Wind Turbine Load Reduction Predictive Control Method and System Based on Mechanism and Neural Network Model

Abstract

This invention discloses a wind turbine load reduction predictive control method and system based on mechanism and neural network models. The method includes the following steps: establishing a wind turbine mechanism analysis and neural network model; linearizing the wind turbine mechanism analysis and neural network model at the steady-state operating point of the wind turbine to construct a linearized model of the wind turbine; constructing a wind turbine load reduction predictive control model based on the linearized model, wherein the load reduction predictive control model aims to minimize the wind turbine load while balancing the wind turbine's power generation efficiency; and solving the load reduction predictive control model to obtain the optimal values ​​of various control parameters of the wind turbine. This invention can achieve a high-precision solution to the wind turbine load reduction control problem.

Description

Technical Field

[0001] This invention relates to the field of wind turbine load reduction control technology, specifically to a wind turbine load reduction predictive control method and system based on mechanism and neural network model. Background Technology

[0002] Traditional PID control strategies for wind turbines can lead to excessive loads under extreme wind speeds, posing a significant challenge to the load-bearing capacity of turbine blades and tower materials. For example, excessive loads can cause fatigue damage to turbine blades, towers, and other structural components, increasing the risk of cracks and fractures. Therefore, wind turbine load reduction control technology is necessary. To improve the accuracy of wind turbine modeling, complex wind turbine dynamic models are typically established. Due to the complexity of these models, directly designing model predictive control strategies is difficult to solve. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a wind turbine load reduction prediction control method and system based on mechanism and neural network model, which addresses the above-mentioned problems in the prior art and achieves high-precision solution to the wind turbine load reduction control problem.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0005] A wind turbine load reduction predictive control method based on mechanism and neural network model includes the following steps:

[0006] S1. Establish a wind turbine mechanism analysis and neural network model, and linearize the wind turbine mechanism analysis and neural network model at the steady-state operating point of the wind turbine to construct a linearized model of the wind turbine.

[0007] S2, Based on the linearized model, construct a load reduction prediction control model for the wind turbine, wherein the load reduction prediction control model aims to minimize the wind turbine load while taking into account the wind turbine's power generation efficiency.

[0008] S3, Solve the load reduction predictive control model to obtain the optimal values ​​of each control parameter of the wind turbine.

[0009] Further, in step S1, when linearizing the wind turbine mechanism analysis and neural network model, the nonlinear aerodynamic model, neural network model, and wind turbine power generation model in the wind turbine mechanism analysis and neural network model are linearized using the first-order Taylor expansion method; wherein, the established expressions of the wind turbine mechanism analysis and neural network model are as follows:

[0010]

[0011]

[0012]

[0013]

[0014] In the formula, These are the various state variables of the fan. The torsion angle of the drive chain shaft. For rotor speed, For generator speed, The distance the tower moves forward and backward. The forward and backward movement speed of the tower. For the angle of the blade's waving, The angular velocity of the blade flapping. For aerodynamic torque, For generator torque, For air thrust, For wind speed, The pitch angle is the propeller angle. , These are the system matrix and input matrix of the discretized wind turbine dynamics mechanism model, respectively. This is a neural network model.

[0015] Furthermore, the nonlinear aerodynamic model is linearized to obtain the expression of the linearized aerodynamic model:

[0016]

[0017] , , , , , , ,

[0018] In the above formula, , These are the aerodynamic torques at the current moment and at the steady-state operating point of the wind turbine, respectively. , These are the rotor speeds at the current moment and at the steady-state operating point of the wind turbine, respectively. These are the wind speeds at the current moment and at the steady-state operating point of the wind turbine, respectively. , These represent the forward and backward movement speeds of the tower at the current moment and at the wind turbine's steady-state operating point, respectively. , These are the pitch angles at the current moment and at the steady-state operating point of the wind turbine, respectively. , These represent the air thrust at the current moment and at the steady-state operating point of the wind turbine, respectively. This indicates the steady-state operating point. These are intermediate parameters.

[0019] Furthermore, linearizing the neural network model yields the following expression for the linearized neural network model:

[0020]

[0021] ,

[0022] In the above formula, For neural network models, Indicates the steady-state operating point and The neural network function value at that location, This is the first neural network matrix. This is the second neural network matrix. These represent the current state variables of the wind turbine and the state variables of the wind turbine at its steady-state operating point, respectively. These represent the model input at the current moment and the model input at the wind turbine's steady-state operating point, respectively.

[0023] Furthermore, the wind turbine power generation model is linearized to obtain the expression for the linearized wind turbine power generation model:

[0024]

[0025] In the above formula, , These represent the generator power at the current moment and at the wind turbine's steady-state operating point, respectively. , These are the generator torques at the current moment and at the wind turbine's steady-state operating point, respectively. , These represent the generator speeds at the current moment and at the wind turbine's steady-state operating point, respectively.

[0026] Furthermore, in step S1, constructing the linearized model of the wind turbine specifically includes the following steps:

[0027] By substituting the linearized aerodynamic model and the linearized neural network model into the wind turbine mechanism analysis model, a wind turbine deviation state signal is defined, and a linearized state-space model based on the wind turbine deviation state signal is constructed. The expression of the linearized state-space model is as follows:

[0028]

[0029]

[0030]

[0031]

[0032] , ,

[0033] , ,

[0034] , , , , , , ,

[0035] In the above formula, k and k+1 represent time k and time k+1, respectively. These are the fan deviation state signals at time k+1 and time k, respectively. This is the fan deviation control signal. The torsion angle of the drive shaft. For the fan rotor speed, For generator speed, For the angle of the blade's waving, The angular velocity of the blade flapping. The pitch angle is the propeller angle. For generator torque, For the deviation wind speed, , These are the system matrix and input matrix of the discretized wind turbine dynamics mechanism model, respectively. This is the first neural network matrix. This is the second neural network matrix. for The 2nd and 3rd columns of the matrix, for The first column of the matrix block, These are intermediate parameters.

[0036] Furthermore, the objective function expression of the load reduction predictive control model is:

[0037]

[0038] In the above formula, k+i represents time k+i. This is a fan deviation status signal. This is the fan deviation control signal. , , , , , , , , , , , These are the weight coefficients of the objective function. For power deviation, This refers to the torsional angle deviation of the drive shaft. For the fan rotor speed deviation, For generator speed deviation, This refers to the deviation in the forward and backward movement distance of the tower. This refers to the deviation in the forward and backward movement speed of the tower. For the deviation of the blade flapping angle, For the deviation of the blade flapping angular velocity, This is a single-step increment for the pitch angle deviation control quantity. N represents the single-step increment of the generator torque deviation control quantity, where N represents the step size of the predictive control.

[0039] Furthermore, in step S3, before solving the load reduction predictive control model, the wind turbine load reduction predictive control problem is established by combining the wind turbine state constraints, wind turbine control constraints, and generator power constraints. The expression is:

[0040]

[0041] The expression for the wind turbine state constraint is as follows:

[0042]

[0043] In the above formula, Minimum shaft torsion angle, The maximum shaft torsion angle, The maximum rotor speed, Minimum rotor speed, This is the maximum generator speed. Minimum generator speed, For the maximum tower displacement, For the minimum tower displacement, The maximum tower moving speed, For the minimum tower moving speed, For the maximum blade flapping angle, For the minimum blade flapping angle, The maximum blade flapping angular velocity, For the minimum blade flapping angular velocity, The shaft torsion angle at the steady-state operating point of the fan. For the torsional angle deviation of the drive shaft, The rotor speed at the steady-state operating point of the fan. For rotor speed deviation, This refers to the generator speed at the steady-state operating point of the wind turbine. For generator speed deviation, The distance the tower moves back and forth at the steady-state operating point of the wind turbine. This refers to the deviation in the forward and backward movement distance of the tower. The forward and backward movement speed of the tower at the steady-state operating point of the wind turbine. This refers to the deviation in the forward and backward movement speed of the tower. This refers to the blade flapping angle at the steady-state operating point of the wind turbine. For the deviation of the blade flapping angle, The blade flapping angular velocity at the steady-state operating point of the wind turbine. This refers to the deviation in the angular velocity of the blade flapping.

[0044] The expression for the control constraints of the wind turbine is:

[0045]

[0046] In the above formula, For minimum generator torque, For maximum generator torque, For the minimum pitch angle, The maximum pitch angle, The pitch angle is the angle of the blades at the steady-state operating point of the wind turbine. This refers to the pitch angle deviation. This represents the generator torque at the steady-state operating point of the wind turbine. This refers to the generator torque deviation.

[0047] The expression for the generator power constraint is:

[0048]

[0049] In the above formula, This represents the actual maximum generator power. The power generation efficiency factor. This represents the generator power.

[0050] Furthermore, the established wind turbine load reduction prediction and control problem is solved based on a convex quadratic programming solver.

[0051] A wind turbine load reduction predictive control system based on mechanistic and neural network models includes an interconnected microprocessor and a memory, the microprocessor being programmed or configured to execute the steps of a wind turbine load reduction predictive control method based on mechanistic and neural network models.

[0052] A computer-readable storage medium storing a computer program / instructions programmed or configured to execute, via a processor, steps of a wind turbine load reduction predictive control method based on mechanistic and neural network models.

[0053] Compared with the prior art, the advantages of the present invention are as follows:

[0054] This invention establishes a wind turbine mechanism analysis and neural network model, and performs linearization processing at the steady-state operating point. This linearization model can reduce the computational burden, making it easier to analyze and design control, and providing a mathematical foundation for subsequent predictive control algorithms. Based on the linearized model, a load reduction predictive control model is constructed. This model can comprehensively consider the balance between power output and structural load within an optimized control framework, achieving comprehensive optimization of wind turbine performance and lifespan. This improves the economic efficiency and reliability of wind turbine operation under complex wind conditions, thereby enhancing the overall operational stability and lifespan of the wind turbine. Attached Figure Description

[0055] Figure 1 This is a flowchart of a wind turbine load reduction prediction control method based on mechanism and neural network model, according to an embodiment of the present invention. Detailed Implementation

[0056] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0057] like Figure 1 As shown, the wind turbine load reduction predictive control method based on mechanism and neural network model in this embodiment includes the following steps:

[0058] S1. Establish a wind turbine mechanism analysis and neural network model, and linearize the wind turbine mechanism analysis and neural network model at the steady-state operating point of the wind turbine to construct a linearized model of the wind turbine.

[0059] S2, Based on the linearized model, construct a load reduction prediction control model for the wind turbine, wherein the load reduction prediction control model aims to minimize the wind turbine load while taking into account the wind turbine's power generation efficiency.

[0060] S3, Solve the load reduction predictive control model to obtain the optimal values ​​of each control parameter of the wind turbine.

[0061] It is understandable that by establishing a wind turbine mechanism analysis and neural network model and performing linearization at the steady-state operating point, the complex wind turbine model can be linearized. The linearized model has a small computational load, is easy to analyze and control design, and provides a mathematical basis for subsequent predictive control algorithms. Based on the linearized model, a load reduction predictive control model can be constructed, which can comprehensively consider the balance between power output and structural load under the optimized control framework, realize the comprehensive optimization of wind turbine performance and life, improve the economic efficiency and reliability of wind turbine operation under complex wind conditions, and thus improve the overall operational stability and life of wind turbine.

[0062] In specific application embodiments, the description of a wind turbine model typically includes multiple aspects, covering the turbine's structure, aerodynamic characteristics, dynamic behavior, and control system. These models play a crucial role in the design, simulation, and optimization of wind power generation. Establishing a wind turbine's mechanistic model includes an aerodynamic model, a transmission chain model, and a blade-tower coupling model. An example is shown below:

[0063]

[0064] in, , , The torsion angle of the drive chain shaft (unit: rad). Rotor speed (unit: rad / s), Generator speed (unit: rad / s) The distance the tower moves forward and backward (unit: m). The forward and backward movement speed of the tower (unit: m / s) The blade flapping angle (unit: rad) The blade flapping angular velocity (unit: rad / s) Pneumatic torque (unit: Nm). Generator torque (unit: Nm). Air thrust (unit: N). , , This refers to the gearbox transmission ratio. The equivalent rotational inertia on the rotor side (including rotor, hub and main shaft) (unit: kg*m2). The equivalent rotational inertia on the generator side (including the gearbox, high-speed shaft, and generator) (unit: kg*m2). The equivalent torsional stiffness coefficient of the transmission chain system (unit: Nm / rad). The equivalent damping coefficient of the transmission chain system (unit: Nms / rad). Total blade mass (unit: kg). Equivalent mass of the tower (unit: kg). The equivalent damping coefficient of the tower (unit: Nms / rad). The equivalent stiffness coefficient of the tower (unit: Nm / rad). This is the distance from the center of gravity of the blade to the center of gravity at the top of the tower (unit: m). This is the distance from the center of thrust to the center of gravity at the top of the tower (unit: m). The blade flapping angle (unit: rad). The blade damping coefficient (unit: Nms / rad). This is the blade stiffness coefficient (unit: Nm / rad).

[0065] pneumatic torque and air thrust The expression is:

[0066]

[0067] in Air density (unit: kg / m3). The pitch angle is expressed in deg. Relative wind speed (unit: m / s) Wind speed (unit: m / s) The wind energy utilization coefficient, For thrust coefficient, define For the tip speed ratio, The rotor radius is given in meters (m). and It was obtained by looking up a table.

[0068] By applying a high-precision fourth-order Runge-Kutta discretization method to model (1), we can obtain:

[0069]

[0070] in , , The control cycle is discretized for the wind turbine model.

[0071] Due to modeling biases in the wind turbine mechanism modeling process, model (3) was corrected by collecting wind turbine operation data and using neural network modeling. Therefore, the mechanism analysis and neural network wind turbine model are established as follows:

[0072]

[0073] in A three-layer neural network structure model (10 input variables, 7 output variables) is used to correct the modeling deviation of the wind turbine mechanism model and improve the accuracy of wind turbine dynamics modeling.

[0074] In this embodiment, in step S1, when linearizing the wind turbine mechanism analysis model, the nonlinear aerodynamic model, neural network model, and wind turbine power generation model in the wind turbine mechanism analysis model are linearized using the first-order Taylor expansion method. Assuming the wind speed is... At that time, the corresponding steady-state fan state is , , , , , , as well as , , When linearizing the nonlinear aerodynamic model (2), the aerodynamic torque and air thrust The expression is:

[0075]

[0076] , , , , , , ,

[0077] In the above formula, , These are the aerodynamic torques at the current moment and at the steady-state operating point of the wind turbine, respectively. , These are the rotor speeds at the current moment and at the steady-state operating point of the wind turbine, respectively. These are the wind speeds at the current moment and at the steady-state operating point of the wind turbine, respectively. , These represent the forward and backward movement speeds of the tower at the current moment and at the wind turbine's steady-state operating point, respectively. , These are the pitch angles at the current moment and at the steady-state operating point of the wind turbine, respectively. , These represent the air thrust at the current moment and at the steady-state operating point of the wind turbine, respectively. This indicates the steady-state operating point. These are intermediate parameters.

[0078] In this embodiment, the linearized neural network model is obtained by linearizing the neural network model, and the expression of the linearized neural network model is as follows:

[0079]

[0080] ,

[0081] In the above formula, For neural network models, Indicates the steady-state operating point and The neural network function value at that location, This is the first neural network matrix. This is the second neural network matrix. These represent the current state variables of the wind turbine and the state variables of the wind turbine at its steady-state operating point, respectively. These represent the model input at the current moment and the model input at the wind turbine's steady-state operating point, respectively.

[0082] In this embodiment, the linearized wind turbine power generation model is obtained by linearizing the wind turbine power generation model as follows:

[0083]

[0084] In the above formula, , These represent the generator power at the current moment and at the wind turbine's steady-state operating point, respectively. , These are the generator torques at the current moment and at the wind turbine's steady-state operating point, respectively. , These represent the generator speeds at the current moment and at the wind turbine's steady-state operating point, respectively.

[0085] In this embodiment, step S1, constructing the linearized model of the wind turbine, specifically includes the following steps:

[0086] By substituting the linearized aerodynamic model and the linearized neural network model into the wind turbine mechanism analysis model, the wind turbine deviation state signal is defined and a linearized state-space model based on the wind turbine deviation state signal is constructed.

[0087] Specifically, let , , , , , , , , , , , , , , We can obtain:

[0088]

[0089] (7) After sorting, we can obtain

[0090]

[0091] in , , .make , , , for The 2nd and 3rd columns of the matrix, for From the first column of the matrix block, (7) we can obtain:

[0092]

[0093]

[0094]

[0095]

[0096] , ,

[0097] , ,

[0098] In the above formula, k and k+1 represent time k and time k+1, respectively. These are the fan deviation state signals at time k+1 and time k, respectively. This is the fan deviation control signal. for The 2nd and 3rd columns of the matrix, for The first column of the matrix block, These are intermediate parameters.

[0099] In this embodiment, when designing the objective function of predictive control, it is necessary to balance power generation efficiency with minimizing wind turbine load. It is assumed that the step size of the predictive control is... The objective function expression of the load reduction predictive control model is:

[0100]

[0101] In the above formula, k+i represents time k+i. This is a fan deviation status signal. This is the fan deviation control signal. , , , , , , , , , , , These are the weight coefficients of the objective function (all non-negative numbers). For power deviation, For the torsional angle deviation of the drive shaft, For the fan rotor speed deviation, For generator speed deviation, This refers to the deviation in the forward and backward movement distance of the tower. This refers to the deviation in the forward and backward movement speed of the tower. For the deviation of the blade flapping angle, For the deviation of the blade flapping angular velocity, This is a single-step increment for the pitch angle deviation control quantity. N represents the single-step increment of the generator torque deviation control quantity, where N represents the step size of the predictive control. In order to reduce fluctuations in generator power and thus ensure power stability, To reduce wear caused by vibration in the transmission system, To reduce rotor speed fluctuations, To reduce fluctuations in generator speed, and To reduce the tower's forward and backward displacement and oscillation, and Reduce the flapping angle and oscillation of the blades. and In order to reduce the oscillation of the pitch angle and thus achieve smooth pitch angle control, and To reduce generator torque fluctuations and achieve smooth generator torque control, the wind turbine's power output, status, and control parameters are smoothly controlled to minimize load on the turbine while maintaining power generation efficiency.

[0102] Construct the control constraints for the wind turbine. Assume... Minimum shaft torsion angle, The maximum shaft torsion angle, The maximum rotor speed, Minimum rotor speed, This is the maximum generator speed. Minimum generator speed, For the maximum tower displacement, For the minimum tower displacement, The maximum tower moving speed, For the minimum tower moving speed, For the maximum blade flapping angle, For the minimum blade flapping angle, The maximum blade flapping angular velocity, To find the minimum blade flapping angular velocity, the state constraints of the wind turbine are as follows:

[0103]

[0104] In the above formula, Minimum shaft torsion angle, The maximum shaft torsion angle, The maximum rotor speed, Minimum rotor speed, This is the maximum generator speed. Minimum generator speed, For the maximum tower displacement, For the minimum tower displacement, The maximum tower moving speed, For the minimum tower moving speed, For the maximum blade flapping angle, For the minimum blade flapping angle, The maximum blade flapping angular velocity, For the minimum blade flapping angular velocity, The shaft torsion angle at the steady-state operating point of the fan. For the torsional angle deviation of the drive shaft, The rotor speed at the steady-state operating point of the fan. For rotor speed deviation, This refers to the generator speed at the steady-state operating point of the wind turbine. For generator speed deviation, The distance the tower moves back and forth at the steady-state operating point of the wind turbine. This refers to the deviation in the forward and backward movement distance of the tower. The forward and backward movement speed of the tower at the steady-state operating point of the wind turbine. This refers to the deviation in the forward and backward movement speed of the tower. This refers to the blade flapping angle at the steady-state operating point of the wind turbine. For the deviation of the blade flapping angle, The blade flapping angular velocity at the steady-state operating point of the wind turbine. This refers to the deviation in the angular velocity of the blade flapping.

[0105] Assumption For minimum generator torque, For maximum generator torque, For the minimum pitch angle, For the maximum pitch angle, the control constraints for the wind turbine are:

[0106]

[0107] In the above formula, For minimum generator torque, For maximum generator torque, For the minimum pitch angle, The maximum pitch angle, The pitch angle is the angle of the blades at the steady-state operating point of the wind turbine. This refers to the pitch angle deviation. This represents the generator torque at the steady-state operating point of the wind turbine. This represents the generator torque deviation.

[0108] Generator power constraints. Assume the maximum allowable actual generator power is... The power generation efficiency factor is Then the generator power constraint condition is:

[0109] .

[0110] In this embodiment, before solving the load reduction predictive control model in step S3, the method further includes establishing a wind turbine load reduction predictive control problem by combining the wind turbine state constraints, wind turbine control constraints, and generator power constraints. The expression of the established wind turbine load reduction model predictive control (MPC) problem is as follows:

[0111]

[0112] It can be seen that the control problem is a convex quadratic programming problem. The problem can be solved by using a high-performance convex quadratic programming solver, and the values ​​of various control variables of the wind turbine (such as pitch angle and generator torque) can be obtained.

[0113] It is understood that this embodiment addresses the problem of difficulty in accurately modeling the complex operating characteristics of wind turbines. It adopts a model based on wind turbine mechanism analysis and neural network modeling. Since the model is a complex nonlinear model, it is linearized. An objective function that takes into account both power generation efficiency and minimizing wind turbine load is designed to construct a wind turbine load reduction MPC control problem. A high-precision convex quadratic programming solver is used to solve the MPC problem, which makes it easy to realize wind turbine load reduction model predictive control.

[0114] A wind turbine load reduction predictive control system based on mechanistic and neural network models includes an interconnected microprocessor and a memory, the microprocessor being programmed or configured to execute the steps of a wind turbine load reduction predictive control method based on mechanistic and neural network models.

[0115] A computer-readable storage medium storing a computer program / instructions programmed or configured to execute, via a processor, steps of a wind turbine load reduction predictive control method based on mechanistic and neural network models.

[0116] The system and medium of the present invention, corresponding to the methods described above, also have the advantages described above.

[0117] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. Computer-readable media include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0118] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A fan load reduction prediction control method based on a mechanism and a neural network model, characterized by, The method comprises the steps of: S1, establishing a fan mechanism analysis and a neural network model, linearizing the fan mechanism analysis and the neural network model at a steady operating point of the fan to construct a linearized model of the fan; S2, constructing a load reduction predictive control model of the fan based on the linearized model, wherein the load reduction predictive control model takes minimizing the load of the fan while taking into account the power generation efficiency of the fan as an optimization objective; S3, solving the load reduction predictive control model to obtain optimal values of various control parameters of the fan; In step S1, when linearizing the fan mechanism analysis and the neural network model, the nonlinear aerodynamic model, the neural network model, and the fan power generation model in the fan mechanism analysis and the neural network model are linearized by a first-order Taylor expansion method; wherein the expression of the established fan mechanism analysis and neural network model is as follows: In the above formula, is each state quantity of the wind turbine, is the transmission chain shaft torsion angle, is the rotor speed, is the generator speed, is the tower cylinder forward and backward movement distance, is the tower cylinder forward and backward movement speed, is the blade flapping angle, is the blade flapping angular velocity, is the aerodynamic torque, is the generator torque, is the air thrust, is the wind speed, is the pitch angle, , are respectively the system matrix and the input matrix of the discretized wind turbine dynamics mechanism model, is a neural network model; The expression of the linearized aerodynamic model obtained by linearizing the nonlinear aerodynamic model is as follows: ， ， ， ， ， ， ， in the above formulae, , are respectively the aerodynamic torque at the current time instant and at the steady state operating point of the wind turbine, , are respectively the rotor speed at the current time instant and at the steady state operating point of the wind turbine, are respectively the wind speed at the current time instant and at the steady state operating point of the wind turbine, , are respectively the tower fore-aft movement speed at the current time instant and at the steady state operating point of the wind turbine, , are respectively the pitch angle at the current time instant and at the steady state operating point of the wind turbine, , are respectively the air thrust at the current time instant and at the steady state operating point of the wind turbine, denotes at the steady state operating point, is an intermediate parameter; The expression of the linearized neural network model obtained by linearizing the neural network model is as follows: ， In the above formula, is a neural network model, denotes a neural network function value at a steady state operating point and denotes a neural network function value at a steady state operating point is a first neural network matrix, is a second neural network matrix, respectively denote a fan state quantity at a current time and a fan state quantity at a steady state operating point of the fan, respectively denote a model input quantity at a current time and a model input quantity at a steady state operating point of the fan. The expression of the linearized fan power generation model obtained by linearizing the fan power generation model is as follows: In the above formula, , These represent the generator power at the current moment and at the wind turbine's steady-state operating point, respectively. , These are the generator torques at the current moment and at the wind turbine's steady-state operating point, respectively. , These represent the generator speeds at the current moment and at the wind turbine's steady-state operating point, respectively.

2. The mechanism and neural network model based fan load reduction prediction control method of claim 1, wherein, In step S1, constructing the linearized model of the fan comprises the steps of: Substituting the linearized aerodynamic model and the linearized neural network model into the fan mechanism analysis model, defining a fan deviation state signal and constructing a linearized state space model based on the fan deviation state signal, the expression of the linearized state space model is as follows: ， ， ， ， ， ， ， ， ， ， ， In the above formula, k and k+1 represent time k and time k+1, respectively. These are the fan deviation state signals at time k+1 and time k, respectively. This is the fan deviation control signal. The torsion angle of the drive shaft. For the fan rotor speed, For generator speed, For the angle of the blade's waving, The angular velocity of the blade flapping. The pitch angle is the propeller angle. For generator torque, For the deviation wind speed, , These are the system matrix and input matrix of the discretized wind turbine dynamics mechanism model, respectively. This is the first neural network matrix. This is the second neural network matrix. for The 2nd and 3rd columns of the matrix, for The first column of the matrix block, These are intermediate parameters.

3. The mechanism and neural network model based fan load reduction predictive control method according to any one of claims 1-2, wherein, The objective function expression of the load reduction predictive control model is as follows: In the above formula, k+i represents the k+i moment, is a fan deviation state signal, is a fan deviation control signal, , , , , , , , , , , , is each weight coefficient of the target function, is a power deviation, is a transmission shaft torsion angle deviation, is a fan rotor speed deviation, is a generator rotational speed deviation, is a tower cylinder forward and backward movement distance deviation, is a tower cylinder forward and backward movement speed deviation, is a blade flapping angle deviation, is a blade flapping angular velocity deviation, is a pitch angle deviation control signal single-step increment, is a generator torque deviation control signal single-step increment, and N represents a prediction control step length.

4. The mechanism and neural network model based fan load reduction predictive control method of claim 3, wherein, Before solving the load reduction predictive control model in step S3, the wind turbine load reduction predictive control problem is established by simultaneously considering the wind turbine state quantity constraint condition, the wind turbine control quantity constraint condition, and the generator power constraint condition, and the expression is as follows: The expression of the wind turbine state quantity constraint condition is as follows: In the above formulae, is the minimum shaft torsion angle, is the maximum shaft torsion angle, is the maximum rotor speed, is the minimum rotor speed, is the maximum generator speed, is the minimum generator speed, is the maximum tower displacement, is the minimum tower displacement, is the maximum tower movement speed, is the minimum tower movement speed, is the maximum blade flap angle, is the minimum blade flap angle, is the maximum blade flap angular speed, is the minimum blade flap angular speed, is the shaft torsion angle at the steady state operating point of the wind turbine, is the deviation of the shaft torsion angle, is the rotor speed at the steady state operating point of the wind turbine, is the deviation of the rotor speed, is the generator speed at the steady state operating point of the wind turbine, is the deviation of the generator speed, is the tower fore-aft displacement at the steady state operating point of the wind turbine, is the deviation of the tower fore-aft displacement, is the tower fore-aft movement speed at the steady state operating point of the wind turbine, is the deviation of the tower fore-aft movement speed, is the blade flap angle at the steady state operating point of the wind turbine, is the deviation of the blade flap angle, is the blade flap angular speed at the steady state operating point of the wind turbine, is the deviation of the blade flap angular speed; The expression of the wind turbine control quantity constraint condition is as follows: In the above formulae, is the minimum generator torque, is the maximum generator torque, is the minimum pitch angle, is the maximum pitch angle, is the pitch angle at the steady state operating point of the wind turbine, is the pitch angle deviation, is the generator torque at the steady state operating point of the wind turbine, is the generator torque deviation; The expression of the generator power constraint condition is as follows: In the above formulae, is the actual maximum generator power, is the power generation efficiency factor, is the generator power.

5. A mechanism and neural network model based fan load reduction predictive control system comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to perform the steps of the fan load reduction predictive control method based on the mechanism and neural network model in any one of claims 1-4.

6. A computer-readable storage medium having stored therein computer programs / instructions, characterized in that, The computer program / instruction is programmed or configured to perform the steps of the fan load reduction predictive control method based on the mechanism and neural network model in any one of claims 1-4 by the processor.

Images