with the characteristics of the most important current periodic disturbances30.ensuring that the optimization algorithm can make more rational decisions based on perturbation information covering a complete cycle. This significantly improves the effect of the suppression of periodic load components caused by wind shear and the deflection effect, and enhances the adaptability of the control strategy under different operating conditions and the consistency of control performance. In another embodiment of the present invention, the time-variable aerodynamic parameters identified online and in real time at Step 2 include at least the equivalent time constant of the dynamic stall model and the relaxation constant of the induction factor of the incident dynamic flow model. When establishing the wind turbine control model, models describing key aerodynamic phenomena such as dynamic stall10 and incident dynamic flow generally include key time constant parameters,For example, the equivalent time constant of the dynamic stall model and the relaxation constant of the induction factor of the incident dynamic flow model. Traditional methods consider these time constants as fixed values. However, the timescale of these aerodynamic processes changes significantly with operating states such as angle of attack, pitch angle, and tip velocity ratio. The use of fixed time constants cannot accurately reflect the actual response rate of aerodynamics under different operating conditions, leading to a dynamic response discrepancy of the model when conditions deviate from the design conditions, severely reducing the accuracy of aerodynamic load calculations and consequently affecting the accuracy of state estimation and performance. control based on this model. In particular, when faced with highly non-linear and transient processes such as dynamic stall,The model error is particularly marked. In the present embodiment, in the joint state and parameter estimation algorithm executed in Step 2, the set of parameters to be identified is extended in a targeted manner, integrating these key aerodynamic time parameters, namely the equivalent time constant of the dynamic stall model and the relaxation constant of the induction factor of the dynamic flow model incident in the vector of parameters to be identified. Using as observation information the continuous forecast wind speed sequence and the structural load data measured in real time, the algorithm estimates not only the system state and current aerodynamic parameters,but also identifies and updates online and in real time these two temporally variable aerodynamic time constants. This allows the part of the model describing the aerodynamic dynamic processes to automatically adjust its response characteristics according to the actual operating conditions, thus making the entire parameterized state-space model reproduce more faithfully the dynamic behavior of the wind turbine's aerodynamic system. Compared to the closest prior technique, traditional methods rely on offline identification or theoretical calculation to define an invariant and compromised time constant for these dynamic processes, their model being static and incapable of capturing changes in the aerodynamic response characteristics during operation. The present embodiment raises these two key dynamic parameters from fixed coefficients of the model to state variables identified online,equipping the aerodynamic model with the ability to describe the temporally variable characteristics of its dynamic response and significantly improving the dynamic accuracy of the model. The technical solution provided by the present embodiment significantly improves the accuracy of the calculation of aerodynamic loads (particularly in complex transient conditions such as dynamic stall), thus providing the state estimator and the prediction controller with a more realistic and reliable system model. It effectively improves the prediction accuracy and control performance of the entire control system in the face of complex nonlinear aerodynamic phenomena, strengthening the operating behavior of the wind turbine across the entire range of operating conditions. In another embodiment of the present invention, before generating the actual control commands for the current period in Step 4,A 30 independent consistency check and a balanced distribution of the optimal incremental control of the offset angle are carried out according to the measured offset angles of the three blades. 2025 / 5809 BE2025 / 5809 30 During the operation of the wind turbine, due to wind shear and the deflection effect, the three blades rotating at different orientations are subjected to different real aerodynamic loads, exhibiting an intrinsic ease of measurement. Furthermore, there are minor differences in response characteristics between the three independent offset actuators. Traditional control methods 5 generally calculate and deliver a unified offset angle command for all three blades, or perform independent offset control but do not sufficiently take into account the return and balance of the actual execution positions of each blade. This leads to the following: after distribution and execution of the optimal incremental offset angle command calculated in a unified manner, 10The imbalance of bending moments at the base of the three blades cannot be completely eliminated. These residual imbalance loads are transmitted to key components such as the hub and main shaft, exacerbating their fatigue damage and limiting the potential advantages of the independent offset control strategy in load balancing.15 The present embodiment is specifically as follows: after obtaining the unified optimal incremental control of the offset angle in Step 3, it is not directly distributed equally among the three blades. Before generating the final actual control commands in Step 4, an independent verification and distribution step is introduced. This step reads in real time the feedback signals20 of the measured offset angles of each of the three blades and compares them with the current control objective for analysis. Its core consists of performing a consistency check and a reassignment of the total incremental command of the offset angle calculated in a unified manner,depending on the actual position states of the three blades and the specific load balancing requirements. The purpose of verification is to determine if the command risks aggravating the imbalance; the distribution consists of dynamically and independently allocating the total amount of command to the three blades in different proportions or offsets. Its distribution strategy aims to actively compensate for the asymmetric aerodynamic loads due to wind shear and the response offsets resulting from the differences between the actuators. Compared to the closest prior technique, although traditional independent offset control generates different commands for the three blades, its core algorithm can still focus on taking into account the inhomogeneity of spatial wind distributions,without forming a closed-loop verification and reassignment mechanism based on feedback of the actual blade positions and having the direct objective of load balancing. This embodiment adds an intelligent decision layer before the final distribution of commands, which, based on real-time feedback, strives to further optimize and guarantee the load balancing effect at the execution level. This embodiment improves the level of refinement and intelligence of the independent offset control. Thanks to the secondary optimization and distribution of the unified command, it can more effectively compensate for asymmetric factors in actual operation, ensuring that the actions of the three blades ultimately achieve the load balancing objective in a more coordinated manner. This further reduces asymmetric fatigue loads acting on key components such as the hub,and optimizes the overall performance and benefits of independent offset control. The number of pieces of equipment and the processing scale described herein are used to simplify the description of the present invention. The applications, modifications, and variations of the present invention are obvious to those skilled in the art. Although embodiments of the present invention have been described above, they are not limited to the applications listed in the specification and implementations, and can be fully applied to various fields suitable for the present invention, and further modifications can be easily implemented by those skilled in the art. Therefore, the present invention is not limited to the specific details and illustrations shown and described herein without departing from the general concept defined by the claims and the equivalent scope. 2025 / 5809 BE2025 / 5809,