Intelligent control system of wind turbine generator

Abstract

The invention relates to the technical field of wind power generation monitoring and control, and discloses an intelligent control system for a wind turbine generator, which comprises a high-precision rotary laser scanning module, an intelligent decision module, a self-adaptive scanning control module, a rapid pitch-variable execution mechanism and a system communication interface, the high-precision rotary laser scanning module is used for acquiring blade deformation data in real time and comprises a mechanical rotary scanning device and a sealing and positive pressure system. According to the method, the LSTM time sequence prediction capacity and the Q-learning online optimization capacity are combined, turbulence intensity changes are effectively adapted, accurate and self-adaptive control over complex turbulence working conditions is finally achieved, the overall response time is shortened from instruction generation to paddle action, extreme gust is coped with in time, the tower sweeping risk and the fatigue load are remarkably reduced, and the working efficiency is improved. And through accurate advanced control and rapid compensation, the power generation loss caused by turbulent flow and improper variable pitch is reduced, and the economic benefit is outstanding.

Description

Technical Field

[0001] This invention relates to the field of wind power generation monitoring and control technology, and in particular to an intelligent control system for wind turbine generators. Background Technology

[0002] As wind turbines develop towards larger sizes and offshore applications, their rotor diameters are constantly increasing, and the problem of blade flexibility deformation is becoming increasingly prominent. Traditional fixed laser scanning or single sensor monitoring solutions are insufficient to achieve real-time, high-precision measurement of blade deformation, especially blade tip displacement.

[0003] A search revealed that Chinese invention patent CN115370537A discloses a wind power frequency modulation load optimization control method and system based on radar feedforward control. Although it proposes load optimization control based on radar feedforward, it does not solve the problem of insufficient measurement angle positioning accuracy. Chinese invention patent CN112031998A discloses a wind turbine independent pitch control optimization method and system based on lidar, achieving independent pitch control based on lidar. However, its method fails to effectively adapt to a wide range of turbulence intensity variations, and the system's reliability under long-term operation in harsh environments such as salt spray is insufficient. Furthermore, existing solutions generally suffer from the following technical defects: Existing technologies lack in-depth modeling of the coupling effect between the blade's own structural response and external turbulence, resulting in delayed pitch response under high turbulence conditions and posing a significant safety hazard of blade sweeping against the tower, leading to severe power generation efficiency loss and energy waste. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent control system for wind turbines, which solves the problems mentioned in the background art, such as low accuracy of blade deformation monitoring, long response delay of pitch control, poor environmental adaptability, and large loss of power generation efficiency in large wind turbines under complex turbulent wind conditions.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A wind turbine intelligent control system includes: a high-precision rotating laser scanning module, which is used to acquire blade deformation data in real time. The high-precision rotating laser scanning module includes a mechanical rotating scanning device with a worm gear transmission mechanism, and a sealing and positive pressure system for protection against harsh environments. The intelligent decision-making module is used to generate the optimal pitch control command based on the blade deformation data and the unit operating status data. The intelligent decision-making module includes a timing prediction unit and a reinforcement learning decision-making unit. An adaptive scanning control module is used to control the rotating laser scanning module to switch between a high-frequency grid scanning mode and a low-frequency spiral scanning mode based on real-time turbulence intensity information. A fast pitch actuator is used to receive and execute pitch control commands to adjust the pitch angle of the wind turbine. The system communication interface is used to connect the intelligent decision-making module and the rapid pitch actuator.

[0006] Furthermore, the high-precision rotating laser scanning module also includes: The multimodal sensing module is used to collect forward-looking information about the wind farm where the wind turbine is located, structural status information of key components of the turbine, and real-time operating status information.

[0007] Furthermore, the sealing and positive pressure system includes a sealing structure with a three-stage gradient gap and a nitrogen supply device for providing positive pressure to the sealed chamber.

[0008] Furthermore, the time-series prediction unit employs a long short-term memory neural network to predict the blade deformation at future moments; The reinforcement learning decision unit employs the Q-learning algorithm, whose reward function R is related to the real-time power generation P, rated power P_max, blade load stress σ, and pitch angle change Δβ, and is in the form of: R= k1*(P / P_max)-k2*|σ|-k3*|Δβ|, Where k1, k2, and k3 are weighting coefficients.

[0009] Furthermore, the weighting coefficients k1=0.6, k2=0.3, and k3=0.1.

[0010] Furthermore, the multimodal sensing module feeds back the turbulence intensity of the wind field to the adaptive scanning control module, and the intelligent decision-making module sets the initial turbulence intensity threshold.

[0011] Furthermore, the adaptive scanning control module is configured to: enable a low-frequency spiral scanning mode when the turbulence intensity is less than a threshold; and enable a high-frequency grid scanning mode when the turbulence intensity is greater than a threshold.

[0012] Furthermore, the high-precision rotating laser scanning module performs non-contact scanning of the wind turbine blades to acquire point cloud data of the blade surface and calculate the blade deformation in real time.

[0013] The deformation monitoring data, the current power generation of the unit, and the load status are input into the intelligent decision-making module. First, the LSTM network is used to predict the deformation trend in the short term. Then, combined with the current state, the optimal pitch control strategy that maximizes long-term benefits (i.e., the reward function) is calculated through the Q-learning algorithm and converted into a specific pitch angle command.

[0014] The adaptive scanning control module dynamically selects either helical scanning mode or grid scanning mode based on the real-time calculated turbulence intensity, optimizing data acquisition efficiency and quality. Pitch control commands are sent to the rapid pitch actuator through the system communication interface, driving the pitch motor to adjust the blades to the target pitch angle in a very short time, thereby suppressing blade vibration, reducing load, and improving wind capture efficiency.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention uses a high-precision worm gear driven rotary lidar to ensure angular positioning accuracy. Combined with high-frequency scanning, it achieves real-time and accurate measurement of blade deformation. Through a three-level gradient sealing and nitrogen positive pressure protection design, it greatly improves the protection level and service life of the system in harsh environments such as sea salt spray and high humidity.

[0016] 2. This invention combines the timing prediction capability of LSTM with the online optimization capability of Q-learning, enabling the system to not only predict the dynamic interaction between wind and blades, but also autonomously learn the optimal control strategy under complex turbulence, effectively adapting to changes in turbulence intensity. Ultimately, it achieves precise and adaptive control under complex turbulent conditions, shortening the overall response time from command generation to blade movement, enabling timely response to extreme gusts, significantly reducing tower sweep risk and fatigue load. Through precise advance control and rapid compensation, it reduces power generation losses caused by turbulence and inappropriate pitch control, resulting in outstanding economic benefits. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the working process of an intelligent control system for wind turbines proposed in this invention; Figure 2 This is a system structure block diagram of an intelligent control system for wind turbines proposed in this invention; Figure 3 This is a block diagram of the intelligent decision-making module algorithm of an intelligent control system for wind turbines proposed in this invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] Please see Figures 1 to 3 This embodiment of the invention provides an intelligent control system for wind turbines, which includes a high-precision rotating laser scanning module. The high-precision rotating laser scanning module is used to acquire blade deformation data in real time. It includes a mechanical rotating scanning device with a worm gear transmission mechanism, a sealing and positive pressure system for harsh environment protection, and a multimodal sensing module. The multimodal sensing module is used to collect forward-looking information of the wind field at the wind turbine, structural status information of key components of the unit, and real-time operating status information.

[0020] The intelligent decision-making module is used to generate optimal pitch control commands based on blade deformation data and unit operating status data. The intelligent decision-making module includes a timing prediction unit and a reinforcement learning decision-making unit.

[0021] The adaptive scanning control module is used to control the rotating laser scanning module to switch between high-frequency grid scanning mode and low-frequency spiral scanning mode based on real-time turbulence intensity information.

[0022] The fast pitch actuator is used to receive and execute pitch control commands to adjust the pitch angle of the wind turbine.

[0023] The system communication interface is used to connect the intelligent decision-making module and the rapid pitch actuator.

[0024] Application Process: The high-precision rotating laser scanning module monitors the blade deformation of the wind turbine in real time. Its core is a mechanical rotating scanning device, which adopts a worm gear transmission mechanism to achieve high-precision angle positioning of the laser emitter. At the same time, the module is equipped with a three-level gradient sealing structure with a gap gradient of 0.5mm→0.2mm→0.1mm, and is supplemented by nitrogen positive pressure protection of 5±0.5kPa to ensure the long-term operation of the core components in harsh environments such as salt spray.

[0025] By employing a high-precision worm gear driven rotary lidar, angular positioning accuracy is ensured. Combined with high-frequency scanning, real-time and accurate measurement of blade deformation is achieved. Through a three-stage gradient sealing and nitrogen positive pressure protection design, the system's protection level and service life are greatly improved in harsh environments such as marine salt spray and high humidity.

[0026] The intelligent decision-making module generates the optimal pitch control command based on the monitoring data. This module integrates a long short-term memory network and a Q-learning reinforcement learning algorithm. The LSTM neural network is responsible for predicting the blade deformation time series data to obtain the deformation trend at future moments. The Q-learning algorithm constructs a state space based on current power generation, blade load stress, and pitch angle changes, aiming to maximize the comprehensive reward function. It learns online and outputs the optimal pitch strategy. An example of the reward function is as follows: R = 0.6*(P / P_max) - 0.3*|σ| - 0.1*|Δβ|, where P is the real-time power, P_max is the rated power, σ is the load stress, and Δβ is the pitch angle change, thus balancing power generation efficiency, structural safety, and control stability.

[0027] The adaptive scanning control module is used to adjust the scanning strategy according to real-time wind conditions. By receiving the anemometer signal, it calculates the current turbulence intensity. If the turbulence intensity is ≤15%, it controls the laser to perform a spiral scanning mode with a frequency of 50Hz. If the turbulence intensity is >15%, it switches to a high-frequency grid scanning mode of 100Hz to obtain denser point cloud data and adapt to rapid changes under strong turbulence.

[0028] The rapid pitch actuator receives pitch control commands from the intelligent decision module and drives the servo system to adjust the blade pitch angle within a response time of ≤80ms, thereby achieving rapid compensation for dynamic wind conditions such as turbulence.

[0029] The sealing and positive pressure system includes a three-stage gradient gap sealing structure and a nitrogen supply device to provide positive pressure to the sealed chamber.

[0030] The time-series prediction unit uses a long short-term memory neural network to predict the blade deformation at future moments; The reinforcement learning decision unit uses the Q-learning algorithm, whose reward function R is related to the real-time power generation P, rated power P_max, blade load stress σ, and pitch angle change Δβ, and is in the form of: R= k1*(P / P_max)-k2*|σ|-k3*|Δβ|, Where k1, k2, and k3 are weighting coefficients.

[0031] The weighting coefficients are k1=0.6, k2=0.3, and k3=0.1.

[0032] During this process, through continuous learning and optimization, the Q-learning module finally outputs the optimal pitch control command Δβ_optimal obtained from the evaluation under the current state. This command is sent to the pitch actuator, and the unit operates at the new pitch angle, generating new monitoring data (deformation, power, etc.). This data is then fed back to the intelligent decision module to start the next control cycle, forming an intelligent closed loop of continuous self-learning and optimization.

[0033] The multimodal sensing module feeds back the turbulence intensity of the wind field to the adaptive scanning control module, and the intelligent decision-making module sets the initial turbulence intensity threshold.

[0034] First, a high-precision rotating laser scanning module is used to perform non-contact scanning on the wind turbine blades to obtain point cloud data on the blade surface and calculate the blade deformation in real time. Then, the deformation monitoring data, the current power generation of the unit, and the load status are input into the intelligent decision-making module. First, the LSTM network is used to predict the deformation trend in the short term. Then, combined with the current state, the optimal pitch control strategy that maximizes long-term benefits (i.e., reward function) is calculated through the Q-learning algorithm and converted into specific pitch angle commands.

[0035] Then, the adaptive scanning control module dynamically selects either the spiral scanning mode or the grid scanning mode based on the real-time calculated turbulence intensity to optimize data acquisition efficiency and quality. The pitch control command is sent to the rapid pitch actuator through the system communication interface, driving the pitch motor to adjust the blades to the target pitch angle in a very short time, thereby suppressing blade vibration, reducing load, and improving wind capture efficiency.

[0036] The system continuously executes the above steps in a loop, forming a closed-loop intelligent control of "monitoring-prediction-decision-execution". By combining the LSTM timing prediction capability with the Q-learning online optimization capability, the system can not only predict the dynamic interaction between wind and blades, but also autonomously learn the optimal control strategy under complex turbulence, effectively adapting to changes in turbulence intensity. Ultimately, it achieves precise and adaptive control of complex turbulent conditions, shortening the overall response time from command generation to blade movement, enabling timely response to extreme gusts, significantly reducing tower sweep risk and fatigue load. Through precise advance control and rapid compensation, it reduces power generation losses caused by turbulence and inappropriate pitch control, resulting in outstanding economic benefits.

[0037] The adaptive scanning control module is configured to enable a low-frequency spiral scanning mode when the turbulence intensity is less than the threshold, and to enable a high-frequency grid scanning mode when the turbulence intensity is greater than the threshold.

[0038] Usage: The high-precision rotating laser scanning module is installed on the top of the cabin. The worm of the worm gear transmission mechanism is made of 20CrMnTi alloy steel to ensure transmission accuracy and wear resistance. The sealed cabin is filled with dry nitrogen.

[0039] The intelligent decision-making module is deployed in an industrial computer inside the engine room control cabinet. The LSTM network has been pre-trained using historical wind conditions and blade response data and fine-tuned during online operation. The learning rate of the Q-learning algorithm is set to 0.01 and the discount factor is set to 0.9. It updates its Q-value table by continuously interacting with the environment.

[0040] When the anemometer detects severe fluctuations in instantaneous wind speed and the system calculates that the turbulence intensity TI = 18%, which is greater than 15%, the adaptive scanning control module immediately switches the scanning mode from spiral scanning to grid scanning and increases the frequency to 100Hz.

[0041] At this moment, based on the latest high-density scan data, the intelligent decision-making module predicts that the blade tip displacement will exceed the limit in 2 seconds. It immediately calculates a pitch control instruction of 2.5 degrees in advance through Q-learning, and sends the instruction to the pitch control cabinet within a 100ms cycle.

[0042] The pitch actuator uses a high-performance servo motor and reducer to complete a 2.5-degree pitch angle adjustment within 75ms after receiving the command, effectively avoiding excessive blade deformation and load peaks.

[0043] As can be seen from the above, the working principle of this application is as follows: First, a high-precision rotating laser scanning module is used to perform non-contact scanning on the wind turbine blades to obtain point cloud data on the blade surface and calculate the blade deformation in real time.

[0044] Next, the deformation monitoring data, the current power generation of the unit, and the load status are input into the intelligent decision-making module. First, the LSTM network is used to predict the deformation trend in the short term. Then, combined with the current state, the optimal pitch control strategy that maximizes long-term benefits (i.e., the reward function) is calculated through the Q-learning algorithm and converted into a specific pitch angle command.

[0045] By combining the timing prediction capability of LSTM with the online optimization capability of Q-learning, the system can not only predict the dynamic interaction between wind and blades, but also autonomously learn the optimal control strategy under complex turbulence, effectively adapt to changes in turbulence intensity, and ultimately achieve precise and adaptive control of complex turbulent conditions. From command generation to blade action, the overall response time is shortened, enabling timely response to extreme gusts.

[0046] Subsequently, the adaptive scanning control module dynamically selects either the spiral scanning mode or the grid scanning mode based on the real-time calculated turbulence intensity, optimizing data acquisition efficiency and quality. The pitch control command is sent to the rapid pitch actuator through the system communication interface, driving the pitch motor to adjust the blades to the target pitch angle in a very short time, thereby suppressing blade vibration, reducing load, and improving wind capture efficiency.

[0047] The system continuously executes the above steps in a loop, forming a closed-loop intelligent control of "monitoring-prediction-decision-execution".

[0048] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. An intelligent control system for wind turbine generators, characterized in that, include: A high-precision rotating laser scanning module is used to acquire blade deformation data in real time. It includes a mechanical rotating scanning device with a worm gear transmission mechanism and a sealing and positive pressure system for protection against harsh environments. The intelligent decision-making module is used to generate the optimal pitch control command based on the blade deformation data and the unit operating status data. The intelligent decision-making module includes a timing prediction unit and a reinforcement learning decision-making unit. An adaptive scanning control module is used to control the rotating laser scanning module to switch between a high-frequency grid scanning mode and a low-frequency spiral scanning mode based on real-time turbulence intensity information. A fast pitch actuator is used to receive and execute pitch control commands to adjust the pitch angle of the wind turbine. The system communication interface is used to connect the intelligent decision-making module and the rapid pitch actuator.

2. The intelligent control system for a wind turbine generator according to claim 1, characterized in that, The high-precision rotating laser scanning module also includes: The multimodal sensing module is used to collect forward-looking information about the wind farm where the wind turbine is located, structural status information of key components of the turbine, and real-time operating status information.

3. The intelligent control system for a wind turbine generator according to claim 1, characterized in that, The sealing and positive pressure system includes a sealing structure with a three-stage gradient gap and a nitrogen supply device that provides positive pressure to the sealed chamber.

4. The intelligent control system for a wind turbine generator according to claim 1, characterized in that, The time-series prediction unit employs a long short-term memory neural network to predict the blade deformation at future moments; The reinforcement learning decision unit employs the Q-learning algorithm, whose reward function R is related to the real-time power generation P, rated power P_max, blade load stress σ, and pitch angle change Δβ, and is in the form of: R= k1*(P / P_max)-k2*|σ|-k3*|Δβ|, Where k1, k2, and k3 are weighting coefficients.

5. The intelligent control system for a wind turbine generator according to claim 4, characterized in that, The weighting coefficients are k1=0.6, k2=0.3, and k3=0.

1.

6. The intelligent control system for a wind turbine generator according to claim 2, characterized in that, The multimodal sensing module feeds back the turbulence intensity of the wind field to the adaptive scanning control module, and the intelligent decision-making module sets the initial turbulence intensity threshold.

7. The intelligent control system for a wind turbine generator according to claim 6, characterized in that, The adaptive scanning control module is configured to: enable a low-frequency spiral scanning mode when the turbulence intensity is less than a threshold; and enable a high-frequency grid scanning mode when the turbulence intensity is greater than a threshold.

Images