A method for pitch control of offshore wind turbine generators based on predefined time adaptive dynamic programming

Abstract

The application discloses a kind of offshore wind turbine pitch control methods based on predefined time adaptive dynamic programming, comprising: establishing single-pile offshore wind power generator full coupling mathematical model;The single-pile offshore wind power generator full coupling mathematical model includes the motion equation of system and the mathematical model of double-mass transmission system;Using the mathematical model of double-mass transmission system constructs rotor speed tracking error, adopts error adjustment method to process the non-affine characteristic of system, obtains the nonlinear affine model of single-pile offshore wind power generation system in combination with motion equation;The control strategy is constructed as the infinite time domain optimal performance index function of the change of pitch angle;Using predefined time nonlinear disturbance estimator estimates disturbance term in nonlinear affine model;To the equivalent form of Hamilton-Jacobi-Bellman equation of infinite time domain optimal performance index function, the approximate optimal control method based on only evaluation type neural network is used to solve equation to obtain optimal control strategy.

Description

Technical Field

[0001] This invention relates to the field of offshore wind turbine control, and in particular to a method for controlling the pitch of offshore wind turbines based on predefined time adaptive dynamic programming. Background Technology

[0002] Offshore wind turbine foundations can be broadly categorized into two types: fixed and floating. Fixed foundations anchor the tower structure directly to the seabed, making them suitable for shallow nearshore areas. Floating foundations, on the other hand, rely on mooring systems to maintain platform stability and can theoretically be deployed in deep-sea areas. Although floating offshore wind power is considered the ultimate form of the future due to its broader resource potential, its levelized cost of electricity (LCOE) remains significantly higher than that of fixed foundations due to the current high platform construction costs and the lack of a mature large-scale operation and maintenance system, making large-scale commercial application still some time away. In contrast, fixed offshore wind power, after nearly two decades of engineering practice and technological iteration, has a highly mature industrial chain and will remain the dominant force in new offshore wind power capacity additions for the foreseeable future.

[0003] Among various fixed foundation types, monopile offshore wind turbines (MOWTs) are subjected to the combined effects of multiple random excitations, including turbulent winds, irregular waves, and ocean currents, resulting in significant cyclic alternating loads on their tower structures in the out-of-plane (forward and backward) direction. Furthermore, the wind direction and wave propagation direction often exhibit a significant angle in actual marine environments, and this wind-wave misalignment further exacerbates the out-of-plane vibration response and fatigue damage accumulation. Therefore, effectively suppressing out-of-plane structural vibration and fatigue loads while achieving stable output power regulation is a key technical issue for ensuring the long-term safe operation and economic viability of MOWTs.

[0004] For MOWTs (Metal Oxide Wind Turbines) exposed to complex marine dynamic environments for extended periods, the convergence speed of the control system is a key performance indicator for evaluating their operational quality and survivability under extreme conditions. Fast convergence means the unit can suppress state deviations caused by external stimuli in a shorter time, thereby reducing the speed overshoot amplitude, power fluctuation variance, and tower structural stress cycle range during transient processes. Traditional asymptotic convergence control can only guarantee that the error approaches zero as time approaches infinity, lacking quantitative constraints on the convergence rate of transient processes. Finite-time control, by introducing fractional power feedback terms, can bring the system state to the equilibrium point within a finite time, but its upper bound on the convergence time depends on the initial state of the system. When the unit encounters extreme gusts that cause large initial errors, the convergence process will be significantly prolonged. Fixed-time control overcomes the dependence on the initial state by carefully constructing Lyapunov functions, but its convergence time function exhibits a complex implicit coupling relationship with the controller parameters, making it difficult for designers to accurately preset the upper bound of the convergence time based on practical engineering requirements such as grid fault transit time windows or pitch actuator response limits.

[0005] To overcome the limitation of uncontrollable convergence time, predefined time stability theory has emerged. This theory allows users to directly embed the desired convergence time as an explicit adjustable parameter into the controller design, enabling the system state to converge to the neighborhood of the equilibrium point within a predefined time upper bound, which is completely independent of the system's initial conditions. This characteristic has extremely high engineering application value for wind power systems requiring strict time-series responses. Existing research mostly focuses on improving power generation in low-wind-speed regions and does not involve multi-objective cooperative optimization under an optimal control framework. The introduction of predefined time stability theory into the field of optimal pitch control for MOWTs has not yet been reported.

[0006] Deeply integrating predefined-time control with the adaptive dynamic programming (ADP) optimal control framework is an ideal path to achieve fast convergence of MOWTs and multi-objective collaborative optimization, but it still faces significant theoretical and technical challenges. On the one hand, ADP relies on neural networks to learn online to approximate the optimal value function and policy, and transient uncertainties during the learning process may undermine the deterministic guarantee of predefined-time convergence. On the other hand, lumped disturbances (including unmodeled dynamics and external disturbances) that cannot be compensated for in a timely and accurate manner will also hinder the convergence of the system state within the predefined time. Disturbance estimator techniques provide an effective way to suppress lumped disturbances, but traditional disturbance estimators can usually only guarantee asymptotic or finite-time convergence of the estimation error. In recent years, the concept of predefined-time disturbance estimator (PTNDE) has been proposed, which can accurately estimate unknown disturbances within a predefined time. Existing PTNDE research is mostly geared towards systems such as spacecraft and drones that do not involve complex fluid-structure interaction and multiphysics interaction. For high-order nonlinear wind-wave coupled systems such as MOWTs, how to organically combine PTNDE with the ADP optimal control framework and theoretically guarantee the predefined time convergence characteristics of the fused closed-loop system remains an open challenge in the field of intelligent control. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by proposing a pitch control method for offshore wind turbines based on predefined time adaptive dynamic programming (PT-ADP). This method aims to achieve coordinated optimization of out-of-plane power generation regulation and structural load suppression of key components in MOWTs, and to ensure that the system convergence time can be explicitly preset by the user.

[0008] The objective of this invention is achieved through the following technical solution: a method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming, the method comprising:

[0009] First, a fully coupled mathematical model of a monopile offshore wind turbine is established; the fully coupled mathematical model of the monopile offshore wind turbine includes the system's equations of motion and the mathematical model of the dual-mass transmission system;

[0010] The rotor speed tracking error is constructed using a mathematical model of a dual-mass transmission system. The non-affine characteristics of the system are handled by an error adjustment method. The non-affine model of the monopile offshore wind power generation system is obtained by combining the equation of motion.

[0011] Construct the optimal infinite time-domain performance index function for the control strategy of pitch angle variation;

[0012] Use a predefined time-nonlinear perturbation estimator to estimate the perturbation term in a nonlinear affine model;

[0013] An equivalent form of the Hamilton-Jacobi-Bellman equation is constructed for the optimal performance index function in the infinite time domain. The optimal control strategy is obtained by solving the equation based on the approximate optimal control method of the evaluation-only neural network.

[0014] Furthermore, the equations of motion of the system are derived using the Euler-Lagrange equations, where the Lagrange operator represents the difference between the system's kinetic and potential energy. The system's kinetic and potential energy in the out-of-plane direction are obtained from the absolute displacement of the computer cabin in the forward and backward directions, the velocity of the cabin in the forward and backward directions, and the absolute velocities of the tower and monopile in the forward and backward directions. Substituting the kinetic and potential energy formulas into the Euler-Lagrange equations yields the system's equations of motion.

[0015] Furthermore, the mathematical model of the dual-mass transmission system includes:

[0016] The rotor speed change rate model is obtained by considering the rotor speed and aerodynamic torque, the rotor's moment of inertia, the damping coefficient and spring constant of the flexible shaft, the gearbox gear ratio, the generator speed, and the torsional angle of the transmission system.

[0017] The rotor aerodynamic torque variation rate model is obtained by considering the damping coefficient and spring constant of the flexible shaft, the moment of inertia, rotational speed and torque of the generator, the rotor speed, and the torsional angle of the transmission system.

[0018] The model of the rate of change of the torsional angle of the dynamic system is obtained by using the rotor speed, generator speed, and gearbox speed ratio.

[0019] Furthermore, the construction process of the nonlinear affine model includes: obtaining the unit structure dynamics relationship based on the motion equation of the monopile offshore wind turbine; obtaining the transmission system dynamics relationship between rotor speed, generator speed and transmission shaft torsion angle based on the mathematical model of the dual-mass transmission system; subsequently, constructing error dynamics based on the tracking error between rotor speed and desired rotor speed, and introducing the pitch angle change rate into the error dynamics using an error adjustment method to handle the non-affine characteristics of the system;

[0020] The nonlinear affine model combines the unit's structural motion state, transmission system state, and rotor speed regulation error into system state variables, and uses the pitch angle change rate as the control input. The dynamic parts of the system without control input are merged into system state functions, the dynamic parts related to control input are merged into control input functions, and wind and wave loads, external disturbances, measurement noise, model uncertainties, and unmodeled dynamics are merged into lumped disturbance terms.

[0021] Furthermore, the optimal infinite time-domain performance index function is obtained by accumulating the utility function over the entire operating cycle. The utility function includes a quadratic weighted term for the state variables and a quadratic weighted term for the control input. The state variable weighted term is used to evaluate the power generation regulation performance and structural load suppression performance, while the control input weighted term is used to evaluate the control input consumption. Finally, the rate of change of the pitch angle is used as the control strategy. The system state and the control strategy are input into the utility function, and the optimal pitch control strategy is obtained by minimizing the infinite time-domain performance index function.

[0022] Furthermore, using a predefined time-nonlinear disturbance estimator includes: introducing an auxiliary system whose disturbance term is derived from the difference between the state of the monopile offshore wind turbine system and that of the auxiliary system. Composition; the estimated value of the lumped disturbance term of the monopile offshore wind turbine system obtained by the estimator: the estimated value of the disturbance term of the auxiliary system and the state of the monopile offshore wind turbine system and The sum of derivatives, the The estimated update rate is obtained from a predefined time constant and the estimation error.

[0023] Furthermore, the equivalent form of the Hamilton-Jacobi-Bellman equation for the infinite time-domain optimal performance index function includes:

[0024] The first Hamilton-Jacobi-Bellman equation is obtained by differentiating the optimal performance index function in the infinite time domain with respect to time. The partial derivative of the first Hamilton-Jacobi-Bellman equation with respect to the optimal control policy is obtained to give the expression of the optimal control policy. Substituting the expression of the optimal control policy into the first Hamilton-Jacobi-Bellman equation, the equivalent form of the Hamilton-Jacobi-Bellman equation is obtained.

[0025] Furthermore, the evaluation-only neural network structure specifically involves: obtaining the estimated value of the optimal performance index function through the activation function, the optimal weight matrix, and the approximation error; performing online estimation of the optimal weight matrix; and constructing the optimal pitch control law based on the evaluation-only neural network structure using predefined time-adaptive dynamic programming.

[0026] On the other hand, the specification also provides an offshore wind turbine pitch control device based on predefined time adaptive dynamic programming, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements the offshore wind turbine pitch control method based on predefined time adaptive dynamic programming.

[0027] On the other hand, the specification also provides a computer-readable storage medium on which a program is stored, which, when executed by a processor, implements the aforementioned method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming.

[0028] The beneficial effects of this invention are:

[0029] 1. This invention solves the problems of strong nonlinearity and difficult controller design of offshore wind turbines under wind and wave coupling by establishing a fully coupled dynamic model of a monopile offshore wind turbine and constructing a nonlinear affine system model based on the error adjustment method. This enables the control system to more accurately characterize the dynamic characteristics of the turbine and provides a unified modeling basis for the subsequent optimal controller design.

[0030] 2. This invention designs a predefined time nonlinear disturbance estimator to perform online estimation and compensation for wind disturbances, wave disturbances, model uncertainties, and lumped disturbances formed by unmodeled dynamics. This solves the problems of slow estimation speed and insufficient compensation accuracy of traditional disturbance observation methods, and improves the system's ability to suppress disturbances in complex marine environments and the robustness of the closed-loop system.

[0031] 3. This invention combines predefined time stability theory with adaptive dynamic programming technology to construct a predefined time adaptive dynamic programming optimal controller based on an evaluation-only neural network. This solves the problems of traditional adaptive dynamic programming methods, such as difficulty in guaranteeing convergence time and poor transient performance during the learning process. It enables the system state to converge quickly within a pre-set time, thereby improving the dynamic response performance of the system.

[0032] 4. This invention solves the problem that existing control strategies cannot simultaneously address power regulation and structural load suppression by constructing an infinite time-domain optimal performance index function based on state performance index and control input performance index, comprehensively considering power generation regulation performance, structural load suppression performance, and control input consumption, thereby achieving synergistic optimization of power generation performance and structural safety.

[0033] 5. This invention uses the pitch angle change rate as the control input and combines it with the optimal control strategy to control the pitch actuator, which solves the problem of large fluctuations in traditional pitch control input and easy fatigue of the actuator, making pitch control smoother, reducing wear of the actuator, and extending the service life of key components.

[0034] 6. Simulation results show that the present invention can effectively improve rotor speed tracking accuracy, reduce power generation fluctuations, and significantly reduce torsional vibration of the drive shaft, forward and backward displacement of the nacelle, and forward and backward bending moment at the tower base. It achieves synergistic optimization of power generation regulation and structural load suppression of offshore wind turbines, thereby improving the safety, reliability, and economy of turbine operation. Attached Figure Description

[0035] Figure 1 The MOWT dynamic model under wind and wave loads provided in this embodiment of the invention;

[0036] Figure 2 A simplified model of single pile-soil interaction provided in this embodiment of the invention;

[0037] Figure 3 A flowchart of the method provided in an embodiment of the present invention;

[0038] Figure 4 This invention provides a wind speed measurement and an irregular wave surface elevation map for embodiments of the invention.

[0039] Figure 5 The performance diagram of PTNDE perturbation estimation provided in the embodiments of the present invention;

[0040] Figure 6 A comparison chart of rotor speed and power generation provided in an embodiment of the present invention;

[0041] Figure 7 The transmission shaft torsional displacement, nacelle front-to-back displacement, tower base front-to-back bending moment, and propeller pitch control input diagrams are provided for embodiments of the present invention.

[0042] Figure 8 This is a schematic diagram of the apparatus provided in an embodiment of the present invention. Detailed Implementation

[0043] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0044] This invention proposes a pitch control method for offshore wind turbines based on predefined time adaptive dynamic programming.

[0045] The specific steps include:

[0046] Step 1: Establish a fully coupled mathematical model of a monopile offshore wind turbine;

[0047] Step 2: Construct a nonlinear affine system for a monopile offshore wind turbine based on filtering and error adjustment technology;

[0048] Step 3: Design a predefined time-nonlinear perturbation estimator;

[0049] Step 4: Design the optimal pitch controller by combining predefined time theory and adaptive dynamic programming techniques;

[0050] Step 5: Prove the actual predefined time stability of the system based on Lyapunov functions.

[0051] Step 1 specifically includes:

[0052] The mathematical model of MOWT is derived using the Euler-Lagrange equations, and its specific form is as follows:

[0053] (1)

[0054] in, Let be a Lagrange operator, representing the kinetic energy of the MOWT system. With potential energy difference. For the system's first One degree of freedom, In order to be with the first Generalized forces related to each degree of freedom. yes Regarding the derivative of time, the dot (·) above the characters in this application represents the derivative of the relevant character. Represented as about Find the partial derivative. Represented as about Find the partial derivative. Represented as Find the derivative with respect to time.

[0055] Figure 1 The dynamic structure of MOWT in the out-of-plane direction and its corresponding coordinate system definition are given. Coordinate system It is fixed at the intersection of the tower centerline and the mean sea level (MSL). The axis... Pointing towards the average wind direction, axis Along the centerline of the tower, pointing from mean sea level to the top of the tower. (Symbol) Indicates the cabin relative to a point Displacement coordinates in the forward and backward directions, and These represent the rotational coordinates of the foundation structure in the translational and rotational directions, respectively. Additionally, the symbols... It indicates the angle between the wind direction and the wave direction.

[0056] A simplified model of single-pile-soil interaction used to describe the influence of soil is as follows: Figure 2 As shown. The soil damping effect is modeled using translational and rotational dampers, with damping coefficients of respectively... and .

[0057] Absolute displacement of the cabin in the longitudinal direction It can be represented as:

[0058] (2)

[0059] in, This refers to the height of the engine room relative to the seabed.

[0060] Then, the cabin's speed in the forward and backward directions. It can be represented as:

[0061] (3)

[0062] The absolute velocity of the tower and monopile in the forward and backward directions is The expression at the point is:

[0063] (4)

[0064] in, These are the basic vibration modes of the tower and monopile in the front-rear direction. Indicates water depth.

[0065] Based on the above equations, the kinetic energy of MOWT in the out-of-plane direction and potential energy It can be represented as:

[0066] (5)

[0067] in, For cabin quality, For the quality of the wheel hub, For the distributed mass of the tower and monopile, The distance from the top of the tower to the mean sea level. This refers to the longitudinal stiffness of the tower and the monopile. Furthermore, and These represent the equivalent spring stiffness coefficients in the translation and rotation directions, respectively.

[0068] By substituting (5) into (1), we obtain the system's equation of motion:

[0069] (6)

[0070] in, , and Let these represent the mass, damping, and stiffness matrices of the MOWT system, respectively. Represents a generalized coordinate vector. and These are the generalized force vectors caused by wind load and wave load, respectively. Represented as The second derivative of the character, and the dots above the characters in this patent are all the second derivatives of the relevant characters.

[0071] The dual-mass drive system of the wind turbine consists of a damping coefficient and spring constant The system consists of flexible shafts used to transmit torque from the rotor to the generator. The mathematical model of this dual-mass transmission system can be described as follows:

[0072] (7)

[0073] in, and These are rotor speed and aerodynamic torque, respectively. and These are the moments of inertia of the rotor and the generator, respectively. , and These are the gearbox gear ratio, generator speed, and generator torque, respectively. This refers to the torsional angle of the transmission system. ,in, Due to model uncertainty, This is an external disturbance.

[0074] Step 2 specifically includes:

[0075] Step 2.1: To improve the power generation regulation performance, define the rotor speed tracking error. for:

[0076] (8)

[0077] in, This represents the desired rotor angular velocity. The error dynamics can be expressed as:

[0078] (9)

[0079] Subsequently, to facilitate controller design, an error adjustment method is used to handle the non-affine characteristics of the system, and the adjustment error is defined as:

[0080] (10)

[0081] in, Indicates adjustment error. It is a positive integer. Clearly, when... When it asymptotically converges to zero, It will also gradually converge to zero.

[0082] Combining (9) and (10), the adjustment error can be obtained. The dynamic equation is:

[0083] (11)

[0084] in, Indicates the pitch angle The rate of change. Furthermore, the system state function Control input function and disturbance It can be represented as:

[0085] (12)

[0086] in, The wind speed is represented. The system state function is used to characterize the coupling relationship between the unit structure dynamics, transmission system dynamics, and rotor speed error dynamics. The control input function is used to characterize the influence of the pitch angle change rate on the system state change. The lumped disturbance term is used to characterize the combined influence of the complex marine environment and model uncertainties on the system state.

[0087] It should be noted that the wind speed in this study Variables that are not directly accessible, and whose measured values ​​can only be obtained. Therefore, we can assume:

[0088] (13)

[0089] in, For bounded measurement noise, satisfy ,and .

[0090] Considering (6), (7), and (11), and combining these three equations into an augmented system, the MOWT system can be expressed in the following nonlinear affine form:

[0091] (14)

[0092] The system status is as follows: The control input is In addition, the system state matrix Control input matrix and concentrated disturbance terms Represented as:

[0093] (15)

[0094] in, Defined as -1 represents the matrix inversion operation, and T represents the matrix transpose operation.

[0095] Accordingly, the nominal form of the MOWT nonlinear affine model (14) can be expressed as:

[0096] (16)

[0097] To facilitate subsequent analysis, the following assumptions are proposed.

[0098] Assumption 1: Assume there exists a constant. Make .

[0099] Step 2.2: This patent studies the pitch control problem of MOWT (Motorcycle Weighing Threshold) under operating conditions above rated wind speed. Its main objectives include power generation regulation and structural load suppression. Existing research shows that control strategies solely aimed at reducing structural loads often sacrifice power generation performance, while strategies primarily focused on power generation regulation may accelerate structural fatigue accumulation in key components such as the tower and pitch actuator. This contradictory characteristic presents a significant challenge to the design of control algorithms that simultaneously improve power generation quality and reduce structural fatigue loads.

[0100] In response to the above problems, the control objectives of this patent for the MOWT system are summarized as follows: (1) To control the rotor speed through pitch control. Track its expected rotation speed (2) By using optimal control theory, reduce the structural fatigue load of the drive shaft system, tower and pitch actuator in the out-of-plane direction; (3) Accurately estimate and compensate for the concentrated disturbances caused by wind disturbance, external interference and model uncertainty, so as to enhance the robustness of the closed-loop system.

[0101] To ensure optimal performance of the controlled MOWT system (16), a control strategy needs to be determined. Minimize the following infinite time-domain performance metrics:

[0102] (17)

[0103] in, Let be a utility function, and satisfy... . and These are positive definite symmetric weight matrices. The infinite time-domain performance index function is obtained by accumulating the utility function over the entire operating cycle. The utility function includes quadratic weighted terms for state variables and quadratic weighted terms for control inputs. The state variable weighted terms evaluate power generation regulation performance and structural load suppression performance, while the control input weighted terms evaluate control input consumption. Finally, the optimal pitch control strategy is obtained by minimizing the infinite time-domain performance index function.

[0104] Step 3 specifically includes:

[0105] For ease of representation, Abbreviated as For the MOWT nonlinear affine model (14), a nonlinear perturbation estimator is designed to compensate for the lumped perturbation term. First, we introduce an auxiliary system of the following form:

[0106] (18)

[0107] in, It is a positive number, and the auxiliary system error Combining (14) and (18), we can obtain:

[0108] (19)

[0109] Subsequently, a predefined time-nonlinear perturbation estimator (PTNDE) is proposed for estimating perturbations. Its expression is:

[0110] (20)

[0111] in, Indicates disturbance The estimated value, express The estimator, whose update law is:

[0112] (twenty one)

[0113] in, To estimate the error, positive constants normal numbers normal numbers normal numbers Need to meet normal numbers normal numbers normal numbers ,as well as This is a predefined time constant. Furthermore, Defined as a function ,in For symbolic functions, It is a positive number.

[0114] Combining (19) and (21), the dynamic equation for the state estimation error can be obtained as follows:

[0115] (twenty two)

[0116] Theorem 1: When using the PTNDE schemes (20) and (21), the disturbance estimation error is Able to predefined time It converges to zero.

[0117] Proof: To analyze the estimation error To assess convergence, the following Lyapunov function is selected:

[0118] (twenty three)

[0119] right Taking the derivative with respect to time, we get:

[0120] (twenty four)

[0121] Note that for any scalar All have as well as Therefore, we can conclude that:

[0122] (25)

[0123] make and noticed Then we can obtain:

[0124] (26)

[0125] therefore, It can be rewritten as:

[0126] (27)

[0127] Will , and Substituting into (27), we can further obtain:

[0128] (28)

[0129] Therefore, when Sometimes, Further, the disturbance estimation error can be obtained. for:

[0130] (29)

[0131] Therefore, when Sometimes, Q.E.D.

[0132] Step 4 specifically includes:

[0133] For the optimal control strategy The infinite time-domain optimal performance index function corresponding to system (16) It can be represented as:

[0134] (30)

[0135] To simplify the representation, Abbreviated as Regarding (30) Differentiating the equations, we obtain the Hamilton-Jacobi-Bellman (HJB) equations as follows:

[0136] (31)

[0137] in, Let represent the gradient operator. Regarding (31)... Taking the partial derivative, the optimal control strategy can be obtained as follows:

[0138] (32)

[0139] Substituting the optimal control strategy (32) into the HJB equation, we obtain the following equivalent form:

[0140] (33)

[0141] Because the MOWT system exhibits significant nonlinear characteristics, directly solving the HJB equations yields the desired results. This is typically difficult to achieve. Therefore, an approximate optimal control method based on an evaluation-only neural network (NN) is employed to estimate the optimal performance index function. This evaluation-only NN can be expressed as:

[0142] (34)

[0143] in, For activation function, The optimal weight matrix is... This represents the NN approximation error. For the sake of brevity, it will be referred to as... and They are abbreviated as and .right about Taking the derivative, we get:

[0144] (35)

[0145] It should be noted that the optimal weight matrix It is unknown and requires online estimation; the estimated value is denoted as... .therefore, An approximate estimate can be expressed as:

[0146] (36)

[0147] Based on the evaluation-only NN structure (34), the optimal control strategy (32) can be rewritten as:

[0148] (37)

[0149] Accordingly, the approximate optimal control strategy can be expressed as:

[0150] (38)

[0151] Based on this, the predefined time-adaptive dynamic programming (PT-ADP) optimal pitch control law can be constructed as follows:

[0152] (39)

[0153] in, Indicates the pseudo-inverse operator, , and A positive integer. Parameter , , , , , , ,as well as This is a predefined time constant.

[0154] Neural network weights The update law can be expressed as:

[0155] (40)

[0156] in, For positive integers, , , .also, ,as well as .

[0157] Assumption 2: Optimal weight matrix It is bounded, that is Furthermore, only the activation function of evaluative neural networks satisfies Its partial derivatives satisfy ,in and It is a positive number.

[0158] Step 5 specifically includes:

[0159] The following theorem proves the actual predefined time stability of the system.

[0160] Theorem 2: Consider the MOWT nonlinear affine model (14) under the influence of external disturbances and internal model uncertainties. With the introduction of PTNDE strategies (20) and (21), if the PT-ADP optimal pitch control law is constructed by (39), and the neural network weights If updated according to (40), the system status will be... and neural network weight approximation error Able to predefined time It converges inward to a small neighborhood near the origin, i.e., satisfies Furthermore, the control law given by (39) It can approximate the optimal control strategy in (37) .

[0161] Proof: Choose the following Lyapunov function:

[0162] (41)

[0163] To simplify the representation, and They are respectively denoted as and Taking the derivative of (41) with respect to time, we get:

[0164] (42)

[0165] Note It is a local Lipschitz function, therefore it has positive constants. Make .according to The inequality properties of functions can be obtained. ,in It is a positive constant. Furthermore, according to Theorem 1, when... Sometimes, Substitute (38) and (40) into Combining assumptions 1 and 2, we can obtain:

[0166] (43)

[0167] in, .

[0168] When the condition is met Then, the above formula can be further written as:

[0169] (44)

[0170] According to Young's inequality, we can obtain:

[0171] (45)

[0172] in, , Therefore, it can be seen that Further organized as follows:

[0173] (46)

[0174] in, .

[0175] Further , , , , and Substituting (46), we get:

[0176] (47)

[0177] The system state can be determined from (47). and NN weight approximation error At a predefined time Converges inward to the following compact set:

[0178] (48)

[0179] in, Therefore, the system equilibrium point is actually stable over a predefined time.

[0180] Combining (37) and (39) further, we can obtain:

[0181] (49)

[0182] Reasonable assumptions and ,in and It is a positive number. Because... This represents a perturbation term in a physical system, and therefore can be considered bounded. Furthermore, since... Bounded, can be assumed And because There exists a function Make Therefore, we can obtain:

[0183] (50)

[0184] Therefore, when and Converging to compact At that time, control strategy Able to approximate the optimal control strategy Therefore, the control law (39) is an approximately optimal controller, thus completing the proof.

[0185] To verify the feasibility of the proposed PT-ADP optimal pitch control strategy based on a disturbance estimator, numerical simulations were performed on the NREL 5-MW OC3 MOWT using the OpenFAST simulation platform. Subsequently, under the conditions of turbulent wind and irregular waves, a comparative simulation analysis was conducted between the proposed PT-ADP optimal pitch controller based on a disturbance estimator (hereinafter referred to as PADPC) and the traditional ADP control method based on a critic-only neural network (hereinafter referred to as TADPC). Figure 3 This is a flowchart of the present invention.

[0186] Figure 4 The upper part shows the turbulent wind field generated using the TurbSim software under the Kaimal turbulence model, with an average wind speed of 16 m / s at the hub height and a turbulence intensity of 6%. The corresponding wind speed measurements are also provided. Meanwhile, irregular waves are generated using the JONSWAP spectrum. Figure 4 The lower half is shown, where the peak period (PWP) is set to 10 s and the effective wave height (SWH) is set to 8 m.

[0187] Figure 5 The estimation results of the lumped disturbance are presented. Figure 5 There are a total of 8 subgraphs, each representing a lumped disturbance. The comparison between the true and estimated values ​​of the eight components is shown in the figure. As can be seen from the figure, the PTNDE proposed in this patent can achieve [something] within a predefined time. Fast and accurate estimation of lumped disturbances in the system .

[0188] Figure 6 The comparison results between rotor speed and power generation are presented. As can be seen from the figure, compared with the TADPC strategy, the proposed PADPC strategy can bring the rotor speed closer to the rated value and exhibits a smaller tracking error. Meanwhile, the rotor speed tracking error... The convergence time is This is significantly smaller than the predefined upper bound of time. Furthermore, Figure 6 The power generation performance of the two control algorithms was also compared. The results show that the PADPC strategy can achieve more stable output power, thus enabling MOWT to output higher quality electrical energy.

[0189] Figure 7The results further compare the two controllers under multiple performance indicators. It can be seen that, compared with TADPC, the PADPC strategy can effectively suppress torsional vibration of the drive train and significantly reduce the fore-and-aft displacement of the nacelle and the fore-and-aft bending moment at the tower base. Furthermore, the pitch angle input generated by the PADPC strategy is smoother and has smaller fluctuations, indicating that the proposed control strategy can effectively reduce the control input intensity.

[0190] Corresponding to the aforementioned embodiment of an offshore wind turbine pitch control method based on predefined time adaptive dynamic programming, the present invention also provides an embodiment of an offshore wind turbine pitch control device based on predefined time adaptive dynamic programming.

[0191] See Figure 8 The present invention provides an offshore wind turbine pitch control device based on predefined time adaptive dynamic programming, comprising a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement an offshore wind turbine pitch control method based on predefined time adaptive dynamic programming as described in the above embodiment.

[0192] The embodiment of the offshore wind turbine pitch control device based on predefined time adaptive dynamic programming provided by this invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 8 The diagram shown is a hardware structure diagram of any data processing-capable device, including a predefined time-adaptive dynamic programming-based offshore wind turbine pitch control device provided by the present invention. (Except for...) Figure 8 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0193] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0194] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0195] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements a method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming as described in the above embodiments.

[0196] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0197] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming.

[0198] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0199] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. This application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming, characterized in that, The method includes: First, a fully coupled mathematical model of a monopile offshore wind turbine is established; the fully coupled mathematical model of the monopile offshore wind turbine includes the system's equations of motion and the mathematical model of the dual-mass transmission system; The rotor speed tracking error is constructed using a mathematical model of a dual-mass transmission system. The non-affine characteristics of the system are handled by an error adjustment method. The non-affine model of the monopile offshore wind power generation system is obtained by combining the equation of motion. Construct the optimal infinite time-domain performance index function for the control strategy of pitch angle variation; Use a predefined time-nonlinear perturbation estimator to estimate the perturbation term in a nonlinear affine model; An equivalent form of the Hamilton-Jacobi-Bellman equation is constructed for the optimal performance index function in the infinite time domain. The optimal control strategy is obtained by solving the equation based on the approximate optimal control method of the evaluation-only neural network.

2. The method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming according to claim 1, characterized in that, The system's equation of motion is derived using the Euler-Lagrange equations, where the Lagrange operator represents the difference between the system's kinetic and potential energies. The system's out-of-plane kinetic and potential energies are obtained from the absolute displacement of the computer cabin in the forward and backward directions, the velocity of the cabin in the forward and backward directions, and the absolute energies of the tower and monopile in the forward and backward directions. Substituting the kinetic and potential energy formulas into the Euler-Lagrange equations yields the system's equation of motion.

3. The method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming according to claim 1, characterized in that, The mathematical model of the dual-mass transmission system includes: The rotor speed change rate model is obtained by considering the rotor speed and aerodynamic torque, the rotor's moment of inertia, the damping coefficient and spring constant of the flexible shaft, the gearbox gear ratio, the generator speed, and the torsional angle of the transmission system. The rotor aerodynamic torque variation rate model is obtained by considering the damping coefficient and spring constant of the flexible shaft, the moment of inertia, rotational speed and torque of the generator, the rotor speed, and the torsional angle of the transmission system. The model of the rate of change of the torsional angle of the dynamic system is obtained by using the rotor speed, generator speed, and gearbox speed ratio.

4. The method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming according to claim 1, characterized in that, The construction process of the nonlinear affine model includes: obtaining the unit structure dynamics relationship based on the motion equation of the monopile offshore wind turbine; obtaining the transmission system dynamics relationship between rotor speed, generator speed and transmission shaft torsion angle based on the mathematical model of the dual-mass transmission system; subsequently, constructing error dynamics based on the tracking error between rotor speed and desired rotor speed, and introducing the pitch angle change rate into the error dynamics using an error adjustment method to handle the non-affine characteristics of the system. The nonlinear affine model combines the unit's structural motion state, transmission system state, and rotor speed regulation error into system state variables, and uses the pitch angle change rate as the control input. The dynamic parts of the system without control input are merged into system state functions, the dynamic parts related to control input are merged into control input functions, and wind and wave loads, external disturbances, measurement noise, model uncertainties, and unmodeled dynamics are merged into lumped disturbance terms.

5. The method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming according to claim 1, characterized in that, The optimal infinite time-domain performance index function is obtained by accumulating the utility function over the entire operating cycle. The utility function includes a quadratic weighted term for the state variables and a quadratic weighted term for the control input. The state variable weighted term is used to evaluate the power generation regulation performance and structural load suppression performance, while the control input weighted term is used to evaluate the control input consumption. Finally, the rate of change of the pitch angle is used as the control strategy. The system state and the control strategy are input into the utility function, and the optimal pitch control strategy is obtained by minimizing the infinite time-domain performance index function.

6. The method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming according to claim 1, characterized in that, Using a predefined time-nonlinear disturbance estimator includes: introducing an auxiliary system whose disturbance term is derived from the difference between the state of the monopile offshore wind turbine system and that of the auxiliary system. Composition; the estimated value of the lumped disturbance term of the monopile offshore wind turbine system obtained by the estimator: the estimated value of the disturbance term of the auxiliary system and the state of the monopile offshore wind turbine system and The sum of derivatives, the The estimated update rate is obtained from a predefined time constant and the estimation error.

7. The method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming according to claim 1, characterized in that, The equivalent form of the Hamilton-Jacobi-Bellman equation for the optimal performance index function in the infinite time domain includes: The first Hamilton-Jacobi-Bellman equation is obtained by differentiating the optimal performance index function in the infinite time domain with respect to time. The partial derivative of the first Hamilton-Jacobi-Bellman equation with respect to the optimal control policy is obtained to give the expression of the optimal control policy. Substituting the expression of the optimal control policy into the first Hamilton-Jacobi-Bellman equation, the equivalent form of the Hamilton-Jacobi-Bellman equation is obtained.

8. The method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming according to claim 1, characterized in that, The evaluation-only neural network structure specifically involves: obtaining the estimated value of the optimal performance index function through the activation function, the optimal weight matrix, and the approximation error; performing online estimation of the optimal weight matrix; and constructing the optimal pitch control law based on the evaluation-only neural network structure using predefined time-adaptive dynamic programming.

9. A pitch control device for offshore wind turbines based on predefined time adaptive dynamic programming, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements a method for controlling the pitch of an offshore wind turbine based on predefined time adaptive dynamic programming as described in any one of claims 1-8.

10. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the pitch control method for offshore wind turbines based on predefined time adaptive dynamic programming as described in any one of claims 1-8.

Images