Generative adversarial network-based wind power generation scene generation system and method

Abstract

The invention discloses a wind power generation scene generation system and method based on a generative adversarial network. A wind power generation scene generation system based on a generative adversarial network comprises a data acquisition module used for acquiring actual operation data and environmental data of a wind power plant, the environmental data comprising geographic information and meteorological information; the data preprocessing module is used for preprocessing the actual operation data and the environment data; the generative adversarial network model construction module is used for constructing a generative adversarial network model; the scene generation module is used for inputting the actual operation data and the environment data into the generative adversarial network model to generate a wind power generation scene, and the scene generation module comprises a wind turbine arrangement optimization unit and is used for optimizing arrangement between wind turbines according to a genetic algorithm. And the optimal solution space can be automatically explored, so that the energy output efficiency of the whole wind power plant can be improved.

Description

Technical Field

[0001] The present invention relates to the field of wind power generation scene generation, and in particular to a wind power generation scene generation system and method based on a generative adversarial network. Background Art

[0002] Wind power generation is one of the fastest growing forms of renewable energy in the world today. Effective wind farm design is crucial to improving wind energy efficiency and reducing construction costs. The design and planning of wind farms is a complex project that involves knowledge from multiple fields, such as meteorology, geographic information systems (GIS), fluid mechanics, and engineering optimization. The core challenge of wind farms is how to maximize the efficiency of wind energy resources while reducing construction and operating costs. Specifically, the layout of wind turbines, terrain adaptability, and airflow interference between wind turbines are key issues in the design process. Existing wind farm planning mostly relies on physical simulation and historical data analysis, and the layout of wind turbines has not been improved, which cannot reduce construction and operating costs. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a wind power generation scenario generation system and method based on a generative adversarial network.

[0004] The present invention provides a wind power generation scenario generation system based on a generative adversarial network, which is characterized by comprising the following: Data acquisition module: used to collect actual operation data and environmental data of the wind farm, the environmental data includes geographical information and meteorological information, and the actual operation data includes wind speed, wind direction, number of wind turbines and power generation; Data preprocessing module: used for preprocessing the actual operation data and the environmental data; Generative adversarial network model building module: used to build a generative adversarial network model; A scenario generation module is used to input the actual operation data and the environmental data into the generative adversarial network model to generate a wind power generation scenario. The scenario generation module includes a wind turbine layout optimization unit. The wind turbine layout optimization unit is used to optimize the layout between wind turbines according to a genetic algorithm.

[0005] In a preferred embodiment, the data preprocessing module comprises Data cleaning unit: used for cleaning the actual operation data and the environmental data to remove noise and outliers, wherein outliers are detected by statistical methods and removed or corrected; Data standardization processing unit: used for standardizing the actual operation data and the environmental data; Feature extraction and dimension reduction unit: used to extract features and reduce dimensions of data with dimensions greater than a preset threshold using principal component analysis; Interpolation and rasterization processing unit: used to fill in incomplete or sparse geographic information using interpolation algorithms and perform rasterization processing on terrain data; Data enhancement unit: used to perform data enhancement using a sliding window method to generate virtual data samples under different environmental data.

[0006] In a preferred embodiment, the generative adversarial network model building module includes: A wind power generation scene image generation unit: used to generate a wind power generation scene image containing a variety of terrain features using a generator, wherein the generator adopts a multi-layer convolutional neural network structure; A wind power generation scene image processing unit: used to extract multi-level features of the wind power generation scene image using a discriminator, and judge the authenticity of the wind power generation scene image; Learning rate adjustment unit: used to adjust the learning rate of the generative adversarial network model using the Adam optimizer according to the adversarial results between the generator and the discriminator; Preset wind power generation scenario generation unit: using conditional embedding, by presetting operation data, the generator can generate a wind power generation scenario that meets the preset operation data; In a preferred embodiment, the scene generation module includes: Scenario generation unit: using a generative adversarial network model to generate a wind power generation scenario according to the actual operation data and the environmental data; Simulation unit: Use the Monte Carlo method to simulate and generate random data of the scene, including wind speed fluctuation data and wind direction change data; Wind turbine layout optimization unit: uses genetic algorithm to optimize the layout between wind turbines; In a preferred embodiment, the fan layout optimization unit comprises: Initial population generation subunit: used to generate the initial layout plan of wind turbines; Fitness evaluation subunit: used to evaluate the wind energy utilization and airflow interference degree of each wind turbine layout scheme as the fitness value of the optimization target; Selection and crossover subunit: used to select individuals for crossover operation according to fitness values ​​to generate the next generation layout plan; Mutation operation subunit: used to randomly mutate the positions or heights of some wind turbines to increase the diversity of the population; Convergence judgment subunit: used to judge the convergence information of the genetic algorithm and determine whether the optimization goal is achieved or further iteration is required.

[0007] The present invention provides a method for generating a wind power generation scenario based on a generative adversarial network, which is characterized by comprising the following steps: S1: Collecting actual operation data and environmental data of the wind farm, wherein the environmental data includes geographical information and meteorological information, and the actual operation data includes wind speed, wind direction, number of wind turbines and power generation; S2: preprocessing the actual operation data and the environmental data; S3: Build a generative adversarial network model; S4: Inputting the actual operation data and the environmental data into the generative adversarial network model to generate wind power generation.

[0008] The present invention discloses a wind power generation scene generation system and method based on a generative adversarial network provided by the present invention, which has the following beneficial effects: through the steps of data cleaning, standardization, feature extraction and dimensionality reduction in a data preprocessing module, the data quality in the input constructed model can be effectively improved, which helps to reduce the error in the model training process, thereby generating a more accurate wind power generation scene; by generating virtual data samples under different environments, the original data set can be expanded, so that the model is exposed to more diverse data during the training process; by generating wind power generation scenes, not only realistic images can be created, but also random factors such as wind speed fluctuations and wind direction changes can be added through simulation units, so that the final scene is closer to the actual situation, which is conducive to subsequent planning and decision support; by realizing the optimization of wind turbine layout, the optimal solution space can be automatically explored, and the best wind turbine arrangement method is given after considering factors such as wind energy utilization and airflow interference, which helps to improve the energy output efficiency of the entire wind farm; by creating a three-dimensional scene model, an intuitive visualization platform is provided for users, which is convenient for understanding and analyzing complex wind farm layouts and their working conditions. By optimizing the design and operation of wind farms, it indirectly promotes the development of clean energy, helps reduce fossil fuel consumption and lowers greenhouse gas emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0010] Figure 1 Generate an architecture diagram of a system for wind power generation scenarios based on generative adversarial networks; Figure 2 A structural diagram of a data preprocessing module in a wind power generation scenario generation system based on a generative adversarial network; Figure 3 A structural diagram of a generative adversarial network model building module in a wind power generation scenario generation system based on a generative adversarial network; Figure 4 This is a structural diagram of a scenario generation module in a wind power generation scenario generation system based on a generative adversarial network; Figure 5 The structural diagram of the wind turbine optimization unit in the scenario generation module; Figure 6 This is a flow chart of a method for generating wind power scenarios based on a generative adversarial network. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0012] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0013] refer to Figure 1 , a wind power generation scenario generation system based on a generative adversarial network, comprising the following: Data acquisition module 11: used to collect actual operation data and environmental data of the wind farm, the environmental data includes geographic information and meteorological information; the actual operation data includes wind speed, wind direction, number of wind turbines and power generation; among them, the GIS system is used to collect geographic information including terrain characteristics, vegetation distribution and other data, and meteorological information includes temperature, humidity and other meteorological data, combined with the meteorological station and real-time weather data interface to ensure the real-time and reliability of meteorological information.

[0014] Data preprocessing module 12: used for preprocessing actual operation data and environmental data; Generative adversarial network model building module 13: used to build a generative adversarial network model; The scenario generation module 14 inputs the actual operation data and the environmental data into the generative adversarial network model to generate a wind power generation scenario. The scenario generation module includes a wind turbine layout optimization unit, which is used to optimize the layout between wind turbines according to a genetic algorithm.

[0015] The present invention provides a wind power generation scenario generation system based on a generative adversarial network, which can effectively improve the data quality in the input constructed model through steps such as data cleaning, standardization, feature extraction and dimensionality reduction in the data preprocessing module, which helps to reduce errors in the model training process, thereby generating a more accurate wind power generation scenario; by realizing wind turbine layout optimization, it can automatically explore the optimal solution space, and provide the best wind turbine arrangement after considering factors such as wind energy utilization and airflow interference, which helps to improve the energy output efficiency of the entire wind farm.

[0016] refer to Figure 2 In a preferred embodiment, the data preprocessing module 12 includes Data cleaning unit 121: cleans the actual operation data and environmental data to remove noise and outliers, wherein outliers are detected through statistical methods such as standard deviation analysis, and outliers are removed or corrected to ensure data quality.

[0017] Standardization processing unit 122: standardizes the actual operation data and environmental data so that they are input into the generative adversarial network at a uniform scale.

[0018] Feature extraction and dimension reduction unit 123: principal component analysis is used to extract features and reduce dimensions of data with dimensions greater than a preset threshold, so as to reduce redundant information and improve computational efficiency. In addition, time series feature extraction is used to analyze dynamic characteristics of wind speed, wind direction, etc. that change over time.

[0019] Interpolation and rasterization processing unit 124: uses an interpolation algorithm such as Kriging interpolation to fill in incomplete or sparse geographic information and performs rasterization processing on terrain data to ensure that the data can adapt to the input format of the deep learning model.

[0020] Data enhancement unit 125: A sliding window method is introduced to perform data enhancement to generate virtual data samples under various environmental conditions such as wind speed, wind direction, and humidity, thereby improving the generalization ability of GAN and its adaptability to different environments.

[0021] refer to Figure 3 In a preferred embodiment, the generative adversarial network model building module 13 includes: Wind power generation scene image generation unit 131: used to use a generator to generate a wind power generation scene image containing multiple terrain features; wherein the generator adopts a multi-layer convolutional neural network structure to generate a wind power generation scene image containing multiple terrain features; a residual network is introduced during the generation process to enhance the detail expression ability of the image. The generator input is random noise and specific conditional parameters such as terrain type and wind speed, and the output is a simulated wind power generation scene.

[0022] Wind power generation scene image processing unit 132: using a discriminator to extract multi-level features of the wind power generation scene image and judge the authenticity of the image; the discriminator uses a method combining a convolutional layer and a fully connected layer to extract multi-level features of the image and judge the authenticity of the image; introducing discriminator regularization such as spectral normalization to improve the stability of the adversarial process and prevent overfitting; Dynamic learning rate adjustment unit 133: Using the Adam optimizer, the learning rate of the generative adversarial network model is adjusted according to the adversarial results between the generator and the discriminator, thereby speeding up the speed, introducing a smooth label strategy, enhancing the robustness of the discriminator to the generated images, and preventing the discriminator from being too powerful and causing the generator to be unable to be effectively trained.

[0023] Preset wind power generation scene generation unit 134: Using conditional embedding, the generator generates wind power generation scenes that meet the preset conditions through preset site conditions such as wind speed, wind direction and terrain features, so as to improve the diversity and pertinence of the generated images. Using conditional embedding, the conditional parameters input by the user are embedded into the middle layer of the network to improve the quality of the generated images.

[0024] refer to Figure 4 In a preferred embodiment, the scene generation module 14 comprises: Scenario generation unit 141: Generates wind power generation scenarios based on actual operation data and environmental data using a generative adversarial network model; wherein the user can set the actual operation data and environmental data through a graphical interface, and a corresponding scenario will be generated based on the set conditions.

[0025] Simulation unit 142: Use the Monte Carlo method to simulate and generate random data of the scene, including wind speed fluctuation data and wind direction change data; to be closer to the actual scene; Monte Carlo simulation is used to evaluate the performance of wind turbines under different conditions and help users predict the long-term performance of power generation scenarios.

[0026] Wind turbine layout optimization unit 143: uses a genetic algorithm to optimize the layout of wind turbines; simulates the natural selection process through the genetic algorithm to iteratively optimize the wind turbine positions to improve the utilization rate of wind resources and reduce mutual interference between wind turbines.

[0027] Visualization unit 144: Generate a three-dimensional scene model of the wind farm using OpenGL. Users can rotate and zoom the entire wind farm through visualization tools to more intuitively evaluate the rationality of the wind turbine layout. The three-dimensional visualization can also display the airflow distribution under different wind speeds and wind directions to help users better understand the interaction between wind turbines.

[0028] refer to Figure 5 In a preferred embodiment, the fan arrangement optimization unit 143 includes Initial population generation subunit 1431: used to generate an initial layout plan for wind turbines, ensure population diversity, and provide a good starting point for optimization.

[0029] Fitness evaluation subunit 1432: used to evaluate the wind energy utilization and airflow interference degree of each wind turbine layout scheme as the fitness value of the optimization target; wherein, computational fluid dynamics simulation is used to accurately simulate the airflow interference between wind turbines during the fitness evaluation process, and the mathematical model of wind energy utilization is combined to quantify the overall efficiency of each layout scheme to ensure the scientificity and accuracy of the evaluation results.

[0030] The selection and crossover subunit 1433 is used to select individuals for crossover operation according to the fitness value and generate the next generation layout scheme to improve the overall wind energy utilization rate.

[0031] Mutation operation subunit 1434: used to randomly mutate the positions or heights of some wind turbines to increase the diversity of the population and prevent falling into a local optimum.

[0032] Convergence judgment subunit 1435: used to judge the convergence information of the genetic algorithm, determine whether the optimization goal is achieved or further iteration is required. If the rate of change of fitness is lower than a preset threshold in multiple consecutive iterations, it is considered that the algorithm has converged.

[0033] refer to Figure 6 The present invention provides a method for generating a wind power generation scenario based on a generative adversarial network, comprising the following steps: S1: Collect the actual operation data and environmental data of the wind farm. The environmental data includes geographic information and meteorological information, and the actual operation data includes wind speed, wind direction, number of wind turbines and power generation; S2: used to pre-process actual operation data and environmental data; S3: Build a generative adversarial network model; S4: Input the actual operation data and environmental data into the generative adversarial network model to generate wind power generation scenarios.

[0034] The present invention discloses a wind power generation scene generation system and method based on a generative adversarial network, which has the following beneficial effects: through the steps of data cleaning, standardization, feature extraction and dimensionality reduction in the data preprocessing module, the data quality in the input constructed model can be effectively improved, which helps to reduce errors in the model training process, thereby generating more accurate wind power generation scenes; by generating virtual data samples in different environments, the original data set can be expanded, so that the model is exposed to more diverse data during the training process; by generating wind power generation scenes, not only realistic images can be created, but also random factors can be added through simulation units Such as wind speed fluctuations and wind direction changes, so that the final scene is closer to the actual situation, which is conducive to subsequent planning and decision support; by realizing wind turbine layout optimization, this method can automatically explore the optimal solution space, and give the best wind turbine arrangement after considering factors such as wind energy utilization and airflow interference, which helps to improve the energy output efficiency of the entire wind farm; by creating a three-dimensional scene model, it provides users with an intuitive visualization platform to facilitate understanding and analysis of complex wind farm layouts and their working conditions; by optimizing the design and operation of wind farms, it indirectly promotes the development of clean energy, helps to reduce fossil fuel consumption and reduce greenhouse gas emissions.

Claims

1. A wind power generation scenario generation system based on generative adversarial network, characterized in that: These include: Data acquisition module: used to collect actual operation data and environmental data of the wind farm, the environmental data includes geographical information and meteorological information, and the actual operation data includes wind speed, wind direction, number of wind turbines and power generation; Data preprocessing module: used for preprocessing the actual operation data and the environmental data; Generative adversarial network model building module: used to build a generative adversarial network model; A scenario generation module is used to input the actual operation data and the environmental data into the generative adversarial network model to generate a wind power generation scenario. The scenario generation module includes a wind turbine layout optimization unit. The wind turbine layout optimization unit is used to optimize the layout between wind turbines according to a genetic algorithm.

2. A wind power generation scenario generation system based on generative adversarial network according to claim 1, characterized in that: The data preprocessing module includes Data cleaning unit: used for cleaning the actual operation data and the environmental data to remove noise and outliers, wherein outliers are detected by statistical methods and removed or corrected; Data standardization processing unit: used for standardizing the actual operation data and the environmental data; Feature extraction and dimension reduction unit: used to extract features and reduce dimensions of data with dimensions greater than a preset threshold using principal component analysis; Interpolation and rasterization processing unit: used to fill in incomplete or sparse geographic information using interpolation algorithms and perform rasterization processing on terrain data; Data enhancement unit: used to perform data enhancement using a sliding window method to generate virtual data samples under different environmental data.

3. A wind power generation scenario generation system based on generative adversarial network according to claim 1, characterized in that: The generative adversarial network model building module includes: A wind power generation scene image generation unit: used to generate a wind power generation scene image containing a variety of terrain features using a generator, wherein the generator adopts a multi-layer convolutional neural network structure; A wind power generation scene image processing unit: used to extract multi-level features of the wind power generation scene image using a discriminator, and judge the authenticity of the wind power generation scene image; Learning rate adjustment unit: used to adjust the learning rate of the generative adversarial network model using the Adam optimizer according to the adversarial results between the generator and the discriminator; Preset wind power generation scenario generation unit: Using conditional embedding, by presetting operating data, the generator can generate a wind power generation scenario that meets the preset operating data.

4. A wind power generation scenario generation system based on generative adversarial network according to claim 3, characterized in that: The scene generation module comprises: Scenario generation unit: using a generative adversarial network model to generate a wind power generation scenario according to the actual operation data and the environmental data; Simulation unit: Use the Monte Carlo method to simulate and generate random data of the scene, including wind speed fluctuation data and wind direction change data; Wind turbine layout optimization unit: uses genetic algorithm to optimize the layout between wind turbines; Visualization unit: Use OpenGL to generate a three-dimensional scene model of the wind farm.

5. A wind power generation scenario generation system based on generative adversarial network according to claim 4, characterized in that: The fan layout optimization unit comprises: Initial population generation subunit: used to generate the initial layout plan of wind turbines; Fitness evaluation subunit: used to evaluate the wind energy utilization and airflow interference degree of each wind turbine layout scheme as the fitness value of the optimization target; Selection and crossover subunit: used to select individuals for crossover operation according to fitness values ​​to generate the next generation layout plan; Mutation operation subunit: used to randomly mutate the positions or heights of some wind turbines to increase the diversity of the population; Convergence judgment subunit: used to judge the convergence information of the genetic algorithm and determine whether the optimization goal is achieved or further iteration is required.

6. A method for generating wind power generation scenarios based on a generative adversarial network, characterized in that: The steps include: S1: Collecting actual operation data and environmental data of the wind farm, wherein the environmental data includes geographical information and meteorological information, and the actual operation data includes wind speed, wind direction, number of wind turbines and power generation; S2: preprocessing the actual operation data and the environmental data; S3: Build a generative adversarial network model; S4: Inputting the actual operation data and the environmental data into the generative adversarial network model to generate a wind power generation scenario.

Images