Floating predictive lockup control method for enhancing performance of wind turbine generator system

Abstract

The application belongs to the technical field of ocean energy, and discloses a floating prediction locking control method for enhancing the performance of a wind wave coupling unit, comprising: (1) PTO state prediction and decoupling; the state prediction adopts a Hankel-DMDc algorithm to obtain the PTO response behavior without control effect in a short future time; (2) control effect definition; the control effect is defined by an input-output system, a PTO motion curve without control effect in a short future time is given, and a control time is given, and a PTO response curve in a controlled state is returned; (3) optimization solving; the control time used for testing is given by an optimization solving algorithm, the energy acquisition performance is evaluated through the PTO response curves in different control times, and then the optimal control time is determined. The application realizes the predictive intervention on the floating body motion through the real-time monitoring and prediction of the historical state of the floating body and the locking control strategy, can effectively improve the wave energy capture efficiency, and can significantly improve the overall performance of the existing coupling unit.

Description

Technical Field

[0001] This invention relates to a marine energy technology, specifically a floating predictive lock-up control method for enhancing the performance of wind-wave coupled units. Background Technology

[0002] With the development of offshore renewable energy, the coupled utilization of floating wind turbines and wave energy devices is gradually becoming a key direction for improving energy capture efficiency and system economy. Wind-wave coupled units can simultaneously convert wind energy and wave energy on the same floating platform, providing efficient energy output for offshore energy systems.

[0003] For offshore floating wind turbine-heavy point absorber coupled unit configurations, lock-in control is a common control method for wave energy devices. This method is based on the principle of phase matching, which improves instantaneous output power by briefly locking the motion of the floating body at a specific moment to make it anti-matched with the incident wave surface.

[0004] Existing control schemes are generally developed for stationary wave energy converters (WECs), and predictive control strategies for WECs in wind-wave coupled units are lacking. Unlike control strategies for stationary single-unit WECs, the main technical challenge of FOHS multibody systems lies in the accurate modeling of the nonlinear state space. In wind-wave coupled units, the floating bodies are simultaneously affected by turbulent winds and random wave disturbances. Existing WEC control strategies generally lack predictive capabilities and cannot identify critical states of the floating bodies in advance for optimal locking operations, leading to unstable energy capture efficiency. Real-time decision-making for control timing requires the integration of wind speed and wave prediction algorithms with the unit's state-space model.

[0005] Chinese invention patent application CN118110621A discloses a wave energy control method based on deep machine learning algorithms, including the following steps: S1, establishing a numerical model of a point-absorbing wave energy converter; S2, establishing a wave force prediction model, which is an artificial neural network; S3, training a multi-layer artificial neural network based on sample data using deep machine learning algorithms to predict short-term wave forces; S4, combining wave force prediction with an optimal control strategy to perform real-time lockout control on the point-absorbing wave energy converter to maximize the extraction of energy from random waves. This invention develops real-time lockout control for single-degree-of-freedom wave energy converters (WECs) using artificial neural network prediction algorithms. This technology can only be applied to fixed WECs and does not have six-degree-of-freedom motion prediction and control capabilities. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a floating predictive lock-up control method for enhancing the performance of wind-wave coupled units. Based on historical measurement data of the unit and a deep understanding of the dynamic mechanism of the power take-off (PTO) device, the method utilizes the Hankel control version (Dynamic Mode Decomposition With Control, Hankel-DMDc) algorithm, which uses real-time decision-making to determine the optimal control timing to improve wave energy extraction efficiency.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A floating predictive lock-up control method is proposed to enhance the performance of wind-wave coupled units. This method dynamically determines the optimal control timing for the wave energy extraction (WEC) in the wind-wave coupled system to maximize wave energy extraction efficiency. The WEC predictive control method mainly comprises three stages: state prediction and decoupling, control effect definition, and optimization solution. State prediction employs the Hankel-DMDc algorithm, primarily acquiring the pulse-to-torque (PTO) response behavior after removing the control effect in the near future. The control effect is defined through an input-output (IO) system. Given the PTO motion curve after removing the control effect in the near future and the control time, the method returns the PTO response curve under controlled conditions. The control time used for testing is given by the optimization algorithm. Energy extraction performance is evaluated through the PTO response curves under different control times, thereby determining the optimal control time. The details are as follows: (1) PTO state prediction and decoupling External input control decoupling and intrinsic behavior evolution prediction of system dynamics are achieved based on Hankel-DMDc technology. For multi-floating body systems in combined wind and wave units, mode decomposition is based on each WEC unit (…). i = 2-4, corresponding k = 1-3) are carried out. Among them, the internal state parameters include WEC sway, heave and pitch displacement ( X i,1 , X i,3 , X i,5 ),speed( ), excitation load ( F exci,1 , F exci,3 , F exci,5 ), and PTO displacement and velocity ( X PTOk , The external input parameters correspond to the PTO speed control signal. Since the Hankel-DMDc algorithm identifies system characteristics based on the least squares (best fit) principle, wave excitation loads, as the dominant factor driving the evolution of system dynamics, must be included in the modeling process. Furthermore, external input parameters can be intuitively understood as interventions that need to be decoupled from system dynamics. Although the rotor controller also participates in system response regulation, Hankel-DMDc treats it as an inherent behavior of the system.

[0008] For dynamical systems subject to external control interference, Hankel-DMDc can separate the effects of external inputs from controlled observation data and extract the system's inherent evolutionary characteristics. Assume the system's observed state and control input at the current moment are x... j With y j Then the future system state X' can be described by Hankel-DMDc as follows: X' = ​​AX + BY (1) In the formula, A is the driving operator of the system's inherent dynamics, and B is the input driving operator. Observed state x j With control input y j All based on time interval Δ t = t j+1 - t j sampling, m This represents the number of snapshots. Y = [y0, y1, …, y m-1 [X'] is the input matrix. = [x1, x2, …, x m ] and X = [x0, x1, …, x m-1 The difference is a unit time interval. In real-world scenarios, operator B is generally not directly obtainable, so formula (1) needs to be rewritten as: (2) C and Z are the augmenting operator and snapshot matrix, respectively. To obtain the eigenvalues ​​and eigenvectors of operator A, singular value decomposition must first be performed on matrix Z. , , † Refers to the Moore-Penrose pseudoinverse; * indicates complex conjugate transpose. Let Z represent the left singular matrix, singular value matrix, and right singular matrix, respectively. Matrix C can then be expressed as: (3) Expanding formula (3) into blocks, we get: (4) and For matrix The block-based results show that the inherent dynamic characteristics of the system are mainly reflected in the measurement output matrix X'. Therefore, singular value decomposition needs to be performed on matrix X' separately to obtain the modal subspace Û dominated only by state evolution. Based on this, the eigenvalue decomposition of operator A can be approximated by reduced-order projection: (5) * indicates complex conjugate transpose. For matrix The block result is Û, which is the orthogonal projection matrix of matrix X'; The approximation operator A preserves the first part of the original operator A. m The dominant eigenvalue Λ = Diag( l 1, l 2, …, l m ).

[0009] By execution ÃW = WΛ (6) The Hankel-DMDc mode matrix Φ = [Φ1, Φ2, …, Φ m The eigenvector matrix W = [w1, w2, …, w] can be approximated by the eigenvector matrix of the operator Ã. m Determine: (7) make Ω = In(Λ) / Δ t (8) A snapshot at any given time can be approximately reconstructed as: (9) α = [ α 1, α 2, …, α m ] T The Hankel-DMDc amplitude vector can be obtained by projecting the initial snapshot onto the Hankel-DMDc mode matrix: α =Φ † x0 (10) x0 is the initial value. t State measurement results at time 0.

[0010] The Hankel-DMDc algorithm requires performing singular value decomposition on the augmented matrix Z and the measurement output matrix X' respectively, with the truncation orders denoted as follows: and r Ultimately, the inherent behavior of the system decoupled from external control can be reconstructed in reduced order using formula (9).

[0011] (2) Definition of control effect In predictive control methods, the effect of latch-up control on the PTO curve is defined through the I / O system. The input parameter is the control time and the predicted PTO motion curve provided by the state predictor, while the output parameter is the PTO response curve after control is applied. The effect of the latch-up strategy is mainly described by phase delay and amplitude scaling, and the mapping relationship between control time and PTO dynamics can be defined using function fitting or neural network methods.

[0012] (3) Optimization solution Based on the uncontrolled PTO motion curves and IO system over the next two wave cycles, the optimal control time is determined using a general optimization algorithm. When the PTO velocity reaches zero, each WEC unit sequentially performs state prediction and optimization decisions, thereby providing the optimal control time for the next motion cycle that satisfies the objective function. The control framework integrates various optimization algorithms from the MATLAB Global Optimization Toolbox, including surrogate models, scattering search, multi-starting point, genetic algorithms, particle swarm optimization, simulated annealing, and their integer-constrained and multi-objective variants. The control dynamic optimization problem is solved using an online surrogate model with integer constraints. Although this model lacks local optimization capabilities, its computational efficiency makes it highly suitable for fast decision-making scenarios.

[0013] The PTO motion curve reconstructed by order reduction is truncated at the end of the next motion cycle, and the intermediate time interval is defined as follows: T M Therefore, the optimization problem can be formulated as determining the motion period (0 ≤ 0 ≤ 1). t ≤ T M The optimal control time within a given range is determined to maximize wave energy extraction. To adapt to integer-constrained algorithms, the optimal control time for time-discrete dynamic systems is equivalent to 0 ≤ [condition missing]. x ≤ L M The optimal integer solution within the range, where L M = T M / d t The number of discrete time steps corresponding to the motion cycle, d t For discrete time intervals; Based on the above definition, the control period coefficient s MAs a key input parameter of the I / O system, it can be represented as s M = x / L M ,in x To optimize the integer test points generated by the solver, the dynamic effects of PTO (Pulse-Track Occurrence) are achieved by embedding a locking strategy. X PTOk ( c )= X PTOk (0), and ( c = 1, 2, …, x The I / O system generates PTO displacement and velocity curves characterized by a quadratic model based on the iteratively updated control cycle coefficients. The PTO response curves are then fed back to the optimization solver, which determines the optimal control timing based on the PTO power performance.

[0014] The beneficial effects of this invention are: by real-time monitoring and prediction of the historical state of the floating body, combined with a locking control strategy, this invention can achieve predictive intervention in the movement of the floating body, which can effectively improve the wave energy capture efficiency and thus significantly improve the overall performance of existing coupling units.

[0015] This invention enables real-time monitoring and prediction of the historical state of the wave energy coupling system (WEC), allowing for early determination of the optimal timing for WEC locking operations. This ensures precise phase matching between the buoy's motion and the incident wave surface, significantly improving the average output power. Compared to traditional timed locking control methods, this invention effectively enhances wave energy utilization.

[0016] This invention combines floating body motion prediction technology, which can fully consider the nonlinear coupling characteristics of turbulent wind and random waves, realize real-time prediction of the key state of wind-wave coupled units, and enable the control strategy to maintain high reliability and stability in a multi-degree-of-freedom and multi-disturbance marine environment.

[0017] This invention achieves coordinated control of the WEC array through the synergistic effect of prediction algorithms and locking strategies. The Hankel-DMDc algorithm used does not require the construction of a complete unit; it can extract the aerodynamic effects of the turbines using only historical WEC state data, thus avoiding the need for wind speed prediction and turbine modeling. Attached Figure Description

[0018] Figure 1 This is a diagram of a typical floating wind-wave coupling system. Figure 2 It is a control decision-making flowchart; Figure 3 It is a wave energy efficiency gain diagram based on the predictive control strategy; Figure 4 It is a WEC predictive control model for wind-wave combined units based on Hankel-DMDc. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] The structures, proportions, and sizes illustrated in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art in understanding and reading the invention. They are not intended to limit the scope of the invention and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, provided they do not affect the effectiveness or purpose of the invention, should still fall within the scope of the technical content disclosed herein. Furthermore, the terms "upper," "lower," "left," "right," "middle," and "one" used in this specification are merely for clarity and not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0021] like Figure 1-Figure 4 This paper presents a floating predictive lock-up control method to enhance the performance of wind-wave coupled units. It dynamically determines the optimal control timing for the wave energy extraction (WEC) in the wind-wave coupled system to maximize wave energy extraction efficiency. The WEC predictive control framework mainly comprises three stages: state prediction and decoupling, control effect definition, and optimization solution. State prediction employs the Hankel-DMDc algorithm, primarily acquiring the PTO response behavior after removing the control effect in the near future. The control effect is defined through an input-output (IO) system. Given the PTO motion curve after removing the control effect in the near future and the control time, it returns the PTO response curve under controlled conditions. The control time used for testing is given by the optimization algorithm. Energy extraction performance is evaluated through the PTO response curves under different control times, thereby determining the optimal control time. The details are as follows: (1) PTO state prediction and decoupling External input control decoupling and intrinsic behavior evolution prediction of system dynamics are achieved based on Hankel-DMDc technology. For multi-floating body systems in combined wind and wave units, mode decomposition is based on each WEC unit (…). i = 2-4, corresponding k = 1-3) are carried out. Among them, the internal state parameters include WEC sway, heave and pitch displacement ( X i,1 , X i,3 , X i,5 ),speed( ), excitation load (F exci,1 , F exci,3 , F exci,5 ), and PTO displacement and velocity ( X PTOk , The external input parameters correspond to the PTO speed control signal. Since the Hankel-DMDc algorithm identifies system characteristics based on the least squares (best fit) principle, wave excitation loads, as the dominant factor driving the evolution of system dynamics, must be included in the modeling process. Furthermore, external input parameters can be intuitively understood as interventions that need to be decoupled from system dynamics. Although the rotor controller also participates in system response regulation, Hankel-DMDc treats it as an inherent behavior of the system.

[0022] For dynamical systems subject to external control interference, Hankel-DMDc can separate the effects of external inputs from controlled observation data and extract the system's inherent evolutionary characteristics. Assume the system's observed state and control input at the current moment are respectively... x j and y j Then the future system state X' can be described by Hankel-DMDc as follows: X' = ​​AX + BY (1) In the formula, A is the driving operator of the system's inherent dynamics, and B is the input driving operator. Observed state x j With control input y j All based on time interval Δ t = t j+1 - t j sampling, m This represents the number of snapshots. Y = [y0, y1, …, y m-1 [X'] is the input matrix. = [x1, x2, …, x m ] and X = [x0, x1, …, x m-1 The time interval between the two units.

[0023] In real-world scenarios, operator B is generally not directly obtainable, so formula (1) needs to be rewritten as: (2) C and Z are the augmenting operator and snapshot matrix, respectively. To obtain the eigenvalues ​​and eigenvectors of operator A, singular value decomposition must first be performed on matrix Z. , , † Refers to the Moore-Penrose pseudoinverse; * indicates complex conjugate transpose. Let Z represent the left singular matrix, singular value matrix, and right singular matrix, respectively. Matrix C can then be expressed as: (3) Expanding formula (3) into blocks, we get: (4) and For matrix The block-based results show that the inherent dynamic characteristics of the system are mainly reflected in the measurement output matrix X'. Therefore, singular value decomposition needs to be performed on matrix X' separately to obtain the modal subspace Û dominated only by state evolution. Based on this, the eigenvalue decomposition of operator A can be approximated by reduced-order projection: (5) * indicates complex conjugate transpose. For matrix The block result is Û, which is the orthogonal projection matrix of matrix X'; The approximation operator A preserves the first part of the original operator A. m The dominant eigenvalue Λ = Diag( l 1, l 2, …, l m ).

[0024] By execution ÃW = WΛ (6) The Hankel-DMDc mode matrix Φ = [Φ1, Φ2, …, Φ m The eigenvector matrix of the approximator operator à can be approximated. Determine: (7) make Ω = In(Λ) / Δ t (8) A snapshot at any given time can be approximately reconstructed as: (9) α = [ α 1, α 2, …, αm ] T The Hankel-DMDc amplitude vector can be obtained by projecting the initial snapshot onto the Hankel-DMDc mode matrix: α =Φ † x0 (10) x0 is the initial value. t State measurement results at time 0.

[0025] The Hankel-DMDc algorithm requires performing singular value decomposition on the augmented matrix Z and the measurement output matrix X' respectively, with the truncation orders denoted as follows: and r Ultimately, the inherent behavior of the system decoupled from external control can be reconstructed in reduced order using formula (9).

[0026] (2) Definition of control effect In predictive control methods, the effect of latch-up control on the PTO curve is defined through the I / O system. The input parameter is the control time and the predicted PTO motion curve provided by the state predictor, while the output parameter is the PTO response curve after control is applied. The effect of the latch-up strategy is mainly described by phase delay and amplitude scaling, and the mapping relationship between control time and PTO dynamics can be defined using function fitting or neural network methods.

[0027] (3) Optimization solution Based on the uncontrolled PTO motion curves and IO system over the next two wave cycles, the optimal control time is determined through an optimization algorithm. When the PTO velocity reaches zero, each WEC unit sequentially performs state prediction and optimization decisions, thereby providing the optimal control time for the next motion cycle that satisfies the objective function. The control framework integrates various optimization algorithms from the MATLAB Global Optimization Toolbox, including surrogate models, scattering search, multi-starting point, genetic algorithms, particle swarm optimization, simulated annealing, and their integer-constrained and multi-objective variants. The control dynamic optimization problem is solved using an online surrogate model with integer constraints. Although this model lacks local optimization capabilities, its computational efficiency makes it highly suitable for fast decision-making scenarios.

[0028] The PTO motion curve reconstructed by order reduction is truncated at the end of the next motion cycle, and the intermediate time interval is defined as follows: T M Therefore, the optimization problem can be formulated as determining the motion period (0 ≤ 0 ≤ 1). t ≤ T M The optimal control time within a given range is determined to maximize wave energy extraction. To adapt to integer-constrained algorithms, the optimal control time for time-discrete dynamic systems is equivalent to 0 ≤ [condition missing]. x ≤ LM The optimal integer solution within the range, where L M = T M / d t The number of discrete time steps corresponding to the motion cycle, d t For discrete time intervals; Based on the above definition, the control period coefficient s M As a key input parameter of the I / O system, it can be represented as s M = x / L M ,in x To optimize the integer test points generated by the solver, the dynamic effects of PTO (Pulse-Track Occurrence) are achieved by embedding a locking strategy. X PTOk ( c )= X PTOk (0), and ( c = 1, 2, …, x The I / O system generates PTO displacement and velocity curves characterized by a quadratic model based on the iteratively updated control cycle coefficients. The PTO response curves are then fed back to the optimization solver, which determines the optimal control timing based on the PTO power performance.

[0029] Wind-wave coupling system ( Figure 1 The performance verification of the WEC predictive control was first conducted based on steady-state wind-regular stripes. The PTO motion response in this test environment exhibits simple harmonic oscillation characteristics, and the optimal control time can be determined through a timing control strategy. The floating prediction is activated when the PTO velocity reaches zero, and the lock-up control method dynamically determines the optimal control timing of the WEC through decoupling prediction and optimization iteration. Figure 2 First, based on the collected WEC displacement, velocity, excitation load, and historical PTO displacement and velocity data, the Hankel-DMDc algorithm is used to predict the PTO free response behavior after removing the control effect in the short term. Then, using this free response curve as input, and combining it with multiple candidate control times provided by the optimization algorithm, the controlled PTO response curves under different control times are generated through the input-output system. Then, the power performance of each response is evaluated, and the optimal locking time is determined with the goal of maximizing wave energy extraction efficiency. Finally, the control time is fed back to the actual PTO system for execution, and the next round of control is executed after the PTO velocity returns to zero.

[0030] For comparison, the WEC array uses a wave period timing control strategy, with the execution time uniformly set to [time value missing].T C = s T T P ( s T = 0-0.25, interval 0.0125; T C To control time, T P (Wave period). Figure 3 The solid line and the dashed line represent the timing strategy and predictive control results, respectively. Compared to passive strategies ( s T = 0), the overall power level of the WEC array in the dynamic control scheme ( m CWR CWR (Capture Width Ratio) represents a 50.6% improvement. Even compared to the optimal energy efficiency of the timing strategy (black circle), it still achieves a 4.6% gain in wave energy extraction. The predictive controller enables WEC 1 to achieve an energy extraction level comparable to the optimal performance of the timing control strategy (red circle), while significantly improving the energy efficiency of WEC 2. WEC 2 and WEC 3 adopt a symmetrical arrangement and have basically the same energy extraction efficiency; only the results of WEC 2 are shown here. A schematic diagram of the overall implementation of predictive latching control is shown below. Figure 4 As shown.

[0031] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A floating predictive lock-up control method for enhancing the performance of wind-wave coupled units, characterized in that, It includes three stages: state prediction and decoupling, control effect definition and optimization solution, as detailed below: (1) PTO state prediction and decoupling: The state prediction adopts the Hankel-DMDc algorithm to obtain the PTO response behavior after removing the control effect in the short term. (2) Definition of control effect: The control effect is defined through the input-output system. Given the PTO motion curve after removing the control effect in the future short time and the control time, the PTO response curve in the controlled state is returned. (3) Optimization solution: The control time used in the test is given by the optimization solution algorithm. The energy gain performance is evaluated by the PTO response curve under different control times, and then the optimal control time is determined.

2. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 1, characterized in that, In the (1) PTO state prediction and decoupling, the external input control decoupling and intrinsic behavior evolution prediction of system dynamics are realized based on Hankel-DMDc technology; for the wind-wave combined unit multi-floating body system, the modal decomposition is carried out based on each WEC unit, wherein the internal state parameters include WEC sway, heave and pitch displacement. X i,1 , X i,3 , X i,5 ,speed Excitation load F exci,1 , F exci,3 , F exci,5 and PTO displacement and velocity X PTOk , The external input parameters correspond to the PTO speed control signal. .

3. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 2, characterized in that, For a dynamic system subject to external control interference, Hankel-DMDc separates the effects of the external input from the controlled observation data and extracts the inherent evolutionary characteristics of the system; assuming that the observed state of the system and the control input at the current moment are respectively... x j and y j Then the future system state X' is described by Hankel-DMDc as follows: X' = ​​AX + BY (1) In the formula, A is the driving operator of the system's inherent dynamics, B is the input driving operator, and the observed state x is... j With control input y j All based on time interval Δ t = t j+1 - t j sampling, m Let Y be the number of snapshots, where Y = [y0, y1, …, y]. m-1 Let X' be the input matrix, where X' = [x1, x2, ..., x]. m ] and X=[x0, x1, …, x m-1 The time interval between the two units is [different].

4. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 3, characterized in that, In real-world scenarios, operator B cannot be directly obtained, and formula (1) is rewritten as: （2） C and Z are the augmentation operator and the snapshot matrix, respectively.

5. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 4, characterized in that, To obtain the eigenvalues ​​and eigenvectors of operator A, we first need to perform singular value decomposition on matrix Z, and matrix C can then be expressed as: （3） † Refers to the Moore-Penrose pseudoinverse; * indicates complex conjugate transpose. Let Z represent the left singular matrix, singular value matrix, and right singular matrix of matrix Z, respectively.

6. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 5, characterized in that, Expanding formula (3) into blocks, we get: （4） and For matrix The block results show that the inherent dynamic characteristics of the system are mainly reflected in the measurement output matrix X'. Therefore, singular value decomposition needs to be performed on matrix X' separately to obtain the modal subspace Û dominated only by state evolution.

7. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 6, characterized in that, The eigenvalue decomposition of operator A is approximated by reduced-order projection: （5） * indicates complex conjugate transpose. For matrix The block result is Û, which is the orthogonal projection matrix of matrix X'; The approximation operator A preserves the first part of the original operator A. m The dominant eigenvalue Λ = Diag( λ 1, λ 2, …, λ m ), By execution ÃW = WΛ (6) The Hankel-DMDc mode matrix Φ = [Φ1, Φ2, …, Φ m The eigenvector matrix W = [w1, w2, …, w] is approximated by the eigenvector matrix of the approximation operator Ã. m Determine: （7） make Ω=In(Λ) / δ t (8) The snapshot at any given time is approximately reconstructed as follows: （9） α = [ α 1, α 2, …, α m ] T Given the Hankel-DMDc amplitude vector, the initial snapshot is projected onto the Hankel-DMDc mode matrix to obtain: a= F † x0 (10) x0 is the initial value. t The state measurement results at time 0 The Hankel-DMDc algorithm requires performing singular value decomposition on the augmented matrix Z and the measurement output matrix X' respectively, with the truncation orders denoted as follows: and r Finally, the inherent behavior of the system decoupled from external control is reconstructed by reducing its order using formula (9).

8. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 1, characterized in that, In the definition of (2) control effect, the effect of lock-up control on the PTO curve in the predictive control method is defined by the IO system. The input parameter is the control time and the control decoupling PTO motion prediction curve provided by the state predictor, while the output parameter is the PTO response curve after the control is applied. The effect of the lock-up strategy is mainly described by phase delay and amplitude scaling. The mapping relationship between control time and PTO dynamics is defined by function fitting or neural network method.

9. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 1, characterized in that, The optimization solution in (3) is based on the uncontrolled PTO motion curve and IO system in the next two wave cycles. The optimal control time is determined by the optimization algorithm. When the PTO speed returns to zero, each WEC unit performs state prediction and optimization decision in sequence, and then gives the optimal control time that satisfies the objective function in the next motion cycle. The control dynamic optimization problem is solved by an online surrogate model with integer constraints. The PTO motion curve reconstructed by order reduction is truncated at the end of the next motion cycle, and the intermediate time interval is defined as follows: T M Therefore, the optimization problem is formulated as determining the motion period 0 ≤ t ≤ T M The optimal control time within the time frame is determined to maximize wave energy extraction; to adapt to integer constraint algorithms, the optimal control time of the time-discrete dynamic system is equivalent to 0 ≤ x ≤ L M The optimal integer solution within the range, where L M = T M / d t The number of discrete time steps corresponding to the motion cycle, d t For discrete time intervals.

10. The floating predictive lock-up control method for enhancing the performance of wind-wave coupled units as described in claim 9, characterized in that, Control period coefficient σ M As a key input parameter of the I / O system, it is represented as σ M = x / L M ,in x To optimize the integer test points generated by the solver; The dynamic effects of PTO through embedded locking strategies X PTOk ( c ) = X PTOk (0), and , c = 1, 2, …, x The IO system generates PTO displacement and velocity curves characterized by a quadratic model based on the iteratively updated control cycle coefficients; the PTO response curves are then fed back to the optimization solver, and the optimal control timing is determined based on the PTO power performance.

Images