Data-driven floating type wind turbine generator model-free control method and system

Abstract

The invention discloses a data-driven floating type wind turbine generator model-free control method and system, and the method comprises the steps: obtaining the operation state and external environment parameters of a wind turbine generator, and constructing sample data; inputting the sample data into a pre-trained controller, and outputting a control instruction to control the wind turbine generator; the specific process of obtaining the controller comprises the steps that the simulation operation state / historical operation state and external environment parameters of the wind turbine generator are obtained, and an initial controller is constructed based on the simulation operation state / historical operation state and the external environment parameters; determining a control target, a wind turbine generator operating environment, an intelligent agent, a state space, an action space, a constraint condition and a reward function in the initial controller; and performing continuous interactive learning training on the intelligent agent and the operation environment of the wind turbine generator through a deep reinforcement learning algorithm to obtain a trained controller. The method has the advantages of reducing the modeling difficulty, dynamically correcting the control strategy, improving the power generation efficiency and the like.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of floating wind turbines, and in particular to a data-driven model-free control method and system for floating wind turbines. Background Art

[0002] Compared with fixed-base wind turbines, floating wind turbines face more challenges. The additional six degrees of freedom of the floating platform will cause more dynamic characteristics in the floating wind turbine system. These disturbances are caused by the combined action of waves, currents, and turbulent wind and waves. Therefore, although the floating wind turbine is moored to the seabed, the whole system is prone to stability problems caused by these environmental loads, which directly affects power generation and increases operation and maintenance costs. Considering that the control system plays a vital role in the power generation performance, stability and reliability of the whole system during operation, in order to meet these challenges, it is urgent to make substantial improvements in control methods.

[0003] At present, the main control methods for floating wind turbines are classical model-based control methods and data-driven model control methods.

[0004] Classical control methods are usually designed based on linear models or simplified models. It is difficult to obtain the optimal control law based on simplified models and linearization processing because they do not model dynamic models and are intertwined with robustness issues. It is difficult to accurately reflect the nonlinear characteristics and time-varying characteristics of floating wind turbines. Therefore, in practical applications, the control accuracy of the control method of the classical model is often subject to certain limitations, and it is difficult to achieve accurate control of the unit state. In addition, due to the computational complexity and control strategy limitations of the classical control method, it often responds slowly when dealing with emergencies or rapidly changing environmental conditions, which may cause the unit to be unable to respond effectively in time when facing an emergency, thereby increasing safety risks.

[0005] Although data-driven model-based control methods are expected to solve the problem of difficult modeling of nonlinear systems, they show obvious limitations. The first is that this method is highly dependent on the quality and quantity of available data. If the training data is insufficient, inaccurate, or cannot represent the different operating conditions of the system, the performance may be affected. At the same time, this method is prone to overfitting, that is, the model will capture noise or anomalies in the training set, resulting in poor generalization ability to new scenarios, which is particularly challenging for floating wind turbines that are subject to uncertainty and change. In addition, these control methods may have difficulty coping when faced with sudden changes or new operating conditions that are not included in the training data, which highlights the challenge of achieving robustness in unforeseen situations.

[0006] Traditional model-based control methods require the establishment of an accurate mathematical model to describe the dynamic characteristics of the system. However, it is difficult to establish an accurate mathematical model for floating wind turbines due to their complex operating environment and dynamic characteristics.

[0007] In order to improve control accuracy, data-driven models often need to include a large number of variables and parameters, which makes the model complex. Complex models not only increase the amount of calculation, but may also introduce additional noise and uncertainty, affecting the stability and reliability of the control system. Summary of the invention

[0008] In view of the technical problems existing in the prior art, the present invention provides a data-driven model-free control method and system for a floating wind turbine generator set, which reduces the difficulty of modeling and improves power generation efficiency.

[0009] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0010] A data-driven model-free control method for a floating wind turbine generator system comprises the following steps:

[0011] Obtain the operating status and external environmental parameters of the wind turbine and construct sample data;

[0012] Input the sample data into a pre-trained controller and output control instructions to control the wind turbine;

[0013] The specific process of obtaining a pre-trained controller is as follows:

[0014] Acquire the simulated operating state / historical operating state and external environmental parameters of the wind turbine generator set, and construct an initial controller based on the simulated operating state / historical operating state and external environmental parameters;

[0015] Determine the control objectives, wind turbine operating environment, agent, state space, action space, constraints and reward function in the initial controller;

[0016] Through the deep reinforcement learning algorithm, the intelligent agent and the wind turbine operating environment are continuously interactively learned and trained to obtain a trained controller.

[0017] Preferably, the operating status and external environmental parameters include one or more of wind speed, wind direction, wave height, current speed, platform pitch angle and unit vibration.

[0018] Preferably, the control objective is one or more of maximizing power generation efficiency, minimizing platform pitch motion, maintaining system stability and reducing unit load.

[0019] Preferably, the state space S is defined as:

[0020] Observe the value of each state s from the state space S and learn the optimal policy that allows the agent in a given state to take the best action a from the action space A;

[0021] Define the state space S as a multidimensional vector:

[0022]

[0023] Among them, ω g is the generator speed; ω g -ω g,rated is the generator speed ω g With rated speedω g,rated The error between y is the platform pitch angle; is the platform pitch angular velocity; β is the blade pitch angle; P g is the generator power; P g,rated is the rated power.

[0024] Preferably, the action space A is defined as:

[0025] The action space A is a scalar that determines the best action a taken by the controller, specifically:

[0026]

[0027] Where β1, β2, β3 are blade pitch angles, is the blade pitch rate, ω g is the generator speed.

[0028] Preferably, the constraint design is:

[0029]

[0030] In the formula, ω g is the generator speed; θ is the pitch angle of the floating platform; is the pitch angular velocity of the floating platform; is the blade pitch rate, P is the generating power of the unit; subscript min is the minimum value, and subscript max is the maximum value.

[0031] Preferably, the reward function r is designed as:

[0032] The reward function is designed to guide the learning process, and the positive / negative rewards are set according to the control objectives; the blade pitch control and generator speed control are aimed at reducing the platform pitch motion of the floating wind turbine while ensuring the tracking of the rated power; therefore, the reward function is defined as:

[0033]

[0034] In the formula, Qi The factors correspond to the rewards and penalties assigned to each reward term. If it is within the constraints, it will be rewarded, and if it is outside the constraints, it will be punished;

[0035] When the generator speed ω g is within the constraints and is consistent with the rated speed ω g,rated The closer it is, the greater the weight of the positive reward will be. If it is outside the constraints, a penalty will be imposed.

[0036] When the platform pitch angle θ y If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed.

[0037] When the platform pitch angular velocity If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed.

[0038] When the blade pitch rate If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed.

[0039] When the generator power P g The closer the power is to the rated power P, the g,rated , the greater the positive reward weight assigned, if it is outside the constraints, a penalty will be imposed.

[0040] Preferably, the operating status and external environmental parameters of the wind turbine generator set are obtained through simulation data of the simulation software bladed / OpenFAST or sensors installed on the floating wind turbine generator set.

[0041] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are executed.

[0042] The present invention further discloses a data-driven model-free control system for a floating wind turbine, comprising a memory and a processor connected to each other, wherein a computer program is stored on the memory, and when the computer program is run by the processor, the steps of the method described above are executed.

[0043] Compared with the prior art, the advantages of the present invention are:

[0044] 1. Reduce the difficulty of modeling: Traditional model-based control methods require the establishment of accurate mathematical models to describe the dynamic characteristics of the system. However, due to the complex operating environment and dynamic characteristics of floating wind turbines, it is very difficult to establish accurate mathematical models. The data-driven floating unit model-free control method does not require the establishment of accurate mathematical models, thereby reducing the difficulty and complexity of modeling.

[0045] 2. Dynamic correction of control strategy: Data-driven model-free control methods control and adjust based on data collected in real time or simulated by simulation software, and can respond more quickly to changes in the external environment, such as wind speed, wind direction, waves, etc.; this helps ensure that floating wind turbines maintain a stable operating state in complex and changeable marine environments.

[0046] 3. Improve power generation efficiency: By adjusting the control strategy in real time, the model-free control method helps to optimize the operating parameters of wind turbines and improve power generation efficiency and energy conversion efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a diagram showing an embodiment of the model-free control method for a floating wind turbine generator system of the present invention in a specific application. DETAILED DESCRIPTION

[0048] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0049] like Figure 1 As shown, the data-driven model-free control method for a floating wind turbine according to an embodiment of the present invention comprises the following steps:

[0050] Obtain the operating status and external environmental parameters of the wind turbine and construct sample data;

[0051] Input the sample data into a pre-trained controller and output control instructions to control the wind turbine;

[0052] The specific process of obtaining a pre-trained controller is as follows:

[0053] Acquire the simulated operating state / historical operating state and external environmental parameters of the wind turbine generator set, and construct an initial controller based on the simulated operating state / historical operating state and external environmental parameters;

[0054] Determine the control objectives, wind turbine operating environment, agent, state space, action space, constraints and reward function in the initial controller;

[0055] Through the deep reinforcement learning algorithm, the intelligent agent and the wind turbine operating environment are continuously interactively learned and trained to obtain a trained controller.

[0056] The data-driven model-free control method of the present invention does not require the establishment of an accurate mathematical model, which can effectively reduce the difficulty and complexity of modeling; the model-free control method can compensate for the impact of faults by adjusting the control strategy, thereby improving the fault tolerance and reliability of the system. It has strong anti-interference ability and fault tolerance; the controller is debugged by interacting with the environment through experimental means or simulation means, which can better adapt to the changes of floating wind turbines under different working conditions and environmental conditions, and improve the adaptability and flexibility of the control system.

[0057] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0058] The data-driven model-free control method for a floating wind turbine generator system of the present invention specifically comprises the following steps:

[0059] S1. Problem definition and goal setting

[0060] The data-driven model-free controller for floating wind turbines is trained through interaction with the environment to reduce the platform pitch motion, reduce the load of the unit, and ensure the tracking of the rated power of the wind turbine. Its goal is to find a suitable control strategy to give the best action a to maximize the expected cumulative reward r. Specifically:

[0061] (1) Clarify control objectives: Determine the specific control objectives of the floating wind turbine. The control objectives are to maximize power generation efficiency, minimize platform pitch motion, maintain system stability, and reduce unit load.

[0062] (2) Environmental analysis: Analyze the working environment of floating wind turbines, including natural factors such as wind speed, waves, ocean currents, and their dynamic changes.

[0063] S2. Environmental Modeling

[0064] (1) Definition of state space S

[0065] Observe the value of each state s from the state space S and learn the optimal policy that allows the agent in a given state to take the best action a from the action space A. Define the input state of the controller, such as generated power, wind speed, generator speed, blade pitch angle, platform attitude, etc.

[0066] Define the state space S as a multi-dimensional vector, such as a 5-dimensional vector:

[0067]

[0068] Among them, ω g is the generator speed; ω g -ω g,rated is the generator speed ωg With rated speedω g,rated The error between y is the platform pitch angle; is the platform pitch angular velocity; β is the blade pitch angle; P g is the generator power; P g,rated is the rated power.

[0069] (2) Definition of action space A

[0070] The action space A is a scalar that determines the best action a taken by the controller, such as adjusting the blade pitch angle β, blade pitch rate Generator speedω g wait:

[0071]

[0072] (3) Constraint design:

[0073] To ensure the feasible operating range of floating wind turbines, the following additional constraints need to be considered:

[0074]

[0075] Wherein, the first constraint limits the operating range of the generator speed; the second constraint limits the pitch angle of the floating platform; the third constraint limits the pitch angular velocity of the floating platform; the fourth constraint limits the blade pitch rate, and the fifth constraint limits the power generation power of the unit.

[0076] (4) Reward function r design:

[0077] The reward function is designed to guide the learning process, and the positive / negative rewards are set according to the control objectives (such as power generation, load fluctuation, platform pitch angle, etc.). The blade pitch control and generator speed control are designed to reduce the platform pitch motion of the floating wind turbine while ensuring the tracking of the rated power. Therefore, the reward function is defined as:

[0078]

[0079] In the formula, Q i The factors correspond to the rewards and penalties assigned to each reward term, with rewards for being within the constraints and penalties for being outside the constraints.

[0080] a) When the generator speed ω g is within the constraints and is consistent with the rated speed ω g,rated The closer it is, the greater the weight of the positive reward will be. If it is outside the constraints, a penalty will be imposed.

[0081] b) When the platform pitch angle θ yIf it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed.

[0082] c) When the platform pitch angular velocity If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed.

[0083] d) When the blade pitch rate If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed.

[0084] e) When the generator power P g The closer the power is to the rated power P, the g,rated , the greater the positive reward weight assigned, if it is outside the constraints, a penalty will be imposed.

[0085] S3. Data collection and preprocessing

[0086] (1) Data acquisition: The operation data of the unit is collected through the simulation data of the simulation software bladed / OpenFAST or the sensors installed on the floating wind turbine to obtain the multidimensional vector required by the state space S.

[0087] (2) Data preprocessing: The collected data is cleaned, denoised, and normalized to meet the input requirements of the deep learning model.

[0088] S4. Deep Neural Network Design

[0089] (1) Network structure design

[0090] According to the complexity of the problem, a suitable deep neural network structure is designed, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or a combination thereof, which includes a critic neural network and an action neural network; wherein the evaluation is performed through the critic neural network, and the best execution action a is obtained through the action neural network.

[0091] (2) Feature extraction: Design network layers to extract effective features of the input state for subsequent control decisions.

[0092] (3) Output layer design: The output layer should be able to generate control actions, such as generator speed or blade pitch angle.

[0093] S5. Model training

[0094] (1) Training set construction: Use the preprocessed data to construct a training set, including the state-action-reward sequence.

[0095] (2) Training process: Use deep reinforcement learning algorithms (such as deep Q network DQN, deep deterministic policy gradient DDPG, trust region policy optimization TRPO, etc.) to train deep neural networks and optimize network parameters to maximize the cumulative reward r. The training of the agent is carried out in episodes. Each training set starts with the system in the initial state.

[0096] (3) Hyperparameter adjustment: Adjust hyperparameters such as learning rate, batch size, and number of iterations through methods such as cross-validation to optimize training results.

[0097] S6. Model application

[0098] The real-time operating status of the wind turbine generator set and external environmental parameters are input into the trained controller, and real-time decision information is output to adjust the operating status of the wind turbine generator set.

[0099] The data-driven model-free control method for floating wind turbines of the present invention has the following advantages:

[0100] 1. Reduce the difficulty of modeling: Traditional model-based control methods require the establishment of accurate mathematical models to describe the dynamic characteristics of the system. However, due to the complex operating environment and dynamic characteristics of floating wind turbines, it is very difficult to establish accurate mathematical models. The data-driven floating unit model-free control method does not require the establishment of accurate mathematical models, thereby reducing the difficulty and complexity of modeling.

[0101] 2. Dynamic correction of control strategy: Data-driven model-free control methods control and adjust based on data collected in real time or simulated by simulation software, and can respond more quickly to changes in the external environment, such as wind speed, wind direction, waves, etc.; this helps ensure that floating wind turbines maintain a stable operating state in complex and changeable marine environments.

[0102] 3. Improve power generation efficiency: By adjusting the control strategy in real time, the model-free control method helps to optimize the operating parameters of wind turbines and improve power generation efficiency and energy conversion efficiency.

[0103] The embodiment of the present invention further provides a data-driven model-free control system for a floating wind turbine generator system, which specifically includes:

[0104] Sensor system: To monitor the operating status of wind turbines and external environmental parameters in real time, such as wind speed, wind direction, wave height, current speed, platform pitch angle, unit vibration, etc. Sensors including anemometers, wave monitors, current meters, accelerometers, displacement sensors, etc. are deployed to provide accurate and real-time data input for the control system.

[0105] Data processing and analysis system: pre-process, filter, extract features, etc. of the data collected by the sensor to extract information useful for control decisions. The system includes data acquisition modules, data processing algorithm libraries, data storage and management systems, etc. Based on the processed data, the control instructions are calculated through the preset control algorithm to adjust the operating status of the wind turbine, optimize the energy conversion process of the wind turbine, and improve the power generation efficiency and economy. The algorithm type can include traditional PID control, fuzzy control, neural network control, and advanced algorithms such as deep reinforcement learning.

[0106] Actuator: According to the controller's instructions, adjust various parameters of the wind turbine, such as blade pitch angle, generator speed, etc. The actuator needs to have high reliability and fast response capabilities to ensure the accurate execution of control instructions.

[0107] Monitoring system: The monitoring system is a key link to ensure the safe operation of wind turbines. It can detect and solve problems in a timely manner to avoid major accidents. It monitors the execution effect of the controller and the operating status of the wind turbine in real time, and detects and handles abnormal situations in a timely manner. The specific components include monitoring software, alarm system, fault diagnosis system, etc.

[0108] Communication system: To ensure the timely transmission of control instructions and remote monitoring of unit status, improve the overall operational efficiency of the wind farm, and realize data transmission and instruction reception between wind turbines and remote monitoring centers. The specific components include wireless communication equipment, wired communication lines, data transmission protocols, etc.

[0109] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are executed.

[0110] The present invention further discloses a data-driven model-free control system for a floating wind turbine, comprising a memory and a processor connected to each other, wherein a computer program is stored on the memory, and when the computer program is run by the processor, the steps of the method described above are executed.

[0111] The medium and system of the present invention correspond to the above method and also have the advantages described in the above method.

[0112] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiment when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. The memory is used to store computer programs and / or modules, and the processor implements various functions by running or executing computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0113] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A data-driven model-free control method for a floating wind turbine, characterized in that: Includes steps: Obtain the operating status and external environmental parameters of the wind turbine and construct sample data; Input the sample data into a pre-trained controller and output control instructions to control the wind turbine; The specific process of obtaining a pre-trained controller is as follows: Acquire the simulated operating state / historical operating state and external environmental parameters of the wind turbine generator set, and construct an initial controller based on the simulated operating state / historical operating state and external environmental parameters; Determine the control objectives, wind turbine operating environment, agent, state space, action space, constraints and reward function in the initial controller; Through the deep reinforcement learning algorithm, the intelligent agent and the wind turbine operating environment are continuously interactively learned and trained to obtain a trained controller.

2. The data-driven model-free control method for floating wind turbines according to claim 1, characterized in that: The operating status and external environmental parameters include one or more of wind speed, wind direction, wave height, current speed, platform pitch angle and unit vibration.

3. The data-driven model-free control method for floating wind turbines according to claim 1, characterized in that: The control objective is one or more of maximizing power generation efficiency, minimizing platform pitch motion, maintaining system stability, and reducing unit load.

4. The data-driven model-free control method for floating wind turbines according to claim 1, 2 or 3, characterized in that: The state space S is defined as: Observe the value of each state s from the state space S and learn the optimal policy that allows the agent in a given state to take the best action a from the action space A; Define the state space S as a multidimensional vector: Among them, ω g is the generator speed; ω g -ω g,rated is the generator speed ω g With rated speedω g,rated The error between y is the platform pitch angle; is the platform pitch angular velocity; β is the blade pitch angle; P g is the generator power; P g,rated is the rated power.

5. The data-driven model-free control method for floating wind turbines according to claim 1, 2 or 3, characterized in that: Action space A is defined as: The action space A is a scalar that determines the best action a taken by the controller, specifically: Where β1, β2, β3 are blade pitch angles, is the blade pitch rate, ω g is the generator speed.

6. The data-driven model-free control method for floating wind turbines according to claim 1, 2 or 3, characterized in that: Design with constraints: In the formula, ω g is the generator speed; θ is the pitch angle of the floating platform; is the pitch angular velocity of the floating platform; is the blade pitch rate, P is the generating power of the unit; subscript min is the minimum value, and subscript max is the maximum value.

7. The data-driven model-free control method for floating wind turbines according to claim 1, 2 or 3, characterized in that: Reward function r design: The reward function is designed to guide the learning process, and the positive / negative rewards are set according to the control objectives; the blade pitch control and generator speed control are aimed at reducing the platform pitch motion of the floating wind turbine while ensuring the tracking of the rated power; therefore, the reward function is defined as: In the formula, Q i The factors correspond to the rewards and penalties assigned to each reward term. If it is within the constraints, it will be rewarded, and if it is outside the constraints, it will be punished; When the generator speed ω g is within the constraints and is consistent with the rated speed ω g,rated The closer it is, the greater the weight of the positive reward will be. If it is outside the constraints, a penalty will be imposed. When the platform pitch angle θ y If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed. When the platform pitch angular velocity If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed. When the blade pitch rate If it is within the constraints and the closer it is to 0, the greater the positive reward weight is assigned. If it is outside the constraints, a penalty is imposed. When the generator power P g The closer the power is to the rated power P, the g,rated , the greater the positive reward weight assigned, if it is outside the constraints, a penalty will be imposed.

8. The data-driven model-free control method for floating wind turbines according to claim 1, 2 or 3, characterized in that: The operating status and external environmental parameters of the wind turbine are obtained through simulation data of the simulation software bladed / OpenFAST or sensors installed on the floating wind turbine.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 8.

10. A data-driven model-free control system for a floating wind turbine, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, wherein: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 8.

Images