Wave pitching motion control method and system for offshore floating type wind turbine generator

Abstract

The invention discloses a wave pitching motion control method and system for an offshore floating type wind turbine generator, and belongs to the technical field of offshore wind power equipment control. The method comprises the following steps: signal acquisition: acquiring unit operation state, marine environment and pitching motion state parameters in real time; wave load prediction: predicting a future wave load based on the improved LSTM neural network model; evaluating a pitching motion state, pre-judging a motion trend through a kinetic model, and judging whether the motion trend exceeds a safety threshold value or not; cooperative control instruction generation: generating cooperative control instructions for fine tuning of a variable pitch angle, a generator torque and a cabin yaw angle based on an MPC algorithm; and control execution and feedback correction: driving an execution mechanism to act and dynamically correcting a control instruction. By accurately predicting the wave load, cooperatively adjusting multiple controllable components and dynamically correcting, efficient and accurate suppression of pitching motion is achieved, and the unit operation stability, the wind energy capturing efficiency and the equipment safety are improved.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power equipment control technology, and in particular to a wave pitch motion control method and system for offshore floating wind turbines. Background Technology

[0002] With the acceleration of the global energy structure transformation, offshore wind power, as an important component of clean and renewable energy, has received widespread attention for its development and utilization. Compared with stationary offshore wind turbines, floating wind turbines can be deployed in deeper waters, offering advantages such as abundant wind energy resources and less impact on the nearshore ecological environment, making them an important direction for the future development of offshore wind power.

[0003] However, the foundations of floating wind turbines (such as semi-submersible, spar, and tension leg turbines) float on the sea surface and are susceptible to multi-degree-of-freedom motion caused by marine environmental loads such as waves, wind, and ocean currents. Among these, pitch motion (rotation around the turbine's lateral axis) has a particularly significant impact on the safe and stable operation of the turbine. On the one hand, severe pitch motion causes the rotor's center position to rise and fall periodically, disrupting the optimal alignment between the rotor and the incoming wind and reducing wind energy capture efficiency. On the other hand, the inertial loads generated by pitch motion can exacerbate fatigue damage to critical components such as the tower, blades, and transmission system, shortening equipment lifespan and potentially leading to structural failures and other safety accidents.

[0004] Existing control methods for the pitch motion of floating wind turbines are mainly divided into two categories: passive control and active control. Passive control suppresses pitch motion by optimizing the foundation structure design (such as increasing ballast, installing damping plates, etc.) or installing passive damping devices, but its control effect is limited and it is difficult to adapt to the complex and ever-changing marine environment. Active control generates a counter-torque by actively adjusting the controllable components of the turbine (such as pitch angle, nacelle yaw angle, generator torque, etc.) to counteract the pitch motion caused by wave loads, and this is currently a hot research topic.

[0005] However, existing active control methods still have many shortcomings: First, most are based on traditional PID control algorithms, which are poorly adapted to nonlinear and time-varying floating wind turbine systems and have low control accuracy under complex wave loads (such as irregular waves); Second, the control strategies mostly adjust the pitch angle or generator torque alone, without fully considering the synergistic effect between various controllable components, resulting in low control efficiency; Third, they do not accurately predict wave loads and rely solely on real-time feedback signals for control, resulting in control lag and difficulty in effectively suppressing high-frequency or sudden pitch motions.

[0006] Therefore, this invention proposes a wave pitch motion control method and system for offshore floating wind turbines that can accurately predict wave loads, coordinate multiple controllable components, and adapt to complex marine environments. Summary of the Invention

[0007] The purpose of this invention is to overcome the defects in the existing technology and provide a wave pitch motion control method and system for offshore floating wind turbines. This method achieves efficient and precise suppression of pitch motion by accurately predicting wave loads and coordinating the adjustment of multiple controllable components, thereby improving the operational stability and safety of floating wind turbines.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A method for controlling the wave pitch motion of a floating offshore wind turbine includes the following steps:

[0010] S1. Signal Acquisition: Real-time acquisition of the unit's operating status parameters, marine environment parameters, and pitch motion parameters is achieved through a sensor array deployed on the floating wind turbine.

[0011] S2. Wave load prediction: Based on the marine environmental parameters collected in step S1, an improved long short-term memory (LSTM) neural network model is used to predict the wave load acting on the wind turbine foundation within a preset time period in the future.

[0012] S3. Pitch motion status assessment: Based on the pitch motion status parameters collected in step S1 and the wave load predicted in step S2, establish a pitch motion dynamic model, assess the future pitch motion trend of the unit, and determine whether the current pitch motion exceeds the preset safety threshold.

[0013] S4. Cooperative control command generation: If step S3 determines that the pitch motion exceeds the preset safety threshold, based on the model predictive control (MPC) algorithm, with the optimization goal of suppressing pitch motion and ensuring wind energy capture efficiency, the pitch angle adjustment command, generator torque adjustment command, and nacelle yaw angle fine-tuning command are generated.

[0014] S5. Control Execution and Feedback Correction: The control commands generated in step S4 are sent to the corresponding actuators to drive their actions. The feedback parameters of the unit's pitch motion are collected in real time through the sensor group to dynamically correct the control commands.

[0015] Further, in step S1, the operating status parameters include wind turbine speed, generator power, current pitch angle, and current nacelle yaw angle; the marine environment parameters include significant wave height, wave period, wave direction, wind speed, and wind direction; the pitch motion status parameters include pitch angle, pitch rate, and pitch acceleration; and the sensor group includes a wave height meter, anemometer, gyroscope, speed sensor, power sensor, and angle sensor.

[0016] Furthermore, in step S2, the construction and training process of the improved LSTM neural network model is as follows:

[0017] S21. Dataset Construction: Collect historical marine environmental parameters and corresponding wave load data, normalize the data, and divide it into training set, validation set, and test set;

[0018] S22. Model Structure Design: The improved LSTM neural network model includes an input layer, a hidden layer, an attention mechanism layer, and an output layer. The input layer receives normalized marine environmental parameters, the hidden layer adopts a two-layer LSTM structure, the attention mechanism layer is used to enhance the weight allocation of key environmental parameters, and the output layer outputs the predicted wave load values ​​for a future preset time period.

[0019] S23. Model Training and Optimization: The model is trained using the training set, and the hyperparameters of the model are adjusted using the validation set. The mean squared error (MSE) is used as the loss function, and the loss function is minimized using the Adam optimization algorithm of adaptive moment estimation to obtain the improved LSTM neural network model after training. The loss function is shown in Equation (1):

[0020]

[0021] Where N is the sample size. The actual wave load value for the i-th sample. Let be the predicted wave load value for the i-th sample.

[0022] Furthermore, in step S3, the pitch motion dynamics model is established based on Newton's second law, and its expression is shown in formula (2):

[0023]

[0024] Where J is the moment of inertia of the wind turbine about the pitch axis. Let C be the pitch acceleration and C be the pitch damping coefficient. Where is the pitch angular velocity, and K is the restoring stiffness coefficient. Let t be the pitch angle and t be the time. The pitching moment generated by wind load. The pitching moment generated by wave load, To generate the reverse pitch moment for the control command;

[0025] Based on formula (2), the pitch motion trend within a preset time period is obtained through numerical integration. The predicted maximum pitch angle is then compared with a preset safety threshold. Compare the results to determine whether the pitch motion exceeds the safe range.

[0026] Furthermore, in step S4, the process of generating cooperative control instructions based on the Model Predictive Control (MPC) algorithm is as follows:

[0027] S41. Determine the prediction time domain With control time domain ,and ;

[0028] S42. Establish a discretized linear prediction model, as shown in formula (3):

[0029]

[0030] Where k is the discrete time step. Let k be the state vector at time k. Let k be the control input vector at time k. Let A be the output vector at time k, B be the state matrix, C be the output matrix, and D be the direct transmission matrix.

[0031] S43. Determine the objective function for optimization, as shown in formula (4):

[0032]

[0033] in, , , These are reference values ​​for pitch angle, pitch rate, and rotor speed, respectively. , , These are the control increments for pitch angle, generator torque, and nacelle yaw angle, respectively. These are the weighting coefficients;

[0034] S44. Under the premise of satisfying the control input constraints and output constraints, the minimum value of the optimization objective function is obtained by solving the quadratic programming algorithm to obtain the optimal control input sequence;

[0035] S45. Extract the first element of the optimal control input sequence as the control command at the current moment.

[0036] Furthermore, in step S5, the dynamic correction process involves: real-time acquisition of the pitch angle after control. and pitch angular velocity Calculate the deviation , If the absolute value of the deviation is greater than the preset correction threshold, the control command is corrected based on the proportional-integral (PI) controller. The corrected control command is shown in formula (5).

[0037]

[0038] in, The corrected control command, This is the proportionality coefficient. The integral coefficient is... This is the deviation vector.

[0039] A wave pitch motion control system for a floating offshore wind turbine includes:

[0040] Signal acquisition module: Composed of multiple sensors, used to collect the unit's operating status parameters, marine environmental parameters, and pitch motion parameters in real time, and transmit them to the data processing module;

[0041] Data processing module: It communicates with the signal acquisition module and is used to preprocess the acquired parameters and transmit the preprocessed data to the wave load prediction module and the pitch motion state evaluation module respectively.

[0042] Wave load prediction module: It communicates with the data processing module and has a built-in improved LSTM neural network model to predict wave loads within a preset time period in the future, and transmits the predictions to the pitch motion state evaluation module.

[0043] Pitch motion state assessment module: It communicates with the data processing module and the wave load prediction module respectively. It has a built-in pitch motion dynamics model to assess the future pitch motion trend of the unit and determine whether it exceeds the preset safety threshold. If it does, it sends a control trigger signal to the collaborative control decision module.

[0044] Collaborative control decision module: It communicates with the pitch motion state assessment module and has a built-in model predictive control (MPC) algorithm. After receiving the control trigger signal, it generates pitch angle adjustment command, generator torque adjustment command, and nacelle yaw angle fine-tuning command, and transmits them to the control execution module.

[0045] Control execution module: Communicates with the collaborative control decision module, including pitch actuator, generator torque regulation mechanism and nacelle yaw actuator, used to receive control commands and drive the corresponding component actions;

[0046] Feedback correction module: It communicates with the signal acquisition module and the collaborative control decision module respectively, and is used to calculate the deviation and dynamically correct the control command based on the control pitch motion feedback parameters.

[0047] Furthermore, the sensors in the signal acquisition module include a wave height meter deployed near the unit foundation, a wind speed and direction meter deployed on the top of the nacelle, a gyroscope deployed at the bottom of the tower, a speed sensor deployed on the wind turbine shaft, a power sensor deployed on the generator, and an angle sensor deployed on the pitch system and yaw system.

[0048] Furthermore, the preprocessing process of the data processing module includes filtering and denoising using the Kalman filter algorithm, and normalizing using the min-max normalization method. The normalization formula is shown in formula (6):

[0049]

[0050] in, Here, x represents the normalized parameter values, and x represents the original parameter values. The minimum value of the parameter. This represents the maximum value of the parameter.

[0051] Furthermore, it also includes a fault diagnosis module, which is connected to the signal acquisition module and the control execution module to monitor the operating status of sensors and actuators in real time. When an abnormality occurs, it issues a fault alarm signal and switches to emergency control mode.

[0052] The beneficial effects of this invention are as follows:

[0053] 1. This invention uses an improved LSTM neural network model to predict wave loads and combines an attention mechanism to enhance the weight allocation of key marine environmental parameters. It can accurately predict wave loads in the future time period and predict pitch motion trends in advance, solving the problem of poor control effect caused by control lag in traditional control methods and providing sufficient response time for active control.

[0054] 2. This invention is based on the Model Predictive Control (MPC) algorithm, which coordinates the adjustment of three controllable components: pitch angle, generator torque, and nacelle yaw angle. It fully leverages the control advantages of each component, resulting in higher control efficiency and more significant suppression effect compared to a single control method. At the same time, by reasonably setting and optimizing the objective function and constraints, it ensures wind energy capture efficiency while suppressing pitch motion, thus achieving a balance between control effect and operational efficiency.

[0055] 3. The present invention incorporates a feedback correction module, which dynamically corrects control commands by collecting pitch motion parameters after control in real time, thereby improving control accuracy and enhancing the system's adaptability to nonlinear, time-varying floating wind turbine systems and complex marine environments. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1This is a flowchart of the control method in an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of the structure of the improved LSTM neural network model in an embodiment of the present invention;

[0059] Figure 3 This is a schematic diagram of the module connection of the control system in an embodiment of the present invention. Detailed Implementation

[0060] 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.

[0061] Example 1

[0062] A method for controlling the wave pitch motion of a floating offshore wind turbine includes the following steps:

[0063] S1. Signal Acquisition: A sensor array deployed on the floating wind turbine collects real-time operating status parameters, marine environmental parameters, and pitch motion parameters. Operating status parameters include rotor speed, generator power, current pitch angle (range 0°~90°), and current nacelle yaw angle. Marine environmental parameters include significant wave height (0~15m), wave period (3~20s), wave direction (0°~360°), wind speed (0~30m / s), and wind direction (0°~360°). Pitch motion parameters include pitch angle (-15°~15°), pitch velocity (-0.5rad / s~0.5rad / s), and pitch acceleration (-0.1rad / s²). 2 ~0.1 rad / s 2 The sensor group includes a wave height meter, an anemometer, a gyroscope, a speed sensor, a power sensor, and an angle sensor, with a sampling frequency of 10Hz.

[0064] S2. Wave Load Prediction: Based on the marine environmental parameters collected in step S1, an improved LSTM neural network model is used to predict the wave load acting on the wind turbine foundation within the next 10 seconds (preset time period). The construction and training process of the improved LSTM neural network model is as follows:

[0065] S21. Dataset Construction: Collect historical marine environmental parameters and corresponding wave load data of floating wind turbines in a certain sea area for one year, totaling 10,000 samples; use the 3σ criterion to remove outliers, perform min-max normalization on the data, and map the parameters to the [0,1] interval; divide the dataset into training set (7,000 sets), validation set (2,000 sets), and test set (1,000 sets) in a ratio of 7:2:1.

[0066] S22. Model Structure Design: The improved LSTM neural network model includes an input layer, a hidden layer, an attention mechanism layer, and an output layer. The input layer has 5 neurons (corresponding to 5 environmental parameters: effective wave height, wave period, wave direction, wind speed, and wind direction). The hidden layer adopts a two-layer LSTM structure, with 64 neurons in the first layer and 32 neurons in the second layer. The attention mechanism layer uses the Bahdanau attention mechanism. The output layer has 1 neuron (corresponding to the predicted wave load value within the next 10 seconds).

[0067] S23. Model Training and Optimization: The model was trained using a training set with a learning rate of 0.001, 100 iterations, and a batch size of 32. The mean squared error (MSE) was used as the loss function, and the Adam optimization algorithm was used to minimize the loss function. During training, the hyperparameters were adjusted using the validation set. Training was stopped when the loss function value on the validation set no longer decreased after 10 consecutive iterations, resulting in the improved LSTM neural network model. The model was validated on the test set, and the prediction error RMSE was 0.05, which is better than the traditional LSTM model.

[0068] S3. Pitch motion state assessment: Based on the pitch motion state parameters collected in step S1 and the wave load predicted in step S2, a pitch motion dynamic model is established to assess the pitch motion trend of the unit in the next 10 seconds and determine whether the current pitch motion exceeds the preset safety threshold (the pitch angle safety threshold is set to 8°); the pitch motion dynamic model is shown in formula (2), where the moment of inertia of the wind turbine about the pitch axis is J = 5 × 10 7 kg·m 2 The pitch motion damping coefficient C = 2 × 10 6 N·m·s / rad, stiffness recovery coefficient K=3×10 5 N·m / rad; The pitch angle and pitch velocity prediction curves for the next 10 seconds are obtained by solving formula (2) using the Runge-Kutta numerical integration method; If the predicted maximum pitch angle is ≥8°, the pitch motion is judged to be out of the safe range and step S4 is executed; otherwise, return to step S1 to continue collecting signals.

[0069] S4. Cooperative Control Command Generation: Based on the Model Predictive Control (MPC) algorithm, generate pitch angle adjustment commands, generator torque adjustment commands, and nacelle yaw angle fine-tuning commands; the specific process is as follows:

[0070] S41. Determine the prediction time domain and control time domain: Set the prediction time domain T_p=10 (corresponding to 10s, consistent with the wave load prediction time period), and the control time domain T_c=5 (the control command is updated every 2s).

[0071] S42. Establishing a prediction model: Based on the pitch motion dynamics model, the matrix parameters of the discretized linear prediction model are obtained through system identification: A∈R 3 × 3 , B∈R 3 × 3 , C∈R 4 × 3 , D∈R 4 × 3 ;

[0072] S43. Determine the objective function: Set the pitch angle reference value θ_ref=0, and the pitch angular velocity reference value... =0, wind turbine speed reference value Ω_ref=12rad / s (optimal wind energy capture speed); weighting coefficients are set to ω1=0.4, ω2=0.3, ω3=0.2, ω4=0.05, ω5=0.03, ω6=0.02;

[0073] S44. Solving the optimization problem: The control input constraints are set as follows: pitch angle β∈[0°,30°], generator torque T_g∈[0,5×10]. 4 [N·m], the yaw angle fine-tuning range of the nacelle is γ∈[-5°,5°]; the output constraint is set as: the wind turbine speed Ω∈[8rad / s,15rad / s]; the minimum value of the optimization objective function is solved by the quadratic programming algorithm (interior point method) to obtain the optimal control input sequence in the control time domain;

[0074] S45. Extract control commands: Take the first element of the optimal control input sequence as the control command at the current moment, such as β×=12°, T_g×=3×10. 4 N·m, γ×=-2°.

[0075] S5. Control Execution and Feedback Correction: Control commands are sent to the pitch actuator, generator torque regulation mechanism, and nacelle yaw actuator to drive the actions of each component; simultaneously, the pitch angle θ_real=3° and pitch velocity are collected in real time by the sensor array. =0.1rad / s; Calculate the deviation e_θ=3°-0°=3°, e_ =0.1rad / s-0=0.1rad / s; the preset correction thresholds are 2° (pitch angle deviation) and 0.2rad / s (pitch angular velocity deviation). The current deviation does not exceed the correction threshold, so no correction is needed; if the subsequently acquired pitch angle deviation is 9°, the control command will be corrected based on the PI controller (K_p=0.5, Ki=0.1), resulting in a corrected pitch angle β_corr=14° and generator torque T_g_corr=3.2×10 4 N·m, cabin yaw angle γ_corr=-2.5°.

[0076] Example 2

[0077] This embodiment provides a wave pitch motion control system for an offshore floating wind turbine, used to implement the control method of Embodiment 1, including:

[0078] Signal acquisition module: It consists of wave height meter, wind speed and direction meter, gyroscope, speed sensor, power sensor and angle sensor. Each sensor is deployed in the corresponding position to collect the unit's operating status parameters, marine environment parameters and pitch motion parameters in real time. The sampling frequency is 10Hz. The collected parameters are transmitted to the data processing module through CAN bus.

[0079] Data processing module: Uses an STM32F407 microcontroller, which communicates with the signal acquisition module via CAN bus; it is used to preprocess the acquired parameters, using Kalman filtering algorithm to filter and denoise, and min-max normalization method to normalize; after preprocessing, the data is transmitted to the wave load prediction module and pitch motion state evaluation module via Ethernet respectively.

[0080] Wave load prediction module: It adopts an industrial control computer with a built-in improved LSTM neural network model (implemented through the Python TensorFlow framework) and communicates with the data processing module via Ethernet; it is used to predict wave loads in the next 10 seconds based on preprocessed marine environmental parameters and transmit the prediction results to the pitch motion state assessment module via Ethernet.

[0081] Pitch motion state assessment module: It adopts an industrial control computer with a built-in pitch motion dynamics model (implemented through MATLAB / Simulink) and communicates with the data processing module and wave load prediction module via Ethernet. It is used to assess the future pitch motion trend of the unit and determine whether it exceeds the preset safety threshold (8°). If it exceeds the threshold, it sends a control trigger signal to the collaborative control decision module via Ethernet.

[0082] The collaborative control decision module employs an industrial control computer with a built-in Model Predictive Control (MPC) algorithm (implemented via MATLAB Model Predictive Control Toolbox). It communicates with the pitch motion state assessment module via Ethernet. Upon receiving control trigger signals, it generates pitch angle adjustment commands, generator torque adjustment commands, and nacelle yaw angle fine-tuning commands, transmitting these commands to the control execution module via the CAN bus. This module also includes a parameter self-tuning unit. When the effective wave height is greater than 5m, ω1 is adjusted to 0.5 and ω2 to 0.35 to enhance pitch motion suppression priority. When the wind speed approaches the rated wind speed (12m / s), ω3 is adjusted to 0.3 to enhance wind energy capture efficiency priority.

[0083] Control execution module: includes pitch actuator (using hydraulic pitch system), generator torque regulation mechanism (using frequency converter) and nacelle yaw actuator (using yaw motor), communicates with the cooperative control decision module via CAN bus; used to receive control commands and drive the corresponding component actions;

[0084] Feedback correction module: Employs an STM32F407 microcontroller, which communicates with the signal acquisition module (CAN bus) and the collaborative control decision module (Ethernet) respectively; it is used to calculate the deviation and dynamically correct the control commands based on the acquired control pitch motion feedback parameters.

[0085] Fault diagnosis module: Employs an STM32F407 microcontroller, which communicates with both the signal acquisition module (CAN bus) and the control execution module (CAN bus). It is used to monitor the operating status of sensors and actuators in real time. When the data collected by the wave height meter exceeds the 0~15m range for 5 consecutive seconds, or when the pitch actuator fails to respond to the control command within 2 seconds, a fault alarm signal is issued through an audible and visual alarm, and the system switches to emergency control mode (fixing the pitch angle to 30° and reducing generator power).

[0086] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for controlling the wave pitch motion of a floating offshore wind turbine, characterized in that, Includes the following steps: S1. Signal Acquisition: Real-time acquisition of the unit's operating status parameters, marine environment parameters, and pitch motion parameters is achieved through a sensor array deployed on the floating wind turbine. S2. Wave load prediction: Based on the marine environmental parameters collected in step S1, an improved long short-term memory (LSTM) neural network model is used to predict the wave load acting on the wind turbine foundation within a preset time period in the future. S3. Pitch motion status assessment: Based on the pitch motion status parameters collected in step S1 and the wave load predicted in step S2, establish a pitch motion dynamic model, assess the future pitch motion trend of the unit, and determine whether the current pitch motion exceeds the preset safety threshold. S4. Cooperative control command generation: If step S3 determines that the pitch motion exceeds the preset safety threshold, based on the model predictive control (MPC) algorithm, with the optimization goal of suppressing pitch motion and ensuring wind energy capture efficiency, the pitch angle adjustment command, generator torque adjustment command, and nacelle yaw angle fine-tuning command are generated. S5. Control Execution and Feedback Correction: The control commands generated in step S4 are sent to the corresponding actuators to drive their actions. The feedback parameters of the unit's pitch motion are collected in real time through the sensor group to dynamically correct the control commands.

2. The control method according to claim 1, characterized in that, In step S1, the operating status parameters include wind turbine speed, generator power, current pitch angle, and current nacelle yaw angle; the marine environment parameters include significant wave height, wave period, wave direction, wind speed, and wind direction; the pitch motion status parameters include pitch angle, pitch rate, and pitch acceleration; and the sensor group includes a wave height meter, an anemometer, a gyroscope, a speed sensor, a power sensor, and an angle sensor.

3. The control method according to claim 1, characterized in that, In step S2, the construction and training process of the improved LSTM neural network model is as follows: S21. Dataset Construction: Collect historical marine environmental parameters and corresponding wave load data, normalize the data, and divide it into training set, validation set, and test set; S22. Model Structure Design: The improved LSTM neural network model includes an input layer, a hidden layer, an attention mechanism layer, and an output layer. The input layer receives normalized marine environmental parameters, the hidden layer adopts a two-layer LSTM structure, the attention mechanism layer is used to enhance the weight allocation of key environmental parameters, and the output layer outputs the predicted wave load values ​​for a future preset time period. S23. Model Training and Optimization: The model is trained using the training set, and the hyperparameters of the model are adjusted using the validation set. The mean squared error (MSE) is used as the loss function, and the loss function is minimized using the Adam optimization algorithm of adaptive moment estimation to obtain the improved LSTM neural network model after training. The loss function is shown in equation (1): Where N is the sample size. The actual wave load value for the i-th sample. Let be the predicted wave load value for the i-th sample.

4. The control method according to claim 1, characterized in that, In step S3, the pitch motion dynamics model is established based on Newton's second law, and the expression is shown in formula (2): Where J is the moment of inertia of the wind turbine about the pitch axis. Let C be the pitch acceleration and C be the pitch damping coefficient. Where is the pitch angular velocity, and K is the restoring stiffness coefficient. Let t be the pitch angle and t be the time. The pitching moment generated by wind load. The pitching moment generated by wave load, To generate the reverse pitch moment for the control command; Based on formula (2), the pitch motion trend within a preset time period is obtained through numerical integration. The predicted maximum pitch angle is then compared with a preset safety threshold. Compare the results to determine whether the pitch motion exceeds the safe range.

5. The control method according to claim 1, characterized in that, In step S4, the process of generating cooperative control instructions based on the Model Predictive Control (MPC) algorithm is as follows: S41. Determine the prediction time domain With control time domain ,and ; S42. Establish a discretized linear prediction model, as shown in formula (3): Where k is the discrete time step. Let k be the state vector at time k. Let k be the control input vector at time k. Let A be the output vector at time k, B be the state matrix, C be the output matrix, and D be the direct transmission matrix. S43. Determine the objective function for optimization, as shown in formula (4): in, , , These are reference values ​​for pitch angle, pitch rate, and rotor speed, respectively. , , These are the control increments for pitch angle, generator torque, and nacelle yaw angle, respectively. These are the weighting coefficients; S44. Under the premise of satisfying the control input constraints and output constraints, the minimum value of the optimization objective function is obtained by solving the quadratic programming algorithm to obtain the optimal control input sequence; S45. Extract the first element of the optimal control input sequence as the control command at the current moment.

6. The control method according to claim 1, characterized in that, In step S5, the dynamic correction process involves: real-time acquisition of the pitch angle after control. and pitch angular velocity Calculate the deviation , If the absolute value of the deviation is greater than the preset correction threshold, the control command is corrected based on the proportional-integral (PI) controller. The corrected control command is shown in formula (5). in, The corrected control command, This is the proportionality coefficient. The integral coefficient is... This is the deviation vector.

7. A wave pitch motion control system for an offshore floating wind turbine, used to implement the control method described in any one of claims 1-6, characterized in that, include: Signal acquisition module: Composed of multiple sensors, used to collect the unit's operating status parameters, marine environmental parameters, and pitch motion parameters in real time, and transmit them to the data processing module; Data processing module: It communicates with the signal acquisition module and is used to preprocess the acquired parameters and transmit the preprocessed data to the wave load prediction module and the pitch motion state evaluation module respectively. Wave load prediction module: It communicates with the data processing module and has a built-in improved LSTM neural network model to predict wave loads within a preset time period in the future, and transmits the predictions to the pitch motion state evaluation module. Pitch motion state assessment module: It is connected to the data processing module and the wave load prediction module respectively. It has a built-in pitch motion dynamics model to assess the future pitch motion trend of the unit and determine whether it exceeds the preset safety threshold. If the limit is exceeded, a control trigger signal is sent to the collaborative control decision module; Collaborative control decision module: It communicates with the pitch motion state assessment module and has a built-in model predictive control (MPC) algorithm. After receiving the control trigger signal, it generates pitch angle adjustment command, generator torque adjustment command, and nacelle yaw angle fine-tuning command, and transmits them to the control execution module. Control execution module: Communicates with the collaborative control decision module, including pitch actuator, generator torque regulation mechanism and nacelle yaw actuator, used to receive control commands and drive the corresponding component actions; Feedback correction module: It communicates with the signal acquisition module and the collaborative control decision module respectively, and is used to calculate the deviation and dynamically correct the control command based on the control pitch motion feedback parameters.

8. The control system according to claim 7, characterized in that, The sensors in the signal acquisition module include a wave height meter deployed near the generator base, a wind speed and direction meter deployed on the top of the nacelle, a gyroscope deployed at the bottom of the tower, a speed sensor deployed on the wind turbine shaft, a power sensor deployed on the generator, and an angle sensor deployed on the pitch system and yaw system.

9. The control system according to claim 7, characterized in that, The preprocessing process of the data processing module includes filtering and denoising using the Kalman filter algorithm, and normalizing using the min-max normalization method. The normalization formula is shown in formula (6): in, Here, x represents the normalized parameter values, and x represents the original parameter values. The minimum value of the parameter. This represents the maximum value of the parameter.

10. The control system according to claim 7, characterized in that, It also includes a fault diagnosis module, which communicates with the signal acquisition module and the control execution module to monitor the operating status of sensors and actuators in real time. When an abnormality occurs, it issues a fault alarm signal and switches to emergency control mode.

Images