Heliostat feature extraction method, heliostat pneumatic simulation method and equipment

Abstract

The invention discloses a heliostat feature extraction method, a heliostat pneumatic simulation method and equipment, and belongs to the technical field of heliostat simulation. An existing heliostat point cloud processing method lacks a feature mining means, cannot extract geometric features of a heliostat, and cannot be used for pneumatic simulation of the heliostat. The heliostat feature extraction method comprises the following steps: sampling heliostat point cloud scanning data by constructing a heliostat preprocessing model, a feature mining model and a feature fusion model to obtain heliostat key points; secondly, performing feature mining on the key points of the heliostat to obtain multi-dimensional geometric features so as to capture local features and global features of the heliostat; and performing nonlinear transformation on the multi-dimensional geometric features to obtain a structural feature quantity, and coupling the structural feature quantity with an environment feature quantity and a heliostat elevation angle to generate a fusion feature vector, so that accurate mining and accurate extraction can be performed on the heliostat features, and the method can be further used for pneumatic simulation of the heliostat.

Description

TECHNICAL FIELD

[0001] The application relates to a heliostat feature extraction method, a heliostat aerodynamic simulation method and equipment, and belongs to the technical field of heliostat simulation. BACKGROUND

[0002] As the core equipment of a tower type photo-thermal power station, the key feature extraction and accurate prediction of the aerodynamic coefficient of a heliostat play a key role in the stable operation and efficient power generation of the power station.

[0003] Further, a Chinese patent application (publication number: CN119756227A) discloses a heliostat back surface type detection method, which comprises the following steps: S1, collecting position information of a plurality of points on the back surface of a to-be-detected heliostat to form back surface point cloud data; the position information comprises a collection distance, a pitch angle and an azimuth angle; S2, sequentially performing coordinate conversion and rotation translation on the back surface point cloud data to obtain rotated point cloud data; the normal vector of the rotated point cloud data is parallel to the coordinate axis, and the center point of the rotated point cloud data coincides with the coordinate axis origin; S3, fitting the back surface type of the to-be-detected heliostat according to the rotated point cloud data to obtain a back surface fitting type, and determining a back surface type detection result of the to-be-detected heliostat according to the back surface fitting type.

[0004] The above-mentioned scheme provides a heliostat back surface point cloud processing method, but the scheme is mainly used for fitting the back surface structure of a heliostat, and therefore lacks feature mining means when processing point cloud data, cannot extract the geometric features of the heliostat, and cannot be used for aerodynamic simulation of the heliostat.

[0005] The information disclosed in the background section merely serves to enhance the understanding of the background of the present inventive concept, and therefore it can include information that does not constitute the prior art. SUMMARY

[0006] In view of the above problems or one of the above problems, the purpose of the present application is to provide a heliostat feature extraction method, a heliostat aerodynamic simulation method and equipment, which can accurately mine and accurately extract the features of a heliostat by constructing a heliostat preprocessing model, a feature mining model and a feature fusion model, sampling heliostat point cloud scanning data to obtain heliostat key points, mining the features of the heliostat key points to obtain multi-dimensional geometric features to capture the local features and global features of the heliostat, performing nonlinear transformation on the multi-dimensional geometric features to obtain structure feature quantities, coupling the structure feature quantities with environmental feature quantities and a heliostat elevation angle to generate a fusion feature vector, and thus the features of the heliostat can be accurately mined and accurately extracted, and the heliostat can be used for aerodynamic simulation, and the scheme is scientific, reasonable and feasible.

[0007] To solve the above problems or one of the above problems, the second object of the present application is to provide a heliostat aerodynamic simulation method, which fully considers the mutual influence and interaction between the multi-modal variables of the heliostat, can effectively model the multi-physical field interaction of the heliostat, can accurately reflect the aerodynamic characteristics of the heliostat under complex environment, and makes the aerodynamic coefficient prediction result close to the actual situation; and through simulation calculation by the algorithm model, the modeling cost can be effectively reduced, the simulation time is shorter, the heliostats of different design schemes can be quickly evaluated and optimized, the real-time requirement is met, the scheme is scientific, reasonable and feasible.

[0008] To achieve one of the above objects, the first technical solution of the present application is: A heliostat feature extraction method, comprising the following steps: Step one, obtaining heliostat point cloud scanning data; Step two, using a pre-constructed heliostat preprocessing model, based on curvature weight, sampling the heliostat point cloud scanning data to obtain heliostat key points; Step three, using a pre-constructed feature mining model, performing feature processing on the heliostat key points to obtain multi-dimensional geometric features to capture local features and global features of the heliostat; Step four, using a pre-constructed feature fusion model, performing nonlinear transformation on the multi-dimensional geometric features to obtain structural feature quantities, coupling the structural feature quantities with environmental feature quantities of the heliostat and the heliostat elevation angle to generate a fusion feature vector, and realizing feature extraction of the heliostat.

[0009] The present application samples the heliostat point cloud scanning data by constructing a heliostat preprocessing model, a feature mining model and a feature fusion model to obtain heliostat key points, then mines features of the heliostat key points to obtain multi-dimensional geometric features to capture local features and global features of the heliostat, and then performs nonlinear transformation on the multi-dimensional geometric features to obtain structural feature quantities, and couples the structural feature quantities with environmental feature quantities and the heliostat elevation angle to generate a fusion feature vector, so that the features of the heliostat can be accurately mined and precisely extracted, and the heliostat aerodynamic simulation can be used, the scheme is scientific, reasonable and feasible.

[0010] Further, while effectively preserving key geometric structure information, the present application can significantly reduce the amount of point cloud data, thereby greatly improving the data processing efficiency; and by using curvature weight and nonlinear transformation mechanism, the aerodynamic response features of the heliostat surface can be fully captured; at the same time, through the multi-modal feature fusion strategy, the coupling and expression ability between different modalities are effectively enhanced, so that the extracted features are more discriminative and representative, and therefore more efficient and accurate data processing and feature expression ability can be achieved.

[0011] As a preferred technical measure: Step two, based on the curvature weight, the method for sampling the heliostat point cloud scanning data to obtain the heliostat key point is as follows by using the pre-constructed heliostat preprocessing model: According to the heliostat point cloud scanning data, a point cloud set is established, which includes a plurality of discrete points, and a selected point set is constructed; Based on a plurality of discrete points, the maximum distance and the maximum curvature value between the discrete points are calculated; The minimum distance from a discrete point to all discrete points in the selected point set is calculated; Based on the covariance matrix, the curvature weight of the discrete point is determined; According to the curvature weight, the minimum distance, the maximum distance and the maximum curvature value of the discrete point, the sampling value of the discrete point is calculated; Based on the sampling values of a plurality of discrete points, a sampling distribution is established, and the maximum value and the discrete point corresponding to the maximum value are selected from the sampling distribution; The discrete point corresponding to the maximum value is taken as the heliostat key point to represent the structure of the high curvature area of the heliostat.

[0012] As a preferred technical measure: Based on the covariance matrix, the curvature weight of the discrete point is determined as follows: According to a discrete point, the points in the neighborhood and the neighborhood point centroid are obtained; According to the points in the neighborhood and the neighborhood point centroid, a covariance matrix of the points in the neighborhood of the discrete point is established; The covariance matrix is subjected to eigenvalue decomposition to obtain three eigenvalues; According to the three eigenvalues, the curvature is calculated; The curvature is taken as the curvature weight of the discrete point.

[0013] As a preferred technical measure: Step three, the method for feature processing of the heliostat key point to obtain multi-dimensional geometric features by using the pre-constructed feature mining model is as follows: The covariance matrix of the points in the neighborhood of a heliostat key point is calculated; The curvature is obtained by eigenvalue decomposition of the covariance matrix; The eigenvector corresponding to the minimum eigenvalue of the covariance matrix is taken as the normal feature; The curvature and the normal feature are taken as the geometric feature, and the feature vector of the heliostat key point is constructed based on the geometric feature; The feature vectors of all heliostat key points are combined to obtain a geometric feature matrix; The multi-dimensional geometric features are obtained by sampling, grouping and feature extraction of the geometric feature matrix.

[0014] As a preferred technical measure: The method for sampling, grouping and feature extraction of the geometric feature matrix to obtain multi-dimensional geometric features is as follows: The geometric feature matrix is processed by fixed time period or sliding window sampling method to obtain several representative features; According to each representative feature, a local coordinate system is constructed, and based on the local feature coordinate system, a plurality of adjacent representative features are found; Based on the plurality of representative features, a plurality of neighborhood feature sets are established to capture the local structural features of the heliostat surface; Using the plurality of neighborhood feature sets, the corresponding heliostat key points are determined; According to the heliostat key points, the representative features in the neighborhood feature set are spliced with the original geometric features to form an enhanced feature representation for retaining basic geometric information and fusing local context semantics; Based on the enhanced feature representation, a field feature region is constructed; The field feature region is subjected to a pooling operation to extract the most significant features representing the field feature region; The plurality of most significant features are aggregated to obtain multi-dimensional geometric features to describe global features.

[0015] As a preferred technical measure: Step four, using a pre-constructed feature fusion model to perform nonlinear transformation on the multi-dimensional geometric features to obtain structural feature quantities, the method is as follows: Obtain multi-dimensional geometric features, the data dimension of which is ; Using nonlinear activation function and normalization operation, the multi-dimensional geometric features are subjected to nonlinear transformation to obtain a first simplified feature vector of dimension, which is used to mine the potential information hidden in the multi-dimensional geometric features; Using nonlinear activation function and normalization operation, the first simplified feature vector is subjected to nonlinear transformation to obtain a second simplified feature vector of dimension; Using nonlinear activation function and normalization operation, the second simplified feature vector is subjected to nonlinear transformation to obtain a third simplified feature vector of dimension, so as to form a structural feature quantity which can retain discriminability and be aligned and fused.

[0016] Further, The value of d is 312 or 256 or 128.

[0017] As a preferred technical measure: The method for extracting the characteristics of the heliostat is as follows The environmental characteristic quantity is obtained, which includes a 3-dimensional wind speed characteristic and a 1-dimensional air density characteristic; The multi-layer perception network for the wind speed characteristic is constructed, the 3-dimensional wind speed characteristic is nonlinearly transformed, and a 3-dimensional wind speed characteristic vector is obtained The multi-layer perception network for the air density characteristic is constructed, the 1-dimensional air density characteristic is nonlinearly transformed, and a 1-dimensional air density characteristic vector is obtained The heliostat elevation angle is obtained, and the heliostat elevation angle is encoded to form a 2-dimensional elevation angle characteristic vector for representing the spatial posture of the heliostat; The multi-layer perception network for the heliostat elevation angle characteristic is constructed, the 2-dimensional elevation angle characteristic vector is nonlinearly transformed, and the multi-layer perception network is guided to extract discriminative representations from the periodicity and directionality of the heliostat elevation angle, and a 2-dimensional heliostat elevation angle characteristic vector is obtained Then, the structural characteristic quantity, the wind speed characteristic vector, the air density characteristic vector and the heliostat elevation angle characteristic vector are spliced in rows to obtain the fusion characteristic vector.

[0018] As a preferred technical measure: The method for obtaining the air density characteristic is as follows: The air density data of the environment around the heliostat is collected; The air density data is sorted to obtain a minimum density value and a maximum density value; Based on the minimum density value and the maximum density value, the air density data is normalized to obtain the air density characteristic with a uniform order of magnitude; Or / and, the method for obtaining the heliostat elevation angle characteristic is as follows: Firstly, based on the points of the heliostat point cloud scanning data, a plane equation is fitted to determine a heliostat plane equation and a reference plane equation; Then, the normal vector one of the heliostat plane equation and the normal vector two of the reference plane equation are calculated according to the heliostat plane equation and the reference plane equation; According to the vector angle formula, the included angle between the normal vector one and the normal vector two is calculated; Based on the vector angle, the elevation angle of the heliostat is determined; The elevation angle of the heliostat is converted into a numerical characteristic by using the trigonometric function coding to obtain a 2-dimensional heliostat elevation angle characteristic.

[0019] To achieve one of the above purposes, a second technical scheme of the present application is: ​​​A heliostat aerodynamic simulation method, applying the heliostat feature extraction method, constructs an aerodynamic coefficient prediction model, which includes the following steps: Step one, process the fusion feature vector of the heliostat, generate multiple modal features; Step two, mine the interaction between modal features, get modal coupling data; Step three, introduce graph convolution operation, represent modal coupling data as graph structure, and perform message propagation on the graph structure to represent the dependency relationship between modal features, form graph embedding representation information; Step four, aggregate and map the graph embedding representation information to get the predicted heliostat aerodynamic coefficient, realize heliostat aerodynamic simulation.

[0020] The present application fully considers the mutual influence and interaction between the multi-modal variables of the heliostat, can effectively model the multi-physical field interaction of the heliostat, can accurately reflect the aerodynamic characteristics of the heliostat under complex environment, so that the prediction result of the aerodynamic coefficient is close to the actual situation; and through the simulation calculation of the algorithm model, the modeling cost can be effectively reduced, the simulation time is shorter, the different design schemes of the heliostat can be quickly evaluated and optimized, the real-time requirement is met, the scheme is scientific, reasonable and feasible.

[0021] Further, while reducing the calculation cost, the adaptability to complex aerodynamic working conditions and the prediction accuracy are significantly improved, and the present application has good engineering application value and promotion prospect.

[0022] To achieve one of the above purposes, the third technical scheme of the present application is: A heliostat aerodynamic simulation device, comprising: One or more processing units; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processing units, the one or more processing units implement the above-mentioned heliostat feature extraction method.

[0023] Compared with the prior art, the present application has the following beneficial effects: The application can accurately mine and accurately extract the heliostat features by constructing a heliostat pretreatment model, a feature mining model and a feature fusion model, sampling the heliostat point cloud scanning data to obtain heliostat key points, then mining the features of the heliostat key points to obtain multi-dimensional geometric features to capture the local features and global features of the heliostat, and then performing nonlinear transformation on the multi-dimensional geometric features to obtain structural feature quantities, and coupling the structural feature quantities with environmental feature quantities and the elevation angle of the heliostat to generate a fusion feature vector, so that the heliostat features can be accurately mined and accurately extracted, and then can be used for aerodynamic simulation of the heliostat, which is scientific, reasonable and feasible.

[0024] Meanwhile, the application fully considers the mutual influence and interaction between the multi-modal variables of the heliostat, can effectively model the multi-physical field interaction of the heliostat, can accurately reflect the aerodynamic characteristics of the heliostat in a complex environment, so that the aerodynamic coefficient prediction result is close to the actual situation, and through simulation calculation by the algorithm model, the modeling cost can be effectively reduced, the simulation time is shorter, and different design schemes of the heliostat can be quickly evaluated and optimized, which meets the real-time requirement, and the scheme is scientific, reasonable and feasible.

[0025] Further, the application can significantly reduce the amount of point cloud data while effectively preserving key geometric structure information, thereby greatly improving data processing efficiency, and using curvature weight and nonlinear transformation mechanism can fully capture the aerodynamic response features of the heliostat surface, and through the multi-modal feature fusion strategy, the coupling and expression ability between different modalities are effectively enhanced, so that the extracted features are more discriminative and representative, and therefore more efficient and accurate data processing and feature expression ability can be achieved.

[0026] Further, the application can significantly improve the adaptability and prediction accuracy of complex aerodynamic working conditions while reducing the calculation cost, and has good engineering application value and promotion prospect. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 It is a first flowchart of the heliostat feature extraction method of the application. Figure 2 It is a second flowchart of the heliostat feature extraction method of the application. Figure 3 It is a structure diagram of the deep cross-attention network of the application. DETAILED DESCRIPTION

[0028] In order to make the person skilled in the art better understand the scheme of the present application, the technical scheme in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor should belong to the scope of protection of the present application. The present application covers any alternative, modification, equivalent method and scheme made within the essence and scope of the present application defined by the claims.

[0029] As shown in Figure 1 , the first specific embodiment of the heliostat feature extraction method of the present application is: A heliostat feature extraction method, comprising the following steps: Step one, obtaining heliostat point cloud scanning data; Step two, using a pre-constructed heliostat preprocessing model, based on curvature weight, sampling the heliostat point cloud scanning data to obtain heliostat key points; Step three, using a pre-constructed feature mining model, performing feature processing on the heliostat key points to obtain multi-dimensional geometric features to capture local features and global features of the heliostat; Step four, using a pre-constructed feature fusion model, performing nonlinear transformation on the multi-dimensional geometric features to obtain structural feature quantities, and coupling the structural feature quantities with environmental feature quantities of the heliostat and the elevation angle of the heliostat to generate a fusion feature vector, realizing feature extraction of the heliostat.

[0030] As shown in Figure 2 , the second specific embodiment of the heliostat feature extraction method of the present application is: A heliostat feature extraction method, comprising the following steps: First, obtaining multi-modal features and pre-processing them, which includes the following contents: Step 11, using a laser scanning device to perform omnidirectional scanning on the heliostat to obtain detailed heliostat point cloud data.

[0031] Step 12, the traditional farthest point sampling (FPS) only considers the distance between points and points, and the present embodiment introduces curvature weight to optimize the sampling distribution. Let the point cloud set be , the set of selected points be S, and the selected point be , then the improved FPS sampling formula is:

[0032] Wherein represents the selected point, P is the point cloud set, S is the set of selected points, represents the point the minimum distance to all points in the set S, and the maximum distance and maximum curvature value in the current stage point cloud, is the curvature weight, and the optimal balance is obtained when a = 0.7 through system experiment evaluation, is the index for finding the maximum value in a given function or array.

[0033] In this embodiment, the calculation method of the curvature weight K( ) is as follows: First, the covariance matrix of the points in the neighborhood of the point is calculated, and the calculation formula is as follows:

[0034] wherein, is the point in the neighborhood, is the centroid of the neighborhood points.

[0035] Then, the eigenvalue decomposition is performed on the covariance matrix to obtain three eigenvalues λ1≥λ2≥λ3, and the curvature is:

[0036] Further, the curvature weight K( ) can be defined as K( )= .

[0037] By this method, 1024 key points are selected, the data amount is reduced and the key information is retained, and especially the details of the high curvature area (such as the mirror edge) can be better retained, which is crucial for aerodynamic characteristic analysis.

[0038] Step 13, normalize the air density data and map it to [0, 1]. Let the original air density data be x, and the normalized air density data be , then the normalization formula is:

[0039] wherein, and are the minimum and maximum values of the original air density data, respectively.

[0040] Normalization processing can eliminate the influence of different orders of magnitude data on the model, so that the model can more stably learn the features.

[0041] Step 14, use computer vision and geometric analysis algorithms to identify the elevation angle of the heliostat from the three-dimensional model of the heliostat. The specific process is as follows:​ Firstly, the equations of the heliostat plane and the reference plane are determined, which can be obtained by fitting the points in the point cloud data.

[0042] Then, the normal vectors of the two planes are calculated, and the normal vector of the heliostat plane is set as and the normal vector of the reference plane is .

[0043] Further, according to the vector angle formula, the angle between the normal vector of the heliostat plane and the normal vector of the reference plane is calculated, and then the elevation angle θ of the heliostat is determined according to the angle, and the calculation formula is as follows:

[0044] Considering that the mirror surface may be deformed, the point cloud data after deformation is processed by the algorithm to restore it to the original shape as much as possible, thereby reducing the elevation angle recognition error caused by deformation.

[0045] Finally, the elevation angle of the heliostat is converted into numerical features by using trigonometric function coding, that is, the elevation angle θ is coded into two values of sin(θ) and cos(θ), which is convenient for subsequent model processing.

[0046] The second step is to extract geometric features and encode them, which includes the following contents: Step 21, multi-dimensional geometric feature extraction is performed on 1024 key points, including curvature and normal feature.

[0047] As described above, by calculating the covariance matrix of the points in the neighborhood of the key point and performing eigenvalue decomposition, the curvature can be obtained. The curvature feature reflects the local bending degree of the heliostat surface, which is closely related to the flow of air on the heliostat surface. For example, larger curvature may cause air flow separation and vortex generation, thereby affecting the aerodynamic coefficient.

[0048] The normal feature is the eigenvector corresponding to the smallest eigenvalue of the covariance matrix. Let the smallest eigenvalue of the covariance matrix be , and the corresponding eigenvector be , then is the normal feature of the key point . The normal feature represents the direction of the heliostat surface at this point, which has an important influence on the flow direction and pressure distribution of the air flow.

[0049] The extracted geometric features are combined into a feature matrix , where is the feature vector of the th key point, which contains the curvature and normal features.

[0050] Step 22, encode the geometric features using the PointNet++ algorithm. PointNet++ includes three core processes: sampling, grouping, and feature extraction.

[0051] The sampling process includes the following: An improved sampling algorithm, FPS, is used to select representative points. This sampling algorithm effectively reduces the amount of data while preserving key information, laying the foundation for efficient processing later.

[0052] The grouping process includes the following: Find the neighborhood point set for each sampling point. Use the k-neighborhood method for grouping. Grouping allows the network to capture local features of point cloud data, better understanding the geometric structure of the heliostat surface.

[0053] The feature extraction process includes the following: Apply the PointNet algorithm to the neighborhood point set for hierarchical sampling and local feature aggregation, and perform concatenation and pooling operations to obtain a 256-dimensional feature vector, effectively capturing local and global features of the point cloud. It mainly includes the following three steps: For each sampling point, the point set in its neighborhood is denoted as , which includes the point and its k nearest neighbors. In each neighborhood, a local coordinate system is constructed, and the geometric features of the point (such as curvature, normal) are used as input feature vectors. These features are input into a shared weight PointNet network for feature extraction.

[0054] In this embodiment, the PointNet network includes the following structure: First layer: fully connected layer FC + activation function ReLU + batch normalization BN; Second layer: fully connected layer FC + activation function ReLU + batch normalization BN; Third layer: use the fully connected layer FC to output local features.

[0055] This structure can effectively model the geometric features of the local structure of the point cloud, especially suitable for irregular heliostat edges or curved deformation areas.

[0056] In this embodiment, the concatenation operation includes the following: For each point, concatenate the local features extracted by the PointNet network with the original geometric features (such as point coordinates, normal, curvature) to form an enhanced feature representation . This not only preserves the basic geometric information, but also integrates local context semantics, which helps subsequent global feature modeling.

[0057] In this embodiment, the pooling operation includes the following: The pooling operation is performed within each region to extract the most salient features representing the region, thereby eliminating the influence of the number of local points. The global feature description is then formed by aggregating the features of multiple regions. The final output is a 256-dimensional feature vector, which is used for subsequent fusion and modeling.

[0058] In the third step, multiple feature vectors are fused to obtain a fused feature vector, which includes the following: In order to further enhance the interaction and fusion effect between features, a nonlinear transformation can be performed on each feature vector before concatenation. For example, a multi-layer perceptron (MLP) is used to transform each feature vector.

[0059] The expression of the point cloud feature transformation is as follows:

[0060] wherein is a multi-layer perceptron network for point cloud features, is the transformed point cloud feature, is the original point cloud feature, which contains local / global geometric features (such as curvature, normal direction) encoded by the point network enhancement algorithm PointNet++, and its distribution structure usually has nonlinearity and spatial dispersion.

[0061] In order to avoid the problem of feature scale imbalance and semantic inconsistency caused by direct concatenation, the following MLP network is designed: The input dimension is the original point cloud feature dimension (such as 256 dimensions); The hidden layer is set to three perceptrons, each equipped with a nonlinear activation function (ReLU) and normalization (BatchNorm); the transformation process is as follows: Layer Layer1: 256→128; Layer Layer2: 128→64; Layer Layer3: 64→32.

[0062] Output dimension: unified to 32 dimensions to align and fuse with other modal features.

[0063] This network can automatically learn the high-order nonlinear relationship in the point cloud feature, so that the original geometric information is preserved in the low-dimensional space while reducing redundant information.

[0064] The expression of the wind speed feature transformation is as follows:

[0065] wherein is a multi-layer perceptron network for wind speed features, is the transformed wind speed feature, is the original wind speed feature.

[0066] The expression of air density feature transformation is as follows: , wherein is a multi-layer perceptron network for air density features, is the transformed air density feature, is the original air density feature.

[0067] Air density and wind speed belong to environmental feature quantities, usually with low dimensions (such as wind speed: 3-dimensional vector, density: 1-dimensional scalar), but their changes have nonlinear effects on aerodynamic characteristics. For example: under different air densities, the flow state on the surface of the heliostat may undergo a critical transition.

[0068] Therefore, an independent MLP network is designed for each physical quantity to extract its implicit nonlinear influencing factors. The structure is as follows: Input dimension: 3 dimensions of wind speed or 1 dimension of air density; Hidden layer structure: Layer Layer1: input dimension → 16; Layer Layer2: 16 → 32; Output dimension: unified to 32 dimensions.

[0069] Thus, the original low-dimensional continuous variables are embedded into a unified feature space, facilitating splicing with point cloud features, and learning high-order effects of variables on aerodynamic effects through the network.

[0070] The expression of heliostat elevation feature transformation is as follows:

[0071] wherein is a multi-layer perceptron network for heliostat elevation features, is the transformed heliostat elevation feature, is the original heliostat elevation feature.

[0072] The heliostat elevation is encoded by in(θ), cos(θ) to form a 2-dimensional vector, which is used to represent the spatial posture of the heliostat. Although this feature has low dimension, it has a great influence on aerodynamic characteristics (such as changes in windward surface, migration of flow separation point, etc.).

[0073] In order to enable the model to better learn the coupling relationship between the elevation feature and the aerodynamic performance, a dedicated MLP is designed, and its structure is as follows: Input dimension: 2; Structure: 2→16→32; Output dimension: 32 dimensions, aligned with other modalities.

[0074] And guide the model to extract discriminative representation from the periodicity and directionality of the elevation angle, avoiding the "elevation angle information dilution" caused by direct concatenation.

[0075] Finally, the transformed feature vectors are then concatenated to obtain a fusion feature vector , whose expression is as follows:

[0076] This embodiment can learn more complex relationships between features through nonlinear transformation, and mine potential information hidden in the features, thereby further improving the quality of the fusion features and the prediction performance.

[0077] The nonlinear transformation of the present application enables the model to have the ability to fit high-order, nonlinear, and multivariate coupling relationships by introducing an activation function. For example, when the point cloud normal changes dramatically and the elevation angle is close to the critical value, the influence of wind speed or air density on aerodynamic response will no longer be linear growth, but may exhibit complex behaviors such as jumping, saturation, and gating. Traditional linear mapping cannot capture such changes, but a multilayer perceptron (MLP) can model these potential nonlinear patterns under the action of multiple activation units.

[0078] In addition, although each modality feature (point cloud, wind speed, density, and elevation angle) is transformed by an MLP network in the early stage, the core purpose of this transformation is to project each type of feature into a unified embedding space, so that they have similar scale and distribution characteristics, providing a semantically aligned representation basis for subsequent feature fusion and interaction modeling. After the fusion feature vector is formed by concatenation, a unified aerodynamic performance prediction model (heliostat aerodynamic simulation method) is further used to model the deep coupling relationship between cross-modalities, thereby improving the overall prediction performance and model generalization ability.

[0079] A specific embodiment of the heliostat aerodynamic simulation method of the present application: A heliostat aerodynamic simulation method, i.e., an aerodynamic coefficient prediction model, includes the following steps: Step 1. Apply the above-mentioned heliostat feature extraction method to process and extract features from the multi-source data of the heliostat, obtaining a fusion feature vector.

[0080] Step 2. Mine the interaction relationship between modal features to obtain modal coupling data; then introduce a graph convolution operation to represent the modal coupling data as a graph structure, and perform message propagation on the graph structure to represent the dependency relationship between modal features, forming a graph embedding representation information.

[0081] In this embodiment, the method for mining the interaction relationships between modal features to obtain modal coupling data is as follows: A deep cross-attention network is constructed to achieve mutual attention between different modalities at multiple levels. This network iterates multiple times, calculating attention scores between features of different modalities and assigning dynamically changing weights to each feature.

[0082] A deep cross-attention network is constructed based on the cross-attention mechanism to more deeply explore the interaction relationships between different modalities of data. Let the input multimodal features be as follows: Each of them The data representing the i-th modal feature (e.g., point cloud, air density, elevation angle, etc.) consists of M modalities, each containing There are positions (or nodes), each position being... 3D eigenvectors.

[0083] First, a linear transformation is performed on each modal feature to obtain the query matrix, key matrix, and value matrix. For the first... Modal features Its query matrix Key matrix Value matrix ,in , The weight matrix is ​​a learnable matrix. For the attention dimension.

[0084] The learnable weight matrix in this embodiment , All features were automatically optimized during the training phase of a deep cross-attention network, with the training objective being to minimize the designed physical constraint loss function. This training process fully considers the modal differences and physical consistency constraints of the input features, ensuring that the fused features retain physical interpretability while possessing strong fitting and generalization abilities.

[0085] Then, the attention scores between different modal features are calculated, for modal feature i and modal feature i. Attention score The calculation formula is as follows:

[0086] in, It is the dimension of the key vector. For activation function, Modal features The query matrix, Modal features The bond matrix.

[0087] In this embodiment The value is 64, which is more suitable for carrying out fine-grained matching and attention calculation between point clouds and physical quantities through experiments.

[0088] Activation function The operation is performed by row, and the expression is as follows:

[0089] wherein, The output value of the i-th node, C is the number of output nodes, that is, the number of categories of classification; e is a natural constant, The output value of the c-th node.

[0090] Then, the value matrix of the modal feature j is weighted and summed according to the attention weight to obtain the fusion feature of the modal feature i, and the calculation formula is as follows:

[0091] wherein, is the fusion feature of the modal feature Each position feature of the modal feature

[0092] In the deep cross-attention network, multi-layer attention calculation is performed. Let the fusion feature of the modal feature of the i-th layer be The query, key and value matrices of the i-th layer continue to be generated based on the fusion feature of the previous layer, and the calculation formula is as follows:

[0093] Similarly, define and The attention score and the fusion feature are updated iteratively, so as to realize the deep interaction between modalities at multiple semantic levels and improve the modeling ability of complex physical coupling relationships.

[0094] In this embodiment, a graph convolution operation is introduced, the modal coupling data is represented as a graph structure, and message propagation is performed on the graph structure to represent the dependency relationship between modal features, and the method of forming a graph embedding representation information is as follows: Different modal data is represented as a graph structure wherein is a node set, representing the features of multi-modal data; is an edge set, representing the relationship between features.

[0095] ​​In this embodiment, the node set V represents the observation points on the heliostat surface, each node contains the geometric modal features (such as coordinates, normal vectors, curvature, etc.) and physical modal features (such as local velocity, pressure, etc.) of its corresponding position, and the fusion forms a node feature vector; the edge set E represents the adjacency relationship (such as Euclidean distance, flow field correlation) between nodes, which is used for information propagation and aggregation in the subsequent graph neural network.

[0096] In the present application, the relationship between features is modeled by constructing edges in the graph structure, and the relationship between features includes geometric adjacency relationship, feature similarity relationship and physical prior relationship.

[0097] The geometric adjacency relationship includes the following contents: According to the position of the point cloud or structural element in space, the K nearest neighbor algorithm is used to establish the edges between nodes, reflecting the local physical interaction.

[0098] The feature similarity relationship includes the following contents: The cosine similarity of the modal features between nodes is calculated, and the high similarity is connected to capture the long-distance but similar behavior coupling characteristics.

[0099] The physical prior relationship includes the following contents: Combined with computational fluid dynamics (CFD) simulation data or structural connection relationship, the physical dependent relationship such as air flow path and heat transfer path is defined to enhance the physical interpretability of the graph structure.

[0100] Finally, the three relationships are fused to construct the graph structure G=(V,E), which provides basic topological support for the graph neural network, supports deep fusion between modalities and physical information propagation.

[0101] Further introduce the graph convolution operation, update the features of the node , and its update formula is as follows:

[0102] Wherein: is the neighbor node set of node ; , are the degrees of nodes and v respectively; is the learnable weight matrix of the th layer; is the activation function (such as ReLU); is the updated feature after fusing neighbor information; is the neighbor information.

[0103] In this embodiment, the nodes in the graph structure represent spatial observation points on the heliostat surface, and the feature vectors of the points consist of features in multiple modalities, including geometric features (such as coordinates, normal vectors, curvature) and physical features (such as local wind speed, air pressure, density, etc.). These modal features are transformed by a linear transformation to obtain the query matrix (Query), key matrix (Key) and value matrix (Value) required by the attention network, and cross-fusion and graph convolution propagation are performed in the graph structure.

[0104] Further, by multiple graph convolution operations and the introduction of information of neighbor nodes, the feature representation of the nodes is constantly updated, further enhancing the modeling ability of the inter-modal physical association, and realizing the ability of local modal feature propagation to the global. At the same time, the deep cross-attention network and the graph convolution network are comprehensively fused, and the interaction relationship and structural association between multi-modal features are respectively modeled.

[0105] Step 3. Training and optimization of the model network, including the following contents: A physical constraint loss function is established, and its expression is as follows:

[0106] wherein, represents the final loss value used for training and optimization, and are weighting coefficients and + =1, satisfying the normalization constraint, and are the mean square errors of the drag coefficient and the pressure coefficient, respectively, and are penalty coefficients. The hyperparameters are determined by grid search , , , to guide the model network to learn features that conform to physical laws; represents the predicted value of the drag coefficient corresponding to the i-th sample or prediction point; represents the predicted value of the pressure coefficient corresponding to the i-th sample or prediction point.

[0107] The true values (benchmark values) of the drag coefficient and the pressure coefficient are provided by computational fluid dynamics (CFD) simulation data as a supervision signal. The mean square error (MSE) between the predicted values output by the model and these true labels is included in the physical constraint loss function to ensure that the model not only fits well at the data level, but also meets the basic physical consistency requirements.

[0108] Step 4. Prediction using the trained model network, including the following contents: The full connection neural network is used as a drag coefficient prediction head, and an input fusion feature vector is used The network includes multiple hidden layers The calculation formula of each layer is as follows:

[0109] wherein, is a weight matrix, is a bias vector, and sigma is an activation function.

[0110] After multiple layers of calculation, the predicted drag coefficient is output.

[0111] The specific value of the bias vector is automatically learned in the model training process, and the learning process is as follows: S1. In the initialization stage, that is, when the network is established, the bias vector of each layer is assigned a value of all zeros, and the initialization operation is completed. S2. In the forward propagation stage, the bias is a vector with the same dimension as the output, which is added to the output variable of each node. S3. In the back propagation stage, that is, in the training process, according to the loss between the network output and the true target, the gradient of the loss with respect to the bias is calculated using the chain rule; these gradients are updated by the optimizer Adam. After multiple iterations of forward propagation and back propagation, the value of the bias vector gradually converges to a set of parameter values that minimize the loss function.

[0112] In this embodiment, the method of multi-layer calculation is as follows: Step one, obtain the input data, which is a plurality of modal feature vectors with physical meaning, and the expression is as follows:

[0113] wherein each represents the data feature of the th modal, for example: is a point cloud shape feature, is an air density value, is an elevation angle parameter, is a wind field feature.

[0114] Step two, perform deep cross-attention mechanism fusion on the input data, which includes the following contents: (1) For each modal feature , use a learnable matrix to perform query matrix (Query), key matrix (Key) and value matrix (Value) conversion, and the expression of the learnable matrix is as follows:

[0115] (2) For the modal and , the cross-attention score, i.e., the attention weight, is calculated The calculation formula of the layer attention weight is as follows:

[0116] (3) Feature fusion is performed to obtain fused features , and the expression is as follows:

[0117] After layer attention mechanism iterative update, the fused feature representation , i.e., the modal coupling data, is obtained.

[0118] Step three, modeling the graph structure and graph convolution processing, which includes the following contents: (1) Construct a graph structure G=(V,E), where the node set Each node represents a modal fused feature vector (i.e., the part obtained in the previous step). The edge set E is obtained from the physical correlation or Euclidean distance / structural dependence between features, and the edge weight can be set based on the correlation coefficient or prior knowledge.

[0119] (2) Perform graph convolution operation to update the graph embedding representation information of each layer node , and the expression is as follows:

[0120] Wherein: is the adjacent node set; is the layer graph convolution learnable matrix; is the initial input of the graph convolution, which is the cross-attention output in the previous step, is the fused feature of the layer node i; is the activation function ReLU.

[0121] After layer graph convolution operation, the final graph embedding representation information is obtained.

[0122] Step four, map all node final features (graph embedding representation information) to resistance coefficient prediction value through aggregation and nonlinear regression layer , and the expression is as follows:

[0123] Wherein: For the characteristic aggregation method, the calculation is performed by a weighted summation method; For the regression function, it can be a fully connected network or a linear layer.

[0124] Step five, supervised training of the loss function, in the training stage, the predicted value is compared with the true value , the mean square error is calculated, and the calculation formula is as follows:

[0125] Among them, the pressure coefficient The prediction process of the resistance coefficient is the same, and the same process is used to calculate , and finally combined with the physical constraint loss function, the calculation formula is as follows:

[0126] In this embodiment, the pressure coefficient prediction head is also a fully connected neural network, and the structure is similar to the resistance coefficient prediction head. The input fusion feature vector is calculated through multiple layers to output the predicted pressure coefficient .

[0127] A specific embodiment of applying the heliostat feature extraction method of the application to heliostat aerodynamic simulation: A large-scale tower type photo-thermal power station plans to upgrade and transform the existing heliostat system to improve power generation efficiency and stability. When designing a new heliostat scheme, it is necessary to quickly and accurately predict the aerodynamic coefficients of the heliostat in order to evaluate and optimize different design schemes.

[0128] The heliostat is aerodynamically simulated by applying the heliostat feature extraction method of the application, and the fusion feature vector is input into the pre-constructed aerodynamic coefficient prediction model to obtain the aerodynamic coefficient prediction result of the heliostat. Compared with the traditional wind tunnel experiment method and CFD numerical simulation method, the application greatly shortens the data processing and feature extraction time from several hours or even several months to tens of minutes, while reducing the calculation cost.

[0129] According to the predicted aerodynamic coefficients, the project team evaluated and optimized the heliostats of different design schemes. By adjusting the geometric shape, elevation angle and other parameters of the heliostat, the aerodynamic performance of the heliostat was improved, thereby improving the power generation efficiency and stability of the photo-thermal power station.

[0130] In this embodiment, the method of inputting the fusion feature vector into the aerodynamic coefficient prediction model to obtain the aerodynamic coefficient prediction result of the heliostat includes the following steps: Step one, the fusion feature vector of the heliostat is processed to generate a plurality of modal features; Step two, the interaction relationship between the modal features is mined to obtain modal coupling data; Step three, the graph convolution operation is introduced, the modal coupling data is represented as a graph structure, and the message propagation is performed on the graph structure to represent the dependency relationship between the modal features, and the graph embedding representation information is formed; Step four, the graph embedding representation information is aggregated and mapped to obtain the predicted aerodynamic coefficients of the heliostat, and the aerodynamic simulation of the heliostat is realized.

[0131] In step one of the embodiment, the method for processing the geometric data and air data of the heliostat to generate a plurality of modal features is as follows: The geometric data of the heliostat is obtained, which includes the three-dimensional point cloud data of the heliostat and the elevation angle data of the heliostat; The air data is obtained, which includes the air density data, and the air density data is constructed in combination with the historical meteorological data; The three-dimensional point cloud data is processed to extract its geometric features to obtain the first modal feature; The angle recognition and coding are performed on the elevation angle data of the heliostat to obtain the second modal feature; The air density data is normalized to obtain the third modal feature; The first modal feature, the second modal feature and the third modal feature are fused to form a plurality of modal features.

[0132] In step two of the embodiment, the method for mining the interaction relationship between the modal features to obtain the modal coupling data is as follows: A plurality of modal features are obtained, which at least include the first modal feature and the second modal feature, each modal feature includes a plurality of positions, and each position is a d-dimensional feature vector; Considering the difference and physical consistency constraint between the modal features, a learnable weight matrix corresponding to each modal feature is set; the learnable weight matrix is obtained by training a deep cross-attention network, and the training target is to minimize a physical constraint loss function; Based on the learnable weight matrix, the modal features are linearly transformed to obtain a query matrix, a key matrix and a value matrix; According to the query matrix and the key matrix, the attention score between the first modal feature and the second modal feature is calculated; Based on the attention score, the value matrix of the second modal feature is weighted and summed to obtain the first fusion feature of the first modal feature; according to the attention score, the value matrix of the first modal feature is weighted and summed to obtain the second fusion feature of the second modal feature; According to the first fusion feature and the second fusion feature, new query matrix, key matrix and value matrix are calculated; Based on the new query matrix and the key matrix, an attention score between the first fusion feature and the second fusion feature is calculated; The attention score and the fusion feature are iteratively updated, so as to realize deep interaction between modal features at multiple semantic levels. Finally, the first fusion feature and the second fusion feature are summarized to obtain modal coupling data.

[0133] As shown in Figure 3 The method for training the deep cross-attention network in the embodiment is as follows: A predicted heliostat aerodynamic coefficient is obtained, and compared with a reference aerodynamic coefficient value to calculate the mean square error of the two; Based on the mean square error, the weighted coefficient and the penalty coefficient, a physical constraint loss function is constructed with a loss value as the target; Global interaction features between different modal features are extracted through the deep cross-attention network; Then, a graph structure between modal features is constructed, and a local dependency relationship and topological association are modeled through graph convolution; Then, based on the deep cross-attention network and the graph structure, an aerodynamic coefficient prediction result is output, and the physical constraint loss function is used as an optimization target to guide the entire deep cross-attention network to maintain physical consistency while fitting the aerodynamic coefficient; a variable dependency and a back propagation mechanism constitute an end-to-end training joint learning network.

[0134] In step three of the embodiment, a graph convolution operation is introduced, the modal coupling data is represented as a graph structure, and message propagation is performed on the graph structure to represent the dependency relationship between modal features, and the method for forming graph embedding representation information is as follows: Based on the observation points on the surface of the heliostat and the corresponding position information, a plurality of nodes are constructed; According to the modal coupling data and the corresponding position information, the nodes are valued, so that each node has a geometric modal and a physical modal; and the geometric modal and the physical modal are fused to form a node feature vector; Based on the modal coupling data, the relationship between the features is obtained; and based on the relationship between the features, a geometric edge is constructed; According to the geometric edge, the node and the node feature vector, a graph structure is drawn; Based on the deep cross-attention network and the graph convolution network, and combined with the node feature vectors of the neighbor nodes, the node feature vectors are updated, so that the local modal features can be propagated globally, and can represent the dependency relationship and structural association between modal features.

[0135] In the embodiment, the relationship between the features includes a geometric adjacency relationship, a feature similarity relationship and a physical prior relationship; Geometric adjacency relationship is established according to the positions of point clouds or structural elements in space and by using a near neighbor algorithm, and is used to reflect local physical interaction; Characteristic similarity relationship is established by calculating the cosine similarity of modal characteristics, and is used to capture remotely similar coupling characteristics; Physical prior relationship is established based on structural connection relationship and the relationship of airflow path and heat transfer path, and is used to enhance the physical interpretability of the graph structure.

[0136] Based on the deep cross-attention network and the graph convolutional network, and in combination with the node feature vectors of neighbor nodes, the method for updating the node feature vectors is as follows: First, the node feature vectors are cross-attended layer by layer by the deep cross-attention network, and in combination with the attention mechanism, multi-modal fusion features are generated, which include a plurality of fusion feature units; The multi-modal fusion features are taken as input nodes of the graph convolutional network to construct a graph structure, which includes a plurality of nodes and a plurality of edges, wherein the nodes represent the fusion feature units, and the edges represent the physical dependency relationship, spatial connection or structural similarity between the feature units; An adjacency matrix is established according to the node feature vectors of neighbor nodes; In the graph convolutional network, the node feature vectors are updated layer by layer through a message propagation mechanism controlled by the adjacency matrix, so that the node feature vectors can interact with local modal characteristics and can be integrated into the topological information of the global physical field.

[0137] In the fourth step of the embodiment, the method for aggregating and mapping the graph embedding representation information to obtain the predicted heliostat aerodynamic coefficients is as follows: Based on the weight matrix, the bias vector and the activation function, a coefficient calculation formula is constructed; According to the coefficient calculation formula, a plurality of hidden layers are established; The graph embedding representation information is input into the plurality of hidden layers for multi-layer calculation to obtain final features of nodes; The final features of nodes are aggregated and weighted summed, and then input into a regression function for calculation to obtain the predicted heliostat aerodynamic coefficients; the heliostat aerodynamic coefficients include a drag coefficient and a pressure coefficient.

[0138] A server embodiment applying the method of the present application is as follows: A server comprises: one or more processing units; a storage device for storing one or more programs; When the one or more programs are executed by the one or more processing units, the one or more processing units implement the above-mentioned heliostat feature extraction method.

[0139] The storage device is an internal memory or an external memory or a cache memory or other special memory. The processing unit can be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a programmable logic device or other programmable logic device.

[0140] An embodiment of a device applying the method of the present application is as follows: An electronic device is provided with a computer readable storage medium, and the computer readable storage medium stores a computer program, which, when executed by a processing unit, implements the above-mentioned heliostat feature extraction method.

[0141] The computer readable storage medium refers to a physical carrier capable of storing computer-recognizable data, instructions or programs, and these media need to meet the core characteristics of "being readable by a computer" (i.e. data exists in the form of electrical, magnetic, optical signals, which can be converted into binary information that can be processed by a computer through corresponding devices). The physical carrier is a magnetic storage medium, an optical storage medium, a semiconductor storage medium or other storage medium.

[0142] The model in the present application is an object that constitutes an objective description of a morphological structure with the aid of an entity or a virtual representation. The object is not equal to an object and is not limited to entities and virtualities. It can be a data processing function, a software program, a processing mode, a use method, an operation method, a work flow, an application process, an electronic hardware, a circuit module, a processing system, a system imitation or a simulation object.

[0143] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the same. The protection scope of the present application is not limited thereto, even though the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some of the technical features, without departing from the technical range disclosed by the present application. Such modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A heliostat feature extraction method, characterized in that: comprising the following steps: Step 1: obtaining heliostat point cloud scanning data; Step 2: using a pre-constructed heliostat preprocessing model, based on curvature weight, sampling the heliostat point cloud scanning data to obtain heliostat key points; Step 3: using a pre-constructed feature mining model, processing the heliostat key points to obtain multi-dimensional geometric features to capture the local and global features of the heliostat; Step 4: using a pre-constructed feature fusion model, performing nonlinear transformation on the multi-dimensional geometric features to obtain structural feature quantities, and coupling the structural feature quantities with the environmental feature quantities of the heliostat and the elevation angle of the heliostat to generate a fusion feature vector, thereby realizing feature extraction of the heliostat. 2.The heliostat feature extraction method of claim 1, characterized in that: Step 2: using a pre-constructed heliostat preprocessing model, based on curvature weight, sampling the heliostat point cloud scanning data to obtain heliostat key points is as follows: According to the heliostat point cloud scanning data, a point cloud set is established, which includes a plurality of discrete points, and a selected point set is constructed; Based on the plurality of discrete points, the maximum distance and the maximum curvature value between the discrete points are calculated; The minimum distance from a discrete point to all discrete points in the selected point set is calculated; Based on the covariance matrix, the curvature weight of the discrete point is determined; According to the curvature weight, the minimum distance, the maximum distance between the discrete points and the maximum curvature value of the discrete point, the sampling value of the discrete point is calculated; Based on the sampling values of the plurality of discrete points, a sampling distribution is established, and the maximum value and the discrete point corresponding to the maximum value are selected from the sampling distribution; The discrete point corresponding to the maximum value is taken as the heliostat key point to represent the structure of the high curvature area of the heliostat. 3.The heliostat feature extraction method of claim 2, characterized in that: Based on the covariance matrix, the curvature weight of the discrete point is determined as follows: According to the discrete point, the points in the neighborhood and the neighborhood point centroid are obtained; According to the points in the neighborhood and the neighborhood point centroid, a covariance matrix of the points in the neighborhood of the discrete point is established; Eigenvalue decomposition is performed on the covariance matrix to obtain three eigenvalues; According to the three eigenvalues, the curvature is calculated; The curvature is taken as the curvature weight of the discrete point. 4.The heliostat feature extraction method of claim 1, characterized in that: Step 3: using a pre-constructed feature mining model, processing the heliostat key points to obtain multi-dimensional geometric features is as follows: The covariance matrix of the points in the neighborhood of a heliostat key point is calculated; Eigenvalue decomposition is performed on the covariance matrix to obtain the curvature; The eigenvector corresponding to the minimum eigenvalue of the covariance matrix is taken as the normal feature; The curvature and the normal feature are taken as the geometric feature, and the eigenvector of the heliostat key point is constructed based on the geometric feature; The eigenvectors of all heliostat key points are combined to obtain a geometric feature matrix; The geometric feature matrix is sampled, grouped and feature extracted to obtain multi-dimensional geometric features. 5.The heliostat feature extraction method of claim 4, characterized in that: The method for sampling, grouping and feature extraction of the geometric feature matrix to obtain multi-dimensional geometric features is as follows: The geometric feature matrix is processed by fixed time period or sliding window sampling method to obtain several representative features; According to each representative feature, a local coordinate system is constructed, and based on the local feature coordinate system, a plurality of adjacent representative features are found; Based on the plurality of representative features, a plurality of neighborhood feature sets are established to capture the local structural features of the heliostat surface; Using the plurality of neighborhood feature sets, the corresponding heliostat key points are determined; According to the heliostat key points, the representative features in the neighborhood feature set are spliced with the original geometric features to form an enhanced feature representation for retaining the basic geometric information and fusing the local context semantics; Based on the enhanced feature representation, a field feature region is constructed; The field feature region is subjected to a pooling operation to extract the most significant features representing the field feature region; The plurality of most significant features are aggregated to obtain multi-dimensional geometric features to describe the global features.

6. The heliostat feature extraction method of claim 1, wherein: In step four, a pre-constructed feature fusion model is used to perform nonlinear transformation on the multi-dimensional geometric features to obtain structural feature quantities, and the method is as follows: The multi-dimensional geometric features are obtained, and the data dimensions thereof are ; By using a nonlinear activation function and a normalization operation, the multidimensional geometric features are nonlinearly transformed to obtain a first simplified feature vector of the plurality of dimensions, for mining potential information hidden in the multidimensional geometric features; By using a nonlinear activation function and normalization operation, a nonlinear transformation is performed on the first simplified feature vector to obtain... The second simplified eigenvector of dimension; The second simplified feature vector is nonlinearly transformed by using a nonlinear activation function and a normalization operation to obtain a third simplified feature vector of dimensionality The third simplified feature vector is used to form a structural feature quantity that can retain discriminativeness and be fused in alignment.

7. The heliostat feature extraction method of claim 6, wherein: The structural feature quantities are coupled with the environmental feature quantities of the heliostat and the heliostat elevation angle to generate a fusion feature vector to realize the feature extraction of the heliostat, and the method is as follows The environmental feature quantities include 3-dimensional wind speed features and 1-dimensional air density features; A multi-layer perception network is constructed for the wind speed characteristics, and the 3-dimensional wind speed characteristics are nonlinearly transformed to obtain a 2-dimensional wind speed characteristic vector. a 2-dimensional wind speed characteristic vector; A multilayer perceptron network is constructed to perform a nonlinear transformation on the 1D air density features, resulting in... A dimensional air density feature vector; The heliostat elevation angle is obtained, and the heliostat elevation angle is encoded to form a 2-dimensional elevation angle feature vector for representing the spatial posture of the heliostat; A multi-layer perceptron network is constructed for the heliostat elevation feature, a 2-dimensional elevation feature vector is nonlinearly transformed, and the multi-layer perceptron network is guided to extract discriminative representations from the periodicity and directionality of the heliostat elevation, obtaining an elevation feature vector of the heliostat in the 2-dimensional space. Then, the structural feature quantities, the wind speed feature vector, the air density feature vector and the heliostat elevation angle feature vector are spliced in rows to obtain the fusion feature vector.

8. The heliostat feature extraction method of claim 7, wherein: The method for obtaining the air density feature is as follows: Air density data of the environment around the heliostat is collected; The air density data is sorted to obtain the minimum density value and the maximum density value; Based on the minimum density value and the maximum density value, the air density data is normalized to obtain air density features with uniform magnitude; Or / and, the method for obtaining the heliostat elevation angle feature is as follows: First, based on the points of the heliostat point cloud scanning data, a plane equation is fitted to determine the heliostat plane equation and the reference plane equation; Then, the normal vector one of the heliostat plane equation and the normal vector two of the reference plane equation are calculated; According to the vector angle formula, the included angle between the normal vector one and the normal vector two is calculated; Based on the vector angle, the elevation angle of the heliostat is determined; The elevation angle of the heliostat is converted into a numerical feature by using trigonometric function coding to obtain a 2-dimensional heliostat elevation angle feature.

9. A heliostat aerodynamic simulation method, characterized in that: A heliostat feature extraction method according to any one of claims 1-8 is applied to construct an aerodynamic coefficient prediction model, which includes the following steps: Step one, processing the fusion feature vector of the heliostat to generate multiple modal features; Step two, mining the interaction relationship between modal features to obtain modal coupling data; Step three, introducing graph convolution operation, representing the modal coupling data as a graph structure, and performing message propagation on the graph structure to represent the dependency relationship between modal features, forming graph embedding representation information; Step four, aggregating and mapping the graph embedding representation information to obtain the predicted heliostat aerodynamic coefficient, realizing the aerodynamic simulation of the heliostat.

10. A heliostat aerodynamic simulation device, characterized in that: It includes: One or more processing units; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processing units, the one or more processing units implement a heliostat feature extraction method as claimed in any one of claims 1-8.

Images