Fan arrangement elevation optimization method and system based on AI combined with height triangulation algorithm

Abstract

This application discloses a method and system for optimizing wind turbine layout elevation based on AI combined with an elevation triangulation algorithm. The method includes: acquiring wind turbine location coordinates, survey point elevations, triangulation network structure, and engineering boundary conditions; constructing a triangulation graph structure containing node and edge features; correcting the initial elevation using a graph convolutional network; determining the initial wind turbine foundation type and identifying critical nodes by combining the corrected elevation and boundary conditions; constructing a spatial adjacency graph of wind turbine locations; optimizing the wind turbine foundation type and identifying abnormal nodes using a graph attention network; subsequently performing elevation perturbation simulation on nodes with inconsistent foundation types; determining the optimal corrected elevation through multi-objective optimization; and marking manually verified nodes; finally, obtaining the optimal elevation and wind turbine foundation type through iterative convergence. This method achieves bidirectional coupled optimization of elevation and wind turbine foundation type, improving the accuracy of elevation calculation and the rationality of wind turbine foundation type selection.

Description

Technical Field

[0001] This application relates to the field of intelligent design technology for new energy engineering, specifically a method and system for optimizing the elevation of wind turbine layout based on AI combined with elevation triangulation algorithm. Background Technology

[0002] In the design of wind turbine foundation layout in wind farm areas, CASS, Excel, and CAD software are often used to calculate the turbine location elevation and select the foundation type. The core process involves using CASS to construct a triangulation network of survey points, interpolating the initial elevation of the turbine location, then using Excel to calculate the embedment depth based on engineering boundary conditions and select the foundation type. Finally, the layout is mapped and annotated in CAD. However, this traditional design process has many technical shortcomings and fails to meet the accuracy and efficiency requirements of engineering design.

[0003] The minute elevation errors generated by triangulation interpolation can easily cause abrupt changes in wind turbine foundation type at critical boundaries, creating a coupled amplification effect between interpolation errors and foundation type changes. Furthermore, the existing process lacks the ability to reverse-engineer elevation and foundation type, making it impossible to verify elevation accuracy through foundation type rationality checks. Simultaneously, relying solely on single-point elevation intervals for foundation type selection lacks spatial consistency analysis capabilities, easily leading to discrepancies between local foundation types and surrounding areas. Moreover, manual verification points are often clustered in complex terrain areas, and the existing process cannot identify this spatial distribution characteristic, resulting in a large workload and low efficiency for manual verification. In addition, when adjusting the foundation locations of wind turbine sites during iterative design iterations, the entire manual verification process must be repeated, resulting in repetitive work and further reducing design efficiency. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for optimizing the elevation of wind turbine layout based on AI combined with the elevation triangulation algorithm, so as to solve the problems mentioned in the background art.

[0005] According to the first aspect of this application, a method for optimizing the elevation of wind turbine layout based on AI combined with elevation triangulation algorithm is provided, including the following steps: S1, acquire the wind turbine location plane coordinate data, survey point elevation data, triangular network structure data constructed from the survey points, and engineering boundary condition parameters for each area; S2, taking the wind turbine location and survey point as nodes, construct a triangular network structure G1 based on the triangular network structure data. The triangular network structure G1 includes node features and edge features. The node features include initial elevation and terrain features. S3, input the triangular network structure G1 into the graph convolutional network model, correct the initial elevation of the nodes, and output the corrected elevation; S4. Based on the corrected elevation and the engineering boundary condition parameters, calculate the soil penetration depth of each wind turbine location node, determine the initial wind turbine foundation type according to the correspondence between the soil penetration depth and the preset foundation design depth, and identify the critical nodes in the critical zone of the wind turbine foundation type boundary. S5. Construct a spatial adjacency graph G2 for wind turbine locations using wind turbine location nodes. The edges of the spatial adjacency graph G2 are determined based on the spatial distance between nodes. Input the spatial adjacency graph G2, the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker into the graph attention network model. Output the spatially optimized wind turbine foundation type and identify abnormal nodes based on the spatially optimized wind turbine foundation type and the initial wind turbine foundation type. S6. For wind turbine location nodes where the initial wind turbine foundation type is inconsistent with the space-optimized wind turbine foundation type, perform elevation perturbation simulation, determine the optimal corrected elevation based on the preset multi-objective optimization function, and mark the nodes corresponding to the corrected elevations that cannot meet the engineering boundary conditions as manual verification nodes. S7. Repeat S4-S6 until the number of abnormal nodes no longer changes after two consecutive iterations. Take the current corrected elevation as the optimal corrected elevation, output the wind turbine foundation type determined according to the optimal corrected elevation as the optimal wind turbine foundation type, and output the manually verified node.

[0006] Preferably, the terrain features include the region number to which the node belongs, the distance from the node to the boundary of the region, the local slope value of the node, and the neighborhood elevation variance of the node; The edge features include the edge length, the absolute value of the elevation difference between the two endpoints of the edge, and a binary identifier indicating whether the edge crosses the boundary of the engineering area.

[0007] Preferably, the graph convolutional network model includes a neighborhood aggregation layer, a boundary constraint embedding layer, and an anomaly suppression layer; the neighborhood aggregation layer performs weighted aggregation of the initial elevations of neighboring nodes based on the edge features between nodes to obtain a first intermediate elevation; the boundary constraint embedding layer performs constraint correction on the first intermediate elevation based on engineering boundary condition parameters to obtain a second intermediate elevation; the anomaly suppression layer performs weighted suppression on the second intermediate elevation based on the variance of the neighborhood elevations of the nodes to obtain the corrected elevation.

[0008] Preferably, the graph attention network model calculates the attention weights between nodes based on the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker. The attention weights are calculated only for the neighboring nodes in the wind turbine location spatial adjacency graph G2, and are used to weighted aggregate the features of the neighboring nodes to output the spatially optimized wind turbine foundation type.

[0009] Preferably, the identification of abnormal nodes includes: calculating the wind turbine foundation type consistency index of each wind turbine location node, wherein the wind turbine foundation type consistency index is the proportion of the number of nodes in the node's spatial neighborhood that have the same initial wind turbine foundation type as the node to the total number of nodes in the neighborhood; if the wind turbine foundation type consistency index is less than a preset threshold and the average elevation difference between the node and the neighboring nodes is less than a preset critical zone elevation threshold, then the node is determined to be an abnormal node.

[0010] Preferably, the multi-objective optimization function includes four sub-objectives: elevation adjustment, number of wind turbine foundation type jumps, number of engineering boundary violations, and spatial dispersion. The elevation adjustment is the absolute value of the difference between the candidate elevation and the corrected elevation of a node. The number of wind turbine foundation type jumps is the number of nodes whose optimized wind turbine foundation type differs from the initial wind turbine foundation type. The number of engineering boundary violations is the number of nodes whose optimized elevation violates the engineering boundary conditions. The spatial dispersion is the variance of the consistency index of wind turbine foundation types for all nodes.

[0011] Preferably, the elevation perturbation simulation includes increasing and decreasing the corrected elevation of the nodes with a preset perturbation step size to generate candidate elevations, and verifying whether the wind turbine foundation type corresponding to the candidate elevation meets the engineering boundary conditions.

[0012] Preferably, the method further includes importing the optimal corrected elevation, the optimal wind turbine foundation type, and the manual verification node information into CAD software, and automatically drawing the wind turbine foundation points and marking the wind turbine foundation type and verification marks in the CAD software.

[0013] In a second aspect, this application also provides a wind turbine layout elevation optimization system based on AI combined with an elevation triangulation network algorithm, comprising: The data acquisition module is used to acquire wind turbine location plane coordinate data, survey point elevation data, triangular network structure data constructed from survey points, and engineering boundary condition parameters for each area; The first graph structure construction module is used to construct a triangular network graph structure G1 based on the triangular network structure data, using the wind turbine location and survey point as nodes. The triangular network graph structure G1 includes node features and edge features. The node features include initial elevation and terrain features. The elevation correction module is used to input the triangular network structure G1 into the graph convolutional network model, correct the initial elevation of the nodes, and output the corrected elevation. The wind turbine foundation type initial determination and critical identification module is used to calculate the soil penetration depth of each wind turbine location node based on the corrected elevation and the engineering boundary condition parameters, determine the initial wind turbine foundation type according to the correspondence between the soil penetration depth and the preset foundation design depth, and identify the critical nodes in the critical zone of the wind turbine foundation type boundary. The spatial optimization module is used to construct a spatial adjacency graph G2 for wind turbine locations based on wind turbine location nodes. The edges of the spatial adjacency graph G2 are determined according to the spatial distance between nodes. The spatial adjacency graph G2, along with the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker, are input into the graph attention network model. The module outputs the spatially optimized wind turbine foundation type and identifies abnormal nodes based on the spatially optimized wind turbine foundation type and the initial wind turbine foundation type. The perturbation optimization module is used to perform elevation perturbation simulation on wind turbine location nodes where the initial wind turbine foundation type is inconsistent with the spatially optimized wind turbine foundation type. Based on the preset multi-objective optimization function, the optimal corrected elevation is determined, and nodes that cannot obtain corrected elevations that meet the engineering boundary conditions are marked as manually verified nodes. The iterative output module is used to repeatedly execute the initial determination and critical identification module for wind turbine foundation type, the spatial optimization module, and the perturbation optimization module until the number of abnormal nodes no longer changes after two consecutive iterations. The current corrected elevation is taken as the optimal corrected elevation, the wind turbine foundation type determined according to the optimal corrected elevation is taken as the optimal wind turbine foundation type, and the manually verified node is output.

[0014] This application combines elevation triangulation network (ETN) algorithms with artificial intelligence technology to achieve bidirectional coupling optimization of wind turbine layout elevation and foundation type. This effectively solves the foundation type jump problem caused by triangulation interpolation errors in traditional design, improving elevation calculation accuracy. By constructing a triangulation network structure and a spatial adjacency graph of wind turbine locations, combined with graph convolutional networks and graph attention networks, elevation correction and foundation type spatial optimization are completed. This overcomes the deficiency of traditional single-point selection lacking spatial consistency analysis, accurately identifying abnormal and critical nodes. The optimal elevation is determined through elevation perturbation simulation and multi-objective optimization, reducing the number of nodes requiring manual verification and significantly decreasing the workload. This invention is compatible with existing engineering design software workflows, requiring no changes to designers' operating habits. The entire process can be quickly and repeatedly executed during scheme iteration, improving the efficiency and rationality of wind turbine foundation layout design. The optimization results can be directly imported into CAD for automatic layout, making it applicable to various new energy engineering wind turbine foundation layout design scenarios. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of this application, 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 this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0016] 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: Figure 1 A schematic diagram illustrating a wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm provided in this application embodiment; Figure 2 This is a schematic diagram illustrating the construction process of the triangular mesh diagram provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of a graph convolutional network model; Figure 4 This is a schematic diagram of the critical node identification process provided in an embodiment of this application; Figure 5 This is a schematic diagram of the process for identifying abnormal nodes provided in an embodiment of this application; Figure 6 This is a schematic diagram of the optimal corrected elevation determination process provided in an embodiment of this application; Figure 7 This is a schematic diagram of a wind turbine layout elevation optimization system based on AI combined with elevation triangulation algorithm, provided for an embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] 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 application are information and data authorized by the user or fully authorized by all parties.

[0019] This method is applicable to the wind turbine foundation or photovoltaic support wind turbine foundation layout design stage, where site survey elevation data has been obtained and a triangulation model of the project area has been constructed using CASS software. Simultaneously, engineering design utilizes Excel, CASS, and CAD software for wind turbine location coordinate processing, elevation calculation, and wind turbine foundation type selection. The executing entity can be an elevation collaborative optimization processing system deployed on the engineering design unit's server or design terminal. This system can call the triangulation data output by CASS software, the wind turbine location coordinate data exported from CAD, and the preset engineering boundary condition parameters in Excel. It incorporates a graph convolutional network model and a graph attention network model, achieving seamless data interaction between various software through standardized data interfaces. The entire process of this method can be repeated when adjusting the wind turbine foundation location. It can be understood that the wind turbine location refers to the core point of the wind turbine layout, which is the actual location of the wind turbine foundation; the wind turbine foundation type refers to the specifications / model of the wind turbine foundation, such as cast-in-place piles, independent foundations, raft foundations, etc.

[0020] The following detailed description, in conjunction with specific embodiments, illustrates the implementation process of the wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm described in this application. It should be noted that this embodiment is merely for explaining this application and not for limiting the scope of protection of this application. Any conventional adjustments or substitutions made by those skilled in the art to the steps without departing from the concept of this application should be included within the scope of protection of this application.

[0021] like Figure 1 As shown in the figure, this application discloses a schematic diagram of a wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm, including the following method steps: S1, acquire the wind turbine location plane coordinate data, survey point elevation data, triangular network structure data constructed from the survey points, and engineering boundary condition parameters for each area; S2, taking the wind turbine location and survey point as nodes, construct a triangular network structure G1 based on the triangular network structure data. The triangular network structure G1 includes node features and edge features. The node features include initial elevation and terrain features. S3, input the triangular network structure G1 into the graph convolutional network model, correct the initial elevation of the nodes, and output the corrected elevation; S4. Based on the corrected elevation and the engineering boundary condition parameters, calculate the soil penetration depth of each wind turbine location node, determine the initial wind turbine foundation type according to the correspondence between the soil penetration depth and the preset foundation design depth, and identify the critical nodes in the critical zone of the wind turbine foundation type boundary. S5. Construct a spatial adjacency graph G2 for wind turbine locations using wind turbine location nodes. The edges of the spatial adjacency graph G2 are determined based on the spatial distance between nodes. Input the spatial adjacency graph G2, the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker into the graph attention network model. Output the spatially optimized wind turbine foundation type and identify abnormal nodes based on the spatially optimized wind turbine foundation type and the initial wind turbine foundation type. S6. For wind turbine location nodes where the initial wind turbine foundation type is inconsistent with the space-optimized wind turbine foundation type, perform elevation perturbation simulation, determine the optimal corrected elevation based on the preset multi-objective optimization function, and mark the nodes corresponding to the corrected elevations that cannot meet the engineering boundary conditions as manual verification nodes. S7. Repeat S4-S6 until the number of abnormal nodes no longer changes after two consecutive iterations. Take the current corrected elevation as the optimal corrected elevation, output the wind turbine foundation type determined according to the optimal corrected elevation as the optimal wind turbine foundation type, and output the manually verified node.

[0022] In some embodiments, for step S1, for example, basic data is automatically obtained from a specified storage location through a standardized data interface. All data is validated for validity and reasonableness, and invalid, erroneous or data that deviates from a reasonable range is removed to ensure that the input data is accurate and complete.

[0023] Specifically, the wind turbine location planar coordinate data can be obtained from exported files of CAD software, which are batch-extracted planar coordinate information from the wind turbine foundation layout drawing. The system automatically verifies whether the coordinate values ​​are within the geographical boundary of the engineering site and removes invalid coordinates that exceed the boundary. The survey point elevation data comes from the surveying results of the engineering site, containing the three-dimensional coordinate information of the survey points. The system automatically verifies the rationality of the elevation data and removes abnormal survey point data that deviates significantly from the terrain features. The triangular network structure data comes from irregular triangular network files constructed by CASS software based on the survey point elevation data, containing information such as the node connection relationship and geometric features of the triangular network. The system automatically parses the file topology to ensure that the triangular network has no topological errors. The engineering boundary condition parameters for each area come from the wind turbine foundation design parameters preset in Excel software, including the minimum embedment depth of the wind turbine foundation. Flood control elevation Pre-set foundation design depth for each type of wind turbine foundation The system automatically matches each engineering partition with its corresponding boundary condition parameters, achieving a one-to-one correspondence between partitions and parameters.

[0024] In some embodiments, for step S2, this step takes the standardized basic dataset obtained in S1 as input, integrates the wind turbine location and survey point as nodes, and constructs a triangular network graph structure G1 containing node features and edge features based on the triangular network structure data, realizing the digital encoding of terrain features and triangular network topological features. This solves the technical problem that traditional triangular networks can only represent elevation information and cannot capture multi-dimensional terrain geometric features. Its principle is to transform the triangular network in geospatial space into a graph structure that can be recognized by graph neural networks.

[0025] Please see Figure 2 , Figure 2 This is a schematic diagram of the triangular network diagram construction process provided in this application embodiment. In S201, the set of wind turbine location points and the set of survey points are merged and uniformly incorporated into the node set of the triangular network diagram structure. The system merges the set of wind turbine location points and the set of survey points and uniformly incorporates them into the node set of the triangular network diagram structure G1, so that the diagram structure includes both the actual measured survey points and the wind turbine location points that need to be optimized, ensuring the benchmark and specificity of elevation optimization. For each node in the node set, the system extracts its multi-dimensional features to construct a node feature vector. The node feature vector includes the initial elevation and terrain features, where the initial elevation... The original elevation value of the node, the survey point To measure the actual elevation, the location of the wind turbine site The initial elevation is obtained by CASS software through triangulation; the terrain features include the region number to which the node belongs. Distance from a node to the boundary of its region Local slope value of nodes neighborhood elevation variance of nodes The calculation methods and parameters for each terrain feature are determined as follows: Node's region number The system uses a point-to-polygon inclusion determination algorithm to automatically determine the engineering partition to which each node belongs based on the aforementioned engineering partition boundary information, and assigns a unique area number to each partition. This feature integrates engineering partition features with terrain features. The number is a sequence of natural numbers with no fixed value range, and is dynamically adjusted according to the number of engineering partitions. Distance from a node to the boundary of its region The system calculates the shortest perpendicular distance from a node to the boundary of its partition using a point-to-polygon perpendicular distance algorithm. The unit is meters, and the calculation accuracy can be configured to 0.01 meters. This feature can characterize the positional features of a node within a partition and provide a basis for subsequent boundary constraint embedding. Local slope value of the node The system takes the node as the center, extracts all adjacent triangles directly connected to the node in the triangulation network, fits the plane equation of each adjacent triangle by the least squares method, calculates the slope value of each plane equation and takes the arithmetic mean as the local slope value of the node, with a value range of 0~100%. This feature can characterize the terrain undulation features of the node's location. neighborhood elevation variance of a node The system extracts all neighboring nodes directly connected to the current node in the triangulation network and calculates the variance of the elevation difference between the neighboring nodes and the current node. This feature can characterize the elevation dispersion of the node's neighborhood. The larger the variance, the more severe the terrain undulation of the node's neighborhood, and the easier it is to generate interpolation errors. The value range is non-negative real numbers.

[0026] For example, the system normalizes all node features, mapping the numerical range of each feature to... To avoid training bias in graph neural networks due to different feature dimensions, the normalization formula adopts the minimum-maximum normalization formula, as shown in the following equation:

[0027] In the formula, For nodes No. Normalized values ​​of dimensional features For nodes No. The original values ​​of the dimensional features. For the first The global minimum of the dimensional feature. For the first The global maximum value of a feature.

[0028] In S202, the edges of the triangulated graph structure G1 are constructed based on all the edges of the triangulated mesh. The edge set of the triangulated graph structure G1 consists of all the edges of the triangulated mesh generated by CASS software; that is, an undirected edge is constructed between any two directly connected nodes in the triangulated mesh, representing the topological connection relationship between the nodes. For each edge in the edge set, the system extracts its three-dimensional features to construct an edge feature vector, and the edge features include the length of the edge. The absolute value of the elevation difference between the two endpoints of the edge A binary identifier indicating whether an edge crosses the boundary of the engineering area. The calculation method and parameters for each edge feature are determined as follows: Side length The system calculates the distance between two points in space using the formula for distance between two points. It is the straight-line distance between the two endpoints of the edge, in meters. The calculation accuracy can be configured to 0.01 meters, and the value range is positive real numbers. absolute value of the elevation difference between the two endpoints of the edge The absolute value of the initial elevation difference between the two endpoints of the calculated edge, in meters, with a configurable accuracy of 0.01 meters and a range of non-negative real numbers. A binary identifier indicating whether an edge crosses the boundary of the engineering area. If the two endpoints of an edge belong to different project zones, then If the two endpoints of an edge belong to the same project partition, then This feature can characterize whether an edge crosses an engineering boundary, providing a weighting basis for subsequent neighborhood aggregation.

[0029] After the system extracts the node feature vectors and edge feature vectors, it generates a complete triangular network graph structure G1. This graph structure includes the terrain geometry features, elevation features, zoning features and topological relationships of the engineering site. It transforms the traditional triangular network elevation data into graph structure data that can be processed by the graph neural network, realizes the fusion extraction of multi-dimensional terrain features, and improves the accuracy of elevation correction.

[0030] In some embodiments, for step S3, the graph convolutional network model corrects the elevation. Using the previously constructed triangular network structure G1 as input, it is imported into the graph convolutional network model to correct the initial elevation of the nodes, and the corrected elevation is output. This invention solves the technical problems of traditional triangulation interpolation errors being amplified in the critical zone of wind turbine foundation type boundaries and the propagation of local elevation anomalies through triangulation networks affecting the judgment of surrounding wind turbine locations. Its principle is to integrate multi-dimensional terrain features and engineering boundary conditions into the elevation correction process through neighborhood aggregation, boundary constraints and anomaly suppression of graph convolutional networks, thereby achieving accurate elevation correction.

[0031] Specifically, such as Figure 3 As shown, Figure 3 This is a schematic diagram of the core structure of a graph convolutional network model. The core structure of the graph convolutional network model is a three-layer sequential connection structure: a neighborhood aggregation layer, a boundary constraint embedding layer, and an anomaly suppression layer. The output of each layer becomes the input of the next layer. The model's input consists of the node feature matrix and adjacency matrix of the triangular graph structure G1, and the output is the corrected elevation of the nodes. .

[0032] The neighborhood aggregation layer, the first layer of the graph convolutional network model, aggregates the features of each node in its neighborhood, capturing the local terrain features of the node and achieving preliminary elevation aggregation and correction. This solves the problem that traditional triangulation interpolation only considers local triangles and does not consider the global neighborhood features of the nodes. For each node in the node set... This layer performs weighted aggregation of the initial elevations of neighboring nodes based on the edge features between nodes to obtain the first intermediate elevation. The calculation is shown in the following formula:

[0033] In the formula, For nodes In the triangular network graph structure G1, there is a set of all neighboring nodes. For nodes The initial elevation, For nodes For nodes Aggregate weights, Indicates that only nodes are included. The set of nodes themselves is added to the summation range, indicating that the initial elevation of each node also participates in the aggregation; node The aggregation range includes all its neighboring nodes as well as the node itself.

[0034] Aggregate weight The value of is defined as the fusion weight of the topological connection features and edge features between nodes, and its calculation is shown in the following formula:

[0035] In the formula, For nodes With nodes The length of the side between It is the arithmetic mean of the side lengths of all edges in the triangular network structure G1. It is calculated by the system by counting the side lengths of all edges. It is a global statistical parameter and its value range is positive real numbers. For nodes With nodes A binary identifier indicating whether the edges between them cross the boundary of the engineering area. Aggregate weights. The range of values ​​is The physical meaning of this is that the shorter the edge length between nodes, the greater the weight of neighborhood aggregation, meaning that the elevation of a node is more affected by its nearest neighbor nodes. If the edge between nodes crosses the boundary of an engineering partition, the aggregation weight is 0, meaning that neighborhood aggregation is not performed between nodes in different partitions, avoiding elevation correction errors caused by differences in boundary conditions between different partitions. For example, the system assigns aggregation weights to each node... Normalization is performed to ensure that the sum of the weights of all neighboring nodes and the node itself is 1, thus avoiding elevation shifts caused by excessively large or small weights.

[0036] The boundary constraint embedding layer is the second layer of the graph convolutional network model. It corrects the first intermediate elevation based on the engineering boundary condition parameters to obtain the second intermediate elevation. This ensures that the corrected elevation meets the engineering requirements for the minimum embedment depth of the wind turbine foundation, and solves the technical problem that the elevation interpolation in traditional design does not take into account engineering constraints, resulting in the foundation design embedment depth not meeting the design requirements.

[0037] This layer outputs the first intermediate elevation from the neighborhood aggregation layer. As input, calculate the nodes based on the engineering boundary condition parameters in S1. The corresponding wind turbine foundation is based on the first intermediate elevation. Calculated depth of penetration The calculation of the embedment depth follows the classic calculation formulas of engineering design and is consistent with the calculation logic of the embedment depth determined by the subsequent wind turbine foundation type. If This leads to the calculated depth of penetration. Less than the minimum embedment depth of the wind turbine foundation Then, the first intermediate elevation of that node is corrected upwards, and the correction formula is as follows:

[0038] In the formula, This is the elevation correction factor, which is a preset empirical factor with a value range of [value range missing]. It can be configured to 0.3, and its value is determined by the engineering designers based on their design experience with similar projects. If the value is too large, it may lead to excessive elevation correction; if the value is too small, it may not effectively meet the minimum soil depth constraint.

[0039] For example, the constraint correction at this level only applies to wind turbine foundation nodes. The elevation of survey point nodes is the measured elevation and is not subject to constraint correction, ensuring the accuracy of the survey points as elevation benchmarks. For nodes crossing regional boundaries, the system automatically matches the minimum embedment depth of the wind turbine foundation in its respective zone. This enables the embedding of boundary constraints that differentiate between partitions.

[0040] The anomaly suppression layer is the third layer of the graph convolutional network model. It applies weighted suppression to the second intermediate elevation based on the neighborhood elevation variance of each node to obtain the corrected elevation. This method suppresses the impact of local elevation anomalies on the elevation of surrounding nodes, solving the technical problem of local elevation anomalies propagating through the triangular network in traditional triangular network interpolation and causing deviations in the elevation judgment of surrounding wind turbine locations.

[0041] This layer embeds the second intermediate elevation output by the boundary constraint layer. Using the given input, calculate the global mean of the neighborhood elevation variances of all nodes in the triangular network structure G1. And set a neighborhood elevation variance threshold. , global mean The value is 1.5 times the average elevation, calculated by the system through statistical analysis of the neighborhood elevation variance of all nodes. It is a global statistical parameter, and its value range is a non-negative real number. For nodes... If its neighborhood elevation variance If the node is determined to be an elevation anomaly node, the system will reduce the elevation weight of that node by a weight suppression coefficient. The preset calibration coefficients have a range of values. It can be configured to 0.5; if If the node is normal, its weight remains at 1.

[0042] Final corrected elevation of the anomaly suppression layer The calculation is shown in the following formula:

[0043] In one embodiment, the graph convolutional network model is a spatial domain-based graph convolutional network, suitable for triangular mesh graph structures G1 with irregular topologies. The model is trained under supervision. The training dataset consists of site survey data, triangular mesh structure data, and measured elevation data of wind turbine foundations from multiple similar new energy projects. The dataset is divided into training, validation, and test sets in a 7:2:1 ratio. The model's training labels are the measured elevations of the survey points and the measured elevation data of the wind turbine foundations; that is, the model's prediction target is the actual elevation of the nodes. The loss function uses the mean squared error loss function, as shown in the following formula:

[0044] In the formula, For mean square error loss, This represents the number of nodes in the training set. The corrected elevation predicted by the model. The actual elevation label for the node.

[0045] The model's optimizer uses a stochastic gradient descent optimizer with a configurable learning rate of 0.001. Iterative training continues until the loss function converges, and the convergence threshold is configurable. During training, the system inputs a validation set into the model to verify the accuracy of elevation correction. If the mean square error of the validation set is less than a preset threshold, training continues; otherwise, the model's hyperparameters are adjusted and training is repeated. After model training is complete, a test set is input into the model for performance testing. If the mean square error of the test set meets the accuracy requirements of the engineering design, the model training is complete, and it is deployed in the elevation collaborative optimization processing system for elevation correction of the new engineering site triangular network structure. If the mean square error of the test set does not meet the accuracy requirements, the training dataset and hyperparameters are readjusted, and training is repeated.

[0046] In some embodiments, for step S4, the soil penetration depth is calculated to determine the initial wind turbine foundation type and critical nodes are identified. This is done using the corrected elevation. Using the aforementioned engineering boundary condition parameters as input, the soil penetration depth of each wind turbine location node is calculated. Based on the correspondence between the soil penetration depth and the preset foundation design depth, the initial wind turbine foundation type is determined, i.e., the wind turbine foundation specifications are initially determined based on the corrected elevation. Critical nodes in the critical zone of the wind turbine foundation type boundary (i.e., the interval where the foundation type is prone to change due to elevation changes) are identified. The principle is to combine the engineering boundary conditions and the corrected elevation, and calculate the soil penetration depth using the classic formula of engineering design to achieve the interval judgment of wind turbine foundation type and the accurate identification of critical zones.

[0047] Specifically, please refer to Figure 4 , Figure 4 This is a schematic diagram of the critical node identification process provided in an embodiment of this application. In S401, the soil penetration depth is calculated for each wind turbine location node in the node set. For each wind turbine location node in the node set... The system is based on its corrected elevation. Flood control elevation Preset foundation design depth corresponding to various types of wind turbine foundations Calculate the actual embedment depth of the wind turbine foundation. The formula for calculating the embedment depth is as follows:

[0048] In the formula, For wind turbine location nodes The depth of the wind turbine foundation embedded in the soil, in meters, with a calculation accuracy configurable to 0.01 meters; The design depth of each type of wind turbine foundation is preset for the zone to which the wind turbine location node belongs, in meters; This refers to the flood control elevation of the zone to which the wind turbine location belongs, in meters. This is the corrected elevation of the node, in meters.

[0049] For example, the system automatically matches the flood control elevation of the zone to which each wind turbine location node belongs. Burial depth with pre-designed foundation To ensure differentiated soil penetration depth calculations for different zones, for survey points and nodes, soil penetration depth calculations and wind turbine foundation type determinations are not performed; they are only used as elevation benchmarks.

[0050] In S402, the initial wind turbine foundation type is determined based on the correspondence between preset wind turbine foundation types and preset foundation design depths. The system has multiple preset wind turbine foundation types, each corresponding to a preset foundation design depth. For wind turbine location node i, based on the previously calculated embedment depth... The system selects one type of foundation from the preset wind turbine foundation types, and its corresponding design burial depth. satisfy ,Right now Deep enough to ensure that the burial depth is at least A, and at the same time This is the minimum design burial depth that meets this condition, taking into account both economic efficiency and cost-effectiveness. The selected type is the initial wind turbine foundation type. Engineering designers can modify the preset foundation design depth in Excel software according to actual design requirements. The system will automatically identify and update the judgment rules based on the correspondence with the type of wind turbine foundation, providing excellent flexibility.

[0051] In S403, the critical zone of the wind turbine foundation type boundary is identified for each wind turbine location node, and the critical node is marked. After completing the initial wind turbine foundation type determination, the system identifies and marks the critical zone of the wind turbine foundation type boundary for each wind turbine location node. First, the elevation of the wind turbine foundation type boundary is calculated by back-calculating the formula for the embedment depth. That is, the embedment depth of both different wind turbine foundation types meets the requirements. The corresponding elevation value is then used to set the critical zone elevation threshold. This threshold is a preset calibration parameter, and its value range is [value range missing]. The threshold can be configured to be 0.1 meters. Its value is determined by the engineering designers based on the accuracy requirements of the wind turbine foundation design. The higher the accuracy requirement, the lower the threshold can be.

[0052] If the corrected elevation of the wind turbine location node Boundary elevation of wind turbine foundation type The absolute value of the difference is less than the critical zone elevation threshold. If the node is located in the critical zone of the wind turbine foundation type boundary, it is marked as a critical node. The determination formula is as follows:

[0053] The marking of critical nodes can accurately identify nodes that are prone to sudden changes in wind turbine foundation type due to slight elevation changes, providing clear optimization targets for subsequent spatial consistency analysis and bidirectional coupling optimization. The technical effect is that based on the accurate corrected elevation, the interval judgment of wind turbine foundation type and the accurate identification of critical zone are realized, solving the problem of failure to identify critical zone in traditional design, which leads to sudden changes in wind turbine foundation type.

[0054] In some embodiments, for step S5, a spatial adjacency graph G2 of the wind turbine location is constructed and spatial optimization of the wind turbine foundation type is output to identify abnormal nodes. This is based on the initial wind turbine foundation type. Critical node markers and corrected elevation of S3 As input, a spatial adjacency graph G2 of wind turbine locations is constructed using the wind turbine location nodes. This graph, along with the corrected elevations of the nodes, the initial wind turbine foundation type, and the critical zone markers, is then input into the graph attention network model. The output is the spatially optimized wind turbine foundation type. It identifies abnormal nodes and solves the technical problems of traditional design, which only selects wind turbine foundations based on single-point elevation intervals, lacks spatial consistency analysis capabilities, and leads to the dispersion of local wind turbine foundation types from the surrounding areas. The principle is to construct a spatial adjacency graph of wind turbine locations and use the attention mechanism of graph attention networks to capture the spatial correlation characteristics between wind turbine locations, thereby achieving spatial optimization of wind turbine foundation types and accurate identification of abnormal nodes.

[0055] Please see Figure 5 , Figure 5 This is a schematic diagram of the abnormal node identification process provided in an embodiment of this application. In S501, a spatial adjacency graph containing only wind turbine location nodes is constructed. The system constructs a wind turbine location spatial adjacency graph G2 containing only wind turbine location nodes. Its node set is the wind turbine location set in S1, and its edge set is determined based on the spatial distance between the wind turbine location nodes, representing the spatial proximity relationship between wind turbine locations. Specifically, a spatial neighborhood radius is set with each wind turbine location node as the center. This parameter is a preset calibration parameter, and its value range is [value range missing]. The radius can be configured to 10 meters. Its value is determined by the engineering designers based on the density of the wind turbine foundations in the site. The denser the wind turbine foundations are, the smaller the radius can be. If another wind turbine location node is within the spatial neighborhood radius of the central node, an undirected edge is constructed between the two nodes to represent them as spatial neighbors.

[0056] For each node in the wind turbine location adjacency graph G2, the system extracts its feature vector, which contains the node's corrected elevation. Initial wind turbine foundation type Critical section marking Critical section marking For binary features, if a node is a critical node, then ,otherwise For each edge of the wind turbine location spatial adjacency graph G2, the system extracts its feature as the spatial Euclidean distance between two wind turbine location nodes, in meters, with a calculation accuracy configurable to 0.01 meters.

[0057] In S502, the graph attention network model is used for feature extraction and spatial optimization of wind turbine foundation type output. The system outputs the wind turbine location spatial adjacency graph G2 and the corrected elevation of the nodes. Initial wind turbine foundation type Critical section marking The input is fed into the graph attention network model for feature extraction and spatial optimization of the wind turbine foundation type output. The core of the graph attention network model is to assign differentiated attention weights to different neighboring nodes through the attention mechanism, accurately capture neighboring nodes that have a significant impact on the wind turbine foundation type of the central node, and achieve spatial consistency optimization of the wind turbine foundation type.

[0058] Specifically, the graph attention network model is a two-layer attention aggregation structure. The first layer is the attention aggregation layer, which performs attention-weighted aggregation on the neighborhood features of each node to extract the local spatial wind turbine foundation type features of the node. The second layer is the wind turbine foundation type classification layer, which performs spatial optimization classification of wind turbine foundation types based on the aggregated features and outputs the spatially optimized wind turbine foundation type. The model takes as input the node feature matrix and adjacency matrix of the spatial adjacency graph G2, and outputs the spatially optimized wind turbine foundation type for each wind turbine location node. The activation function used is the Softmax activation function to ensure that the output type classification result is a probability distribution, and the wind turbine foundation type with the highest probability is selected as the space-optimized wind turbine foundation type.

[0059] For example, the training process of the graph attention network model is supervised training. The training dataset consists of wind turbine foundation layout data, elevation data, and actual wind turbine foundation type design data from similar projects. The loss function is the cross-entropy loss function, and the optimizer is the Adam optimizer. Iterative training continues until the loss function converges. The convergence threshold can be configured as follows: The trained model is deployed in the elevation collaborative optimization processing system to extract spatial consistency features and optimize the wind turbine foundation type from the wind turbine location spatial adjacency graph G2.

[0060] In S503, abnormal nodes are identified by combining the average elevation difference between a node and its neighboring nodes. The system optimizes the type of wind turbine foundation based on the spatial optimization of the graph attention network model output. The initial wind turbine foundation type determined above The consistency index of the wind turbine foundation type for each wind turbine location node, i.e. the consistency ratio of the foundation type of neighboring locations, is calculated. Abnormal nodes are identified by combining the average elevation difference between the node and its neighboring nodes. This solves the technical problem of not being able to identify discrete nodes of local wind turbine foundation type in traditional design.

[0061] Calculate the wind turbine foundation type consistency index for each wind turbine location node. This index is the proportion of nodes in a node's spatial neighborhood that have the same initial wind turbine foundation type as that node, out of the total number of nodes in the neighborhood. It is calculated as follows:

[0062] In the formula, For nodes Spatial neighborhood and initial wind turbine foundation type Same number of nodes For nodes The total number of spatial neighborhood nodes, and the consistency index of wind turbine foundation types. The range of values ​​is , The closer it is to 0, the higher the degree of dispersion of the wind turbine foundation type of the node from its neighborhood, and the more likely it is to be an abnormal node; The closer a value is to 1, the higher the consistency between the node's machine type and its neighborhood, indicating that it is a normal node.

[0063] Subsequently, a threshold for the consistency index of wind turbine foundation types was set. This parameter is a preset calibration parameter, and its value range is [value range missing]. The threshold can be configured to 0.2, and its value is determined by the engineering designers based on the design requirements for spatial consistency of wind turbine foundation types. The higher the consistency requirement, the lower the threshold can be. This indicates the consistency index of the wind turbine foundation type at the wind turbine location node. Less than the wind turbine foundation type consistency index threshold Furthermore, the average elevation difference between the node and its neighboring nodes is less than the set critical zone elevation threshold. If a node is identified as an abnormal node, the system will add all abnormal nodes to the abnormal node set. .

[0064] The identification of abnormal nodes can accurately locate discrete nodes of wind turbine foundation types caused by elevation interpolation errors, unreasonable boundary condition settings, or excessive rigidity of critical zone boundaries. The spatial consistency analysis of wind turbine foundation types is realized through graph attention network model, which accurately captures the spatial correlation characteristics between wind turbine locations and effectively identifies abnormal nodes.

[0065] In some embodiments, for step S6, elevation perturbation simulation and multi-objective optimization determine the optimal corrected elevation. This embodiment performs elevation perturbation simulation on wind turbine location nodes where the initial wind turbine foundation type differs from the spatially optimized wind turbine foundation type. Based on a preset multi-objective optimization function, the optimal corrected elevation is determined. Nodes whose corrected elevations cannot meet the engineering boundary conditions are marked as manually verified nodes. This solves the technical problem in traditional design where elevation and wind turbine foundation type lack reverse reasoning capabilities and cannot achieve bidirectional coupled optimization. The principle is to simulate the impact of minute elevation changes on the wind turbine foundation type through elevation perturbation simulation, and combine this with multi-objective optimization to achieve a comprehensive optimization of indicators such as elevation adjustment amount and the number of wind turbine foundation type jumps (i.e., the number of locations where the foundation type changes after optimization). Finally, under the premise of meeting the engineering boundary conditions, bidirectional coupled optimization of elevation and wind turbine foundation type is achieved.

[0066] Please see Figure 6 , Figure 6 This is a schematic diagram of the optimal corrected elevation determination process provided in an embodiment of this application. In S601, the initial wind turbine foundation type for each wind turbine location node is determined. Space-optimized fan foundation types Compare the results to determine whether elevation perturbation simulation and multi-objective optimization are necessary. This indicates that the wind turbine foundation type of this node is consistent with the results of single-point judgment and spatial consistency analysis, and there is no type discrepancy problem. Therefore, its corrected elevation should be maintained. No change, no optimization needed; if This indicates that the wind turbine foundation type at this node exhibits spatial dispersion, requiring further investigation to determine whether a slight change in elevation causes a jump in the wind turbine foundation type. If the corrected elevation of this node... Fine-tuning within a small range Is it possible to make To restore spatial consistency, if the condition is met, the node is included in the set of nodes to be optimized for subsequent elevation perturbation simulation; otherwise, it is directly marked as a manually verified node.

[0067] In S602, elevation perturbation simulation is performed on all nodes in the set of nodes to be optimized. The system performs elevation perturbation simulation on all nodes in the set of nodes to be optimized, that is, it performs elevation perturbation simulation on the nodes after correction. Small-scale vertical adjustments were made to simulate changes in wind turbine foundation types at different elevations, verifying whether the adjusted elevations could restore spatial consistency of wind turbine foundation types and meet engineering boundary conditions.

[0068] Set the perturbation step size This parameter is a preset calibration parameter, and its value range is [value range missing]. The value of the elevation calculation is calibrated by the engineering designers based on the accuracy of the elevation calculation, and is consistent with the accuracy of the elevation calculation. The system optimizes each node... Calculate their elevations and adjust them accordingly. and Subsequent candidate elevations and The soil depth corresponding to the candidate elevation is recalculated, and the type of wind turbine foundation is re-determined based on the correspondence between the soil depth and the preset foundation design depth.

[0069] For a generated candidate elevation, it is determined to be a feasible candidate elevation if the following three conditions are met: the adjusted wind turbine foundation type is consistent with the space-optimized wind turbine foundation type, that is, the re-determined type is equal to... The adjusted elevation corresponds to a soil depth greater than or equal to the minimum soil depth of the wind turbine foundation. The adjusted elevation meets the engineering boundary conditions and does not violate other engineering boundary conditions, such as flood control elevation constraints. and If all elevations are feasible candidate elevations, the system selects the one with the smaller elevation adjustment amount as the final candidate elevation for that node; if only one elevation is feasible candidate, that elevation is selected as the final candidate elevation; if none of them are feasible elevations, the node is marked as a node for manual review.

[0070] In S603, multi-objective optimization determines the optimal corrected elevation. Multi-objective optimization is performed on the final candidate elevations of all nodes to be optimized. A multi-objective optimization function is defined, and the optimal corrected elevation for each node is determined by minimizing this function. This enables bidirectional coupling optimization of elevation and wind turbine foundation type.

[0071] The multi-objective optimization function includes four sub-objectives: elevation adjustment, number of wind turbine foundation type jumps, number of engineering boundary violations, and spatial dispersion. All objectives are minimized. The multi-objective optimization function is shown in the following equation:

[0072] In the formula, To optimize the function value for multiple objectives, , , , This refers to the weighting coefficients for each sub-objective. These coefficients are preset calibration parameters, and their values ​​range from [value range missing]. And satisfy Its value is determined by the engineering designers according to the design requirements. For example, if the project has higher requirements for the accuracy of elevation correction, then... A larger value can be taken; if the control requirements for the abrupt changes in the wind turbine foundation type are higher, then... A larger value can be taken.

[0073] The definitions and calculation methods for each sub-objective are as follows: The elevation adjustment is represented by the final candidate elevation of the node to be optimized and the original corrected elevation. The absolute value of the difference represents the magnitude of the elevation adjustment. The smaller the value, the gentler the elevation correction and the more it matches the actual characteristics of the terrain. This represents the number of jumps in the wind turbine foundation type, and the number of nodes where the optimized wind turbine foundation type differs from the initial wind turbine foundation type. The smaller the value, the fewer the type jumps, and the higher the consistency of the wind turbine foundation design; Here, represents the number of violations of the engineering boundary conditions, and represents the number of nodes whose elevations violate the engineering boundary conditions after optimization. The smaller the value, the better the optimized elevation meets the engineering design requirements; Let be the spatial dispersion, and be the variance of the consistency index of all node wind turbine foundation types after optimization. The smaller the value, the higher the spatial consistency of the optimized wind turbine foundation type and the lower the degree of dispersion.

[0074] For example, the system uses a particle swarm optimization algorithm to solve a multi-objective optimization function. In this algorithm, the particle position is the elevation value of the node to be optimized, and the particle velocity is the elevation adjustment amount. Through iterative search, the algorithm finds the optimal solution for the objective function. The position of the smallest particle is the optimal corrected elevation of the node to be optimized. The hyperparameters of the particle swarm optimization algorithm can be configured as follows: number of particles 50-200, number of iterations 100-500, learning factor 1.5-2.0, and inertia weight 0.5-0.9.

[0075] In S604, manual verification node marking is implemented. For nodes to be optimized that have no feasible candidate elevations after elevation perturbation simulation, the system marks them as manual verification nodes and records information such as the node number, 3D coordinates, corrected elevation, initial wind turbine foundation type, and spatially optimized wind turbine foundation type. This provides complete data support for manual verification by engineering designers. Marking manually verified nodes can accurately locate discrete nodes of wind turbine foundation types that cannot be resolved through algorithm optimization, avoiding the limitations of algorithm optimization and improving the rationality of wind turbine foundation design.

[0076] In some embodiments, for step S7, S4 to S6 are repeated until the number of abnormal nodes no longer changes after two consecutive iterations. The current corrected elevation is taken as the optimal corrected elevation, and the wind turbine foundation type determined based on the optimal corrected elevation is taken as the optimal wind turbine foundation type. Manual verification nodes are also output to achieve iterative optimization of elevation and wind turbine foundation type, ensuring the stability and rationality of the optimization results.

[0077] Specifically, the system will determine the optimal corrected elevation in S6. Substitute the values ​​from S4, recalculate the embedment depth of each wind turbine location node, determine the new initial wind turbine foundation type, and identify new critical nodes; then mark the new initial wind turbine foundation type, critical nodes, and optimal corrected elevation. Substitute S5, reconstruct the wind turbine location spatial adjacency graph G2, output a new spatially optimized wind turbine foundation type, and identify a new set of abnormal nodes. And count the number of new abnormal nodes. ,in This represents the current iteration number. The system will then... Number of abnormal nodes in the next iteration With the Number of abnormal nodes in the next iteration To make a comparison, if If the iteration converges, the optimization is stopped; if Then the new set of abnormal nodes will be... As input, repeat the elevation perturbation simulation and multi-objective optimization of S6 for the next iteration.

[0078] After iterative convergence, the system will use the optimal corrected elevation obtained from the last round of multi-objective optimization. As the final optimal corrected elevation, the wind turbine foundation type determined after calculating the embedment depth based on this optimal corrected elevation will be taken as the optimal wind turbine foundation type. The system then aggregates information from all manually verified nodes and finally outputs the optimal corrected elevation. Optimal wind turbine foundation type Together with the set of manually reviewed nodes, it provides data support for subsequent automatic CAD layout.

[0079] For example, the system can also adjust the optimal elevation. Optimal wind turbine foundation type The system imports manually verified node information into CAD software, automatically drawing the wind turbine foundation locations. It also labels the corresponding foundation type information based on the optimal foundation type and adds unique verification marks to manually verified nodes, enabling rapid and accurate layout of wind turbine foundations. Simultaneously, Excel software can quickly summarize and statistically analyze the specifications of similar wind turbine foundations and import the results into CAD as a data overview, facilitating viewing and use by engineering designers.

[0080] Therefore, this method achieves bidirectional coupled collaborative optimization of elevation and wind turbine foundation type, effectively suppressing the amplification effect of triangular network interpolation error in the critical region of wind turbine foundation type, solving the problem of local wind turbine foundation type dispersion caused by single-point selection, accurately identifying and marking abnormal nodes and manually verified nodes, and significantly reducing the workload of manual verification. Simultaneously, the method is compatible with the existing CASS-Excel-CAD engineering design process, adding only an AI collaborative optimization layer in the middle, without changing the designers' operating habits. The entire process can be quickly and repeatedly executed during scheme iteration, improving the efficiency and accuracy of wind turbine and photovoltaic power plant foundation design. The optimized elevation and wind turbine foundation type results can be directly imported into CAD for automatic layout, and Excel can perform rapid statistical summarization, achieving efficient management of design data. It is applicable to wind turbine foundation layout design scenarios for various new energy projects.

[0081] It should be noted that although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0082] Please see Figure 7 , Figure 7 This application provides a structural block diagram of a wind turbine layout elevation optimization system based on AI combined with an elevation triangulation network algorithm, as an embodiment of the present application. The system specifically includes: The data acquisition module 701 is used to acquire wind turbine location plane coordinate data, survey point elevation data, triangular network structure data constructed from survey points, and engineering boundary condition parameters for each area; The first graph structure construction module 702 is used to construct a triangular network graph structure G1 based on the triangular network structure data, using the wind turbine location and survey point as nodes. The triangular network graph structure G1 includes node features and edge features. The node features include initial elevation and terrain features. The elevation correction module 703 is used to input the triangular network structure G1 into the graph convolutional network model, correct the initial elevation of the nodes, and output the corrected elevation. The wind turbine foundation type initial determination and critical identification module 704 is used to calculate the soil penetration depth of each wind turbine location node based on the corrected elevation and the engineering boundary condition parameters, determine the initial wind turbine foundation type according to the correspondence between the soil penetration depth and the preset foundation design depth, and identify the critical nodes in the critical zone of the wind turbine foundation type boundary. The spatial optimization module 705 is used to construct a spatial adjacency graph G2 for wind turbine locations based on wind turbine location nodes. The edges of the spatial adjacency graph G2 are determined according to the spatial distance between nodes. The spatial adjacency graph G2, along with the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker, are input into the graph attention network model. The module outputs the spatially optimized wind turbine foundation type and identifies abnormal nodes based on the spatially optimized wind turbine foundation type and the initial wind turbine foundation type. The perturbation optimization module 706 is used to perform elevation perturbation simulation on wind turbine location nodes where the initial wind turbine foundation type is inconsistent with the spatially optimized wind turbine foundation type. Based on the preset multi-objective optimization function, the optimal corrected elevation is determined, and nodes that cannot obtain the corrected elevation that meets the engineering boundary conditions are marked as manually verified nodes. The iterative output module 707 is used to repeatedly execute the initial determination and critical identification module for wind turbine foundation type, the spatial optimization module, and the perturbation optimization module until the number of abnormal nodes no longer changes after two consecutive iterations. The current corrected elevation is taken as the optimal corrected elevation, the wind turbine foundation type determined according to the optimal corrected elevation is taken as the optimal wind turbine foundation type, and the manually verified node is output.

[0083] Preferably, the terrain features include the region number to which the node belongs, the distance from the node to the boundary of the region, the local slope value of the node, and the neighborhood elevation variance of the node; The edge features include the edge length, the absolute value of the elevation difference between the two endpoints of the edge, and a binary identifier indicating whether the edge crosses the boundary of the engineering area.

[0084] It should be noted that the working process of each module in the wind turbine layout elevation optimization system based on AI combined with elevation triangulation algorithm described in this embodiment can refer to the working process of the wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm described in the above embodiment. The technical effect achieved is also the same as that of the wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm described in the above embodiment, and will not be repeated here.

[0085] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for optimizing wind turbine layout elevation based on AI combined with elevation triangulation algorithm, characterized in that, Includes the following steps: S1, acquire the wind turbine location plane coordinate data, survey point elevation data, triangular network structure data constructed from the survey points, and engineering boundary condition parameters for each area; S2, taking the wind turbine location and survey point as nodes, construct a triangular network structure G1 based on the triangular network structure data. The triangular network structure G1 includes node features and edge features. The node features include initial elevation and terrain features. S3, input the triangular network structure G1 into the graph convolutional network model, correct the initial elevation of the nodes, and output the corrected elevation; S4. Based on the corrected elevation and the engineering boundary condition parameters, calculate the soil penetration depth of each wind turbine location node, determine the initial wind turbine foundation type according to the correspondence between the soil penetration depth and the preset foundation design depth, and identify the critical nodes in the critical zone of the wind turbine foundation type boundary. S5. Construct a spatial adjacency graph G2 for wind turbine locations using wind turbine location nodes. The edges of the spatial adjacency graph G2 are determined based on the spatial distance between nodes. Input the spatial adjacency graph G2, the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker into the graph attention network model. Output the spatially optimized wind turbine foundation type and identify abnormal nodes based on the spatially optimized wind turbine foundation type and the initial wind turbine foundation type. S6. For wind turbine location nodes where the initial wind turbine foundation type is inconsistent with the space-optimized wind turbine foundation type, perform elevation perturbation simulation, determine the optimal corrected elevation based on the preset multi-objective optimization function, and mark the nodes corresponding to the corrected elevations that cannot meet the engineering boundary conditions as manual verification nodes. S7. Repeat S4-S6 until the number of abnormal nodes no longer changes after two consecutive iterations. Take the current corrected elevation as the optimal corrected elevation, output the wind turbine foundation type determined according to the optimal corrected elevation as the optimal wind turbine foundation type, and output the manually verified node.

2. The wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm as described in claim 1, characterized in that, The terrain features include the region number to which the node belongs, the distance from the node to the boundary of the region, the local slope value of the node, and the elevation variance of the node's neighborhood. The edge features include the edge length, the absolute value of the elevation difference between the two endpoints of the edge, and a binary identifier indicating whether the edge crosses the boundary of the engineering area.

3. The method for optimizing wind turbine layout elevation based on AI combined with elevation triangulation algorithm as described in claim 1, characterized in that, The graph convolutional network model includes a neighborhood aggregation layer, a boundary constraint embedding layer, and an anomaly suppression layer. The neighborhood aggregation layer performs weighted aggregation of the initial elevations of neighboring nodes based on the edge features between nodes to obtain a first intermediate elevation. The boundary constraint embedding layer performs constraint correction on the first intermediate elevation based on engineering boundary condition parameters to obtain a second intermediate elevation. The anomaly suppression layer performs weighted suppression on the second intermediate elevation based on the variance of the neighborhood elevation of the nodes to obtain the corrected elevation.

4. The wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm according to claim 1, characterized in that, The graph attention network model calculates the attention weights between nodes based on the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker. The attention weights are calculated only for the neighboring nodes in the wind turbine location spatial adjacency graph G2, and are used to weighted aggregate the features of the neighboring nodes to output the spatially optimized wind turbine foundation type.

5. The wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm according to claim 4, characterized in that, The identification of abnormal nodes includes: calculating the wind turbine foundation type consistency index for each wind turbine location node, wherein the wind turbine foundation type consistency index is the proportion of the number of nodes in the node's spatial neighborhood that have the same initial wind turbine foundation type as the node to the total number of nodes in the neighborhood; if the wind turbine foundation type consistency index is less than a preset threshold and the average elevation difference between the node and its neighboring nodes is less than a preset critical zone elevation threshold, then the node is determined to be an abnormal node.

6. The wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm according to claim 1, characterized in that, The multi-objective optimization function includes four sub-objectives: elevation adjustment, number of wind turbine foundation type jumps, number of engineering boundary violations, and spatial dispersion. The elevation adjustment is the absolute value of the difference between the candidate elevation and the corrected elevation of a node. The number of wind turbine foundation type jumps is the number of nodes whose optimized wind turbine foundation type differs from the initial wind turbine foundation type. The number of engineering boundary violations is the number of nodes whose optimized elevation violates the engineering boundary conditions. The spatial dispersion is the variance of the consistency index of wind turbine foundation types across all nodes.

7. The wind turbine layout elevation optimization method based on AI combined with elevation triangulation algorithm according to claim 1, characterized in that, The elevation perturbation simulation includes increasing and decreasing the corrected elevation of the nodes with a preset perturbation step size to generate candidate elevations, and verifying whether the wind turbine foundation type corresponding to the candidate elevation meets the engineering boundary conditions.

8. The method for optimizing wind turbine layout elevation based on AI combined with elevation triangulation algorithm according to claim 1, characterized in that, It also includes importing the optimal corrected elevation, optimal wind turbine foundation type, and manual verification node information into CAD software, and automatically drawing the wind turbine foundation points and marking the wind turbine foundation type and verification marks in the CAD software.

9. A wind turbine layout elevation optimization system based on AI combined with elevation triangulation algorithm, characterized in that, include: The data acquisition module is used to acquire wind turbine location plane coordinate data, survey point elevation data, triangular network structure data constructed from survey points, and engineering boundary condition parameters for each area; The first graph structure construction module is used to construct a triangular network graph structure G1 based on the triangular network structure data, using the wind turbine location and survey point as nodes. The triangular network graph structure G1 includes node features and edge features. The node features include initial elevation and terrain features. The elevation correction module is used to input the triangular network structure G1 into the graph convolutional network model, correct the initial elevation of the nodes, and output the corrected elevation. The wind turbine foundation type initial determination and critical identification module is used to calculate the soil penetration depth of each wind turbine location node based on the corrected elevation and the engineering boundary condition parameters, determine the initial wind turbine foundation type according to the correspondence between the soil penetration depth and the preset foundation design depth, and identify the critical nodes in the critical zone of the wind turbine foundation type boundary. The spatial optimization module is used to construct a spatial adjacency graph G2 for wind turbine locations based on wind turbine location nodes. The edges of the spatial adjacency graph G2 are determined according to the spatial distance between nodes. The spatial adjacency graph G2, along with the corrected elevation of the nodes, the initial wind turbine foundation type, and the critical zone marker, are input into the graph attention network model. The module outputs the spatially optimized wind turbine foundation type and identifies abnormal nodes based on the spatially optimized wind turbine foundation type and the initial wind turbine foundation type. The perturbation optimization module is used to perform elevation perturbation simulation on wind turbine location nodes where the initial wind turbine foundation type is inconsistent with the spatially optimized wind turbine foundation type. Based on the preset multi-objective optimization function, the optimal corrected elevation is determined, and nodes that cannot obtain corrected elevations that meet the engineering boundary conditions are marked as manually verified nodes. The iterative output module is used to repeatedly execute the initial determination and critical identification module for wind turbine foundation type, the spatial optimization module, and the perturbation optimization module until the number of abnormal nodes no longer changes after two consecutive iterations. The current corrected elevation is taken as the optimal corrected elevation, the wind turbine foundation type determined according to the optimal corrected elevation is taken as the optimal wind turbine foundation type, and the manually verified node is output.

10. The wind turbine layout elevation optimization system based on AI combined with elevation triangulation algorithm according to claim 9, characterized in that, The terrain features include the region number to which the node belongs, the distance from the node to the boundary of the region, the local slope value of the node, and the elevation variance of the node's neighborhood. The edge features include the edge length, the absolute value of the elevation difference between the two endpoints of the edge, and a binary identifier indicating whether the edge crosses the boundary of the engineering area.

Images