Wind power plant arrangement method and system based on deep learning
Through deep learning technology, combining wind speed, wind direction, terrain and transmission line data, wind direction stability and wake effect are calculated, and wind farm layout is optimized, which solves the problem of insufficient evaluation of wind direction stability and wake effect in traditional methods, and improves wind power operational efficiency and economic benefits of the field.
Patent Information
- Application Number
- CN202411275941.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-12
- Publication Date
- 2024-12-17
AI Technical Summary
The traditional wind farm layout method has shortcomings in wind direction stability and wake effect evaluation, which affects the operating efficiency and life of the wind turbine, and fails to fully consider the impact of wake effect on the overall efficiency of the wind farm.
Using a deep learning-based method, by acquiring and preprocessing wind speed, wind direction, terrain and transmission line data, calculate the wind direction stability characteristics and wake effect wind speed attenuation, perform multi-level wind power score calculations, optimize the wind farm layout, and select reasonable ones Fan position to reduce wake interference.
The operation efficiency and economic benefits of the wind farm are improved, and the layout of the wind farm is optimized through more sophisticated wind direction stability evaluation and precise wake effect simulation, and the scientificity and rationality of the wind farm are enhanced.
Smart Images
Figure CN119150495A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wind farm layout, and in particular to a wind farm layout method and system based on deep learning. Background Art
[0002] Wind farm layout is a key link in the development of wind power projects. It aims to find areas with abundant wind energy resources, stable wind direction, suitable topography and good construction conditions to ensure that the wind farm can operate efficiently, safely and sustainably. Traditional wind farm layout methods mainly rely on comprehensive consideration of multiple factors such as meteorological data, topographic analysis, geological exploration and socio-economic conditions assessment. However, traditional wind farm layout methods still have shortcomings in the evaluation of wind direction stability and wake effect. The stability of wind direction directly affects the operating efficiency and life of wind turbines. Although traditional methods can evaluate wind direction, they often focus on the dominant wind direction and do not consider the stability of wind direction enough. In terms of wake effect, traditional methods mostly consider the impact of wake effect on power generation, but often only make rough estimates in location selection, and fail to fully consider the impact of wake effect on wind turbine layout and the overall efficiency of wind farms. In recent years, deep learning technology, with its powerful data processing and pattern recognition capabilities, has provided a possible solution to this problem. By analyzing massive amounts of meteorological data through deep learning technology, subtle changes in wind direction and stability characteristics can be captured, providing wind farms with more detailed wind direction stability assessments. At the same time, it can accurately simulate the dynamic changes of wake effects, and then adjust the spacing and orientation of units to minimize wake interference and improve the overall operating efficiency and economic benefits of wind farms. Therefore, it is necessary to introduce deep learning technology to optimize wind farm layout methods and improve the rationality and scientificity of the layout. Summary of the invention
[0003] In view of the deficiencies in the prior art, the present invention provides a method and system for arranging a wind farm based on deep learning; wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data are obtained, and preprocessed to obtain preprocessed wind speed grid data, wind direction grid data and terrain grid data; the frequencies of each wind direction are counted according to the preprocessed wind direction grid data, the dominant wind direction is determined, and the wind direction stability characteristics are extracted; a first wind power score is calculated according to the preprocessed wind speed grid data, wind direction grid data, terrain grid data and wind direction stability characteristics; the first wind power score is updated according to the transmission line distribution data to obtain a second wind power score; the wind speed attenuation of the wake effect is calculated according to the wind direction stability characteristics, and the second wind power score is updated to obtain a third wind power score; the third wind power score is sorted in positive order, and the grid with the highest third wind power score is selected as the new wind turbine position, and then the wind speed attenuation of the wake effect is updated; this step is repeated until all wind turbine positions are selected. In addition to the traditional wind speed and prevailing wind direction, the present invention also comprehensively considers the influence of wind direction stability, transmission distance cost and wake effect, thereby improving the rationality of wind farm layout.
[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: Obtain wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data of the wind farm, and perform preprocessing to obtain preprocessed wind speed grid data, wind direction grid data and terrain grid data; According to the preprocessed wind direction grid data, the frequency of each wind direction is counted, the dominant wind direction is determined, and the wind direction stability characteristics are extracted; Calculating a first wind power score based on the preprocessed wind speed grid data, wind direction grid data, terrain grid data, and wind direction stability characteristics; According to the transmission line distribution data, the first wind power score is updated to obtain a second wind power score; Calculate the wind speed attenuation of the wake effect according to the wind direction stability characteristics, and update the second wind power score to obtain a third wind power score; The third wind power scores are sorted in positive order, and the grid with the highest third wind power score is selected as the new wind turbine position, and then the wake effect wind speed attenuation is updated; this step is repeated until all wind turbine positions are selected.
[0005] Further, wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data of the wind farm are obtained and preprocessed to obtain preprocessed wind speed grid data, wind direction grid data and terrain grid data, including: Obtain wind speed grid data, wind direction grid data and terrain grid data of the wind farm; the wind speed grid data, wind direction grid data and terrain grid data of the wind farm are numerical scalar data, which respectively represent the wind speed, wind direction angle and elevation value of the grid position; obtain transmission line distribution data, which are in linear form and contain geographic coordinate information of the transmission line; Outliers are removed from wind speed grid data, wind direction grid data, and terrain grid data, and Kriging interpolation method is used to fill in empty grids to obtain preprocessed wind speed grid data, wind direction grid data, and terrain grid data; The transmission line distribution data is converted from Shapefile format to segment set format using the format conversion tool in ArcGIS.
[0006] Furthermore, the frequencies of each wind direction are counted according to the preprocessed wind direction grid data, the dominant wind direction is determined, and the wind direction stability characteristics are extracted, including: Traverse the preprocessed wind direction grid data and count the number of times each wind direction angle appears; The frequency of each wind direction is calculated based on the number of times each wind direction angle appears, and the wind direction with the highest frequency is taken as the dominant wind direction. The calculation method is: in, is the frequency of the i-th wind direction, i is the wind direction index, is the number of times the i-th wind direction angle appears, N is the total number of wind direction angles, B is the dominant wind direction, To obtain the maximum value, To make The value of i when it reaches the maximum; Calculate the standard deviation of wind direction angles , calculated as: in, is the angle value of the i-th wind direction angle, is the weighted average of wind direction angles; Construct a wind direction feature extraction network to extract wind direction stability features. The calculation method is: in, is the wind direction encoding feature, For the full convolutional network operation, is the angle value of the wind direction angle of the nth grid in the preprocessed wind direction grid data, n is the grid index, is the wind direction stability characteristic, A convolutional neural network operation is performed, wherein the size of the feature map output by the convolutional neural network operation is the same as the size of the wind direction grid data; The present invention first determines the dominant wind direction according to the wind direction frequency, and then calculates the standard deviation of the wind direction angle to quantify the volatility of the wind direction. The standard deviation of the wind direction angle is then used as an enhanced feature and input into a fully convolutional network as an encoder. The wind direction encoding feature output by the fully convolutional network is then input into a convolutional neural network to process the spatial correlation of the raster data and decode to obtain the wind direction stability feature. Compared with the traditional wind direction stability evaluation method based on statistics, the present invention uses deep learning technology for feature fusion, which can better represent the nonlinear change law of the wind direction stability feature. The wind direction stability feature and the dominant wind direction calculated in this step will also be used as input data for subsequent wind power scoring calculations, enhancing the model's understanding of wind direction changes in wind farms.
[0007] Further, a first wind power score is calculated according to the preprocessed wind speed grid data, wind direction grid data, terrain grid data, and wind direction stability characteristics; including: The pre-processed wind speed grid data and terrain grid data are superimposed, and the wind speed is corrected and calculated using the terrain factor correction method to obtain the corrected wind speed grid data; The corrected wind speed grid data is superimposed with the preprocessed wind direction grid data, and combined with the dominant wind direction, the preliminary wind power score is calculated as follows: in, is the absolute value of the angle between the nth grid wind direction and the dominant wind direction, To obtain the minimum value, is the wind direction deviation characteristic, For the preliminary wind power rating, is the fully connected layer operation, is the wind speed in the nth grid in the corrected wind speed grid data, is the maximum wind speed in the corrected wind speed grid data; Compared with the traditional method of evaluating wind power potential by relying on single wind speed data or empirical formula, this step first corrects the wind speed grid data through terrain factors to make the wind speed data closer to the real environment; secondly, it organically integrates multi-source data such as wind speed grid data, wind direction grid data, and wind direction stability characteristics, calculates the wind direction deviation by the angle between the wind direction and the dominant wind direction, and combines it with a deep learning model (fully connected layer and convolutional neural network) to enable the model to perform adaptive learning based on a large amount of data, thus realizing a comprehensive evaluation of multi-dimensional information; at the same time, the calculated preliminary wind power score also provides a data basis for the subsequent calculation of the first, second, and third wind power scores; According to the preliminary wind power score, combined with the wind direction stability characteristics, the first wind power score is calculated as follows: in, Score the First Wind Power.
[0008] Further, the first wind power score is updated according to the transmission line distribution data to obtain a second wind power score, including: Based on the distribution data of transmission lines, the buffer zone analysis tool in ArcGIS is used to create transmission line buffer zones. Based on the transmission line buffer, the transmission cost score is calculated as follows: in, Score the transmission cost, is a natural constant, is the distance coefficient, is the straight-line distance from the target grid to the transmission line, is the radius of the transmission line buffer; According to the transmission cost score, combined with the first wind power score, the second wind power score is calculated as follows: in, Scoring the second wind power, is the distance penalty function, is the penalty coefficient; In this step, the present invention introduces the transmission cost score to comprehensively consider the spatial relationship between wind farm construction and transmission lines; in the grid within the transmission line buffer zone, wind turbines cannot be built in this area due to electromagnetic interference, safety distance requirements and other factors, so The exponential decay function gives a larger negative cost score, and for cells outside the buffer zone, the transmission wiring cost gradually increases with the increase of distance. , taking the cost into consideration; when calculating the second wind power score, by combining the first wind power score and the transmission cost score and adjusting it using the distance penalty function, the wind power score is closer to the actual situation.
[0009] Furthermore, the wind speed attenuation of the wake effect is calculated according to the wind direction stability characteristics, and the second wind power score is updated to obtain a third wind power score, including: The Jensen model is selected as the wake effect evaluation model, and the wind speed attenuation of the wake effect is calculated according to the wind direction stability characteristics. The calculation method is: in, is the wind speed attenuation due to the wake effect, is the wake influence coefficient, is the wake diffusion coefficient, is the downstream distance, is the center position of the wind wheel; According to the wind speed attenuation of the wake effect, the second wind power score is updated to obtain the third wind power score, which is calculated as follows: in, Scoring the third wind power; Traditional methods usually ignore or simplify the calculation of wake effect, or only use static wind speed attenuation models, which cannot accurately reflect the complex interactions between wind turbines in a wind farm. The present invention introduces the Jensen model and combines the wind direction stability characteristics to calculate the wind speed attenuation of the wake effect. It not only considers the physical process of wake diffusion, but also incorporates the influence of wind direction changes on the wake effect, realizes a comprehensive evaluation of multi-dimensional information, and improves the accuracy and dynamics of the wake effect evaluation. Based on the third wind power score, the wind energy resource potential and the wake effect impact of each location in the wind farm can be more accurately evaluated, thereby optimizing the layout design of the wind farm, reducing the mutual influence between wind turbines, and improving the overall power generation efficiency.
[0010] The present invention also discloses a wind farm layout system based on deep learning, comprising: Data pre-operation unit: obtains wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data of the wind farm, and performs pre-processing to obtain pre-processed wind speed grid data, wind direction grid data and terrain grid data; Wind direction stability feature extraction unit: count the frequency of each wind direction according to the preprocessed wind direction grid data, determine the dominant wind direction, and extract the wind direction stability feature; A first wind power score calculation unit: calculating a first wind power score according to the preprocessed wind speed grid data, wind direction grid data, terrain grid data, and wind direction stability characteristics; A second wind power score calculation unit: updating the first wind power score according to the transmission line distribution data to obtain a second wind power score; A third wind power score calculation unit: calculates the wind speed attenuation of the wake effect according to the wind direction stability characteristics, and updates the second wind power score to obtain a third wind power score; Iterative calculation unit: sort the third wind power scores in positive order, select the grid with the highest third wind power score as the new wind turbine position, and then update the wake effect wind speed attenuation; repeat this step until all wind turbine positions are selected.
[0011] The present invention has the following beneficial effects: 1. The present invention integrates multi-source data such as wind speed grid data, wind direction grid data, wind direction stability and transmission line cost, designs a multi-level wind power scoring calculation unit, and gradually refines the wind power scoring to ensure that various factors affecting the location of the wind power station can be fully considered; at the same time, when selecting the location of each new wind turbine, the wake effect of the existing wind turbines is considered, realizing a comprehensive consideration of the site selection of the wind power station.
[0012] 2. The present invention proposes a wind direction stability feature extraction unit, which determines the dominant wind direction by counting the frequencies of each wind direction, and then calculates the standard deviation of the wind direction angle to quantify the volatility of the wind direction. The standard deviation of the wind direction angle is used as an enhanced feature and input into a fully convolutional network as an encoder. The wind direction encoding feature output by the fully convolutional network is then input into a convolutional neural network to process the spatial correlation of the raster data and decode it to obtain the wind direction stability feature to evaluate the stability of wind resources.
[0013] 3. In the third wind power scoring calculation unit, the wake effect between wind turbines is taken into account, and the wind speed attenuation is calculated based on the wind direction stability characteristics. Not only the physical process of wake diffusion is considered, but also the influence of wind direction change on the wake effect is incorporated, realizing a comprehensive evaluation of multi-dimensional information, improving the prediction accuracy of wind resource distribution within the wind farm, and reducing the negative impact of the wake effect on the power generation efficiency of the wind farm. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 It is a schematic diagram of the flow of the wind farm layout method based on deep learning of the present invention; Figure 2 This is a schematic diagram of the wind direction stability feature extraction algorithm flow of the present invention; Figure 3 It is a schematic diagram of the third wind power score calculation process of the present invention. DETAILED DESCRIPTION
[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.
[0016] Embodiment 1: A wind farm layout method based on deep learning provided by the present invention, such as Figure 1 As shown, including: Step 1: Obtain wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data of the wind farm within unit time, and preprocess them to obtain preprocessed wind speed grid data, wind direction grid data and terrain grid data; the unit time can be determined according to the required design time, which can be 1 year or 2 years. It is not specifically limited in the present invention and is only for demonstration.
[0017] The step 1 includes the following contents: Step 101, obtaining wind speed grid data, wind direction grid data and terrain grid data of the wind farm; the wind speed grid data, wind direction grid data and terrain grid data of the wind farm are numerical scalar data, respectively representing the wind speed, wind direction angle and elevation value of the grid position; obtaining transmission line distribution data, the transmission line distribution data is in linear form, and contains geographic coordinate information of the transmission line; Step 102, removing outliers from the wind speed grid data, wind direction grid data, and terrain grid data, and using Kriging interpolation to fill in empty grids to obtain pre-processed wind speed grid data, wind direction grid data, and terrain grid data; Step 103: The transmission line distribution data is converted from the Shapefile format to the line segment set format using a format conversion tool in ArcGIS. Step 2: Count the frequencies of each wind direction based on the preprocessed wind direction grid data, determine the dominant wind direction, and extract the wind direction stability characteristics; The step 2 includes the following contents: Figure 2 As shown: Step 201, traverse the pre-processed wind direction grid data, and count the number of times each wind direction angle appears; Step 202: Calculate the frequency of each wind direction according to the number of times each wind direction angle appears, and take the wind direction with the highest frequency as the dominant wind direction. The calculation method is: in, is the frequency of the i-th wind direction, i is the wind direction index, is the number of times the i-th wind direction angle appears, N is the total number of wind direction angles, B is the dominant wind direction, To obtain the maximum value, To make The value of i when it reaches the maximum; Step 203: Calculate the standard deviation of the wind direction angle , calculated as: in, is the angle value of the i-th wind direction angle, is the weighted average of wind direction angles; Step 204: construct a wind direction feature extraction network to extract wind direction stability features. The calculation method is: in, is the wind direction encoding feature, For the full convolutional network operation, is the angle value of the wind direction angle of the nth grid in the preprocessed wind direction grid data, n is the grid index, is the wind direction stability characteristic, A convolutional neural network operation is performed, wherein the size of the feature map output by the convolutional neural network operation is the same as the size of the wind direction grid data; The present invention first determines the dominant wind direction according to the wind direction frequency, and then calculates the standard deviation of the wind direction angle to quantify the volatility of the wind direction. The standard deviation of the wind direction angle is then used as an enhanced feature and input into a fully convolutional network as an encoder. The wind direction encoding feature output by the fully convolutional network is then input into a convolutional neural network to process the spatial correlation of the raster data and decode to obtain the wind direction stability feature. Compared with the traditional wind direction stability evaluation method based on statistics, the present invention uses deep learning technology for feature fusion, which can better represent the nonlinear change law of the wind direction stability feature. The wind direction stability feature and the dominant wind direction calculated in this step will also be used as input data for subsequent wind power scoring calculations, enhancing the model's understanding of wind direction changes in wind farms.
[0018] Step 3: Calculate a first wind power score based on the preprocessed wind speed grid data, wind direction grid data, terrain grid data, and wind direction stability characteristics; The step three includes the following contents: Step 301, superimposing the pre-processed wind speed grid data and terrain grid data, and using the terrain factor correction method to correct the wind speed, to obtain the corrected wind speed grid data; Step 302: Superimpose the corrected wind speed grid data with the preprocessed wind direction grid data, and calculate the preliminary wind power score in combination with the dominant wind direction. The calculation method is: in, is the absolute value of the angle between the nth grid wind direction and the dominant wind direction, To obtain the minimum value, is the wind direction deviation characteristic, For the preliminary wind power rating, is the fully connected layer operation, is the wind speed in the nth grid in the corrected wind speed grid data, is the maximum wind speed in the corrected wind speed grid data; Compared with the traditional method of evaluating wind power potential by relying on single wind speed data or empirical formula, this step first corrects the wind speed grid data through terrain factors to make the wind speed data closer to the real environment; secondly, it organically integrates multi-source data such as wind speed grid data, wind direction grid data, and wind direction stability characteristics, calculates the wind direction deviation by the angle between the wind direction and the dominant wind direction, and combines it with a deep learning model (fully connected layer and convolutional neural network) to enable the model to perform adaptive learning based on a large amount of data, thus realizing a comprehensive evaluation of multi-dimensional information; at the same time, the calculated preliminary wind power score also provides a data basis for the subsequent calculation of the first, second, and third wind power scores; Step 303: Calculate the first wind power score based on the preliminary wind power score and the wind direction stability characteristics. The calculation method is: in, Scoring the First Wind Power; Step 4: Update the first wind power score according to the transmission line distribution data to obtain a second wind power score; The step 4 includes the following contents: Step 401: Create a transmission line buffer using the buffer analysis tool in ArcGIS based on the transmission line distribution data; Based on the transmission line buffer, the transmission cost score is calculated as follows: in, Score the transmission cost, is a natural constant, is the distance coefficient, is the straight-line distance from the target grid to the transmission line, is the radius of the transmission line buffer; Step 402: Calculate a second wind power score based on the transmission cost score and the first wind power score. The calculation method is: in, Scoring the second wind power, is the distance penalty function, is the penalty coefficient; In this step, the present invention introduces the transmission cost score to comprehensively consider the spatial relationship between wind farm construction and transmission lines; in the grid within the transmission line buffer zone, wind turbines cannot be built in this area due to electromagnetic interference, safety distance requirements and other factors, so The exponential decay function gives a larger negative cost score, and for cells outside the buffer zone, the transmission wiring cost gradually increases with the increase of distance. , taking the cost into consideration; when calculating the second wind power score, by combining the first wind power score and the transmission cost score and adjusting it using the distance penalty function, the wind power score is closer to the actual situation.
[0019] In addition, if the number of newly added wind turbines is greater than 3, a safety distance network should be constructed, and the wind turbine location grid data should be input into it to update the second wind power score. The calculation method of the safety distance network is: Where, u is the safe distance of the wind turbine, Wind turbine location grid data, including the location of each wind turbine in the wind farm; is terrain raster data, is the set of distances between the newly added fans; After calculating the safe distance of the wind turbine, the distance penalty function needs to be Modified to , so that it is affected not only by the transmission cost score, but also by the safety distance between wind turbines, calculated as: in, is the distance penalty function considering the safe distance of the wind turbine, is the safety distance penalty coefficient, The second wind power score considering the safe distance of wind turbines; When calculated After that, in the subsequent iteration process, replace .
[0020] Step 5: Calculate the wind speed attenuation of the wake effect according to the wind direction stability characteristics, and update the second wind power score to obtain a third wind power score; The step five includes the following contents: Step 501: Select the Jensen model as the wake effect evaluation model, and calculate the wake effect wind speed attenuation according to the wind direction stability characteristics. The calculation method is: in, is the wind speed attenuation due to the wake effect, is the wake influence coefficient, is the wake diffusion coefficient, is the downstream distance, is the center position of the wind wheel; Step 502: Figure 3 As shown, according to the wind speed attenuation of the wake effect, the second wind power score is updated to obtain the third wind power score, which is calculated as follows: in, Scoring the third wind power; Traditional methods usually ignore or simplify the calculation of wake effect, or only use static wind speed attenuation models, which cannot accurately reflect the complex interactions between wind turbines in a wind farm. The present invention introduces the Jensen model and combines the wind direction stability characteristics to calculate the wind speed attenuation of the wake effect. It not only considers the physical process of wake diffusion, but also incorporates the influence of wind direction changes on the wake effect, realizes a comprehensive evaluation of multi-dimensional information, and improves the accuracy and dynamics of the wake effect evaluation. Based on the third wind power score, the wind energy resource potential and the wake effect impact of each location in the wind farm can be more accurately evaluated, thereby optimizing the layout design of the wind farm, reducing the mutual influence between wind turbines, and improving the overall power generation efficiency.
[0021] For example, if the prevailing wind direction is 30° north by east (i.e. the wind direction angle is 30°), and the wind direction angle at a certain location is 45° north by east (i.e. The wind speed before correction is 8 m / s. After terrain factor correction, the wind speed becomes 9 m / s. The maximum wind speed in the entire wind farm area after correction is 12 m / s. The wind direction stability characteristic is calculated from the historical wind direction data to be 0.8 (indicating that the wind direction at this location is relatively stable). Then: The calculated preliminary wind power score for this location is 0.6, and because the wind direction at this location is relatively stable, the final first wind power score is 0.7; Then, after considering various factors such as safety distance and electromagnetic interference, the radius of the transmission line buffer is 700 meters, and the straight-line distance from the grid at this location to the transmission line is 800 meters, which is outside the buffer zone. The distance coefficient =0.01, penalty coefficient =0.1, then: Then, according to the Jensen model, the wake influence coefficient C = 0.9, the wake diffusion coefficient γ = 0.02, the downstream distance y = 500 meters, and the center position of the wind wheel = 0 meters, we can calculate: Finally, after MLP and CNN processing and calculation, the third wind power score is obtained. , indicating that the location is more suitable for building a new wind power station, but whether this grid position will be used as the new wind turbine location when building a wind farm still depends on the calculation results of other grids; the grid with the highest third wind power score should be taken as the new wind turbine location.
[0022] Step 6: sort the third wind power scores in positive order, select the grid with the highest third wind power score as the new wind turbine position, and then update the wake effect wind speed attenuation; repeat this step until all wind turbine positions are selected.
[0023] Embodiment 2: The present invention also discloses a wind farm layout system based on deep learning, comprising: Data pre-operation unit: obtains wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data of the wind farm, and performs pre-processing to obtain pre-processed wind speed grid data, wind direction grid data and terrain grid data; Wind direction stability feature extraction unit: count the frequency of each wind direction according to the preprocessed wind direction grid data, determine the dominant wind direction, and extract the wind direction stability feature; A first wind power score calculation unit: calculating a first wind power score according to the preprocessed wind speed grid data, wind direction grid data, terrain grid data, and wind direction stability characteristics; A second wind power score calculation unit: updating the first wind power score according to the transmission line distribution data to obtain a second wind power score; A third wind power score calculation unit: calculates the wind speed attenuation of the wake effect according to the wind direction stability characteristics, and updates the second wind power score to obtain a third wind power score; Iterative calculation unit: sort the third wind power scores in positive order, select the grid with the highest third wind power score as the new wind turbine position, and then update the wake effect wind speed attenuation; repeat this step until all wind turbine positions are selected.
[0024] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.
[0025] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0026] If the functions are implemented in the form of software functional units 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 the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.
[0027] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.
Claims
1. A wind farm layout method based on deep learning, characterized in that: include: Obtain wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data of the wind farm, and perform preprocessing to obtain preprocessed wind speed grid data, wind direction grid data and terrain grid data; According to the preprocessed wind direction grid data, the frequency of each wind direction is counted, the dominant wind direction is determined, and the wind direction stability characteristics are extracted; Calculating a first wind power score based on the preprocessed wind speed grid data, wind direction grid data, terrain grid data, and wind direction stability characteristics; According to the transmission line distribution data, the first wind power score is updated to obtain a second wind power score; Calculate the wind speed attenuation of the wake effect according to the wind direction stability characteristics, and update the second wind power score to obtain a third wind power score; The third wind power scores are sorted in positive order, and the grid with the highest third wind power score is selected as the new wind turbine position, and then the wake effect wind speed attenuation is updated; this step is repeated until all wind turbine positions are selected.
2. The wind farm layout method based on deep learning according to claim 1, characterized in that: Obtain wind speed grid data, wind direction grid data and terrain grid data of the wind farm; the wind speed grid data, wind direction grid data and terrain grid data of the wind farm are all numerical scalar data, respectively representing the wind speed magnitude, wind direction angle and elevation value of the grid position; obtain transmission line distribution data, the transmission line distribution data is in linear form and contains geographic coordinate information of the transmission line; Outliers are removed from wind speed grid data, wind direction grid data, and terrain grid data, and Kriging interpolation method is used to fill in empty grids to obtain preprocessed wind speed grid data, wind direction grid data, and terrain grid data; The transmission line distribution data is converted from Shapefile format to segment set format using the format conversion tool in ArcGIS.
3. The wind farm layout method based on deep learning according to claim 2, characterized in that: Traverse the preprocessed wind direction grid data and count the number of times each wind direction angle appears; The frequency of each wind direction is calculated based on the number of times each wind direction angle appears, and the wind direction with the highest frequency is taken as the dominant wind direction. The calculation method is: in, is the frequency of the i-th wind direction, i is the wind direction index, is the number of times the i-th wind direction angle appears, N is the total number of wind direction angles, B is the dominant wind direction, To obtain the maximum value, To make The value of i when it reaches the maximum; Calculate the standard deviation of wind direction angles , calculated as: in, is the angle value of the i-th wind direction angle, is the weighted average of wind direction angles; Construct a wind direction feature extraction network to extract wind direction stability features. The calculation method is: in, is the wind direction encoding feature, For the full convolutional network operation, is the angle value of the wind direction angle of the nth grid in the preprocessed wind direction grid data, n is the grid index, is the wind direction stability characteristic, It is a convolutional neural network operation, and the size of the feature map output by the convolutional neural network operation is the same as the size of the wind direction grid data.
4. The wind farm layout method based on deep learning according to claim 3, characterized in that: The pre-processed wind speed grid data and terrain grid data are superimposed, and the wind speed is corrected and calculated using the terrain factor correction method to obtain the corrected wind speed grid data; The corrected wind speed grid data is superimposed with the preprocessed wind direction grid data, and combined with the dominant wind direction, the preliminary wind power score is calculated as follows: in, is the absolute value of the angle between the nth grid wind direction and the dominant wind direction, To obtain the minimum value, is the wind direction deviation characteristic, For the preliminary wind power rating, is the fully connected layer operation, is the wind speed in the nth grid in the corrected wind speed grid data, is the maximum wind speed in the corrected wind speed grid data; According to the preliminary wind power score, combined with the wind direction stability characteristics, the first wind power score is calculated as follows: in, Score the First Wind Power.
5. The wind farm layout method based on deep learning according to claim 4, characterized in that: Based on the distribution data of transmission lines, the buffer zone analysis tool in ArcGIS is used to create transmission line buffer zones. Based on the transmission line buffer, the transmission cost score is calculated as follows: in, Score the transmission cost, is a natural constant, is the distance coefficient, is the straight-line distance from the target grid to the transmission line, is the radius of the transmission line buffer; According to the transmission cost score, combined with the first wind power score, the second wind power score is calculated as follows: in, Scoring the second wind power, is the distance penalty function, is the penalty coefficient.
6. The wind farm layout method based on deep learning according to claim 5, characterized in that: The Jensen model is selected as the wake effect evaluation model, and the wind speed attenuation of the wake effect is calculated according to the wind direction stability characteristics. The calculation method is: in, is the wind speed attenuation due to the wake effect, is the wake influence coefficient, is the wake diffusion coefficient, is the downstream distance, is the center position of the wind wheel; According to the wind speed attenuation of the wake effect, the second wind power score is updated to obtain the third wind power score, which is calculated as follows: in, Score the third wind power.
7. A wind farm layout system based on deep learning, characterized in that: include: Data pre-operation unit: obtains wind speed grid data, wind direction grid data, terrain grid data and transmission line distribution data of the wind farm, and performs pre-processing to obtain pre-processed wind speed grid data, wind direction grid data and terrain grid data; Wind direction stability feature extraction unit: count the frequency of each wind direction according to the preprocessed wind direction grid data, determine the dominant wind direction, and extract the wind direction stability feature; A first wind power score calculation unit: calculating a first wind power score according to the preprocessed wind speed grid data, wind direction grid data, terrain grid data, and wind direction stability characteristics; A second wind power score calculation unit: updating the first wind power score according to the transmission line distribution data to obtain a second wind power score; A third wind power score calculation unit: calculates the wind speed attenuation of the wake effect according to the wind direction stability characteristics, and updates the second wind power score to obtain a third wind power score; Iterative calculation unit: sort the third wind power scores in positive order, select the grid with the highest third wind power score as the new wind turbine position, and then update the wake effect wind speed attenuation; repeat this step until all wind turbine positions are selected.