An artificial intelligence-based fan point automatic arrangement method and system

Abstract

The application belongs to the technical field of fan arrangement, and discloses a fan point position automatic arrangement method and system based on artificial intelligence, which comprises the following steps: first, receiving multiple space constraint layers and deployment parameters and unifying coordinates; second, constructing a conflict relation graph based on the processed constraint layers, wherein the graph nodes are different constraint layers, and the edges are the conflict intensity between the constraints; third, inputting the graph into a pre-trained graph attention network to infer a conflict release factor quantifying the relaxable degree of the constraint; fourth, feeding the factor back to a point position search algorithm, and evaluating the candidate point positions by using a scoring function fused with the factor; and fifth, outputting a final arrangement scheme according to the scoring results. The application solves the problem of excessive reduction of areas in traditional methods by using artificial intelligence middleware such as a graph attention network and a search algorithm based on artificial intelligence, and improves the deployable area and arrangement rationality.

Description

Technical Field

[0001] This invention relates to the field of automatic wind turbine location layout technology based on artificial intelligence, specifically, to an automatic wind turbine location layout method and system based on artificial intelligence. Background Technology

[0002] In the development of wind power projects, the scientific layout of wind turbine sites directly affects the project's power generation efficiency, construction feasibility, and safety compliance. Wind turbine layout must simultaneously meet multiple spatial constraints, including project boundary restrictions, avoidance of unusable land, building safety distances, altitude thresholds, mountain top differences, and requirements for turbine elliptical spacing. These constraints can easily lead to local conflicts in spatial distribution, such as overlaps between suitable layout areas and safety distance zones, or overlaps between compliant elevation zones and building avoidance zones, posing significant challenges to site planning.

[0003] Traditional wind turbine site layout methods often employ conventional GIS overlay technology, using a simplistic "intersection-and-removal" logic to handle multiple constraint conflicts, directly deleting entire conflicting areas. This approach fails to consider the adjustability of constraint conflicts, leading to the excessive reduction of many areas with minor conflicts, resulting in a severe underestimation of usable layout area and ultimately sparse and irrationally distributed wind turbine sites. Furthermore, traditional methods lack intelligent judgment and coordination capabilities regarding constraint priority, failing to dynamically adjust constraint requirements based on actual scenarios. This reduces the flexibility and practicality of wind turbine layout, making it difficult to meet the efficient development needs of wind power projects in complex terrain conditions. Summary of the Invention

[0004] In response, the present invention provides an automatic wind turbine location layout method and system based on artificial intelligence, so as to at least partially solve the above-mentioned technical problems.

[0005] This invention provides an automatic wind turbine location layout method based on artificial intelligence, comprising the following steps:

[0006] S1, receive multiple spatial constraint layers and deployment parameters related to the wind turbine layout, and perform coordinate unification processing on each of the spatial constraint layers;

[0007] S2, Based on the processed spatial constraint layer, a conflict relationship graph is constructed, wherein the nodes of the conflict relationship graph represent different types of spatial constraint layers, and the edges of the conflict relationship graph represent the conflict intensity between any two spatial constraint layers.

[0008] S3, input the conflict relationship graph into a pre-trained graph attention network model, and use the model to perform attention allocation and reasoning on the importance of each node in the graph, and calculate the conflict mitigation factor corresponding to each spatial constraint layer. The conflict mitigation factor is used to quantify the degree to which the constraint is allowed to be relaxed during the arrangement process.

[0009] S4, the conflict mitigation factor is fed back to the wind turbine location search algorithm. When generating candidate wind turbine locations, a scoring function that incorporates the conflict mitigation factor is used to evaluate the candidate locations. The scoring function is weighted according to the degree of default of each constraint and its corresponding conflict mitigation factor.

[0010] S5. Based on the evaluation results of the scoring function, search and output the final wind turbine location layout scheme.

[0011] Preferably, in step S2, calculating the conflict intensity between the two spatial constraint layers includes,

[0012] Calculate the mask overlap ratio between the satisfied region of the first constraint layer and the unsatisfied region of the second constraint layer;

[0013] And / or,

[0014] Calculate the average geometric distance between the satisfied region boundary of the first constraint layer and the unsatisfied region boundary of the second constraint layer;

[0015] And / or,

[0016] Calculate the degree of difference between the first constraint layer and the second constraint layer on the corresponding terrain gradient;

[0017] The results of one or more of the above conflict metrics are weighted and summed to comprehensively quantify the conflict intensity.

[0018] Preferably, a node feature vector is constructed for each node, and the node feature vector includes at least one of constraint type encoding, constraint strength value, constraint available area ratio, and elevation and slope features extracted from terrain data.

[0019] Preferably, the graph attention network model takes the feature vector of the node and the conflict intensity of the edge as input, and its inference goal is to output the conflict mitigation factor corresponding to each node. The smaller the conflict mitigation factor, the less likely the constraint represented by the node is to be violated during the arrangement process, and vice versa, it means that it can be relaxed within a certain range.

[0020] Preferably, in step S4, the scoring function specifically involves calculating the degree of default of the candidate wind turbine location on each spatial constraint layer and the corresponding conflict mitigation factor, summing the calculation results for all constraint layers, and using the summation result as the final score of the candidate location to guide the search algorithm in location selection and optimization.

[0021] Preferably, the spatial constraint layer includes at least one of the following: project boundary constraints, unusable land avoidance constraints, safety distance constraints, altitude threshold constraints, mountain top difference constraints, and wind turbine elliptical spacing constraints.

[0022] In another aspect, this application also provides an automatic wind turbine location layout system based on artificial intelligence, comprising:

[0023] The data receiving module is used to receive multiple spatial constraint layers and deployment parameters related to the wind turbine layout, and to perform coordinate unification processing on each of the spatial constraint layers;

[0024] The conflict relationship graph construction module is used to construct a conflict relationship graph based on the processed spatial constraint layer, wherein the nodes of the conflict relationship graph represent different types of spatial constraint layers, and the edges of the conflict relationship graph represent the conflict intensity between any two spatial constraint layers.

[0025] The conflict mitigation factor calculation module is used to input the conflict relationship graph into a pre-trained graph attention network model. Through this model, attention is allocated and reasoned about the importance of each node in the graph, and the conflict mitigation factor corresponding to each spatial constraint layer is calculated. The conflict mitigation factor is used to quantify the degree to which the constraint is allowed to be relaxed during the arrangement process.

[0026] The candidate wind turbine location evaluation module is used to feed back the conflict mitigation factor to the wind turbine location search algorithm. When generating candidate wind turbine locations, a scoring function that incorporates the conflict mitigation factor is used to evaluate the candidate locations. The scoring function is weighted according to the degree of violation of each constraint and its corresponding conflict mitigation factor.

[0027] The wind turbine location layout scheme output module is used to search for and output the final wind turbine location layout scheme based on the evaluation results of the scoring function.

[0028] In another aspect, this application also provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the artificial intelligence-based automatic wind turbine location arrangement method as described above.

[0029] In another aspect, this application provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor to implement the artificial intelligence-based automatic wind turbine location arrangement method described above.

[0030] Another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the artificial intelligence-based automatic wind turbine location arrangement method described above.

[0031] This application achieves multi-constraint collaborative optimization by constructing a conflict relationship graph and combining it with an artificial intelligence middleware, namely a graph attention network. It breaks through the limitation of the traditional GIS overlay method of "intersection and elimination". The conflict relationship graph quantifies the conflict intensity between constraints, and the graph attention network intelligently allocates the priority of constraints and generates conflict mitigation factors, which significantly reduces excessive deletion, significantly increases the deployable area, and solves the problem of no points to place due to local conflicts. At the same time, by replacing binary judgment with a scoring function that integrates conflict mitigation factors, the evaluation of candidate points is more in line with the actual engineering situation, ensuring that the distribution of wind turbines is both safe and reasonable, improving construction feasibility, and providing key technical support for the efficient planning of wind power projects in complex terrain. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0033] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0034] Figure 1 This is a schematic diagram of an automatic wind turbine location arrangement method based on artificial intelligence, provided in an embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of the graph attention network model structure provided in an embodiment of the present invention.

[0036] Figure 3 This is a schematic diagram of the model inference process provided in an embodiment of the present invention.

[0037] Figure 4 This is a schematic diagram illustrating the process of selecting and optimizing candidate wind turbine locations according to an embodiment of the present invention.

[0038] Figure 5 This is a schematic diagram of an automatic wind turbine location layout system based on artificial intelligence, provided in an embodiment of the present invention.

[0039] Figure 6 This is a schematic diagram of the structure of a device provided in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0041] It should be noted that all user information (including but not limited to user device information, user personal information, object information corresponding to device usage data, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, device usage data, etc.) involved in all embodiments of this disclosure are information and data authorized by the user or fully authorized by all parties.

[0042] To address the aforementioned issues, this application proposes an automatic wind turbine location layout method based on artificial intelligence. The technical solution of this application will be described in detail below with reference to various embodiments.

[0043] like Figure 1 As shown in the diagram, this invention discloses a method for automatically arranging wind turbine locations based on artificial intelligence, comprising the following steps:

[0044] S1, receive multiple spatial constraint layers and deployment parameters related to the wind turbine layout, and perform coordinate unification processing on each of the spatial constraint layers;

[0045] S2, Based on the processed spatial constraint layer, a conflict relationship graph is constructed, wherein the nodes of the conflict relationship graph represent different types of spatial constraint layers, and the edges of the conflict relationship graph represent the conflict intensity between any two spatial constraint layers.

[0046] S3, input the conflict relationship graph into a pre-trained graph attention network model, and use the model to perform attention allocation and reasoning on the importance of each node in the graph, and calculate the conflict mitigation factor corresponding to each spatial constraint layer. The conflict mitigation factor is used to quantify the degree to which the constraint is allowed to be relaxed during the arrangement process.

[0047] S4, the conflict mitigation factor is fed back to the wind turbine location search algorithm. When generating candidate wind turbine locations, a scoring function that incorporates the conflict mitigation factor is used to evaluate the candidate locations. The scoring function is weighted according to the degree of default of each constraint and its corresponding conflict mitigation factor.

[0048] S5. Based on the evaluation results of the scoring function, search and output the final wind turbine location layout scheme.

[0049] In some embodiments, for step S1, for example, in this step, the specific categories of multiple spatial constraint layers and deployment parameters related to the wind turbine layout are first clarified to ensure the comprehensiveness and accuracy of the constraint conditions in the subsequent layout process.

[0050] Specifically, the spatial constraint layer includes, but is not limited to, the project layout boundary Shapefile corresponding to the project boundary constraint, the unusable land Shapefile corresponding to the unusable land avoidance constraint, the safety distance reference layer corresponding to the safety distance constraint, the elevation reference layer corresponding to the elevation threshold constraint, the mountain top terrain reference layer corresponding to the mountain top difference constraint, and the wind turbine spacing reference layer corresponding to the wind turbine ellipse spacing constraint.

[0051] The unusable land shapefile can cover various areas unsuitable for wind turbine placement, such as residential areas, factories, water bodies, and photovoltaic zones, storing the spatial range and boundary information of each unusable area in vector data form. The project placement boundary shapefile clarifies the overall geographical range of the wind turbine placement, serving as the basic boundary constraint for subsequent point search. The elevation reference layer provides elevation information within the placement area in raster data form, providing data support for determining elevation threshold constraints. The mountaintop topography reference layer, through processing digital elevation model (DEM) data, marks the location of mountaintops within the area and the elevation differences between mountaintops, used for calculating mountaintop difference constraints. The wind turbine spacing reference layer, combined with the wind turbine design parameters, presets the relevant reference range for the elliptical spacing between wind turbines, providing a basis for wind turbine elliptical spacing constraints.

[0052] Optionally, the deployment parameters include a minimum altitude threshold, a building safety distance threshold, a wind turbine foundation radius, wind turbine ellipse parameters (major axis L, minor axis S, azimuth angle θ), and a mountain peak difference constraint Δh. The minimum altitude threshold is determined based on the wind turbine's operational requirements and the terrain characteristics of the deployment area. For example, if the wind turbine needs to operate in an area with an altitude of at least 500 meters, the minimum altitude threshold is set to 500 meters. The building safety distance threshold needs to refer to relevant industry standards and safety regulations to ensure that the distance between the wind turbine and surrounding buildings, such as residential areas and factories, meets safety requirements; for example, it can be set to 300 meters.

[0053] The radius of the wind turbine foundation is determined based on the wind turbine model and structural parameters. It is used to calculate the space occupied by the wind turbine foundation and avoid spatial conflicts between wind turbine foundations or between the wind turbine foundation and other facilities. In the wind turbine ellipse parameters, the major axis L and minor axis S are determined based on factors such as the wind turbine blade length and the safe operating range. The azimuth angle θ is combined with the prevailing wind direction of the layout area to ensure that the spacing between wind turbines can meet the dual requirements of wind power utilization efficiency and safe operation. The mountain peak difference constraint Δh is set according to the terrain stability of the mountain peak area and the construction requirements of the wind turbine layout. For example, if Δh is set to 50 meters, wind turbines are only allowed to be placed in the area when the elevation difference between two mountain peaks does not exceed 50 meters.

[0054] According to an embodiment of the present invention, after obtaining the aforementioned spatial constraint layers and deployment parameters, coordinate unification processing is performed on each of the spatial constraint layers to eliminate coordinate differences between different layers and ensure the accuracy of subsequent conflict relationship calculations and point search. The specific steps are as follows:

[0055] First, the coordinate system is identified. Specifically, the original coordinate system of each spatial constraint layer is identified to determine the current coordinate system used by each layer. For example, some layers may use the WGS84 geographic coordinate system, while others may use the Beijing 54 coordinate system or the Xi'an 80 coordinate system. Then, computer vision software, such as professional geographic information processing software like ArcGIS and QGIS, is used to read the coordinate information of each Shapefile or raster data, recording key parameters such as the coordinate origin, projection method, and coordinate units of each layer.

[0056] Next, the coordinates are transformed. Specifically, preferably, all spatial constraint layers are uniformly transformed to the Gaussian three-dimensional projection coordinate system. The Gaussian three-dimensional projection coordinate system can effectively reduce projection distortion, is suitable for geographic information processing over large areas, and meets the needs of wind turbine layout, which typically involves a large geographical area.

[0057] During the conversion process, the corresponding Gaussian 3-degree zone number is determined based on the longitude of the central meridian of the area. For example, if the longitude of the central meridian of the area is 117°, the corresponding Gaussian 3-degree zone number is 39. The calculation method for the Gaussian 3-degree zone number is: zone number = (central meridian longitude + 1.5°) / 3°.

[0058] Subsequently, using the coordinate transformation tool in the geographic information processing software, the coordinates of each spatial constraint layer are transformed from the original coordinate system to the target Gaussian three-dimensional projection coordinate system. During the transformation process, it is necessary to ensure the accuracy of the transformation parameters. This can be achieved by selecting multiple common control points within the area to correct the error of the transformation results, so that the coordinate error of each layer after transformation is controlled within the allowable range, such as the plane coordinate error not exceeding 0.5 meters.

[0059] In some embodiments, for step S2, according to the present invention, the conflict relationship graph is represented in the form of graph G=(C,E), where C is the set of nodes in the conflict relationship graph, each node represents a type of spatial constraint layer; E is the set of edges in the conflict relationship graph, each edge represents the conflict intensity between any two spatial constraint layers.

[0060] Specifically, each node Ci in the node set C, where i is the node number (i = 1, 2, ..., n) and n is the number of spatial constraint layers, corresponding to one type of spatial constraint layer. For example, node C1 corresponds to the project boundary constraint layer, node C2 corresponds to the unusable land avoidance constraint layer, node C3 corresponds to the safety distance constraint layer, node C4 corresponds to the altitude threshold constraint layer, node C5 corresponds to the mountain peak difference constraint layer, and node C6 corresponds to the wind turbine elliptical spacing constraint layer. Each edge Eij in the edge set E, where i and j are node numbers (i ≠ j), corresponds to the conflict intensity between the two spatial constraint layers represented by nodes Ci and Cj. Its value range is [0, 1]. The closer the value is to 1, the more severe the conflict between the two spatial constraint layers; the closer the value is to 0, the milder the conflict between the two spatial constraint layers.

[0061] In some embodiments, a node feature vector Fi is constructed for each node Ci. The node feature vector Fi comprehensively describes the attribute information of the corresponding spatial constraint layer, providing feature support for subsequent inference in the graph attention network. The node feature vector Fi includes, but is not limited to, the following feature components:

[0062] Constraint type encoding: A unique type encoding is assigned to each spatial constraint layer. For example, the project boundary constraint is encoded as 1, the unusable land avoidance constraint as 2, the safety distance constraint as 3, the altitude threshold constraint as 4, the mountain peak difference constraint as 5, and the wind turbine elliptical spacing constraint as 6. This encoding is used as the first feature component of the node feature vector Fi, which is then used by the graph attention network to distinguish different types of constraint layers.

[0063] Constraint strength value: The constraint strength value quantifies the importance of the corresponding spatial constraint layer. Its value is determined based on the nature of the constraint and the safety and efficiency requirements of the wind turbine layout. For example, project boundary constraints, as the basic boundary restrictions for wind turbine layout, cannot be breached; their constraint strength value is set to ∞. In actual calculations, an extremely large value can be used, such as... The safety distance constraint is directly related to the safety of surrounding personnel and facilities, and its constraint strength value is set to 1000. The constraint strength values ​​for the altitude threshold constraint, mountain top difference constraint, and wind turbine elliptical spacing constraint are set to 800, 700, and 600, respectively. The constraint strength value for the unusable land avoidance constraint is set to 900. This constraint strength value is used as the second feature component of the node feature vector Fi, reflecting the importance and priority of the corresponding constraint in the layout process.

[0064] Constrained usable area ratio: The constrained usable area ratio is the ratio of the area of ​​the area that meets the constraint requirements in the corresponding spatial constraint layer to the total area of ​​the project layout boundary.

[0065] The specific calculation process is as follows: First, calculate the total area S_total of the project layout boundary using geographic information processing software; then, for each spatial constraint layer Ci, calculate the area S_available_i of the region that meets the constraint requirements in that constraint layer; finally, use the ratio of S_available_i to S_total as the proportion of the constraint available area, that is, the proportion of the constraint available area = S_available_i / S_total.

[0066] Elevation and slope features: Elevation and slope features of the corresponding spatial constraint layer area are extracted from topographic data such as DEM data and used as supplementary feature components of the node feature vector Fi. The elevation feature can be the average elevation value of the area where the constraint layer is located, and the slope feature can be the average slope value of the area where the constraint layer is located.

[0067] Specifically, the DEM data is processed using geographic information processing software. The slope value of each raster cell is calculated through neighborhood analysis. Then, based on the range of the spatial constraint layer Ci, the elevation and slope values ​​of all raster cells within that range are extracted, and their average values ​​are calculated as the elevation and slope feature values.

[0068] For example, if the average elevation of the area where a certain safety distance constraint layer is located is 600 meters and the average slope is 15°, then 600 and 15 are used as the values ​​of elevation and slope features in the feature vector of that node. By introducing elevation and slope features, the graph attention network can more accurately determine the conflict relationship between constraints by combining terrain factors.

[0069] According to an embodiment of the present invention, the conflict intensity between two spatial constraint layers is calculated in step S2. One or more conflict measurement methods can be used, and the measurement results are weighted and summed to comprehensively quantify the conflict intensity.

[0070] 1. Mask overlap ratio

[0071] Specifically, the mask overlap ratio is used to quantify the degree of spatial overlap between the satisfied regions of the first constraint layer and the unsatisfied regions of the second constraint layer. The calculation process is as follows:

[0072] First, constraint masks are constructed. Specifically, constraint masks are constructed for the first constraint layer Ci and the second constraint layer Cj, respectively. and The constraint mask is represented in the form of raster data. The value of the raster cell is 1 or 0. A raster cell with a value of 1 means that the position meets the requirements of the corresponding constraint layer and belongs to the satisfied region. A raster cell with a value of 0 means that the position does not meet the requirements of the corresponding constraint layer and belongs to the unsatisfied region.

[0073] For example, for an altitude threshold constraint layer Ci, if the minimum altitude threshold is 500 meters, the grid cells with an altitude ≥ 500 meters in the DEM data are set to 1, and the grid cells with an altitude < 500 meters are set to 0, thus obtaining the constraint mask. For the unusable land avoidance constraint layer Cj, set the value of the grid cells within the unusable land range to 0, and set the value of the grid cells within the non-unusable land range to 1, to obtain the constraint mask. .

[0074] Then, the area of ​​the overlapping region is calculated, specifically for the constraint mask. The mask for the unsatisfied region of the second constraint layer Cj, i.e., for The mask obtained by performing the inversion operation , The position where the value of the middle grid cell is 1 corresponds to For the locations where the value is 0, representing the unsatisfied areas of the second constraint layer, spatial overlay analysis is performed. The number of grid cells with a value of 1 in both masks is counted. Combined with the area of ​​each grid cell, the area of ​​the overlapping region between the satisfied area of ​​the first constraint layer and the unsatisfied area of ​​the second constraint layer is calculated. .

[0075] Finally, calculate the mask overlap ratio using the following formula. :

[0076]

[0077] Among them, Area ( ) represents the area of ​​the satisfied region of the first constraint layer Ci, i.e., the constraint mask. The total area corresponding to the grid cells with a value of 1. For example, if the area Area(Ci) of the first constraint layer satisfies the condition... The area of ​​the region is 80 km², and the area of ​​the region overlapping with the unsatisfied region of the second constraint layer Cj is... For a radius of 20 km², the mask overlap ratio is... =20 / 80=0.25, indicating that 25% of the area in the satisfied region of the first constraint layer overlaps with the unsatisfied region of the second constraint layer, indicating a certain degree of conflict.

[0078] 2. Average geometric distance

[0079] Optionally, the average geometric distance is used to quantify the proximity between the satisfying region boundary of the first constraint layer and the non-satisfied region boundary of the second constraint layer. The closer the boundaries are, the more potential or severe the conflict between the two constraint layers is. The calculation process is as follows:

[0080] First, the boundary lines are extracted. Specifically, using the boundary extraction tool of geographic information processing software, the boundary lines are extracted from the satisfying region mask of the first constraint layer Ci. Mask of unsatisfied regions with the second constraint layer Cj Extract the boundary lines to obtain the region boundary. and not satisfying the area boundary ⁻. Boundary lines are represented as vector line segments, containing coordinate information of multiple vertices.

[0081] Then, the distance between the boundary lines is calculated. Specifically, the Euclidean distance calculation method is used to calculate the distances that satisfy the regional boundary conditions. Up to each vertex to the region boundary that does not meet the condition The shortest distance, and the area boundary not being satisfied. ⁻Each vertex on the region boundary satisfies The shortest distance. Specifically, for Given the vertex P(x1, y1) on the y-axis, traverse... For all line segments on the line, calculate the perpendicular distance from point P to each line segment, and take the minimum value as the perpendicular distance from point P to the line segment. The shortest distance d1; similarly, for Given a vertex Q(x2,y2) on the x-axis, calculate its distance to... The shortest distance d2.

[0082] Finally, the average geometric distance is calculated, specifically according to the following formula. :

[0083]

[0084] in, This represents the set of distances that satisfy and do not satisfy the region boundary. The Mean() function takes the average of all distance values ​​in this set. For example, if the set of distances satisfying and not satisfying the region boundary is {50m, 80m, 60m, 70m, 90m}, then the average geometric distance is... =(50+80+60+70+90) / 5=70m. To facilitate subsequent weighted summation calculations, the average geometric distance is normalized, mapping its value range to [0,1]. The normalization formula can be:

[0085]

[0086] in, The preset maximum distance threshold is, for example, 500m. The preset minimum distance threshold is, for example, 0m. If... =70m, , Then the normalized =(500-70) / (500-0)=430 / 500=0.86, indicating that the boundaries of the two constraint layers are relatively close, and the risk of conflict is relatively high.

[0087] 3. Degree of difference in terrain gradient

[0088] In some embodiments, the degree of terrain gradient difference is used to quantify the degree of difference between the first constraint layer and the second constraint layer on the corresponding terrain gradient. It is particularly suitable for conflict measurement between terrain-related constraint layers such as elevation threshold constraint layers and mountain peak difference constraint layers. The calculation process is as follows:

[0089] First, the terrain gradient is calculated. Specifically, based on DEM data, slope calculation algorithms such as the D8 algorithm and multi-flow direction algorithm are used to calculate the terrain gradient value, i.e., the slope value, of each grid cell within the layout area, resulting in a terrain gradient raster map. The terrain gradient value is expressed in angles, reflecting the steepness of the terrain at that location.

[0090] Then, the terrain gradient values ​​corresponding to the constraint layers are extracted. Based on the spatial range of the first constraint layer Ci and the second constraint layer Cj, the terrain gradient values ​​of all grid cells within the range of the two constraint layers are extracted from the terrain gradient raster map to obtain the gradient value sets Gi and Gj.

[0091] Finally, the degree of terrain gradient difference is calculated using the following formula. :

[0092]

[0093] in, This represents the set of terrain gradient values ​​within the range of the first constraint layer Ci. This represents the set of terrain gradient values ​​within the range of the second constraint layer Cj. |·| represents the absolute value operation, and the Mean() function represents the average of the absolute values ​​of the differences among all gradient values.

[0094] For example, given the set of terrain gradient values ​​Gi = {10°, 15°, 20°, 12°, 18°} within the first constraint layer and the set of terrain gradient values ​​Gj = {18°, 22°, 25°, 20°, 23°} within the second constraint layer, the set of absolute values ​​of the gradient difference is {8°, 7°, 5°, 8°, 5°}. The degree of terrain gradient difference... =(8+7+5+8+5) / 5=6.6°. Similarly, this value needs to be normalized to the range [0,1]. The normalization formula is:

[0095]

[0096] in, The preset maximum gradient difference threshold is, for example, 30°. The preset minimum gradient difference threshold is 0°. =6.6°, then the normalized value is... =(6.6-0) / (30-0)=6.6 / 30=0.22, indicating that the difference between the two constraint layers in terms of terrain gradient is small, and the conflict caused by the difference in terrain gradient is relatively mild.

[0097] According to an embodiment of the present invention, preferably, after obtaining one or more conflict measurement results (after normalization), these results are weighted and summed to obtain the comprehensive conflict intensity between the two spatial constraint layers. The weighted summation formula is as follows:

[0098]

[0099] Where w1, w2, and w3 are weighting coefficients for the mask overlap ratio, the normalized average geometric distance, and the normalized terrain gradient difference, respectively, and satisfying w1 + w2 + w3 = 1. The values ​​of the weighting coefficients can be determined according to the importance of different conflict measurement methods to conflict assessment. For example, if it is believed that the mask overlap ratio has the greatest impact on conflict assessment, followed by the average geometric distance, and the terrain gradient difference has the least impact, then w1 = 0.5, w2 = 0.3, and w3 = 0.2 can be set.

[0100] For example, assuming that the mask overlap ratio is calculated for the first constraint layer Ci and the second constraint layer Cj. =0.25, normalized mean geometric distance =0.86, normalized topographic gradient difference =0.22, weighting coefficients w1=0.5, w2=0.3, w3=0.2, then the overall conflict intensity is... =0.5×0.25 + 0.3×0.86 +0.2×0.22=0.125 + 0.258 + 0.044=0.427, indicating a moderate degree of conflict between the two constraint layers.

[0101] Therefore, the comprehensive conflict intensity between all pairs of spatial constraint layers is calculated sequentially, and the edge set E of the conflict relationship graph is constructed, ultimately forming a complete conflict relationship graph G=(C,E).

[0102] In some embodiments, for step S3, according to the present invention, before inputting the conflict relationship graph into the pre-trained graph attention network model, the model is preprocessed to ensure that the model can accurately receive and process the conflict relationship graph data. The preprocessing process includes two parts: model structure determination and model training.

[0103] Specifically, see Figure 2 , Figure 2 This is a schematic diagram of the graph attention network model structure provided in an embodiment of the present invention. The graph attention network model adopts a multi-layer graph attention layer stacked structure, including an input layer, a graph attention layer (at least one layer), a fully connected layer and an output layer.

[0104] The number of neurons in the input layer is consistent with the dimension of the node feature vector Fi. For example, if the node feature vector Fi contains five feature components: constraint type encoding, constraint strength value, constraint available area ratio, average elevation, and average slope, then the number of neurons in the input layer is 5.

[0105] The graph attention layer employs a multi-head attention mechanism, which uses multiple attention heads to compute attention weights between nodes in parallel, enhancing the model's ability to capture node importance. The number of attention heads can be set according to actual needs, for example, to 8.

[0106] The fully connected layer is used to perform nonlinear transformations and integration of the features output by the graph attention layer. The number of neurons in this layer can be adjusted according to the model complexity requirements, for example, set to 64. The number of neurons in the output layer is consistent with the number of nodes in the conflict graph. Each neuron corresponds to one node, and the output layer represents the conflict mitigation factor of the spatial constraint layer. .

[0107] Alternatively, the training process of the graph attention network model is as follows:

[0108] A large amount of wind turbine layout case data was collected. Each case data includes spatial constraint layer information, a constructed conflict relationship map, and manually labeled conflict mitigation factors corresponding to each spatial constraint layer. The manual labeling was determined based on the experience of wind turbine layout experts, taking into account the importance of constraints and the adjustability of conflicts. The collected case data was divided into training and testing sets in a 7:3 ratio for model training and performance evaluation.

[0109] The mean squared error loss function (MSE) is used as the model's loss function to measure the difference between the conflict mitigation factor output by the model and the manually labeled true conflict mitigation factor. The formula for the loss function is as follows:

[0110]

[0111] Where N is the number of cases in the training set. This represents the predicted conflict mitigation factor for a node in the i-th case output by the model. This is the actual manually labeled value corresponding to this node.

[0112] The graph attention network model was trained using a training set, with the Adam optimizer used to update the model parameters. The learning rate was set to 0.001, and the training iterations were set to 100 epochs. During each training epoch, the conflict relationship graph from the training set was input into the model. The predicted conflict mitigation factor was calculated through forward propagation, and the loss value was calculated based on the loss function. The model's weight parameters, including the attention weights of the graph attention layers and the weights of the fully connected layers, were updated through backpropagation. Every certain number of training epochs, such as 10 epochs, the model performance was evaluated using a test set. If the model's loss value on the test set no longer decreased or overfitting occurred, training was stopped, and the optimal model parameters were saved, resulting in the pre-trained graph attention network model.

[0113] According to an embodiment of the present invention, the conflict relationship graph G=(C,E) constructed in step S2 is input into a pre-trained graph attention network model. The model calculates the conflict mitigation factor corresponding to each spatial constraint layer by allocating and reasoning attention to the importance of each node in the graph. Please see Figure 3 , Figure 3 This is a schematic diagram of the model inference process provided in an embodiment of the present invention. The specific inference process is as follows:

[0114] First, in S301, the node feature matrix and adjacency matrix of the conflict relationship graph are loaded into the model. The node feature matrix X has a dimension of N×D, where N is the number of nodes (i.e., the number of spatial constraint layers), and D is the dimension of the node feature vectors (i.e., the number of feature components contained in each node feature vector Fi). Each row in X corresponds to the feature vector Fi of one node. The adjacency matrix A has a dimension of N×N, and the elements in A... The conflict intensity value of the edge in the corresponding conflict relationship graph, i.e. = , is used to represent the connection relationship and conflict level between node Ci and node Cj.

[0115] In S302, the graph attention layer is calculated. In this layer, for each node Ci, the model calculates the attention weight of node Ci to node Cj based on its connection relationship with other nodes Cj (i.e., the adjacency matrix A) and the node features (i.e., the node feature matrix X), thus allocating attention based on node importance. The specific calculation process is as follows:

[0116] First, a linear transformation is performed on the feature matrix X. Specifically, a linear transformation is applied to the node feature matrix X by using a weight matrix W with dimensions D×F, where F is the feature dimension after the linear transformation. This maps the feature vector Fi of each node from D dimensions to F dimensions, resulting in the transformed node feature matrix. =X×W.

[0117] Then, the attention coefficient is calculated, specifically the attention coefficient between node Ci and node Cj. Attention coefficient is used to measure the importance of node Cj to node Ci. The calculation uses the following formula:

[0118]

[0119] in, A learnable weight vector of dimension 2F×1 is used to map the concatenated feature vector to one dimension. LeakyReLU is the activation function used to introduce non-linearity. This represents the transformed feature vector of node Ci. With the feature vector of node Cj The pieces are stitched together, with a dimension of 2F×1.

[0120] Attention weights are normalized. To ensure the interpretability of attention weights, a softmax function is used to normalize all attention coefficients of node Ci. After normalization, the attention weights of node Ci to node Cj are obtained. The normalization formula is as follows:

[0121]

[0122] Where Ni is the set of neighboring nodes of node Ci, i.e., nodes that are connected to node Ci, determined by the adjacency matrix A. If >0, then node Ck belongs to exp() is an exponential function used to map attention coefficients to positive values. The denominator is the exponential sum of the attention coefficients of all neighboring nodes of node Ci, ensuring the normalized attention weights. The sum is 1.

[0123] A multi-head attention mechanism is adopted, in which attention weights are calculated by K independent attention heads, and the node features calculated by each attention head are concatenated to obtain the output feature matrix Z of the graph attention layer. The concatenation formula is as follows:

[0124]

[0125] in, The attention weights of node Ci to node Cj calculated by the k-th attention head, where σ is an activation function such as ReLU, used to introduce nonlinearity. This represents the feature concatenation operation. Multi-head attention fusion enhances the model's robustness in capturing node importance and improves inference accuracy.

[0126] In S303, the fully connected layer and the output layer calculate the output conflict mitigation factor. The feature matrix Z output by the graph attention layer is input to the fully connected layer. The fully connected layer performs a nonlinear transformation on the feature matrix Z through a weight matrix W1 with dimensions K×F×H, where H is the output dimension of the fully connected layer and the bias vector b1, to obtain the transformed feature matrix Z1=σ(Z×W1 + b1), where σ is an activation function such as ReLU.

[0127] Subsequently, the feature matrix Z1 is input to the output layer. The output layer maps the feature matrix Z1 to one dimension through a weight matrix W2 of dimension H×1 and a bias vector b2, thus obtaining the conflict mitigation factor corresponding to each node. The output formula is as follows:

[0128]

[0129] in, The eigenvector of the corresponding node Ci in the eigenmatrix Z1. The sigmoid activation function is used to mitigate the conflict. The value range of is restricted to [0,1], which conforms to the definition of the degree to which the quantitative constraint of the conflict mitigation factor can be relaxed. The smaller the value, the less likely the constraint is to be violated; The larger the value, the more the constraints can be relaxed.

[0130] In some embodiments, for step S4, according to the present invention, the wind turbine location search algorithm used in step S4 can be implemented by artificial intelligence middleware, such as a hybrid heuristic algorithm, a genetic algorithm, a particle swarm optimization algorithm, etc., and preferably a search algorithm combining a hybrid heuristic and a genetic algorithm. This algorithm has both the local search capability of the heuristic algorithm and the global search capability of the genetic algorithm, and can efficiently search for the optimal wind turbine location.

[0131] Specifically, based on the project layout boundary Shapefile after coordinate unification in step S1, the search space for wind turbine locations is determined, that is, candidate wind turbine locations are searched only within the project layout boundary, excluding areas outside the boundary. At the same time, combined with the unusable land Shapefile corresponding to the unusable land avoidance constraint, the unusable land range is further excluded in the search space to obtain a preliminary usable search area.

[0132] For the genetic algorithm part, the population is first initialized. Each individual in the population represents a set of wind turbine placement schemes. Each individual consists of multiple genes, and each gene corresponds to the coordinates (X, Y, Z) of a wind turbine location. X and Y are planar coordinates based on a Gaussian three-dimensional projected coordinate system, and Z is the elevation of that location, obtained from an elevation reference layer. The population size is determined based on the size of the placement area and the required number of wind turbines. For example, if the placement area is 100 km² and 50 wind turbines are planned, the population size can be set to 100, that is, initializing 100 different wind turbine placement schemes as the initial population. During initialization, it is necessary to ensure that each wind turbine location is within the initial usable search area and meets the spatial requirements corresponding to the wind turbine foundation radius, i.e., the distance between two wind turbine locations is not less than twice the wind turbine foundation radius, avoiding foundation overlap.

[0133] In some embodiments, candidate wind turbine locations are generated using a wind turbine location search algorithm. The specific generation process is as follows:

[0134] First, a heuristic search method, such as a greedy algorithm, is used to generate initial candidate wind turbine locations within the available search area. Specifically, starting from any point within the available search area, the degree to which that point satisfies each spatial constraint layer when used as a wind turbine location is calculated, and the point with the highest degree of satisfaction is selected as the first candidate wind turbine location. Then, with this point as the center, the influence range of the wind turbine is determined based on the major axis L, minor axis S, and direction angle θ of the wind turbine elliptical spacing constraint. Within the available search area outside the influence range, the search continues to find the next point with the highest degree of satisfaction as a candidate wind turbine location. This process is repeated until a predetermined number of candidate wind turbine locations are generated or no suitable location is found within the available search area.

[0135] The initial candidate wind turbine sites generated by heuristic search are used as part of the initial population of the genetic algorithm, combined with other randomly generated individuals to form a complete initial population. Then, the population is iteratively optimized through selection, crossover, and mutation operations of the genetic algorithm to generate new candidate wind turbine sites.

[0136] Selection operation: The roulette wheel selection method is adopted. Based on the fitness value of each individual, i.e. the wind turbine location layout scheme, the fitness value is related to the score of the candidate location. The higher the score, the greater the fitness value. The probability of the individual being selected is determined. The higher the fitness value, the greater the probability of the individual being selected to participate in subsequent crossover and mutation operations, so as to retain high-quality candidate location schemes.

[0137] Crossover operation: For the two selected parent individuals, a crossover point is randomly chosen, and the genes (wind turbine location coordinates) of the two parent individuals after the crossover point are swapped to generate two offspring individuals. For example, the gene sequence of parent individual A is [(X1,Y1,Z1), (X2,Y2,Z2), (X3,Y3,Z3)], and the gene sequence of parent individual B is [(X4,Y4,Z4), (X5,Y5,Z5), (X6,Y6,Z6)]. If the crossover point is the second gene, then the gene sequence of offspring individual A is [(X1,Y1,Z1), (X2,Y2,Z2), (X6,Y6,Z6)], and the gene sequence of offspring individual B is [(X4,Y4,Z4), (X5,Y5,Z5),(X3,Y3,Z3)].

[0138] Mutation operation: For the offspring individuals generated by the crossover operation, one or more genes are randomly selected for mutation with a certain mutation probability (e.g., 0.05). This means that the X and Y coordinates of the point are slightly adjusted. The adjustment range is determined according to the search accuracy, such as ±50 meters. The Z coordinate is re-obtained from the elevation reference layer based on the adjusted X and Y coordinates to increase the diversity of the population and avoid the algorithm from getting stuck in local optima.

[0139] Through multiple rounds of genetic algorithm iteration, such as 50 rounds, candidate wind turbine locations are gradually optimized to generate a batch of high-quality candidate wind turbine locations for subsequent evaluation and selection.

[0140] According to an embodiment of the present invention, the scoring function in step S4 specifically involves calculating the degree of default at each spatial constraint layer of the candidate wind turbine location and the corresponding conflict mitigation factor, summing the calculation results for all constraint layers, and using this summation result as the final score for the candidate location to guide the search algorithm in location selection and optimization. Please refer to... Figure 4 , Figure 4 This is a schematic diagram of the candidate wind turbine location selection and optimization process provided in an embodiment of the present invention. The specific calculation process is as follows:

[0141] In S401, the degree of default at candidate sites is calculated. Specifically, for each candidate wind turbine site P(X,Y,Z), the degree of default Si(P) at each spatial constraint layer Ci is calculated. The degree of default Si(P) quantifies the extent to which candidate site P does not meet the requirements of constraint layer Ci, and its value ranges from [0,+∞). Si(P)=0 indicates that candidate site P fully meets the requirements of constraint layer Ci. The larger Si(P) is, the more severe the degree of default by candidate site P. The calculation methods for the degree of default at different spatial constraint layers are as follows:

[0142] (1) The degree of breach of project boundary constraints (C1) S1(P)

[0143] If the candidate point P is within the project layout boundary, then S1(P)=0; if P is outside the project layout boundary, then calculate the shortest distance d1 from P to the project layout boundary, S1(P)=d1. The farther the distance, the greater the degree of breach of contract.

[0144] (2) The degree of breach of the unusable land avoidance constraint (C2) S2(P)

[0145] If candidate point P is located within the non-unusable land area, then S2(P) = 0; if P is located within the unusable land area, then calculate the shortest distance d2 from P to the unusable land boundary, S2(P) = d2. The farther the distance, the deeper into the unusable land area, and the greater the degree of breach of contract.

[0146] (3) The degree of breach of the safety distance constraint (C3) S3(P)

[0147] Calculate the shortest distance d3 from candidate point P to surrounding buildings such as residential areas and factories. If d3 ≥ building safety distance threshold, then S3(P) = 0; if d3 < building safety distance threshold, then S3(P) = building safety distance threshold - d3. The larger the difference, the greater the degree of breach of contract.

[0148] (4) The degree of breach of the altitude threshold constraint (C4) S4(P)

[0149] If the elevation Z of candidate point P is greater than or equal to the minimum elevation threshold, then S4(P) = 0; if Z is less than the minimum elevation threshold, then S4(P) = minimum elevation threshold - Z. The larger the difference, the greater the degree of breach.

[0150] (5) The degree of breach of the mountain peak difference constraint (C5) S5(P)

[0151] Determine the elevation difference between the mountain peak where candidate point P is located and the surrounding adjacent mountain peaks. ,like If the difference constraint at the mountain peak is Δh, then S5(P) = 0; if ,but The larger the difference in Δh, the greater the degree of default.

[0152] (6) The degree of violation of the wind turbine elliptical spacing constraint (C6) S6(P)

[0153] Calculate the candidate site P and other generated candidate wind turbine sites. Elliptical distance between The calculation of the elliptical distance is based on the wind turbine ellipse parameters (major axis L, minor axis S, and direction angle θ), and the formula is as follows:

[0154]

[0155] Where (X,Y) are the coordinates of candidate point P, , Other candidate wind turbine locations The coordinates. If (Indicating that the distance between the two wind turbine locations meets the elliptical spacing requirement), then S6(P) = 0; if Then S6(P) = 1 - The larger the difference, the greater the degree of default.

[0156] In S402, the scoring function is calculated, specifically, the degree of default Si(P) of the candidate wind turbine location P at each spatial constraint layer Ci and the corresponding conflict mitigation factor are obtained. Then, the final score Score(P) of the candidate point P is calculated according to the following scoring function:

[0157]

[0158] Where n is the number of spatial constraint layer types, (1- ) represents the weighting coefficient of the constraint. The smaller, (1- The larger the value of the constraint, the greater its importance, and the greater the impact of the degree of default on the final score; conversely, the smaller the value of the constraint, the greater its importance. The larger, (1- The smaller the value, the lower the importance of the constraint, and the smaller the impact of the degree of default on the final score.

[0159] For example, suppose the default severity of candidate point P in the six spatial constraint layers is S1(P)=0, S2(P)=0, S3(P)=20m, S4(P)=30m, S5(P)=10m, and S6(P)=0.2, respectively, and the corresponding conflict mitigation factors are respectively... =0.01、 =0.05、 =0.08、 =0.3、 =0.2、 =0.4, then the weight coefficients of each constraint (1- The scores were 0.99, 0.95, 0.92, 0.7, 0.8, and 0.6, respectively. The final score, Score(P), was calculated as follows: Score(P) = 0 × 0.99 + 0 × 0.95 + 20 × 0.92 + 30 × 0.7 + 10 × 0.8 + 0.2 × 0.6 = 0 + 0 + 18.4 + 21 + 8 + 0.12 = 47.52.

[0160] In S403, candidate sites are selected and optimized. Specifically, the generated candidate wind turbine sites are sorted according to the final score (Score(P). The smaller the Score(P), the higher the degree to which candidate site P satisfies the constraints, and the more suitable it is as a wind turbine site. In the wind turbine site search algorithm, the batch of candidate sites with the smallest Score(P), such as selecting the top 20% of the lowest-scoring sites, are selected as high-quality sites to participate in subsequent iterative optimization or as the final candidate set of wind turbine sites. At the same time, the parameters of the search algorithm, such as the crossover probability and mutation probability of the genetic algorithm, are adjusted based on the feedback of the scoring results. If the current high-quality site has a high score, that is, the overall degree of constraint satisfaction is low, the mutation probability is increased to increase population diversity in order to search for better candidate sites; if the current high-quality site has a low score, that is, the overall degree of constraint satisfaction is high, the mutation probability is decreased to stabilize the high-quality sites and accelerate the convergence speed of the algorithm.

[0161] In some embodiments, for step S5, according to the present invention, based on the evaluation results of the scoring function in step S4, a final wind turbine location layout scheme is further searched to ensure that the scheme can meet all key constraints and that the overall rationality and economy of the wind turbine layout are optimal.

[0162] Specifically, firstly, the set of high-quality candidate points obtained in step S4 is subjected to constraint satisfaction screening, and those points that fully satisfy the rigid constraints are selected. Candidate points required to be ≈0. For example, project boundary constraints ( ≈0), safety distance constraints ( ≈0), Unusable land avoidance constraint ( Rigid constraints such as ≈0 mean that candidate points must fully satisfy the requirements of these constraints. =0, otherwise the candidate site is directly excluded. This screening process ensures that the final wind turbine site will not exceed critical safety and boundary limits;

[0163] Then, the overall layout of the selected candidate sites is optimized, taking into account factors such as the uniformity of spacing between wind turbines, wind resource utilization efficiency, and construction convenience. Specifically, the spacing distribution between all wind turbine sites in the candidate site set is calculated to ensure that the spacing distribution meets the requirements of the wind turbine elliptical spacing constraint, avoiding situations where the spacing is too large or too small; combined with the prevailing wind direction and wind resource distribution data of the layout area, the positions of the wind turbine sites are adjusted so that the wind turbines can utilize wind resources to the maximum extent and reduce wind shading effects; at the same time, considering factors such as the accessibility of construction roads and the cost of laying transmission lines, the wind turbine sites are adjusted to locations that are convenient for construction and have lower costs;

[0164] Finally, a multi-objective optimization evaluation method was used to evaluate the candidate site schemes after overall layout optimization. The evaluation indicators included the number of wind turbines, total installed capacity, average score (P) for meeting constraints, construction cost, and operation and maintenance cost. The weight of each evaluation indicator was determined by the Analytic Hierarchy Process (AHP), the comprehensive evaluation value of each scheme was calculated, and the scheme with the highest comprehensive evaluation value was selected as the final wind turbine site layout scheme.

[0165] According to an embodiment of the present invention, after determining the final wind turbine location layout scheme, the relevant data and documents of the scheme are output to provide a basis for subsequent engineering design and construction.

[0166] Specific output content may include a DXF file, containing full-layer information. This file includes satellite imagery, contour layers, unavailable map layers, project boundary layers, and wind turbine layout layers. The satellite imagery provides real-world geographic information of the layout area, allowing for a visual assessment of the surrounding environment of the wind turbine locations. The contour layers, generated from DEM data, reflect the terrain undulations of the area, providing topographic reference for wind turbine foundation design and construction. The unavailable map layers and project boundary layers clearly define the limitations of the wind turbine layout, ensuring construction stays within permissible limits. The wind turbine layout layer marks the location, number, and coordinates of each wind turbine location, using different graphic symbols to distinguish the type and status of the turbines, such as confirmed locations and alternative locations. The DXF file can be directly imported into engineering design software such as AutoCAD for subsequent wind turbine foundation design and construction drawing creation.

[0167] Therefore, this invention can efficiently and accurately complete the automatic layout of wind turbine locations, solving problems such as unreasonable handling of multiple constraint conflicts and excessive reduction of available areas in traditional methods, improving the rationality and economy of wind turbine layout, and providing technical support for the smooth implementation of wind power projects.

[0168] Please see Figure 5 , Figure 5This is a schematic diagram of an artificial intelligence-based automatic wind turbine location layout system provided in an embodiment of this application. Figure 5 As shown, system 500 includes:

[0169] The data receiving module 501 is used to receive multiple spatial constraint layers and deployment parameters related to the wind turbine layout, and to perform coordinate unification processing on each of the spatial constraint layers.

[0170] The conflict relationship graph construction module 502 is used to construct a conflict relationship graph based on the processed spatial constraint layer, wherein the nodes of the conflict relationship graph represent different types of spatial constraint layers, and the edges of the conflict relationship graph represent the conflict intensity between any two spatial constraint layers.

[0171] The conflict mitigation factor calculation module 503 is used to input the conflict relationship graph into a pre-trained graph attention network model. Through this model, attention is allocated and reasoned about the importance of each node in the graph, and the conflict mitigation factor corresponding to each spatial constraint layer is calculated. The conflict mitigation factor is used to quantify the degree to which the constraint is allowed to be relaxed during the arrangement process.

[0172] The candidate wind turbine location evaluation module 504 is used to feed back the conflict mitigation factor to the wind turbine location search algorithm. When generating candidate wind turbine locations, a scoring function that incorporates the conflict mitigation factor is used to evaluate the candidate locations. The scoring function is weighted according to the degree of violation of each constraint and its corresponding conflict mitigation factor.

[0173] The wind turbine location layout scheme output module 505 is used to search for and output the final wind turbine location layout scheme based on the evaluation results of the scoring function.

[0174] Those skilled in the art will clearly understand that the technical solutions of the embodiments of this application can be implemented by means of software and / or hardware. In this specification, "unit" and "module" refer to software and / or hardware that can independently complete or cooperate with other components to complete a specific function, wherein the hardware may be, for example, a field-programmable gate array (FPGA), an integrated circuit (IC), etc.

[0175] Each processing unit and / or module in the embodiments of this application can be implemented by an analog circuit that implements the functions described in the embodiments of this application, or by software that executes the functions described in the embodiments of this application.

[0176] Please see Figure 6 It shows a schematic diagram of the structure of an electronic device according to an embodiment of this application, which can be used to implement... Figure 1The method in the illustrated embodiment. (As shown) Figure 6 As shown, the electronic device 600 may include:

[0177] The system includes at least one processor 601, at least one network interface 604, a user interface 603, a memory 605, and at least one communication bus 602. The communication bus 602 is used to enable connection and communication between the components. The user interface 603 may include buttons, and optionally include a standard wired or wireless interface. The network interface 604 may include, but is not limited to, a Bluetooth module, an NFC module, a Wi-Fi module, etc.

[0178] The processor 601 may include one or more processing cores and connect to various parts within the device 600 via various interfaces and lines. It implements the various functions and data processing of the device 600 by running or executing instructions, programs, code sets, or instruction sets stored in the memory 605, and by accessing data in the memory 605. Optionally, the processor 601 may be implemented using at least one hardware form of DSP, FPGA, or PLA. The processor 601 may also integrate one or more combinations of CPU, GPU, and modem. The CPU is mainly used to handle the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem is used for wireless communication. It is understood that the modem may not be integrated into the processor 601, but may be implemented through a separate chip.

[0179] Memory 605 may include random access memory (RAM) or read-only memory (ROM). Optionally, memory 605 includes a non-transitory computer-readable medium for storing instructions, programs, code, code sets, or instruction sets. Memory 605 may be divided into a program storage area and a data storage area, wherein the program storage area may be used to store instructions for implementing an operating system, instructions for implementing at least one function (such as touch functionality, audio playback functionality, image playback functionality, etc.), and instructions for implementing the foregoing method embodiments; the data storage area may be used to store data involved in the relevant method embodiments. Memory 605 may also be at least one storage device located remotely from processor 601. Figure 6 As shown, the memory 605, which serves as a computer storage medium, may contain an operating system, a network communication module, a user interface module, and program instructions.

[0180] In particular, the methods and / or embodiments in this application can be implemented as computer software programs. For example, the embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. When the computer program is executed by processor 601, it performs the functions defined in the methods of this application.

[0181] Another embodiment of this application provides a computer-readable storage medium having computer program instructions stored thereon, which can be executed by a processor to implement the methods and / or technical solutions of any one or more embodiments of this application described above.

[0182] The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0183] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0184] Furthermore, it should be noted that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in the system claims may also be implemented by a single unit or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

Claims

1. A method for automatically arranging wind turbine locations based on artificial intelligence, characterized in that, The methods and steps include the following: S1, receive multiple spatial constraint layers and deployment parameters related to the wind turbine layout, and perform coordinate unification processing on each of the spatial constraint layers; S2, Based on the processed spatial constraint layer, a conflict relationship graph is constructed, wherein the nodes of the conflict relationship graph represent different types of spatial constraint layers, and the edges of the conflict relationship graph represent the conflict intensity between any two spatial constraint layers; wherein, calculating the conflict intensity between two spatial constraint layers includes calculating the mask overlap ratio between the satisfied region of the first constraint layer and the unsatisfied region of the second constraint layer; and / or, calculating the average geometric distance between the boundary of the satisfied region of the first constraint layer and the boundary of the unsatisfied region of the second constraint layer; and / or, calculating the degree of difference between the first constraint layer and the second constraint layer on the corresponding terrain gradient; and weighting and summing the above one or more conflict measurement results to comprehensively quantify the conflict intensity; S3, input the conflict relationship graph into a pre-trained graph attention network model, and use the model to perform attention allocation and reasoning on the importance of each node in the graph, and calculate the conflict mitigation factor corresponding to each spatial constraint layer. The conflict mitigation factor is used to quantify the degree to which the constraint is allowed to be relaxed during the arrangement process. S4, the conflict mitigation factor is fed back to the wind turbine location search algorithm. When generating candidate wind turbine locations, a scoring function that incorporates the conflict mitigation factor is used to evaluate the candidate locations. The scoring function performs a weighted calculation based on the degree of default at each spatial constraint layer and its corresponding conflict mitigation factor. Specifically, the scoring function calculates the degree of default at each spatial constraint layer of the candidate wind turbine location with the corresponding conflict mitigation factor, sums the calculation results for all constraint layers, and uses the sum as the final score of the candidate location to guide the search algorithm in location selection and optimization. S5. Based on the evaluation results of the scoring function, search and output the final wind turbine location layout scheme.

2. The method for automatic wind turbine location layout based on artificial intelligence according to claim 1, characterized in that, It also includes, For each node, a node feature vector is constructed, which includes at least one of constraint type encoding, constraint strength value, constraint available area ratio, and elevation and slope features extracted from terrain data.

3. The method for automatic wind turbine location layout based on artificial intelligence according to claim 2, characterized in that, The graph attention network model takes the feature vectors of the nodes and the conflict intensity of the edges as inputs. Its reasoning goal is to output the conflict mitigation factor corresponding to each node. The smaller the conflict mitigation factor, the less likely the constraint represented by the node is to be violated during the arrangement process. Conversely, the larger the factor, the more likely it is to be relaxed within a certain range.

4. The method for automatic wind turbine location layout based on artificial intelligence according to claim 1, characterized in that, Also includes: The spatial constraint layer includes at least one of the following: project boundary constraints, unusable land avoidance constraints, safety distance constraints, altitude threshold constraints, mountain top difference constraints, and wind turbine elliptical spacing constraints.

5. An automatic wind turbine location layout system based on artificial intelligence, characterized in that, include: The data receiving module is used to receive multiple spatial constraint layers and deployment parameters related to the wind turbine layout, and to perform coordinate unification processing on each of the spatial constraint layers; A conflict relationship graph construction module is used to construct a conflict relationship graph based on the processed spatial constraint layers. Nodes in the conflict relationship graph represent different types of spatial constraint layers, and edges represent the conflict intensity between any two spatial constraint layers. Calculating the conflict intensity between two spatial constraint layers includes: calculating the mask overlap ratio between the satisfied region of the first constraint layer and the unsatisfied region of the second constraint layer; and / or calculating the average geometric distance between the boundary of the satisfied region of the first constraint layer and the boundary of the unsatisfied region of the second constraint layer; and / or calculating the degree of difference between the first constraint layer and the second constraint layer on the corresponding terrain gradient. The above one or more conflict measurement results are weighted and summed to comprehensively quantify the conflict intensity. The conflict mitigation factor calculation module is used to input the conflict relationship graph into a pre-trained graph attention network model. Through this model, attention is allocated and reasoned about the importance of each node in the graph, and the conflict mitigation factor corresponding to each spatial constraint layer is calculated. The conflict mitigation factor is used to quantify the degree to which the constraint is allowed to be relaxed during the arrangement process. The candidate wind turbine location evaluation module is used to feed back the conflict mitigation factor to the wind turbine location search algorithm. When generating candidate wind turbine locations, a scoring function that incorporates the conflict mitigation factor is used to evaluate the candidate locations. The scoring function performs a weighted calculation based on the degree of default at each spatial constraint layer and its corresponding conflict mitigation factor. Specifically, the scoring function calculates the degree of default at each spatial constraint layer of the candidate wind turbine location with the corresponding conflict mitigation factor, sums the calculation results for all constraint layers, and uses the sum as the final score for the candidate location to guide the search algorithm in location selection and optimization. The wind turbine location layout scheme output module is used to search for and output the final wind turbine location layout scheme based on the evaluation results of the scoring function.

6. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-4.

7. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, The computer program instructions can be executed by a processor to implement the method as described in any one of claims 1-4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-4.

Images