Training method and device of single-fan wake proxy model, equipment and storage medium

Abstract

The application discloses a training method and device of a single-wind-turbine wake proxy model, equipment and a storage medium. Standard examples are extracted from preset wind farm data, the wind farm data including incoming flow wind speed and yaw angle of a single wind turbine, parameter configuration is performed on the standard examples, simulation examples are generated, the simulation examples are arranged, simulation results are saved in a distributed file structure, simulation data are generated, the simulation results including wake distribution parameters of the wind farm, a preset autoencoder model is trained based on a data set constructed based on the simulation data, and a single-wind-turbine wake proxy model is obtained, so that the single-wind-turbine wake proxy model can be used to quickly optimize wind turbine parameters, and optimization difficulty is reduced and optimization efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of energy storage technology, and in particular to a training method, apparatus, equipment and storage medium for a single wind turbine wake proxy model. Background Technology

[0002] Because wind turbines upwind of a wind farm absorb wind energy to do work, the incoming wind speed to downstream turbines decreases, thus affecting their power generation efficiency. Therefore, wake optimization of the wind farm is necessary. Wake optimization requires analyzing the wake mechanism of the wind turbines, establishing a wake model, and developing a wake optimization control algorithm.

[0003] Currently, common wake models include the Jensen model, the Frandsen model, and the EPFL Gaussian wake model. These models require a large number of approximate calculations and assumptions, which leads to significant errors in the calculation results, making it difficult to optimize wind turbine parameters. Summary of the Invention

[0004] This application provides a training method, apparatus, equipment, and storage medium for a single wind turbine wake proxy model to solve the technical problem of high difficulty in optimizing wind turbine parameters.

[0005] To address the aforementioned technical problems, firstly, this application provides a training method for a single-fan wake surrogate model, comprising:

[0006] Standard calculation examples are extracted from preset wind farm data, which includes the incoming wind speed and yaw angle of a single wind turbine.

[0007] Configure the parameters of the standard case to generate a simulation case;

[0008] The simulation examples are organized and the simulation results are saved in a distributed file structure to generate example data. The simulation results include the wake distribution parameters of the wind farm.

[0009] The dataset constructed based on the example data is used to train the preset autoencoder model to obtain a single-fan wake proxy model.

[0010] In some implementations, the extraction of standard examples from preset wind farm data includes:

[0011] Acquire the preset wind farm data and input the preset wind farm data into the SOWFA simulation platform;

[0012] Based on the SOWFA simulation platform, the preset wind farm data is used to extract calculation examples to obtain the standard calculation examples.

[0013] In some implementations, configuring parameters for the standard simulation case to generate a simulation case includes:

[0014] Determine the direction of the incoming flow from a single fan in the standard example;

[0015] Based on the incoming flow direction, the incoming wind speed and yaw angle of the single fan in the standard example are configured to generate simulation examples with different parameter configurations.

[0016] In some implementations, the step of organizing the simulation examples, storing the simulation results in a distributed file structure, and generating example data includes:

[0017] The wake distribution parameters are stored in a distributed file structure for each time period within the preset simulation time range;

[0018] Based on the wake distribution parameters, the simulation example is sliced ​​to obtain slice data;

[0019] The sliced ​​data is generated into the example data in CSV format.

[0020] In some implementations, the dataset constructed based on the example data is used to train a preset autoencoder model to obtain a single-fan wake proxy model, including:

[0021] The example data is used to construct a dataset.

[0022] The dataset is input into the preset autoencoder model for training until the preset autoencoder model reaches the preset convergence condition, thereby obtaining a single-fan wake surrogate model with the incoming wind speed and yaw angle as control parameters and the wake distribution parameters as output results.

[0023] In some implementations, constructing the example data into a dataset includes:

[0024] Based on a preset editing script, the example data is processed to generate a tensor file;

[0025] The tensor file is used to construct the dataset.

[0026] In some implementations, the step of inputting the dataset into the preset autoencoder model for training until the preset autoencoder model reaches a preset convergence condition, thereby obtaining a single-fan wake surrogate model with the incoming wind speed and yaw angle as control parameters and the wake distribution parameters as output results, includes:

[0027] Based on the encoder and decoder of the preset autoencoder model, the feature function of the dataset is determined.

[0028] The feature function is mapped to the hidden vector space, and linear evolution features are generated under the influence of the control parameters.

[0029] The linear evolution features are combined with wind farm data from the dataset to train a feature embedder;

[0030] The feature embedder is mapped into the linear vector space required by the encoder to obtain the wake distribution parameters;

[0031] Based on the wake distribution parameters, the preset autoencoder model is updated until the preset autoencoder model reaches the preset convergence condition, thereby generating the single-fan wake proxy model.

[0032] Secondly, this application provides a training device for a single-fan wake surrogate model, comprising:

[0033] The extraction module is used to extract standard calculation examples from preset wind farm data, which includes the incoming wind speed and yaw angle of a single wind turbine.

[0034] The generation module is used to configure the parameters of the standard case and generate simulation cases;

[0035] The processing module is used to process the simulation examples, save the simulation results in a distributed file structure, and generate example data. The simulation results include the wake distribution parameters of the wind farm.

[0036] The training module is used to train a preset autoencoder model on a dataset constructed based on the example data to obtain a single-fan wake surrogate model.

[0037] Thirdly, this application provides a computer device including a processor and a memory, the memory being used to store a computer program, which, when executed by the processor, implements the training method for the single-fan wake surrogate model as described in the first aspect.

[0038] Fourthly, this application provides a computer-readable storage medium, characterized in that it stores a computer program, which, when executed by a processor, implements the training method for the single-fan wake surrogate model as described in the first aspect.

[0039] Compared with the prior art, this application has at least the following beneficial effects:

[0040] By extracting standard simulation examples from preset wind farm data, including the incoming wind speed and yaw angle of a single wind turbine, the parameters of the standard simulation examples are configured to generate simulation examples. The simulation examples are then organized, and the simulation results are saved in a distributed file structure to generate simulation example data. The simulation results include the wake distribution parameters of the wind farm. Based on the dataset constructed from the simulation example data, a preset autoencoder model is trained to obtain a single wind turbine wake proxy model. This allows for rapid optimization of wind turbine parameters using the single wind turbine wake proxy model, reducing optimization difficulty and improving optimization efficiency. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating the training method of a single-fan wake proxy model according to an embodiment of this application;

[0042] Figure 2 This is a schematic diagram of the structure of the training device for the single-fan wake proxy model shown in an embodiment of this application;

[0043] Figure 3 This is a schematic diagram of the structure of a computer device shown in an embodiment of this application. Detailed Implementation

[0044] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0045] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a training method for a single-fan wake proxy model provided in an embodiment of this application. The training method for the single-fan wake proxy model in this application can be applied to computer devices, including but not limited to smartphones, laptops, tablets, desktop computers, physical servers, and cloud servers. Figure 1 As shown, the training method for the single-fan wake proxy model in this embodiment includes steps S101 to S104, which are detailed below:

[0046] Step S101: Extract standard calculation examples from preset wind farm data, wherein the wind farm data includes the incoming wind speed and yaw angle of a single wind turbine.

[0047] In this step, standard calculation examples are extracted from wind farm data through a preset simulation calculation platform to determine typical incoming wind speeds and yaw angles from multiple incoming wind speeds and yaw angles. These results serve as the basis for subsequent modifications to control parameters, thereby expanding the data of the standard calculation examples.

[0048] In some implementations, step S101 includes:

[0049] Acquire the preset wind farm data and input the preset wind farm data into the SOWFA simulation platform;

[0050] Based on the SOWFA simulation platform, the preset wind farm data is used to extract calculation examples to obtain the standard calculation examples.

[0051] In this embodiment, SOWFA is used as the simulation platform. It can be accessed by connecting to a VPN in a specified network environment. The SOWFA simulation platform runs in a DOCKER container, which performs file and port mapping and communicates with the outside world through the Secure Shell protocol SSH.

[0052] Optionally, to facilitate interaction, a data interface can be developed using Python tools to simplify parameter settings and automate the various steps of submitting and executing simulations. By calling this interface script, simulations can be submitted and run as required.

[0053] The complete execution process of a SOWFA example is quite complex, requiring copying the example file, modifying example parameters, executing multiple calculation instructions sequentially, and waiting for completion. After execution, the parallel-stored result files need to be packaged and downloaded locally for viewing and analysis. The optimization process requires batch execution of examples to obtain sufficient sample data; therefore, the above steps need to be automated and an interface developed to connect with the optimization analysis engine. In this SRT project, an automated data interface program based on Python was developed to automate example generation, execution, and post-processing.

[0054] Step S102: Configure the parameters of the standard example to generate a simulation example.

[0055] In this step, the required sample test can be quickly generated by modifying the parameter configuration file of the standard test case using SOWFA.

[0056] In some implementations, step S102 includes:

[0057] Determine the direction of the incoming flow from a single fan in the standard example;

[0058] Based on the incoming flow direction, the incoming wind speed and yaw angle of the single fan in the standard example are configured to generate simulation examples with different parameter configurations.

[0059] In this embodiment, a single wind turbine is used as the target, so two blank standard simulation files for the wind turbine are preset, which are used to perform ABL atmospheric boundary layer simulation calculations and ALM (Actuation Line Model) / ADM (Disk Model) wake simulation calculations, respectively. For example, considering the symmetry of the single wind turbine scenario and the control object for yaw angle optimization, the standard simulation files fix the incoming flow direction as due west, and use the incoming wind speed and wind turbine yaw angle as adjustable parameters. When executing a specific simulation file, the standard simulation file will be copied and the parameters modified.

[0060] It should be noted that a basic communication module is implemented based on the paramiko library to enable communication with the SOWFA software running in a Docker container on the server via the Secure Shell protocol (SSH). Based on this basic communication module, each step of executing a SOWFA case is automated and encapsulated, including case duplication and parameter setting, preprocessing, numerical computation execution, post-processing, and packaging of output data. This encapsulation results in two classes: ABLConnecter and ALM / ADMConnecter. Since wake calculations depend on atmospheric boundary layer simulation results, the instantiation of the latter requires an instance of the former. Under specific incoming wind speeds, the yaw angle can be changed to obtain different cases; therefore, one ABLConnecter can be applied to multiple ADMConnecters.

[0061] Step S103: Organize the simulation examples, save the simulation results in a distributed file structure, and generate example data. The simulation results include the wake distribution parameters of the wind farm.

[0062] In this step, the simulation results are saved in a distributed file structure. The simulation results include the distribution of various physical parameters in the entire simulation domain at various time periods within the simulation set time range. The physical parameters include wake distribution parameters. Based on the distribution of physical parameters, the simulation case is viewed and the flow field is sliced. The slices are saved locally in CSV format, and the case data is obtained by sorting them.

[0063] In some implementations, step S103 includes:

[0064] The wake distribution parameters are stored in a distributed file structure for each time period within the preset simulation time range;

[0065] Based on the wake distribution parameters, the simulation example is sliced ​​to obtain slice data;

[0066] The sliced ​​data is generated into the example data in CSV format.

[0067] In this embodiment, after simulating the preset wind farm data based on the SOWFA simulation platform, the simulation results are saved in the cloud storage using a distributed file structure. This file structure includes the distribution of various physical parameters across the entire simulation domain at different time segments within the simulation's set time range. For further analysis and processing, the data needs to be organized, packaged, and retrieved locally. Locally, the data is viewed and the flow field is sliced ​​using the third-party software Paraview. The slices are stored locally in CSV format. Local storage facilitates faster data retrieval, thereby improving computational speed and reducing computational time costs.

[0068] Step S104: Based on the dataset constructed from the example data, train the preset autoencoder model to obtain a single-fan wake surrogate model.

[0069] In this step, the example data is constructed into a dataset, and the dataset is input into the preset autoencoder model for training until the preset autoencoder model reaches the preset convergence condition, thereby obtaining a single-fan wake surrogate model with the incoming wind speed and yaw angle as control parameters and the wake distribution parameters as output results.

[0070] Optionally, constructing the example data into a dataset includes:

[0071] Based on a preset editing script, the example data is processed to generate a tensor file;

[0072] The tensor file is used to construct the dataset.

[0073] In this embodiment, a large number of simulation examples are retrieved locally, and the simulation examples are post-processed using a preset Python script to generate tensor files and save them.

[0074] In some implementations, the step of inputting the dataset into the preset autoencoder model for training until the preset autoencoder model reaches a preset convergence condition, thereby obtaining a single-fan wake surrogate model with the incoming wind speed and yaw angle as control parameters and the wake distribution parameters as output results, includes:

[0075] Based on the encoder and decoder of the preset autoencoder model, the feature function of the dataset is determined.

[0076] The feature function is mapped to the hidden vector space, and linear evolution features are generated under the influence of the control parameters.

[0077] The linear evolution features are combined with wind farm data from the dataset to train a feature embedder;

[0078] The feature embedder is mapped into the linear vector space required by the encoder to obtain the wake distribution parameters;

[0079] Based on the wake distribution parameters, the preset autoencoder model is updated until the preset autoencoder model reaches the preset convergence condition, thereby generating the single-fan wake proxy model.

[0080] In this embodiment, after the dataset is constructed, an Encoder / Decoder pair is trained to find a set of feature functions for the original wind farm data (ALM atmospheric boundary layer simulation and ADM wind turbine wake simulation data), mapping them to a linear vector space that exhibits linear evolution characteristics under the influence of control parameters. Simultaneously, this autoencoder model trains an Embedder for the yaw angle and incoming wind speed, which serve as control parameters, mapping their influence to the linear vector space sought by the Encoder. This superimposes the influence of control parameters onto the ABL atmospheric boundary layer conditions, yielding wake distribution parameters. After training, a surrogate model is obtained with incoming wind speed and yaw angle as control variables and the near-field wake distribution of the wind turbine as the output.

[0081] Optionally, in the network communication part of this embodiment, the SSH network communication function is set up using the jsch communication library, the configuration file is sent through sftp file transfer, remote connection and remote command invocation are achieved through SSH command form, and the results are pulled to the local machine.

[0082] Optionally, in the file operation part of this embodiment, file operation utility classes are set up using Apache's open-source utility library to implement basic file operations such as compression, decompression (packaging and unpacking of distributed test case result files), and adding, deleting, and modifying files.

[0083] Optionally, in this embodiment, the inter-process communication part utilizes Java's process-related libraries to set up the process module. By combining input / output stream redirection with file-based interaction, operations such as calling the Paraview third-party software and executing training code written in Python are achieved.

[0084] Furthermore, to prevent interface blocking and improve the user experience, a multi-threaded approach is adopted when performing time-consuming tasks.

[0085] For example, the program starts running, continuously checks the server connectivity, submits a computational example, sets computational example parameters on the computational example submission interface, and sends a computational example submission command to the cloud. The interface script running on the server executes the computational example command, submits the computational example to the SOWFA platform for calculation, and returns the progress in real time.

[0086] Click the "Local Processing" button to open the local processing interface. Click the "Read Examples" button to retrieve simulation examples from the cloud and local machine, which will be displayed in lists on the left and right sides. Click the "Pull" button, and the system will compare the server and local example libraries, pulling any missing items to the local machine. Clicking other buttons will perform subsequent processing steps in sequence, generating tensor files for model training. Select the local example you want to preview and click the "Preview" button to view the distribution plot.

[0087] Click the "Model Training" button to enter the model training interface, where you can set the model training parameters. After training begins, the error graph on the right will be continuously updated, and the training progress will be updated in the status bar below. Once training is complete, the model will be persistently saved as a file in the project folder. The results visualization interface allows you to view the model training results; input the required parameters to view the wake distribution under that condition.

[0088] It should be noted that, in this embodiment of the application, a dataset is generated by organizing simulation examples, and the autoencoder model is trained using the dataset to learn the wake characteristics of a single wind turbine under different incoming wind speeds and yaw conditions, thereby constructing an efficient surrogate model, and the wind turbine parameters are quickly optimized based on the surrogate model.

[0089] To implement the training method for the single-fan wake proxy model corresponding to the above method embodiments, in order to achieve the corresponding functions and technical effects. See also Figure 2 , Figure 2 This diagram illustrates a structural block diagram of a training device for a single-fan wake proxy model according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The training device for a single-fan wake proxy model provided in this embodiment includes:

[0090] Extraction module 201 is used to extract standard calculation examples from preset wind farm data, wherein the wind farm data includes the incoming wind speed and yaw angle of a single wind turbine;

[0091] The generation module 202 is used to configure the parameters of the standard case and generate a simulation case;

[0092] The processing module 203 is used to process the simulation examples, save the simulation results in a distributed file structure, and generate example data. The simulation results include the wake distribution parameters of the wind farm.

[0093] Training module 204 is used to train a preset autoencoder model on a dataset constructed based on the example data to obtain a single wind turbine wake proxy model.

[0094] In some implementations, the extraction module 201 is used to:

[0095] Acquire the preset wind farm data and input the preset wind farm data into the SOWFA simulation platform;

[0096] Based on the SOWFA simulation platform, the preset wind farm data is used to extract calculation examples to obtain the standard calculation examples.

[0097] In some implementations, the generation module 202 is used to:

[0098] Determine the direction of the incoming flow from a single fan in the standard example;

[0099] Based on the incoming flow direction, the incoming wind speed and yaw angle of the single fan in the standard example are configured to generate simulation examples with different parameter configurations.

[0100] In some implementations, the sorting module 203 is used for:

[0101] The wake distribution parameters are stored in a distributed file structure for each time period within the preset simulation time range;

[0102] Based on the wake distribution parameters, the simulation example is sliced ​​to obtain slice data;

[0103] The sliced ​​data is generated into the example data in CSV format.

[0104] In some implementations, the training module 204 includes:

[0105] A construction unit is used to construct the example data into a dataset;

[0106] The training unit is used to input the dataset into the preset autoencoder model for training until the preset autoencoder model reaches the preset convergence condition, thereby obtaining a single-fan wake surrogate model with the incoming wind speed and yaw angle as control parameters and the wake distribution parameters as output results.

[0107] In some implementations, the building unit is used for:

[0108] Based on a preset editing script, the example data is processed to generate a tensor file;

[0109] The tensor file is used to construct the dataset.

[0110] In some implementations, the training unit is used for:

[0111] Based on the encoder and decoder of the preset autoencoder model, the feature function of the dataset is determined.

[0112] The feature function is mapped to the hidden vector space, and linear evolution features are generated under the influence of the control parameters.

[0113] The linear evolution features are combined with wind farm data from the dataset to train a feature embedder;

[0114] The feature embedder is mapped into the linear vector space required by the encoder to obtain the wake distribution parameters;

[0115] Based on the wake distribution parameters, the preset autoencoder model is updated until the preset autoencoder model reaches the preset convergence condition, thereby generating the single-fan wake proxy model.

[0116] The training device for the single-fan wake proxy model described above can implement the training method for the single-fan wake proxy model in the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining contents of this application embodiment can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0117] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 3 As shown, the computer device 3 of this embodiment includes: at least one processor 30 ( Figure 3 (Only one is shown in the diagram) a processor, a memory 31, and a computer program 32 stored in the memory 31 and executable on the at least one processor 30, wherein the processor 30 executes the computer program 32 to implement the steps in any of the above method embodiments.

[0118] The computer device 3 can be a smartphone, tablet, desktop computer, cloud server, or other computing device. This computer device may include, but is not limited to, a processor 30 and a memory 31. Those skilled in the art will understand that... Figure 3 The computer device 3 is merely an example and does not constitute a limitation on the computer device 3. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0119] The processor 30 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0120] In some embodiments, the memory 31 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 31 may be an external storage device of the computer device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 3. Furthermore, the memory 31 may include both internal and external storage units of the computer device 3. The memory 31 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0121] In addition, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above method embodiments.

[0122] This application provides a computer program product that, when run on a computer device, enables the computer device to execute the steps described in the various method embodiments above.

[0123] In the several embodiments provided in this application, it will be understood that each block in the flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the figures. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved.

[0124] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application for those skilled in the art.

Claims

1. A training method for a single-fan wake surrogate model, characterized in that, include: Standard calculation examples are extracted from preset wind farm data, which includes the incoming wind speed and yaw angle of a single wind turbine. Configure the parameters of the standard case to generate a simulation case; The wake distribution parameters are stored in a distributed file structure for each time period within the preset simulation time range; Based on the wake distribution parameters, the simulation example is sliced ​​to obtain slice data; The sliced ​​data is generated into the example data in CSV format; The dataset constructed based on the example data is used to train the preset autoencoder model to obtain a single-fan wake proxy model.

2. The training method for the single-fan wake surrogate model as described in claim 1, characterized in that, The extraction of standard examples from preset wind farm data includes: Acquire the preset wind farm data and input the preset wind farm data into the SOWFA simulation platform; Based on the SOWFA simulation platform, the preset wind farm data is used to extract calculation examples to obtain the standard calculation examples.

3. The training method for the single-fan wake surrogate model as described in claim 1, characterized in that, The step of configuring parameters for the standard simulation case and generating a simulation case includes: Determine the direction of the incoming flow from a single fan in the standard example; Based on the incoming flow direction, the incoming wind speed and yaw angle of the single fan in the standard example are configured to generate simulation examples with different parameter configurations.

4. The training method for the single-fan wake surrogate model as described in claim 1, characterized in that, The dataset constructed based on the example data is used to train a preset autoencoder model to obtain a single-fan wake surrogate model, including: The example data is used to construct a dataset. The dataset is input into the preset autoencoder model for training until the preset autoencoder model reaches the preset convergence condition, thereby obtaining a single-fan wake surrogate model with the incoming wind speed and yaw angle as control parameters and the wake distribution parameters as output results.

5. The training method for the single-fan wake surrogate model as described in claim 4, characterized in that, The step of constructing the dataset from the example data includes: Based on a preset editing script, the example data is processed to generate a tensor file; The tensor file is used to construct the dataset.

6. The training method for the single-fan wake surrogate model as described in claim 5, characterized in that, The step of inputting the dataset into the preset autoencoder model for training until the preset autoencoder model reaches the preset convergence condition, to obtain a single-fan wake surrogate model with the incoming wind speed and yaw angle as control parameters and the wake distribution parameters as output results, includes: Based on the encoder and decoder of the preset autoencoder model, the feature function of the dataset is determined. The feature function is mapped to the hidden vector space, and linear evolution features are generated under the influence of the control parameters. The linear evolution features are combined with wind farm data from the dataset to train a feature embedder; The feature embedder is mapped into the linear vector space required by the encoder to obtain the wake distribution parameters; Based on the wake distribution parameters, the preset autoencoder model is updated until the preset autoencoder model reaches the preset convergence condition, thereby generating the single-fan wake proxy model.

7. A training device for a single-fan wake surrogate model, characterized in that, include: The extraction module is used to extract standard calculation examples from preset wind farm data, which includes the incoming wind speed and yaw angle of a single wind turbine. The generation module is used to configure the parameters of the standard case and generate simulation cases; The processing module is used to store the wake distribution parameters of each time period within a preset simulation time range in a distributed file structure; based on the wake distribution parameters, the simulation example is sliced ​​to obtain slice data; The sliced ​​data is generated into the example data in CSV format; The training module is used to train a preset autoencoder model on a dataset constructed based on the example data to obtain a single-fan wake surrogate model.

8. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, which, when executed by the processor, implements the training method for the single-fan wake surrogate model as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the training method for the single-fan wake surrogate model as described in any one of claims 1 to 6.

Images