Intelligent control system for suppressing vibration of wind generating set

Abstract

The invention discloses an intelligent control system for restraining vibration of a wind generating set. The intelligent control system comprises an intelligent vibration restraining device, a vibration state monitoring system, a flow field state monitoring system and a main control system. The master control system comprises a data collector and a processor, data of the vibration state monitoring system and data of the flow field state monitoring system are subjected to timestamp alignment through the data collector and then serve as input, the processor is compared with design parameters of the wind generating set for learning judgment, and the intelligent vibration suppression device is controlled to act according to a judgment result. The vibration state suppression of the wind generating set is realized; according to the method, the vibration data of the wind generating set are monitored, the deep reinforcement learning and the model prediction control are combined for active control and vibration suppression, the problems existing in the prior art can be effectively solved, and convenience is provided for more users.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind turbine vibration suppression, and particularly to an intelligent control system for suppressing wind turbine vibration. BACKGROUND

[0002] The wind turbine adopts a tower as a support structure, and a wind wheel and a transmission system are installed on the top of the tower. Wind energy drives the wind wheel to rotate, and the wind wheel drives the transmission system to generate electricity. Since the wind turbine is composed of multiple mechanical systems, vibration of the unit will inevitably occur when it is subjected to environmental excitation. The vibration will affect the service life of the components of the wind turbine. When the vibration is too large, the wind turbine will be shut down, affecting the power generation capacity. When the blade and the tower are subjected to vortex-induced vibration, there is a risk of tower collapse.

[0003] At present, the main solutions to wind turbine vibration are 1) increasing the structural strength, but adjusting the structural strength of the blade and the tower will increase the material cost; 2) adjusting the overall structure damping of the unit through a damper to suppress vibration, but the arrangement of the damper is limited by the size of the space where the component is located, and needs to be designed according to different components, which will also increase the cost of the whole machine. SUMMARY

[0004] The present application aims to overcome the shortcomings of the prior art and provides an intelligent control system for suppressing wind turbine vibration. The vibration data of the wind turbine are monitored, and deep reinforcement learning and model predictive control are combined for active control to suppress vibration, which can effectively solve the problems existing in the prior art and provide convenience for more users.

[0005] The present application is achieved by the following technical scheme: an intelligent control system for suppressing wind turbine vibration, comprising:

[0006] An intelligent vibration suppression device is used to determine the high-vibration energy area of the structure surface, and the intelligent vibration suppression device is uniformly distributed on the structure surface of the wind turbine or installed along the vibration modal node of the wind turbine.

[0007] A vibration state monitoring system is used to record the vibration state data of the wind turbine and transmit the vibration state data of the wind turbine to the main control system.

[0008] A flow field state monitoring system is used to measure the flow velocity and vortex shedding frequency in real time according to the preset hot-wire anemometer and laser radar wind measuring device, measure the surface pressure pulsation in real time according to the preset surface pressure sensor, and transmit the flow velocity, vortex shedding frequency and surface pressure pulsation data to the main control system.

[0009] The main control system includes a data acquisition unit and a processor. The data acquisition unit timestamps and aligns the data from the vibration state monitoring system and the flow field state monitoring system as input. The processor compares the data with the design parameters of the wind turbine generator set to learn and make judgments. Based on the judgment results, it controls the intelligent vibration suppression device to suppress the vibration state of the wind turbine generator set. The processor includes a data analysis system and a vibration suppression control system.

[0010] The data analysis system is used for parameter feature extraction and data dimensionality reduction. It extracts features by obtaining the vibration dominant frequency and vortex shedding frequency through frequency domain analysis, identifies the dominant vibration mode through intrinsic orthogonal decomposition, tracks the structural vibration mode in real time, and locates the high-energy vibration area of ​​the wind turbine generator. It compresses the pressure sensor data into a low-dimensional feature vector through a convolutional autoencoder, maps the multi-channel vibration signal to the amplitude of the first three modes, organizes the flow field data and design parameters, and couples the calculated output parameters to the vibration suppression control system.

[0011] The vibration suppression control system combines deep reinforcement learning with model predictive control. Deep reinforcement learning provides the initial control strategy, while model predictive control performs local optimization. The application control strategy controls the intelligent vibration suppression device to achieve adaptive vibration suppression.

[0012] Furthermore, the intelligent vibration damping device includes a piezoelectric actuator made of a high-voltage coefficient material, a piezoelectric energy harvesting module, and a hybrid power supply. The response frequency range of the piezoelectric actuator covers the main vibration frequency of the wind turbine generator structure. The piezoelectric energy harvesting module converts vibration energy into electrical energy to charge the power supply of the intelligent vibration damping device. The hybrid power supply is a combination of a lithium battery and a supercapacitor.

[0013] Furthermore, the vibration suppression control system includes:

[0014] The deep reinforcement learning training first performs environmental modeling, defining the state space as including vibration amplitude, frequency, flow velocity, and pressure distribution characteristics, with a dimension set to be no less than 20; the action space defines the voltage amplitude and phase of the piezoelectric actuator as a continuous space, with a dimension equal to the number of piezoelectric actuators; the response function is R=-(α*|A vib |+β||V control ||), where A vib V is the vibration amplitude. control To control the voltage, α and β are weighting coefficients that balance vibration suppression and energy consumption.

[0015] Furthermore, the deep reinforcement learning training includes the following steps:

[0016] 1) Simulation pre-training: use the numerical model of CFD and FEA combined simulation to generate training data, define the coupling boundary conditions of fluid-structure interaction, and combine the preset control strategy to design the intelligent vibration suppression device control strategy;

[0017] 2) Transfer learning: transfer the simulation training intelligent vibration suppression device control strategy to the intelligent vibration suppression device, fine-tune the control strategy parameters with the measured data, and complete the application control strategy;

[0018] 3) Online learning: apply the application control strategy to the actual wind turbine generator set, and continuously update the application control strategy according to the actual operation state of the wind turbine generator set to adapt to environmental changes.

[0019] Further, the vibration suppression control system comprises:

[0020] The model predictive control first establishes a dynamic model, and the corresponding state equation x k+1 = Ax k + Bu k + w k , wherein x k is the system state, i.e. the vibration modal amplitude or the flow field characteristic; u k is the control input, i.e. the piezoelectric actuator voltage; w k is the background noise; A and B are correction coefficients; rolling optimization is performed, and the objective function is wherein Q and R are weight matrices, and the vibration suppression and energy consumption are balanced.

[0021] Further, the vibration suppression control system comprises:

[0022] The application control strategy controls the action of the intelligent vibration suppression device based on the real-time phase of the vibration signal, generates a synchronous control voltage, dynamically adjusts the voltage amplitude according to the action value output by the deep reinforcement learning training, activates the main control system when the vibration energy exceeds the threshold, allocates the control amount of each piezoelectric actuator, and realizes the destructive interference of the vibration wave through phase control.

[0023] Further, the vibration state monitoring system is set to have a sampling frequency greater than 3 times the natural frequency of the vibration structure generated by the wind turbine generator set.

[0024] An intelligent control method for suppressing vibration of a wind turbine generator set, which is realized by a processor calling an intelligent vibration suppression device, a vibration state monitoring system, a flow field state monitoring system and a main control system in the intelligent control system for suppressing vibration of a wind turbine generator set.

[0025] A non-transitory computer readable medium storing instructions, which, when executed by a processor, perform the intelligent control method for suppressing vibration of a wind turbine generator set.

[0026] A computing device comprises a processor and a memory for storing a program executable by the processor, and the processor implements the intelligent control method for suppressing vibration of a wind turbine generator set when executing the program stored in the memory.

[0027] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0028] 1. The present application uses the design control strategy of the wind turbine generator set and the operating state data of the wind farm as input, intelligently learns and judges through the combination of deep reinforcement learning and model predictive control, generates a synchronous control voltage based on the real-time phase of the vibration signal, dynamically adjusts the voltage amplitude according to the action value output by the deep reinforcement learning, activates the main control system when the vibration energy exceeds the threshold, distributes the control amount of each actuator, realizes the destructive interference of the vibration wave through phase regulation, avoids the occurrence of vibration overrun during operation, and ensures the power generation and safety of the wind turbine generator set.

[0029] 2. The intelligent vibration suppression device is a piezoelectric actuator, which can be installed at the vibration structure of the wind turbine generator set such as the blade and the tower, uses a high-voltage coefficient material, and covers the main frequency of the mechanism vibration in the response frequency range; provides a fast response voltage signal; uses a piezoelectric energy harvesting module to convert vibration energy into electrical energy to charge the power supply of the intelligent vibration suppression device; and uses a hybrid power supply combining a lithium battery and a super capacitor to ensure the stability and reliability of the intelligent vibration suppression device. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 FIG. 1 is a structural schematic diagram of an intelligent control system.

[0031] Figure 2 FIG. 2 is a schematic diagram of the overall architecture of an intelligent control system.

[0032] Figure 3 FIG. 3 is a workflow diagram of a data analysis system.

[0033] Figure 4 FIG. 4 is a workflow diagram of a suppression control system. DETAILED DESCRIPTION

[0034] The present application will be further described below in conjunction with specific embodiments.

[0035] Embodiment 1

[0036] Referring to FIG. 1, an intelligent control system for suppressing vibration of a wind turbine generator set provided by the present embodiment comprises: Figures 1 to 2

[0037] ​The intelligent vibration suppression device 1 is installed on the structure generating vibration such as the blade 5 and the tower 6 by using the piezoelectric actuator, the response frequency range covers the main frequency of the structure vibration such as the blade 5 and the tower 6, the vibration energy is converted into electric energy by using the piezoelectric energy collection module to charge the power supply of the intelligent vibration suppression device, the mixed power supply is the combination of the lithium battery and the super capacitor; the vibration state monitoring system 2 is used for recording the vibration state data of the structure such as the blade 5 and the tower 6 of the wind turbine generator set, the sampling frequency needs to be higher than 3 times of the natural frequency of each structure, the theorem is met, and the vibration state data of the wind turbine generator set is transmitted to the main control system 4.

[0038] The flow field state monitoring system 3 is used for measuring the flow velocity and vortex shedding frequency in real time by using the hot-wire anemometer or the laser radar wind measuring device, calculating the turbulence intensity and wind shear and the like, and transmitting the flow field state data of the wind turbine generator set to the main control system 4.

[0039] The main control system 4 includes the data collector 401 and the processor 402, the measured and calculated data of the vibration state monitoring system 2 and the flow field state monitoring system 3 are time-stamped and aligned as input parameters by the data collector 401, the input parameters and the design parameters of the unit are combined for learning and judgment by the processor 402, the intelligent vibration suppression device 1 is controlled according to the judgment result, and the vibration state suppression of the unit is realized.

[0040] The processor includes the data analysis system 402a and the vibration suppression control system 402b, wherein:

[0041] Referring to Figure 3 The data analysis system 402 is responsible for parameter feature extraction and data dimension reduction, the vibration main frequency and the vortex shedding frequency are obtained by frequency domain analysis, the dominant vibration mode is identified by intrinsic orthogonal decomposition, the vibration mode of the structure such as the blade 5 and the tower 6 is tracked in real time, and the high-energy vibration area is located; the data dimension reduction compresses the pressure sensor data into a low-dimensional feature vector by using the convolution autoencoder, maps the multi-channel vibration signal into the first three order modal amplitudes, arranges the flow field data and the design parameters, and outputs the coupling calculation parameters to the vibration suppression control system 402b;

[0042] Referring to Figure 4 The vibration suppression control system 402b combines the deep reinforcement learning and the model predictive control, the deep reinforcement learning provides the initial control strategy, the model predictive control performs local optimization, the application control strategy controls the action of the intelligent vibration suppression device, and the adaptive vibration suppression is realized.

[0043] The deep reinforcement learning training needs to model the environment, determine the state space including vibration amplitude, frequency, flow rate and pressure distribution characteristics, etc., and set the dimension to be no less than 20; the action space defines the voltage amplitude and phase of the piezoelectric actuator as a continuous space, and the dimension is equal to the number of piezoelectric actuators; the response function R = -(α * |A vib |+β||V control ||), wherein A vib is the vibration amplitude, V control is the control voltage, α and β are weight coefficients, and the vibration suppression and energy consumption are balanced. The algorithm uses the proximal policy optimization method suitable for continuous action space, and has high stability; the SAC (Soft Actor-Critic) method supports maximum entropy optimization and enhances the exploration ability.

[0044] The deep reinforcement learning training process is divided into three steps, the first step is simulation pre-training: using the numerical model of computational fluid dynamics CFD and finite element analysis FEA to generate training data, defining the coupling boundary conditions of fluid-structure interaction, combining the project design control strategy to design the intelligent vibration suppression device control strategy; the second step is transfer learning: migrating the control strategy of simulation training to the intelligent vibration suppression device, fine-tuning the control strategy parameters using the measured data, and completing the application control strategy; the third step is online learning: applying the application control strategy in the actual wind turbine generator set, and continuously updating the application control strategy according to the actual running state of the wind turbine generator set to adapt to environmental changes.

[0045] The model predictive control first establishes a dynamic model, and the corresponding state equation is x k+1 = Ax k + Bu k + w k , wherein x k is the system state, i.e. vibration modal amplitude, flow field characteristics, etc.; u k is the control input, i.e. piezoelectric actuator voltage; w k is the background noise; A and B are correction coefficients; rolling optimization is performed, and the objective function is wherein Q and R are weight matrices, and the vibration suppression and energy consumption are balanced.

[0046] The application control strategy controls the intelligent vibration suppression device action based on the real-time phase of the vibration signal, generates a synchronous control voltage, dynamically adjusts the voltage amplitude according to the action value output by the deep reinforcement learning, activates the main control system when the vibration energy exceeds the threshold, allocates the control amount of each actuator, and realizes the destructive interference of vibration waves through phase control.

[0047] Example 2

[0048] The embodiment discloses an intelligent control method for inhibiting vibration of a wind turbine generator set, and the method is realized by calling an intelligent vibration-inhibiting device, a vibration state monitoring system, a flow field state monitoring system and a main control system in the intelligent control system for inhibiting vibration of the wind turbine generator set according to the embodiment 1 through a processor.

[0049] Embodiment 3

[0050] The embodiment discloses a non-transitory computer readable medium storing instructions, when the instructions are executed by a processor, the intelligent control method for inhibiting vibration of a wind turbine generator set according to the embodiment 2 is executed.

[0051] The non-transitory computer readable medium in the embodiment can be a disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), a U disk, a mobile hard disk and the like.

[0052] Embodiment 4

[0053] The embodiment discloses a computing device, comprising a processor and a memory for storing a program executable by the processor, when the processor executes the program stored in the memory, the intelligent control method for inhibiting vibration of a wind turbine generator set according to the embodiment 2 is realized.

[0054] The computing device in the embodiment can be a desktop computer, a notebook computer, a smart phone, a PDA handheld terminal, a tablet computer, a programmable logic controller (PLC) or other terminal devices with processor functions.

[0055] The above-mentioned embodiments are only the preferred embodiments of the present application, and do not limit the scope of the present application, so that any changes made according to the shape and principle of the present application should be covered in the protection scope of the present application.

Claims

1. An intelligent control system for suppressing vibration of wind turbine generator sets, characterized in that, include: Intelligent vibration damping devices are used to identify high vibration energy areas on the surface of a structure. The intelligent vibration damping devices are evenly distributed on the surface of the wind turbine generator structure or installed along the vibration mode nodes of the wind turbine generator. The vibration status monitoring system is used to record the vibration status data of the wind turbine generator set and transmit the vibration status data of the wind turbine generator set to the main control system; The flow field state monitoring system measures the flow velocity and vortex shedding frequency in real time using a pre-set hot-wire anemometer and lidar wind measurement device, and measures the surface pressure pulsation in real time using a pre-set surface pressure sensor, and transmits the flow velocity, vortex shedding frequency and surface pressure pulsation data to the main control system. The main control system includes a data acquisition unit and a processor. The data acquisition unit timestamps and aligns the data from the vibration state monitoring system and the flow field state monitoring system as input. The processor compares the data with the design parameters of the wind turbine generator set to learn and make judgments. Based on the judgment results, it controls the intelligent vibration suppression device to suppress the vibration state of the wind turbine generator set. The processor includes a data analysis system and a vibration suppression control system. The data analysis system is used for parameter feature extraction and data dimensionality reduction. It extracts features by obtaining the vibration dominant frequency and vortex shedding frequency through frequency domain analysis, identifies the dominant vibration mode through intrinsic orthogonal decomposition, tracks the structural vibration mode in real time, and locates the high-energy vibration area of ​​the wind turbine generator. It compresses the pressure sensor data into a low-dimensional feature vector through a convolutional autoencoder, maps the multi-channel vibration signal to the amplitude of the first three modes, organizes the flow field data and design parameters, and couples the calculated output parameters to the vibration suppression control system. The vibration suppression control system combines deep reinforcement learning with model predictive control. Deep reinforcement learning provides the initial control strategy, while model predictive control performs local optimization. The application control strategy controls the intelligent vibration suppression device to achieve adaptive vibration suppression.

2. The intelligent control system for suppressing vibration of a wind turbine generator set according to claim 1, characterized in that, The intelligent vibration damping device includes a piezoelectric actuator made of a high-voltage coefficient material, a piezoelectric energy harvesting module, and a hybrid power supply. The response frequency range of the piezoelectric actuator covers the main vibration frequency of the wind turbine generator structure. The piezoelectric energy harvesting module converts vibration energy into electrical energy to charge the power supply of the intelligent vibration damping device. The hybrid power supply is a combination of a lithium battery and a supercapacitor.

3. The intelligent control system for suppressing vibration of a wind turbine generator set according to claim 2, characterized in that, The vibration suppression control system includes: The deep reinforcement learning training first performs environmental modeling, defining the state space as including vibration amplitude, frequency, flow velocity, and pressure distribution characteristics, with a dimension of not less than 20; the action space defines the voltage amplitude and phase of the piezoelectric actuator as a continuous space, with a dimension equal to the number of piezoelectric actuators; the response function is R=-(α*|Avib|+β||Vcontrol||), where Avib is the vibration amplitude, Vcontrol is the control voltage, and α and β are weighting coefficients, balancing vibration suppression and energy consumption.

4. The intelligent control system for suppressing vibration of a wind turbine generator set according to claim 3, characterized in that, The deep reinforcement learning training includes the following steps: 1) Simulation pre-training: Training data is generated using a numerical model of computational fluid dynamics (CFD) and finite element analysis (FEA) co-simulation, the coupling boundary conditions of fluid-structure interaction are defined, and the control strategy of the intelligent vibration suppression device is designed in combination with the preset control strategy. 2) Transfer learning: The control strategy of the intelligent vibration damping device trained in simulation is transferred to the intelligent vibration damping device. The control strategy parameters are fine-tuned using measured data to complete the application of the control strategy. 3) Online learning: Apply the control strategy to the actual wind turbine generator set, and continuously update the control strategy according to the actual operating status of the wind turbine generator set to adapt to environmental changes.

5. The intelligent control system for suppressing vibration of a wind turbine generator set according to claim 2, characterized in that, The vibration suppression control system includes: The model predictive control first establishes a dynamic model, with corresponding state equations xk. +1 =Axk + Buk + wk, where xk is the system state, i.e., the amplitude of the vibration mode or the flow field characteristics; uk is the control input, i.e., the piezoelectric actuator voltage; wk is the background noise; A and B are correction coefficients; rolling optimization is performed, and the objective function is... Q and R are weight matrices that balance vibration suppression and energy consumption.

6. The intelligent control system for suppressing vibration of a wind turbine generator set according to claim 1, characterized in that, The vibration suppression control system includes: The application control strategy controls the intelligent vibration suppression device based on the real-time phase of the vibration signal, generates a synchronous control voltage, dynamically adjusts the voltage amplitude according to the action value output by deep reinforcement learning training, and activates the main control system when the vibration energy exceeds the threshold, allocates the control quantity of each piezoelectric actuator, and achieves destructive interference of vibration waves through phase modulation.

7. The intelligent control system for suppressing vibration of a wind turbine generator set according to claim 1, characterized in that, The sampling frequency set by the vibration state monitoring system is more than three times the natural frequency of the vibration structure generated by the wind turbine generator.

8. An intelligent control method for suppressing vibration of wind turbine generator sets, characterized in that, This method is implemented by the processor calling the intelligent vibration suppression device, vibration state monitoring system, flow field state monitoring system and main control system in the intelligent control system for suppressing the vibration of wind turbine generator sets as described in any one of claims 1-7.

9. A non-transitory computer-readable medium storing instructions, characterized in that, When the instruction is executed by the processor, the intelligent control method for suppressing the vibration of the wind turbine generator set as described in claim 8 is executed.

10. A computing device, comprising a processor and a memory for storing a processor-executable program, characterized in that, When the processor executes the program stored in the memory, it implements the intelligent control method for suppressing the vibration of the wind turbine generator set as described in claim 8.

Images