Wind generating set yaw control method and device based on neural network
By adopting a neural network-based yaw control method in wind turbines, using multiple deep learning algorithms to predict yaw angles and dynamically adjusting the model weights, the problem of low yaw angle prediction accuracy and stability in the prior art is solved, and more efficient and reliable wind power generation is achieved.
Patent Information
- Application Number
- CN202411729281.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-28
- Publication Date
- 2025-02-28
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the existing yaw control method based on deep learning, the prediction accuracy and stability of the yaw angle are relatively low, and are easily affected by environmental factors and sensor accuracy, resulting in a decrease in measurement accuracy and response speed.
The yaw control method of wind turbine sets is adopted based on neural networks. By obtaining the state parameters of the wind turbine, the yaw angle prediction model trained by two different deep learning algorithms is used to predict. Combined with the prediction of yaw angle difference, the target predicted yaw angle is determined, and the model weight is dynamically adjusted to improve the prediction accuracy.
It effectively improves the accuracy of yaw control of wind turbine units, improves the efficiency and reliability of wind power generation, and reduces the dependence on environmental factors and sensor accuracy.
Smart Images
Figure CN119532108A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of wind turbine yaw control, and specifically to a wind turbine yaw control method based on a neural network, a wind turbine yaw control device based on a neural network, a computer-readable storage medium and a terminal device. Background Art
[0002] At present, the yaw control of wind turbines usually adopts the following method: a wind direction sensor is installed on the wind turbine to monitor the change of wind direction in real time. When the wind direction changes, the wind direction sensor transmits a signal to the control system. The control system determines whether a yaw operation is required based on the angle between the wind direction and the axis of the wind turbine cabin. If the angle exceeds a certain threshold, the control system will start the yaw motor to rotate the cabin until the angle between the wind direction and the axis of the cabin is within the allowable range. The existing system can obtain the basic operating parameters of the yaw control system, such as pressure, flow and temperature, for basic status monitoring. However, due to the problems of sensor accuracy, response time, reliability and installation location, the traditional yaw control method is easily affected by obstacles such as ambient temperature, humidity, air pressure, electromagnetic interference, lightning strikes, vibration, surrounding buildings, trees, etc., thereby reducing the measurement accuracy and response speed. At present, through deep learning and analysis of massive data, real-time monitoring and control of system status can also be achieved. For example, by learning a large amount of historical data and real-time operation data, the neural network can accurately perceive changes in wind direction, wind turbine operation status, and the working conditions of yaw control, so as to adjust the yaw angle in real time according to the external environment and wind turbine operation status to adapt to different wind directions and wind conditions, thereby improving the power generation efficiency of the wind turbine. However, the existing yaw control based on deep learning is usually implemented with a single model, and the prediction accuracy and stability of the yaw angle are low. Summary of the invention
[0003] The purpose of the embodiments of the present application is to provide a wind turbine yaw control method based on a neural network, a wind turbine yaw control device based on a neural network, a computer-readable storage medium and a terminal device to solve the above problems.
[0004] In order to achieve the above-mentioned object, the first aspect of the present application provides a yaw control method of a wind turbine generator set based on a neural network, comprising:
[0005] Acquire state parameters of a target wind turbine generator set, wherein the state parameters include at least wind speed, wind direction, temperature, humidity, and air pressure of an environment in which the target wind turbine generator set is located, and at least one of a rotation speed, torque, and output power of the target wind turbine generator set;
[0006] Taking the state parameters of the target wind turbine set as input, a first predicted yaw angle of the wind turbine set is output through a first yaw angle prediction model, and taking the state parameters of the target wind turbine set as input, a second predicted yaw angle of the wind turbine set is output through a second yaw angle prediction model, wherein the first yaw angle prediction model and the second yaw angle prediction model are respectively obtained by training different deep learning algorithms through historical state parameters and corresponding yaw angles of different wind turbine sets;
[0007] Determine a predicted yaw angle difference between the first predicted yaw angle and the second predicted yaw angle, and determine a target predicted yaw angle of the target wind turbine generator set based on the first predicted yaw angle and the second predicted yaw angle according to the predicted yaw angle difference;
[0008] The target wind turbine generator set is controlled to operate at the target predicted yaw angle.
[0009] Optionally, the first yaw angle prediction model is obtained by training a first deep learning algorithm with historical state parameters and corresponding yaw angles of different wind turbines, and the training process of the first deep learning algorithm includes:
[0010] Determine the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as a first historical state parameter;
[0011] Obtaining first historical state parameters and corresponding actual yaw angles of different wind turbines at multiple sampling moments within a preset historical period;
[0012] Taking the first historical state parameters of multiple sampling moments within the preset historical period as input, the first deep learning algorithm outputs the corresponding predicted yaw angle, calculates the error between the predicted yaw angle output by the first deep learning algorithm and the corresponding actual yaw angle, and adjusts the algorithm parameters of the first deep learning algorithm according to the comparison result until the error between the predicted yaw angle output by the first deep learning algorithm and the actual yaw angle is lower than a preset error threshold or reaches a maximum number of iterations, so as to obtain the first yaw angle prediction model.
[0013] Optionally, the second yaw angle prediction model is obtained by training a second deep learning algorithm with historical state parameters and corresponding yaw angles of different wind turbines, and the training process of the second deep learning algorithm includes:
[0014] Determine the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as a second historical state parameter, wherein the second historical state parameter is different from at least one of the first historical state parameters;
[0015] Obtaining second historical state parameters and corresponding actual yaw angles of different wind turbines at multiple sampling moments within a preset historical period;
[0016] Taking the second historical state parameters of multiple sampling moments within the preset historical period as input, the second deep learning algorithm outputs the corresponding predicted yaw angle, calculates the error between the predicted yaw angle output by the second deep learning algorithm and the corresponding actual yaw angle, and adjusts the algorithm parameters of the second deep learning algorithm according to the comparison result until the error between the predicted yaw angle output by the second deep learning algorithm and the actual yaw angle is lower than a preset error threshold or reaches the maximum number of iterations, so as to obtain the second yaw angle prediction model.
[0017] Optionally, determining the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as the first historical state parameter includes:
[0018] Determine the weight coefficients of wind speed, wind direction, temperature, humidity, air pressure, rotation speed, torque and output power, determine the state parameter whose weight coefficient belongs to a first weight coefficient interval as a first state parameter, determine the state parameter whose weight coefficient belongs to a second weight coefficient interval as a second state parameter, and the minimum value of the first weight coefficient interval is not less than the maximum value of the second weight coefficient interval;
[0019] At least one first state parameter and at least one second state parameter are determined as first historical state parameters.
[0020] Optionally, determining the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as the second historical state parameter includes:
[0021] At least one first state parameter and at least one second state parameter are determined to be second historical state parameters, and at least one first state parameter or second state parameter in the second historical state parameters is different from the first state parameter and the second state parameter in the first historical state parameters.
[0022] Optionally, determining a target predicted yaw angle of the target wind turbine generator system based on the first predicted yaw angle and the second predicted yaw angle according to the predicted yaw angle difference includes:
[0023] If the predicted yaw angle difference is lower than the first difference, taking the first predicted yaw angle or the second predicted yaw angle as the target predicted yaw angle of the target wind turbine;
[0024] If the predicted yaw angle difference is not lower than the first difference, obtain the first model weight of the first yaw angle prediction model and the second model weight of the second yaw angle prediction model, and perform weighted summation of the first predicted yaw angle and the second predicted yaw angle based on the first model weight and the second model weight to obtain the target predicted yaw angle of the target wind turbine group.
[0025] Optionally, after determining the target predicted yaw angle of the target wind turbine group, the method further includes:
[0026] In response to the weight adjustment instruction, randomly acquiring a plurality of historical state parameters and corresponding actual yaw angles of the target wind turbine group within a preset period before a current moment;
[0027] Taking the obtained multiple historical state parameters as input, outputting a third predicted yaw angle through the first yaw angle prediction model, and taking the obtained multiple historical state parameters as input, outputting a fourth predicted yaw angle through the second yaw angle prediction model;
[0028] Determining a first accuracy rate of the first yaw angle prediction model according to the obtained multiple third predicted yaw angles and the corresponding actual yaw angles, and determining a second accuracy rate of the second yaw angle prediction model according to the obtained multiple fourth predicted yaw angles and the corresponding actual yaw angles;
[0029] If the first accuracy is greater than the second accuracy, increase the first model weight by a set step size, and decrease the second model weight by the set step size;
[0030] If the first accuracy is less than the second accuracy, the first model weight is reduced by the set step size, and the second model weight is increased by the set step size.
[0031] In a second aspect of the present application, a yaw control device for a wind turbine generator set based on a neural network is provided, comprising:
[0032] A data acquisition module is configured to obtain state parameters of a target wind turbine generator set, wherein the state parameters include at least wind speed, wind direction, temperature, humidity, air pressure of an environment in which the target wind turbine generator set is located, and at least one of a rotation speed, torque, and output power of the target wind turbine generator set;
[0033] an initial yaw angle prediction model, configured to take the state parameters of the target wind turbine set as input, and output a first predicted yaw angle of the wind turbine set via a first yaw angle prediction model, and to take the state parameters of the target wind turbine set as input, and output a second predicted yaw angle of the wind turbine set via a second yaw angle prediction model, wherein the first yaw angle prediction model and the second yaw angle prediction model are obtained by training different deep learning algorithms using historical state parameters and corresponding yaw angles of different wind turbine sets, respectively;
[0034] a target yaw angle prediction model, configured to determine a predicted yaw angle difference between the first predicted yaw angle and the second predicted yaw angle, and determine a target predicted yaw angle of the target wind turbine based on the first predicted yaw angle and the second predicted yaw angle according to the predicted yaw angle difference;
[0035] The control module is configured to control the target wind turbine group to operate at the target predicted yaw angle.
[0036] In a third aspect of the present application, a computer-readable storage medium is provided, which stores a computer program which, when executed by a processor, causes the processor to execute the above-mentioned neural network-based wind turbine yaw control method.
[0037] In a fourth aspect of the present application, a terminal device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned neural network-based wind turbine yaw control method when executing the computer program.
[0038] The embodiments provided in this application have the following beneficial effects:
[0039] The present application effectively improves the accuracy of yaw control of a wind turbine generator set, thereby improving the efficiency and reliability of wind power generation.
[0040] Other features and advantages of the embodiments or implementations of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:
[0042] Figure 1 A method flow chart of a wind turbine generator set yaw control method based on a neural network according to an embodiment of the present application is schematically shown;
[0043] Figure 2A schematic diagram of the system structure of a wind turbine yaw control system according to an embodiment of the present application is shown;
[0044] Figure 3 A schematic block diagram of a wind turbine generator set yaw control device based on a neural network according to an embodiment of the present application is schematically shown;
[0045] Figure 4 A schematic diagram of the structure of a terminal device according to an embodiment of the present application is shown schematically.
[0046] Description of Reference Numerals
[0047] 10-terminal device, 100-processor, 101-memory, 102-computer program. DETAILED DESCRIPTION
[0048] The specific implementation of the embodiment of the present application is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present application, and is not used to limit the embodiment of the present application.
[0049] The airfoil of wind turbine blades has a direct impact on the operation of wind turbines. At present, when designing the airfoil of wind turbine blades, when simulating the load under ice-covered conditions, only the impact of ice on the blade mass is often considered, while the impact of ice on the aerodynamic performance of the blade is not considered. Therefore, when the designed wind turbine blades are operated under ice-covered conditions, it may affect the operation of the wind turbine and pose a safety hazard.
[0050] In order to solve the above problems, Figure 1 As shown, the first aspect of the present application provides a yaw control method for a wind turbine generator set based on a neural network, comprising: obtaining state parameters of a target wind turbine generator set, the state parameters at least including wind speed, wind direction, temperature, humidity, air pressure of the environment in which the target wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the target wind turbine generator set; using the state parameters of the target wind turbine generator set as input, outputting a first predicted yaw angle of the wind turbine generator set through a first yaw angle prediction model, and using the state parameters of the target wind turbine generator set as input, outputting a second predicted yaw angle of the wind turbine generator set through a second yaw angle prediction model, wherein the first yaw angle prediction model and the second yaw angle prediction model are obtained by respectively training different deep learning algorithms through historical state parameters and corresponding yaw angles of different wind turbine generator sets; determining a predicted yaw angle difference between the first predicted yaw angle and the second predicted yaw angle, and determining a target predicted yaw angle of the target wind turbine generator set based on the first predicted yaw angle and the second predicted yaw angle according to the predicted yaw angle difference; and controlling the target wind turbine generator set to operate at the target predicted yaw angle.
[0051] In this way, the present application effectively improves the accuracy of yaw control of the wind turbine generator set, thereby improving the efficiency and reliability of wind power generation.
[0052] It is understandable that the yaw system is an important component of a wind turbine. When the wind speed vector changes to a certain extent, the yaw system will drive the wind turbine to face the wind. However, wind is uncertain, and the wind direction will fluctuate frequently. Therefore, whether the yaw system can face the wind has become a key factor in increasing power generation and reducing unit load. This application constructs a yaw angle prediction model through a deep learning algorithm, which can predict the optimal yaw angle of the wind turbine according to the real-time state parameters of the wind turbine, and control the wind turbine to operate at the optimal yaw angle. Figure 2 As shown, the method of the present application is implemented based on the control system constructed by the present application. The system includes a server, which is connected to the PLC of the entire wind farm through a network to achieve real-time acquisition of the main parameters of the yaw control system of the entire wind farm. The server is used to process, store and analyze the data obtained from the PLC of the entire field. The system of the present application can realize the rapid storage of large amounts of data, and perform real-time analysis of data through efficient algorithms, providing support for the training and model optimization of neural network algorithms. In addition, the solution stores the results of real-time analysis in the cloud, realizing remote access and management of data.
[0053] In the present application, the first yaw angle prediction model is obtained by training the first deep learning algorithm with historical state parameters and corresponding yaw angles of different wind turbines. The training process of the first deep learning algorithm includes:
[0054] Determine the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine is located, and at least one of the rotation speed, torque and output power of the wind turbine as the first historical state parameter, for example, use the wind speed, wind direction, rotation speed and output power as input parameters of the deep learning algorithm; obtain the first historical state parameters of different wind turbines at multiple sampling moments in a preset historical period and the corresponding actual yaw angles, for example, obtain the historical state parameters of different wind turbines in the past six months and the actual yaw angles corresponding to each parameter as training samples of the deep learning algorithm, so as to train the deep learning algorithm. It can be understood that the actual yaw angle corresponding to the state parameter is the yaw angle of the yaw system accurately facing the wind under the state parameter;
[0055] Taking the first historical state parameters at multiple sampling moments in a preset historical period as input, the first deep learning algorithm outputs the corresponding predicted yaw angle, calculates the error between the predicted yaw angle output by the first deep learning algorithm and the corresponding actual yaw angle, and adjusts the algorithm parameters of the first deep learning algorithm according to the comparison result until the error between the predicted yaw angle output by the first deep learning algorithm and the actual yaw angle is lower than the preset error threshold or reaches the maximum number of iterations, thereby obtaining a first yaw angle prediction model. The first deep learning algorithm includes but is not limited to a BP neural network, a CNN neural network, a long-short time neural network, and the like.
[0056] The second yaw angle prediction model is obtained by training the second deep learning algorithm with historical state parameters and corresponding yaw angles of different wind turbines. The training process of the second deep learning algorithm includes:
[0057] Determine that the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine is located, and at least one of the rotation speed, torque and output power of the wind turbine are the second historical state parameters, wherein the second historical state parameter is different from at least one of the first historical state parameters. For example, if the first historical state parameter is wind speed, wind direction, rotation speed and output power, then the second historical state parameter may be wind speed, wind direction, temperature, rotation speed, torque and output power, that is, at least one parameter of the second historical state parameter is different from the first historical state parameter;
[0058] Obtain the second historical state parameters and corresponding actual yaw angles of different wind turbines at multiple sampling moments within a preset historical period; use the second historical state parameters at multiple sampling moments within a preset historical period as input, output the corresponding predicted yaw angle through a second deep learning algorithm, calculate the error between the predicted yaw angle output by the second deep learning algorithm and the corresponding actual yaw angle, and adjust the algorithm parameters of the second deep learning algorithm according to the comparison result until the error between the predicted yaw angle output by the second deep learning algorithm and the actual yaw angle is lower than a preset error threshold or reaches a maximum number of iterations, thereby obtaining a second yaw angle prediction model. The second deep learning algorithm includes but is not limited to a BP neural network, a CNN neural network, a long-short time neural network, etc., and the second deep learning algorithm is different from the first deep learning algorithm. For example, the first deep learning algorithm is a BP neural network, and the second deep learning algorithm can be a long-short time neural network.
[0059] In the present application, determining the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine is located, and at least one of the speed, torque and output power of the wind turbine is determined as the first historical state parameter, including: determining the weight coefficients of the wind speed, wind direction, temperature, humidity, air pressure, speed, torque and output power, determining the state parameter whose weight coefficient belongs to the first weight coefficient interval as the first state parameter, determining the state parameter whose weight coefficient belongs to the second weight coefficient interval as the second state parameter, the minimum value of the first weight coefficient interval is not less than the maximum value of the second weight coefficient interval; determining at least one first state parameter and at least one second state parameter as the first historical state parameter. For example, a corresponding weight coefficient can be configured in advance for each parameter according to the importance of the influence of the above-mentioned parameters on the yaw angle, and each parameter can be divided into multiple groups according to the size of the weight coefficient. For example, the weight coefficient of wind speed and wind direction is the largest, and the wind speed and wind direction are divided into 1 group. The weight coefficients of speed, torque and output power are smaller than those of wind speed and wind direction, so the speed, torque and output power are divided into 1 group. The weight coefficients of temperature, humidity and air pressure are the smallest, and temperature, humidity and air pressure are divided into 1 group; or, the wind speed, wind direction, speed and output power of the wind turbine are divided into 1 group, and the temperature, humidity, air pressure and torque are divided into 1 group. There is no limitation on this here. Then, when determining the first historical state parameter, each group determines at least one parameter as the first historical state parameter. For example, in the first group, wind speed and wind direction have the greatest impact on the yaw system. Therefore, wind speed and wind direction are determined as the first historical state parameters. In the second group, rotation speed and output power are determined as the first historical state parameters. In the third group, air pressure is determined as the first historical state parameter. The first historical state parameters finally determined are wind speed, wind direction, rotation speed, output power and air pressure.
[0060] Similarly, determining the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine is located, and at least one of the speed, torque and output power of the wind turbine as the second historical state parameter includes: determining at least one first state parameter and at least one second state parameter as the second historical state parameter, and at least one first state parameter or second state parameter in the second historical state parameter is different from the first state parameter and the second state parameter in the first historical state parameter. For example, the second historical state parameter may include wind speed, wind direction, speed, torque, output power, and temperature. In this way, different deep learning algorithms use different parameters as training data, which can effectively improve the prediction accuracy of the final target yaw angle.
[0061] In the present application, according to the predicted yaw angle difference, the target predicted yaw angle of the target wind turbine is determined based on the first predicted yaw angle and the second predicted yaw angle, including: if the predicted yaw angle difference is lower than the first difference, the first predicted yaw angle or the second predicted yaw angle is used as the target predicted yaw angle of the target wind turbine; if the predicted yaw angle difference is not lower than the first difference, the first model weight of the first yaw angle prediction model and the second model weight of the second yaw angle prediction model are obtained, and the first predicted yaw angle and the second predicted yaw angle are weighted and summed based on the first model weight and the second model weight to obtain the target predicted yaw angle of the target wind turbine. For example, different weights are configured for different prediction models in advance, wherein the model weight of the prediction model can be configured according to the prediction accuracy of the yaw angle according to different deep learning algorithms, and the model with higher accuracy has a higher model weight. If the difference between the first predicted yaw angle and the second predicted yaw angle is small, any predicted yaw angle can be used as the target yaw angle; if the difference between the two is large, the output results are weighted and summed to eliminate the error.
[0062] In the present application, after determining the target predicted yaw angle of the target wind turbine group, the method further includes:
[0063] In response to the weight adjustment instruction, a plurality of historical state parameters and the corresponding actual yaw angles of the target wind turbine within a preset time period before the current moment are randomly obtained; during the operation of the wind turbine, the present application also dynamically adjusts the model weights of each prediction model, wherein the weight adjustment instruction can be a timing instruction, for example, triggering the weight adjustment once a week; after activating the weight adjustment, the historical state parameters and the corresponding actual yaw angles of the target wind turbine in the recent period of time are randomly obtained, and the obtained data are used as test data. It can be understood that the historical state parameters and the corresponding actual yaw angles of the correct wind can be determined according to the wind direction collected by the wind vane sensor at the same sampling moment. For example, at a certain historical moment, the actual yaw angle corresponding to the state parameter group 1 is X1°, and the optimal yaw angle determined according to the wind direction collected by the wind vane sensor at this time is X2°. If the difference between X1 and X2 is lower than the threshold value, it is considered that the yaw system is correctly facing the wind at this time;
[0064] Taking the obtained multiple historical state parameters as input, outputting a third predicted yaw angle through the first yaw angle prediction model, and taking the obtained multiple historical state parameters as input, outputting a fourth predicted yaw angle through the second yaw angle prediction model; inputting the obtained historical state parameters into the two models to obtain corresponding prediction results;
[0065] Determine a first accuracy rate of a first yaw angle prediction model according to the obtained multiple third predicted yaw angles and the corresponding actual yaw angles, and determine a second accuracy rate of a second yaw angle prediction model according to the obtained multiple fourth predicted yaw angles and the corresponding actual yaw angles; determine the accuracy rate of each prediction model according to the actual yaw angle corresponding to each historical state parameter in the test data and the prediction result;
[0066] If the first accuracy rate is greater than the second accuracy rate, the weight of the first model is increased by a set step size, and the weight of the second model is reduced by a set step size; if the first accuracy rate is less than the second accuracy rate, the weight of the first model is reduced by a set step size, and the weight of the second model is increased by a set step size. That is, according to the accuracy rate, the model weight of the model with lower accuracy rate is reduced by a set step size, and the model weight of the model with higher accuracy rate is increased by a set step size. In this way, different prediction models are tested according to actual data in the recent period of time, and the model weight of each model is dynamically adjusted, which can further improve the prediction accuracy of the yaw angle.
[0067] like Figure 3 As shown, in a second aspect of the present application, a yaw control device for a wind turbine generator set based on a neural network is provided, comprising:
[0068] A data acquisition module is configured to obtain state parameters of a target wind turbine generator set, wherein the state parameters include at least wind speed, wind direction, temperature, humidity, air pressure of an environment in which the target wind turbine generator set is located, and at least one of a rotation speed, torque, and output power of the target wind turbine generator set;
[0069] The initial yaw angle prediction model is configured to take the state parameters of the target wind turbine set as input, output a first predicted yaw angle of the wind turbine set through a first yaw angle prediction model, and take the state parameters of the target wind turbine set as input, output a second predicted yaw angle of the wind turbine set through a second yaw angle prediction model, wherein the first yaw angle prediction model and the second yaw angle prediction model are obtained by training different deep learning algorithms using historical state parameters and corresponding yaw angles of different wind turbine sets, respectively;
[0070] A target yaw angle prediction model is configured to determine a predicted yaw angle difference between the first predicted yaw angle and the second predicted yaw angle, and determine a target predicted yaw angle of the target wind turbine based on the first predicted yaw angle and the second predicted yaw angle according to the predicted yaw angle difference;
[0071] The control module is configured to control the target wind turbine to operate at a target predicted yaw angle.
[0072] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.
[0073] In a third aspect of the present application, a computer-readable storage medium is provided, which stores a computer program which, when executed by a processor, causes the processor to execute the above-mentioned neural network-based wind turbine yaw control method.
[0074] In a fourth aspect of the present application, a terminal device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned neural network-based wind turbine yaw control method when executing the computer program.
[0075] like Figure 4 FIG. 1 is a schematic diagram of a terminal device provided in an embodiment of the present application. Figure 4 As shown, the terminal device 10 of this embodiment includes: a processor 100, a memory 101, and a computer program 102 stored in the memory 101 and executable on the processor 100. When the processor 100 executes the computer program 102, the steps in the above method embodiment are implemented. Alternatively, when the processor 100 executes the computer program 102, the functions of each module / unit in the above device embodiments are implemented.
[0076] Exemplarily, the computer program 102 may be divided into one or more modules / units, one or more modules / units are stored in the memory 101, and are executed by the processor 100 to complete the present application. One or more modules / units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 102 in the terminal device 10.
[0077] The terminal device 10 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device 10 may include, but is not limited to, a processor 100 and a memory 101. Those skilled in the art will appreciate that Figure 4 It is only an example of the terminal device 10 and does not constitute a limitation of the terminal device 10. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.
[0078] The processor 100 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
[0079] The memory 101 may be an internal storage unit of the terminal device 10, such as a hard disk or memory of the terminal device 10. The memory 101 may also be an external storage device of the terminal device 10, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 10. Further, the memory 101 may also include both an internal storage unit of the terminal device 10 and an external storage device. The memory 101 is used to store computer programs and other programs and data required by the terminal device 10. The memory 101 may also be used to temporarily store data that has been output or is to be output.
[0080] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0081] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0082] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.
Claims
1. A yaw control method for a wind turbine generator set based on a neural network, characterized in that: include: Acquire state parameters of a target wind turbine generator set, wherein the state parameters include at least wind speed, wind direction, temperature, humidity, and air pressure of an environment in which the target wind turbine generator set is located, and at least one of a rotation speed, torque, and output power of the target wind turbine generator set; Taking the state parameters of the target wind turbine set as input, a first predicted yaw angle of the wind turbine set is output through a first yaw angle prediction model, and taking the state parameters of the target wind turbine set as input, a second predicted yaw angle of the wind turbine set is output through a second yaw angle prediction model, wherein the first yaw angle prediction model and the second yaw angle prediction model are respectively obtained by training different deep learning algorithms through historical state parameters and corresponding yaw angles of different wind turbine sets; Determine a predicted yaw angle difference between the first predicted yaw angle and the second predicted yaw angle, and determine a target predicted yaw angle of the target wind turbine generator set based on the first predicted yaw angle and the second predicted yaw angle according to the predicted yaw angle difference; The target wind turbine generator set is controlled to operate at the target predicted yaw angle.
2. The wind turbine yaw control method based on neural network according to claim 1, characterized in that: The first yaw angle prediction model is obtained by training a first deep learning algorithm with historical state parameters and corresponding yaw angles of different wind turbines. The training process of the first deep learning algorithm includes: Determine the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as a first historical state parameter; Obtaining first historical state parameters and corresponding actual yaw angles of different wind turbines at multiple sampling moments within a preset historical period; Taking the first historical state parameters of multiple sampling moments within the preset historical period as input, the first deep learning algorithm outputs the corresponding predicted yaw angle, calculates the error between the predicted yaw angle output by the first deep learning algorithm and the corresponding actual yaw angle, and adjusts the algorithm parameters of the first deep learning algorithm according to the comparison result until the error between the predicted yaw angle output by the first deep learning algorithm and the actual yaw angle is lower than a preset error threshold or reaches a maximum number of iterations, so as to obtain the first yaw angle prediction model.
3. The yaw control method of wind turbine generator set based on neural network according to claim 2, characterized in that: The second yaw angle prediction model is obtained by training a second deep learning algorithm with historical state parameters and corresponding yaw angles of different wind turbines. The training process of the second deep learning algorithm includes: Determine the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as a second historical state parameter, wherein the second historical state parameter is different from at least one of the first historical state parameters; Obtaining second historical state parameters and corresponding actual yaw angles of different wind turbines at multiple sampling moments within a preset historical period; Taking the second historical state parameters of multiple sampling moments within the preset historical period as input, the second deep learning algorithm outputs the corresponding predicted yaw angle, calculates the error between the predicted yaw angle output by the second deep learning algorithm and the corresponding actual yaw angle, and adjusts the algorithm parameters of the second deep learning algorithm according to the comparison result until the error between the predicted yaw angle output by the second deep learning algorithm and the actual yaw angle is lower than a preset error threshold or reaches the maximum number of iterations, so as to obtain the second yaw angle prediction model.
4. The yaw control method of wind turbine generator set based on neural network according to claim 3, characterized in that: Determining the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as the first historical state parameter includes: Determine the weight coefficients of wind speed, wind direction, temperature, humidity, air pressure, rotation speed, torque and output power, determine the state parameter whose weight coefficient belongs to a first weight coefficient interval as a first state parameter, determine the state parameter whose weight coefficient belongs to a second weight coefficient interval as a second state parameter, and the minimum value of the first weight coefficient interval is not less than the maximum value of the second weight coefficient interval; At least one first state parameter and at least one second state parameter are determined as first historical state parameters.
5. The yaw control method of wind turbine generator set based on neural network according to claim 4, characterized in that: Determining the wind speed, wind direction, temperature, humidity, air pressure of the environment in which the wind turbine generator set is located, and at least one of the rotation speed, torque and output power of the wind turbine generator set as the second historical state parameter includes: At least one first state parameter and at least one second state parameter are determined to be second historical state parameters, and at least one first state parameter or second state parameter in the second historical state parameters is different from the first state parameter and the second state parameter in the first historical state parameters.
6. The neural network wind turbine yaw control method according to claim 1, characterized in that: Determining a target predicted yaw angle of the target wind turbine generator set according to the predicted yaw angle difference and based on the first predicted yaw angle and the second predicted yaw angle includes: If the predicted yaw angle difference is lower than the first difference, taking the first predicted yaw angle or the second predicted yaw angle as the target predicted yaw angle of the target wind turbine; If the predicted yaw angle difference is not lower than the first difference, obtain the first model weight of the first yaw angle prediction model and the second model weight of the second yaw angle prediction model, and perform weighted summation of the first predicted yaw angle and the second predicted yaw angle based on the first model weight and the second model weight to obtain the target predicted yaw angle of the target wind turbine group.
7. The neural network-based wind turbine yaw control method according to claim 6, characterized in that: After determining the target predicted yaw angle of the target wind turbine generator set, the method further includes: In response to the weight adjustment instruction, randomly acquiring a plurality of historical state parameters and corresponding actual yaw angles of the target wind turbine group within a preset period before a current moment; Taking the obtained multiple historical state parameters as input, outputting a third predicted yaw angle through the first yaw angle prediction model, and taking the obtained multiple historical state parameters as input, outputting a fourth predicted yaw angle through the second yaw angle prediction model; Determining a first accuracy rate of the first yaw angle prediction model according to the obtained multiple third predicted yaw angles and the corresponding actual yaw angles, and determining a second accuracy rate of the second yaw angle prediction model according to the obtained multiple fourth predicted yaw angles and the corresponding actual yaw angles; If the first accuracy is greater than the second accuracy, increase the first model weight by a set step size, and decrease the second model weight by the set step size; If the first accuracy is less than the second accuracy, the first model weight is reduced by the set step size, and the second model weight is increased by the set step size.
8. A wind turbine yaw control device based on neural network, characterized in that: include: A data acquisition module is configured to obtain state parameters of a target wind turbine generator set, wherein the state parameters include at least wind speed, wind direction, temperature, humidity, air pressure of an environment in which the target wind turbine generator set is located, and at least one of a rotation speed, torque, and output power of the target wind turbine generator set; an initial yaw angle prediction model, configured to take the state parameters of the target wind turbine set as input, and output a first predicted yaw angle of the wind turbine set via a first yaw angle prediction model, and to take the state parameters of the target wind turbine set as input, and output a second predicted yaw angle of the wind turbine set via a second yaw angle prediction model, wherein the first yaw angle prediction model and the second yaw angle prediction model are obtained by training different deep learning algorithms using historical state parameters and corresponding yaw angles of different wind turbine sets, respectively; a target yaw angle prediction model, configured to determine a predicted yaw angle difference between the first predicted yaw angle and the second predicted yaw angle, and determine a target predicted yaw angle of the target wind turbine based on the first predicted yaw angle and the second predicted yaw angle according to the predicted yaw angle difference; The control module is configured to control the target wind turbine group to operate at the target predicted yaw angle.
9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program which, when executed by a processor, enables the processor to execute the wind turbine generator set yaw control method based on a neural network as claimed in any one of claims 1 to 7.
10. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the neural network-based yaw control method for a wind turbine generator set described in any one of claims 1 to 7 is implemented.