A method and system for selecting solar array cells based on three-dimensional layout

Abstract

This invention relates to the field of 3D model processing technology, and more particularly to a method and system for selecting solar array cells based on 3D layout. The method includes the following steps: acquiring 3D data of the solar array, and constructing a 3D substrate model based on the 3D data to obtain a 3D model of the solar array; performing structure-driven deployment strategy simulation on the 3D model of the solar array to obtain fixed simulation data, and performing perturbation-based layout scheme generation on the 3D model of the solar array using the fixed strategy simulation to obtain random simulation data; performing regional structural domain analysis on the fixed simulation data and random simulation data to obtain regional structural domain data, and performing optimal layout sorting on the regional structural domain data to obtain layout sorting data; and exporting the layout sorting data as 2D drawings to obtain solar array cell selection data. This invention can effectively expand the diversity of schemes, improve local adaptability, and achieve high coverage and optimal sorting output while ensuring structural constraints.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional model processing technology, and in particular to a method and system for selecting solar array cells based on three-dimensional layout. Background Technology

[0002] A solar array structure is a deployable or fixed energy supply device installed on spacecraft, satellites, or high-altitude aircraft. Its core function is to receive solar radiation and convert it into electrical energy through solar cells deployed on its surface, providing a continuous power supply to the carrier platform. With the widespread application of flexible energy systems in aerospace, high-performance unmanned platforms, and satellite deployment scenarios, the complexity and diversity of solar array structures are constantly increasing, placing higher demands on the efficiency of cell deployment, structural adaptability, and manufacturing coordination. Traditional solar array cell selection processes often adopt two-dimensional template mapping or manually driven layout methods, relying on static arrangement strategies and experience-based graphic design, which are difficult to accommodate the actual needs of densely deploying irregularly shaped substrates, heterogeneous structural areas, or functional zones. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a method and system for selecting solar array cells based on a three-dimensional layout, thereby resolving at least one of the aforementioned technical problems.

[0004] This application provides a method for selecting solar array cells based on a three-dimensional layout, including the following steps:

[0005] Step S1: Obtain the three-dimensional data of the solar array and construct a three-dimensional substrate model based on the three-dimensional data of the solar array to obtain the three-dimensional model of the solar array;

[0006] Step S2: Perform structure-driven deployment strategy simulation on the 3D model of the solar array to obtain fixed simulation data, and perform fixed strategy simulation on the 3D model of the solar array to generate perturbation layout scheme to obtain random simulation data.

[0007] Step S3: Perform regional structure domain analysis on the fixed simulation data and random simulation data to obtain regional structure domain data, and perform optimal layout sorting on the regional structure domain data to obtain layout sorting data;

[0008] Step S4: Export the layout sorting data as two-dimensional drawings to obtain the solar panel battery selection data.

[0009] In this invention, step S1 acquires and constructs a 3D model of the solar array, achieving a high degree of fidelity to the actual structural morphology and providing an accurate geometric basis for subsequent array simulation. Step S2 introduces a perturbation-based layout generation mechanism on the basis of fixed-strategy simulation, effectively expanding the diversity and robustness of array schemes and avoiding getting trapped in local optima. Step S3 uses regional structural domain analysis to jointly model the global and local characteristics of the array scheme, improving the fine-grained discrimination capability of layout sequencing. Finally, step S4 automatically generates 2D drawings based on the 3D simulation results, ensuring seamless integration between 3D optimization and production processes. Compared with traditional 2D manual array placement or regular templates, this method can achieve intelligent array placement, data-driven evaluation, and standardized output under complex irregular substrates, with higher layout adaptability, visual interactivity, and manufacturing friendliness, significantly improving the efficiency of solar array cell selection and assembly quality.

[0010] Preferably, step S1 specifically includes:

[0011] Acquire three-dimensional data of the solar array;

[0012] Heterogeneous three-dimensional structure perception is performed based on the three-dimensional data of the solar array to obtain heterogeneous three-dimensional structure data;

[0013] Adaptive domain sensing is performed on heterogeneous 3D structural data to obtain adaptive domain data;

[0014] The substrate model is generated based on the deployment adaptation domain data, resulting in a three-dimensional model of the solar array.

[0015] In this invention, this step not only acquires basic geometric data, but also uses heterogeneous 3D structure perception technology to deeply analyze the non-homogeneous features in the solar array, such as edge grooves, clamping points, and hole distribution, breaking through the limitations of traditional modeling that only constructs a unified mesh model based on contour lines. By deploying an adaptive domain perception algorithm, it can automatically identify and mark the spatial regions and forbidden zones suitable for solar cell deployment, significantly improving the model's adaptability to actual assembly constraints. The generated 3D substrate model has structural layering and regional adaptability, which can directly support high-precision cell deployment simulation and strategy matching, avoiding problems such as manual division errors and improper rule configuration.

[0016] Preferably, the heterogeneous three-dimensional structure perception specifically refers to:

[0017] Heterogeneous data is obtained by performing heterogeneous data analysis based on the three-dimensional data of the solar array;

[0018] Structural semantic mapping is performed on heterogeneous data to obtain structural semantic data;

[0019] Three-dimensional structural representations are generated based on structural semantic data to obtain heterogeneous three-dimensional structural data.

[0020] This invention breaks away from the dependence on single-format or rule-based modeling data, enabling compatibility with multi-source data input (such as CAD models, point clouds, scanned layers, etc.). Through heterogeneous data parsing, it uniformly converts data into structurally recognizable geometric elements. Based on structural semantic mapping technology, it establishes a mapping relationship between geometric elements and functional semantics (such as "compression points," "perforation areas," "curved surface corners," etc.), thereby endowing the model with logical structural labels. Through the generation of a 3D structural representation, the model not only possesses a geometric shape but also carries higher-order semantic information such as structural purpose and assembly constraints. This invention exhibits extremely high scalability and interpretability, providing a solid foundation for deployment strategy decisions and area deployability analysis. Compared to traditional methods that rely solely on geometric coordinate points or boundary lines for modeling, this technology has significant advantages in structural understanding depth, semantic discrimination capability, and automatic analysis accuracy, making it suitable for non-standard configurations or multi-functional integrated solar panel component design tasks.

[0021] Preferably, the deployment of adaptive domain sensing specifically includes:

[0022] Surface geometric pavingability is calculated based on heterogeneous 3D structural data to obtain preliminary geometric paving domain data.

[0023] Structural functional regions are removed from the preliminary geometrically layable domain data to obtain functional mask data;

[0024] Semantic mask fusion is performed based on functional mask data and heterogeneous 3D structural data to obtain semantically enhanced deployment domain data.

[0025] Regional connectivity filtering is performed on the semantically enhanced deployment domain data to obtain deployment adaptation domain data.

[0026] This invention calculates surface geometric tiling feasibility based on heterogeneous 3D structural data, comprehensively considering geometric features such as local curvature, normal variation, and unevenness to identify areas with actual deployment feasibility, avoiding false tiling planning caused by surface undulations in traditional methods. The introduction of structural functional area elimination effectively shields untileable areas with assembly / electrical functions, such as clamping points, bolt holes, and cable channels, improving deployment safety and functional compatibility. A semantic mask fusion step integrates functional semantics and geometric information to construct a deployment layer with tiling priority and risk level, enhancing the system's ability to perceive deployment strategies for different structural areas. Regional connectivity filtering ensures that selected deployment areas have spatial continuity and wiring accessibility, avoiding isolated, broken, or unconnectable tiling configurations.

[0027] Preferably, the simulation of the structure-driven unfolding strategy specifically includes:

[0028] Structural constraint diagrams were constructed from the 3D model of the solar array to obtain structural constraint diagram data.

[0029] The structural constraint diagram data is processed by laying behavior diagram scheduling to obtain laying behavior diagram data, which includes horizontal filling strategy, vertical filling strategy, central symmetric filling strategy and S-shaped filling strategy.

[0030] The legality of the laying behavior diagram data is verified to obtain fixed simulation data.

[0031] This invention utilizes a structural constraint graph to comprehensively express multi-source constraint information such as geometric boundaries, local restrictions, functional partitions, and forbidden areas in the three-dimensional structure of the solar array, forming a unified structural constraint semantic graph. A multi-strategy graphical model, including horizontal filling, vertical filling, central symmetry, and S-shaped patterns, is introduced through a laying behavior graph scheduling mechanism, effectively accommodating different structural forms and layout requirements, achieving highly adaptable graphic laying driven by strategies. As a graph theory structure, the laying behavior graph possesses advantages such as high visualization, node coherence, and traceable behavior sequences, making strategy generation not only regular but also flexible in combination and optimization. A legality verification operation performs collision detection, boundary verification, and electrical reachability checks on each strategy-generated laying pattern, ensuring that each layout possesses physical implementability and technological feasibility.

[0032] Preferably, the generation of the perturbation layout scheme specifically involves:

[0033] The structural constraint diagram data is divided into local diagrams to obtain local diagram data;

[0034] Perturbation selection is performed on the local plot data to obtain perturbation plot data;

[0035] Based on the disturbance map data, the paving behavior map data is subjected to disturbance space selection to obtain disturbance space data;

[0036] Perturbation processing is performed on the perturbation spatial data to obtain perturbation result data;

[0037] The validity of the perturbation results data is verified to obtain random simulation data.

[0038] This invention utilizes a structural constraint graph for local graph partitioning. While maintaining the semantic consistency of the overall structure, it focuses on key areas such as edge regions, corner regions, or functional transition areas for targeted perturbation operations, enhancing the flexibility of local structural layout. The perturbation graph selection mechanism enables the perturbation strategy to possess graph theory structural control capabilities, allowing for the selection of the intervention range based on node importance, edge connectivity, or layout density, avoiding structural damage or local disorder caused by blind perturbation. Through perturbation space selection, spatial local replacement, rotation, offset, and misalignment of the laying behavior graph are performed to generate a controllable and variable structural layout. After perturbation processing, a legality verification is performed to ensure that the randomly generated layout still satisfies structural constraints, electrical reachability, and manufacturing rules, guaranteeing practicality.

[0039] Preferably, the regional structure domain analysis specifically includes:

[0040] Structural domain response tensors are constructed from fixed and random simulation data to obtain structural domain response tensor data.

[0041] The coverage of the deployable region is calculated based on the structural domain response tensor data to obtain the coverage data of the deployable region.

[0042] Local symmetry offset is calculated based on the structural domain response tensor data to obtain local symmetry offset data;

[0043] Region label data is obtained by performing region label inference based on the available area coverage data and local symmetry offset data;

[0044] Based on the region label data, candidate maps are constructed by sorting and dividing the fixed simulation data and random simulation data into regions to obtain the regional structure domain data.

[0045] This invention constructs a structural domain response tensor, which not only integrates multi-dimensional spatial information on patch distribution, structural constraints, and layability adaptability in simulation data, but also provides a unified tensor expression framework for index calculation. By calculating the layable area coverage using this tensor, the actual utilization rate of patches within each domain can be accurately measured, reflecting the space-filling effect. The calculation of local symmetry offset can identify the degree of deviation of patches in left-right, top-bottom, or centrally symmetrical structures, effectively capturing local structural aesthetics and patch balance. Based on these two key indicators, the system further performs region label inference, assigning semantic labels such as high coverage and high symmetry, high coverage and low symmetry, and low coverage and high symmetry to different layout domains, forming structured region profiles. These region labels drive the construction and sorting of candidate maps for fixed and perturbed simulation data, generating regional structural domain data that is optimal both globally and locally.

[0046] Preferably, the preferred layout order is as follows:

[0047] Regional weight tensor data is obtained by fusing regional weight tensors into regional structural domain data.

[0048] The region weight tensor data is nonlinearly mapped and sorted to obtain preliminary layout sorting data;

[0049] The preliminary layout ranking data is enhanced by semantic annotation of candidate schemes to obtain the layout ranking data.

[0050] In this invention, the system constructs a regional weight tensor for the regional structural domain data, considering multiple spatial semantic indicators such as the availability, symmetry offset, connectivity reachability, and patch continuity of each region. Information weighting and aggregation are achieved within the tensor structure, overcoming the limitations of traditional single-indicator scoring models. By introducing a nonlinear mapping ranking algorithm, such as Softmax normalization, Tanh compression, or a ranking-aware network, dynamic weight transformation and score ranking are performed on each candidate scheme, making the ranking process less sensitive to extreme values ​​and resulting in more stable and balanced ranking results. The system combines regional structural semantic tags to enhance the semantic annotation of candidate schemes in the ranking results, ensuring that each layout result not only has a score but also includes semantic descriptions such as "boundary balance priority type," "center filling dense type," and "structural symmetry optimization type," facilitating human understanding and interactive filtering.

[0051] Preferably, step S4 specifically includes:

[0052] Perform a 3D layout projection based on the layout sorting data to obtain the 3D layout projection data;

[0053] The unfolded surface data is obtained by unfolding the surface based on the 3D layout projection data.

[0054] Based on the unfolded data of the facets, drawing elements are generated to obtain the drawing element data;

[0055] Export the drawing format based on the drawing element data to obtain the solar panel battery selection data.

[0056] In this invention, the system performs a 3D layout projection operation based on layout sorting data, which can accurately preserve the spatial topological relationship of the solar panel fabric on curved or irregular structures and project it onto a 2D processing reference plane, ensuring consistency between the drawing and the actual assembly. Through the panel unfolding operation, the fabric area with spatial curvature is geometrically unfolded, automatically solving the 2D deformation error caused by bending, cornering, or multi-curvature interference, and improving the consistency between the drawing and the physical substrate. The system generates drawing elements including numbers, boundary lines, docking ports, and annotation symbols based on the unfolded panel data, automatically completing the embedding of standardized engineering elements and avoiding errors and omissions in the manual drawing process. Through the drawing format export module, the drawing element data can be quickly output to industrial common formats such as DWG and DXF, supporting direct calls from downstream CAD, electrical wiring, and manufacturing systems.

[0057] Preferably, this application also provides a solar panel battery selection system based on a three-dimensional layout, used to perform the solar panel battery selection method based on a three-dimensional layout as described above. The three-dimensional layout solar panel battery selection system includes:

[0058] The 3D substrate modeling module is used to acquire 3D data of the solar array and construct a 3D substrate model based on the 3D data of the solar array to obtain the 3D model of the solar array.

[0059] The layout simulation generation module is used to perform structure-driven deployment strategy simulation on the 3D model of the solar array to obtain fixed simulation data, and to perform perturbation-type layout scheme generation on the 3D model of the solar array using fixed strategy simulation to obtain random simulation data.

[0060] The structural region analysis and optimization sorting module is used to perform regional structural domain analysis on fixed simulation data and random simulation data to obtain regional structural domain data, and to perform optimization layout sorting on the regional structural domain data to obtain layout sorting data.

[0061] The drawing generation and configuration output module is used to export two-dimensional drawings from the layout sorting data to obtain solar panel battery configuration data.

[0062] The beneficial effects of this invention are as follows: In the data modeling stage, by performing heterogeneous analysis and semantic recognition on the three-dimensional data of the solar array, a three-dimensional representation model capable of expressing local functional constraints such as clamping points, hole areas, and boundary interference is constructed, solving the problem that traditional methods struggle to accurately perceive complex boundaries and functional areas. Simultaneously, the system automatically identifies pavingable areas and generates semantically enhanced paving masks through adaptive domain perception, providing high-precision support for subsequent simulation optimization. In the simulation stage, a structural constraint graph is constructed and various paving behavior graphs are scheduled to generate fixed simulation samples that conform to physical constraints; a local mutation mechanism is introduced through a perturbation-based strategy generation module, effectively expanding the layout scheme space. In the optimization stage, a structural domain response tensor is constructed, integrating coverage and symmetry features, and combining region label inference and weight tensor sorting to improve the interpretability of the selection and the ability to adapt to multiple objectives. The system automatically completes three-dimensional projection, patch unfolding, and drawing generation, forming standard drawing elements and supporting industrial format export, realizing a closed-loop selection process from modeling to manufacturing, possessing technical advantages of strong structural adaptability and intelligent evaluation. Attached Figure Description

[0063] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings:

[0064] Figure 1 A flowchart illustrating the steps of a solar panel cell selection method based on a three-dimensional layout is shown in one embodiment.

[0065] Figure 2 A flowchart illustrating the steps of a three-dimensional substrate modeling method according to an embodiment is shown.

[0066] Figure 3A flowchart illustrating the steps of a structure-driven unfolding strategy simulation method according to one embodiment is shown.

[0067] Figure 4 A flowchart illustrating the steps of a perturbation-based layout scheme generation method according to an embodiment is shown.

[0068] Figure 5 A flowchart illustrating the steps of a drawing generation and selection output method according to an embodiment is shown. Detailed Implementation

[0069] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0070] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0071] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0072] Please see Figures 1 to 5 This application provides a method for selecting solar array cells based on a three-dimensional layout, including the following steps:

[0073] Step S1: Obtain the three-dimensional data of the solar array and construct a three-dimensional substrate model based on the three-dimensional data of the solar array to obtain the three-dimensional model of the solar array;

[0074] In one embodiment, three-dimensional structural data of the solar array is acquired. This three-dimensional data can be obtained in two ways: first, by using a structured light scanner or LiDAR system to perform high-precision point cloud scanning on the physical solar array to obtain its spatial geometry; second, by extracting three-dimensional model data from existing computer-aided design (CAD) models, where the model format may include standard industrial file formats such as .stp (STEP) and .igs (IGES). A three-dimensional substrate model of the solar array is constructed based on the point cloud data. In this embodiment, a Poisson surface reconstruction algorithm or an α-Shape method is preferably used to generate a continuous surface mesh. The former is suitable for smooth surface modeling of dense point cloud data, while the latter is suitable for structural restoration scenarios with boundary preservation requirements. After construction, the main load-bearing structural surfaces of the solar array are further extracted, and a unified reference coordinate system is established based on its geometric center and laying direction, thereby generating a normalized, structurally complete three-dimensional geometric substrate model.

[0075] Step S2: Perform structure-driven deployment strategy simulation on the 3D model of the solar array to obtain fixed simulation data, and perform fixed strategy simulation on the 3D model of the solar array to generate perturbation layout schemes to obtain random simulation data.

[0076] In one embodiment, the generated 3D substrate model is input into a finite element simulation platform, such as ANSYS, ABAQUS, or other commercial simulation tools that support multibody dynamics and nonlinear material properties. In the simulation settings, the deployment start point and direction of the solar array are set, gravity load parameters are applied appropriately, and the stiffness coefficients of each structural unit are set. Based on the actual structural characteristics of the solar array arms, key hinge points in the mechanism's motion are defined, and the connection and activation sequences between nodes are set to reflect the dynamic behavior of the solar array's gradual deployment. During the simulation, the system solves the structural evolution in a time-step manner, outputting data such as configuration changes, nodal stress distribution, displacement vector fields, and deployment progress status at each time step. The above simulation results are collectively referred to as fixed simulation data, which can be used to describe standard deployment behavior under ideal structural control conditions. A perturbation-based layout scheme generation process is then executed. This process, based on the fixed strategy simulation, considers non-ideal perturbation factors during the deployment process, extending the analysis to assess deployment stability and layout robustness. Specifically, small-amplitude perturbation parameters are introduced at fixed simulation boundaries or initial node states, including displacement perturbation δx ranging from ±2 mm and angular perturbation δθ ranging from ±3 degrees. A multi-round stochastic simulation approach (e.g., Monte Carlo simulation strategy) is employed to construct multiple perturbation input samples, and the same dynamic spreading process as the fixed strategy is executed. In each round of perturbation simulation, the system outputs the configuration evolution path after perturbation and calculates whether it satisfies structural stability constraints (such as stress limits, spreading integrity, and no node overlap). Only perturbation samples that meet stability requirements are retained as stochastic simulation data, while key evaluation indicators such as the actual layout position, energy dissipation value, and peak response value of each sample are recorded.

[0077] Step S3: Perform regional structure domain analysis on the fixed simulation data and random simulation data to obtain regional structure domain data, and perform optimal layout sorting on the regional structure domain data to obtain layout sorting data;

[0078] In one embodiment, a spatial partitioning method is used to divide the surface of the 3D model into several regional units. The partitioning method can be either an octree partitioning method or an equal-voxel mesh partitioning strategy. For each partitioned region, the system calculates the following structural evaluation indicators, such as the load-bearing uniformity coefficient, defined as the sum of the stress values ​​at all nodes within that region divided by the region volume. Where μ i σ is the uniformity coefficient of the load-bearing capacity of the i-th region element, j is the node order term, the upper limit is the number of nodes, and σ is the load-bearing capacity of the region element. j V is the equivalent stress value at the j-th node in the region. i Let be the volume or area of ​​the element belonging to the j-th node. Calculate the structural deformation threshold, defined as the maximum difference between the displacement offsets of all nodes in the region and the reference displacement in the initial state, i.e., Δ.i =max(|d ij ―d i0 |), Δ i Let d be the structural deformation offset threshold for the i-th region element, max be the function to maximize the value, and d be the value of the structural deformation offset threshold for the i-th region element. ij Let d be the displacement value of the j-th node in the current unfolding state of this region. i0 This is its reference displacement value in the initial state. Based on the above two indicators, a regional screening rule is set. If the bearing uniformity coefficient μ of a certain region... i The stress value is higher than the set average stress threshold, and the structural deformation Δ i If the deformation is less than the preset maximum allowable limit, the area is determined to be a suitable distribution area. A ranking index, S, is calculated for the suitable distribution areas. i =w1·μ i +w2·(1 / Δ i )+w3·A i S i The distribution area ranking index is defined as w1, which is the weight data for the uniformity coefficient data, with a value of 0.4. i For the load uniformity coefficient data, w2 is the weight data of the load uniformity coefficient data, with a value of 0.3, Δ i For structural deformation data, w3 represents the weighted data of the effective paving area in this region, with a value of 0.3. A i The effective paving area data for this region is obtained by fitting empirical data with weighted data. According to S... i Sort in descending order to form the optimal layout sequence.

[0079] Step S4: Export the layout sorting data as two-dimensional drawings to obtain the solar panel battery selection data.

[0080] In one embodiment, each preferred region in the 3D layout sorting data is mapped to a unified drawing reference plane, and projection modeling is performed to form a series of structured drawing views, including but not limited to the following: optional mesh wiring diagram, showing the grid boundaries, connection relationships, and spatial alignment relationships of the cell layout units in each region; wiring diagram, indicating the electrical connection methods between cell modules, such as series or parallel logic and wire routing; installation point coordinate diagram, marking the specific assembly coordinate position of each cell module or component, facilitating processing positioning and on-site installation. The above drawing files can be exported to standard engineering file formats, including .dwg, .dxf, .svg, etc., supporting seamless integration with process planning systems and manufacturing systems. Solar panel cell selection data is generated. For each preferred layout area, the system automatically matches the most suitable photovoltaic cell module type based on its structural parameters, variation deviation, and area size, and generates a set of structured selection parameters, including: area number (a unique identifier for the area); area centroid coordinates (using three-dimensional coordinates [x, y, z] to represent the area's spatial position in the substrate model); effective laying area (the actual surface area available for cell placement, in square meters); variation deviation (the maximum structural displacement of the area under laying conditions, in millimeters); matched cell type, such as "PERC-158×158-72" high-efficiency crystalline silicon module; and expected power value (estimated based on the cell module's power density and laying area, in watts). After summarizing the above data, the system can output a cell selection recommendation table containing the selection results for all areas, and simultaneously generate a corresponding configuration atlas to assist the engineering implementation unit in subsequent process review, electrical layout, and module procurement.

[0081] Preferably, step S1 specifically includes:

[0082] Step S11: Acquire three-dimensional data of the solar array;

[0083] In one embodiment, 3D model data can be directly extracted from existing engineering design drawings, including but not limited to common CAD file formats such as IGES, STEP, and STL. Alternatively, a 3D modeling operation can be performed on a physical solar array sample, preferably using a laser scanner (e.g., FARO Focus) or a structured light scanner (e.g., Artec Leo) to acquire high-precision point cloud data, covering the surface contour and key structural details of the solar array. If an image-based 3D reconstruction scheme is adopted, image sequences of the solar array from different angles can be acquired, and a structural self-motion algorithm can be executed sequentially to estimate the camera pose and sparse 3D structure; a multi-view stereo reconstruction algorithm can be used to acquire dense point cloud data, achieving complete 3D reconstruction.

[0084] Step S12: Perform heterogeneous three-dimensional structure perception based on the solar array three-dimensional data to obtain heterogeneous three-dimensional structure data;

[0085] In one embodiment, the heterogeneous three-dimensional structure refers to a composite structural unit composed of multiple different materials in the overall structure of the solar array, typically including heterogeneous material components such as carbon fiber reinforced plastic (CFRP) substrates, flexible photovoltaic cells, and aluminum alloy support frames. These components exhibit significant differences in geometric features, reflectivity, and spectral response. The system integrates multi-source sensors to collect material features from each point cloud segment in the solar array's three-dimensional model, primarily including RGB color reflectivity to identify surface color or coating differences; laser reflection intensity to reflect the reflectivity of different materials to laser light; and infrared response characteristics to identify material regions with differences in thermal radiation, such as flexible photovoltaic units. Based on the collected multimodal point cloud attributes, feature vectors including color distribution, texture features, and surface normal direction are extracted. Subsequently, clustering algorithms (such as K-Means or density-based DBSCAN algorithms) are used to classify the point cloud data into material categories, identifying each heterogeneous material region and completing coarse-grained material partitioning. The geometric shape features of the heterogeneous regions in the material classification results are then identified. For example, flexible battery cells in solar panels are often rectangular or curved sheets, while the connecting structures are elliptical rods or conical shafts. To achieve this identification process, the system employs edge contour-based geometric modeling algorithms, such as RANSAC plane fitting or cylindrical surface fitting methods, to perform high-confidence modeling and segmentation of typical geometric segments in point cloud data. A structural annotation set is established for each identified structural unit. The annotation content includes, but is not limited to, the following elements: component type (e.g., "main wing frame," "connecting pin"); material (e.g., "carbon fiber reinforced composite material," "aluminum alloy"); geometric contour type (e.g., 3D cuboid, 3D cylinder); and spatial coordinate range, i.e., the 3D bounding box range occupied by the component in a unified coordinate system.

[0086] Step S13: Perform adaptive domain sensing on the heterogeneous 3D structure data to obtain adaptive domain data;

[0087] In one embodiment, the deployment adaptation domain refers to an effective area on the solar panel surface that meets the requirements for solar cell deployment. This area must meet multiple deployment conditions, including geometric flatness, illumination adaptability, and deployment feasibility assessment, to ensure high reliability of photovoltaic cell deployment in terms of spatial structure and energy efficiency. Geometric flatness analysis specifically involves the system dividing the solar panel surface into multiple local small units (triangular meshes or equidistant sampling elements), and calculating the angle θ between the normal vector of each small unit and the reference normal vector of the overall solar panel span. If the normal deviation angle θ of a small unit is less than a preset threshold (e.g., 10°), the unit is determined to be a flat area, meeting basic deployment stability requirements. Deployment feasibility assessment involves further performing local meshing processing (e.g., Delaunay triangulation) on the flat area, calculating the average surface curvature κ on each mesh segment to assess whether there is excessive undulation or deformation in the area. When the average curvature κ of a certain area is less than or equal to a preset threshold (e.g., 0.05m), the deployment feasibility assessment is performed. ―1 Based on the above three dimensions, the system determines that the region has sufficient geometric flatness and structural ductility, and can be included in the deployment adaptation domain as a feasible area for deployment. With a set light source direction (e.g., a typical solar incidence angle), the system calculates the angle α between this direction and the surface normal vector of each unit to evaluate the region's light reception efficiency. If the incidence angle α of a certain unit is less than a preset light adaptation threshold (e.g., 45°), the region is identified as a high-efficiency light region, suitable for priority deployment of photovoltaic units to improve energy utilization. Based on the joint judgment of the above three dimensions, the system marks a set of deployment adaptation domains that meet the requirements of deployment flatness, deformation tolerance, and light efficiency.

[0088] Step S14: Generate a substrate model based on the deployment adaptation domain data to obtain a three-dimensional model of the solar array.

[0089] In one embodiment, all local geometric fragments identified as deployable regions within the deployment adaptation domain are fused and stitched together to reconstruct an integrated substrate structure. This fusion process ensures that the spatial positional relationships and edge matching relationships of each region are consistent within a unified reference coordinate system, guaranteeing the geometric consistency of the generated model in terms of macroscopic structure. For boundary discontinuities or gaps existing during the region stitching process, a repair strategy based on boundary curve interpolation is employed. Specifically, this includes geometric stretching and smoothing methods to connect regions with similar curvatures; and Bezier surface interpolation techniques to insert continuous surface fragments at boundary transitions, improving the smoothness and structural integrity of the substrate surface. Based on the generated continuous substrate geometry, a topological structure representation is constructed, for example, using a Half-Edge data structure to construct a "node-edge-face" ternary topological network. This topological structure facilitates efficient patch operations, local editing, and attribute binding. For each facet region in the topological mesh, its key attribute information is labeled, including the material type (e.g., flexible battery layer, supporting base plate, etc.); the deployment level (e.g., high, medium, and low levels based on lighting adaptability or deployment priority); and local deformation constraints (e.g., thermal expansion compensation regions, edge tension regions, etc.). Based on the spatial distribution density of the deployment adaptation domain, the system performs mesh refinement processing on the 3D model. Specifically, in regions with high deployment density and complex structural features, the mesh resolution is increased to enhance the model's expressive power; in regions with gentle structures, the mesh density can be appropriately reduced to optimize model volume and computational efficiency. The output is a standard 3D file (e.g., .obj, .stl) or a program-parseable format (e.g., .vtk, .glTF).

[0090] Preferably, the heterogeneous three-dimensional structure perception specifically refers to:

[0091] Heterogeneous data is obtained by performing heterogeneous data analysis based on the three-dimensional data of the solar array;

[0092] In one embodiment, the system performs heterogeneous attribute analysis on the three-dimensional data of the solar array to extract multi-source structural information. The three-dimensional data can be in industry-standard formats such as .STL, IGES, and STEP, or in the form of a custom mesh model, containing metadata such as material identifiers, assembly information, and naming rules. Its structure can be an attribute hierarchy tree in a CAD system. For geometric structure analysis, the system extracts mesh patches or volume elements (such as boundary representation B-rep) from the model and constructs a spatial index structure based on their spatial coordinates to support accurate identification of local features and adjacency relationships. For material attribute extraction, if color coding or explicit material labels exist in the model, the corresponding material type (such as aluminum alloy, carbon fiber reinforced composite material, etc.) is directly recorded; if label information is lacking, a preliminary determination of the material type is made by combining density parameters, reflectivity characteristics, and thickness data (based on a knowledge engine built from historical experience data). A machine learning classification model (built by deep learning training based on historical data and corresponding labels) can be introduced to enhance the accuracy of material identification. In terms of assembly hierarchy extraction, the system reads the assembly structure description information from the model and constructs the component hierarchy relationship, including the main wing assembly, secondary components, and connecting brackets. This structure adopts a tree-like or graph-like representation, supporting structural functional dependency analysis and layout logic modeling.

[0093] Structural semantic mapping is performed on heterogeneous data to obtain structural semantic data;

[0094] In one embodiment, the system constructs a structural semantic mapping model based on heterogeneous structural data to generate structural semantic data. This process aims to assign semantic tags to each three-dimensional structural component, thereby realizing the knowledge-based expression of structural units at the functional and material levels. The semantic tag system includes the following three dimensions: (1) structural role tags, used to identify the functional positioning of components in the overall structure, such as main load-bearing components, support frames, hinged ends, battery embedding areas, etc.; (2) functional area feature tags, used to describe the functional attributes of structural sections, such as unfolded areas, fixed areas, and adjustable areas; (3) material semantic tags, used to reflect the physical properties of structural units, such as lightweight high-stiffness areas, flexible areas, and shielding areas, etc. In the tag assignment process, the system can adopt two semantic mapping methods. The first is a method based on expert knowledge rules, which maps component attributes to semantic tags according to predefined condition rules. For example, when a component is made of carbon fiber reinforced composite material and its main side length exceeds 500 mm, it can be labeled as a main load-bearing component. The second method is a machine learning-based classification approach. This involves training on historically labeled data to build a structural component identification model (such as a model based on random forest, graph convolutional network, or support vector machine). The model input includes the component's geometric features (such as aspect ratio and thickness), material type, and their spatial relationships. The output is the probability distribution of each semantic label, and the system selects the label with the highest confidence as the semantic classification result for that component.

[0095] Three-dimensional structural representations are generated based on structural semantic data to obtain heterogeneous three-dimensional structural data.

[0096] In one embodiment, the system constructs a structural representation based on structural semantic data and a 3D mesh model, thereby generating a 3D structural representation model with heterogeneous semantics, spatial boundaries, and physical properties, referred to as a structural representation. The structural representation refers to a geometric unit with clearly defined spatial boundaries, structural semantic labels, and behavioral attributes, which can serve as input units for tasks such as layout calculation, collision detection, and path simulation. During the construction process, the system establishes a spatial representation object for each structural component, specifically including a local coordinate system to describe the component's relative spatial position and orientation within the overall system; a bounding volume structure, such as using a minimum bounding rectangle, convex hull, or composite volume to describe its physical boundary range; and contact surface information, recording the boundary surfaces that contact or interfere with other structural units, used to support collision detection and connection modeling. The system binds the geometric information of each component with its corresponding semantic labels and physical feature parameters to form a complete structural representation unit. This unit can be abstracted as a structural triple, represented as [geometric entity, semantic label, physical attribute]. The semantic labels include structural role, functional region type, and material semantics, while the physical attributes include density, stiffness coefficient, and thermal expansion characteristics. During the relationship modeling phase of the representations, the system determines structural connections based on spatial proximity rules. If the length of the shared edge region between two representations exceeds a preset proportion (e.g., 5%) of their boundary lengths, a connecting edge is added to the structural connection graph. This structural connection graph serves as a representation of the spatial coupling between the representations.

[0097] Preferably, the deployment of adaptive domain sensing specifically includes:

[0098] Surface geometric pavingability is calculated based on heterogeneous 3D structural data to obtain preliminary geometric paving domain data.

[0099] In one embodiment, based on heterogeneous 3D structural data, a geometrical layability analysis of the solar array surface is performed to identify areas with structural layability and generate preliminary geometrical layability domain data. This process includes the following specific steps: dividing the 3D surface model of the solar array into patches. Preferably, a triangular meshing algorithm is used to divide the entire surface, forming a patch set T = t1, t2, ..., t n Each triangular facet t i It is considered the basic unit of layability analysis. For each facet t... i Extracting geometric pavingability indices includes the following three key parameters, such as calculating the normal deviation angle θ. i This is used to measure the angle between the normal direction of the solar panel and the main deployment direction of the overall solar array. It is calculated as follows: For the direction of this film, The overall principal direction is used. The local curvature κ is calculated. iThis is used to assess the geometric curvature of a patch within its neighborhood. The rate of change of the angle between the normal vectors of adjacent patches is preferably used to approximate the Gaussian curvature or mean curvature value. A smaller curvature value indicates a smoother surface in that area, which is beneficial for tiling. The standard deviation of the patch area σ is calculated. a This is used to evaluate the uniformity of local mesh generation. By statistically analyzing the standard deviation of the area of ​​all faces within a certain neighborhood around each facet, a smaller standard deviation indicates a more uniform mesh density in that area, which is beneficial for layout control and ensuring manufacturing accuracy. If: ―θ i <10° (small directional deviation), ―κ i <0.05 (locally relatively flat), ―σ a If the grid size is less than 20 (uniform grid), the triangular facet is marked as initially paved.

[0100] Structural functional regions are removed from the preliminary geometrically layable domain data to obtain functional mask data;

[0101] In one embodiment, the goal of this process is to identify and exclude structural functional areas on the solar panel surface that are unsuitable for placing photovoltaic cells, such as connector mounting areas, hinge areas, locking mechanism areas, and inspection hole areas. The method for identifying structural functional areas is based on the structural annotation information extracted from the heterogeneous 3D structural data in the preceding steps. This information includes the structural type labels, geometric contour descriptions, and spatial boundaries of each component in the 3D model. In specific operation, the system first extracts marked areas of structural components with functional attributes, including but not limited to "connection areas," "hinge areas," "locking mechanism areas," and "inspection hole locations." Each type of functional area corresponds to a structural unit with a clear physical meaning and arrangement constraints in 3D space. Based on the geometric envelope information of these functional components, the system constructs the corresponding functional area projection in the model surface coordinate system. Preferably, an axis-aligned bounding box or polygonal projection contour is used to map the 3D boundary to a 2D surface mesh, thereby achieving accurate coverage on the original layout area. Based on the above, the system constructs a binary functional mask matrix corresponding to the solar panel surface. In this mask matrix, each surface unit location (e.g., a triangular facet or mesh node) is assigned a binary state flag based on whether it is located within a functional area: if the location belongs to a non-layable functional area, the corresponding mask value is set to 0, indicating that cell placement is not allowed at this location; if the location is not covered by a functional area, the corresponding mask value is set to 1, indicating that the location can remain in the layable domain. The output functional mask data is used to filter out points from the initial geometrically layable domain, thereby retaining only structurally layable, high-availability areas.

[0102] Semantic mask fusion is performed based on functional mask data and heterogeneous 3D structural data to obtain semantically enhanced deployment domain data.

[0103] In one embodiment, each surface triangular patch unit in the 3D model is assigned multiple semantic enhancement variables, including but not limited to the following types: material type, such as fiberglass, flexible copper-based materials, etc.; thermal deformation level, which is divided into high, medium, and low thermal stability levels based on the thermal stress response results of the region in the tiling simulation; and electromagnetic shielding area marking, used to identify whether it is a high electromagnetic interference area, where it is not recommended to place sensitive electronic devices such as photovoltaic cell modules. Based on the introduction of the above semantic variables, the system constructs a fusion function for each patch to fuse the geometric adaptability evaluation results and semantic credibility results, such as S. i =w g GS i +w s ·SS i S i To integrate index data, w g With a geometric score weight of 0.7, GS i For the geometrically plausible score (normalized from the scores of the first two steps), w s With a semantic score weight of 0.3, SS i The system assigns semantic relevance scores (e.g., avoiding high-heat areas and electromagnetic interference areas); after the fusion index is calculated, a unified judgment threshold is set (e.g., a score greater than 0.6), and patches with fusion index results higher than this threshold are judged as semantically enhanced deployment patches. The system further aggregates all patches that meet the conditions to generate semantically enhanced deployment domain data.

[0104] Regional connectivity filtering is performed on the semantically enhanced deployment domain data to obtain deployment adaptation domain data.

[0105] In one embodiment, a connected component extraction operation is performed. Using the triangular facets of the laid surface as graph nodes, if two facets share a common edge (i.e., share a boundary edge), a connecting edge is established between the nodes. The resulting connectivity graph reflects the topological adjacency of the laid area in geometric space. After the graph structure is constructed, a standard graph traversal algorithm (such as Breadth-First Search (BFS) or Depth-First Search (DFS)) is used to identify the connectivity of the connected graph, extracting all structurally connected sub-regions, denoted as the connected subgraph set C = c1, c2, ..., c m , where each subgraph c j This represents a locally continuous distribution region. For each connected subgraph c... j Calculate the following three key geometric parameters, including the effective area A. j , representing the total area that can actually be deployed in the connected region, in square meters; maximum connected diameter D j , representing the maximum Euclidean distance between any two faces within a connected region, used to evaluate the farthest boundary length of the paving path, in millimeters; boundary complexity Bj , defined as the ratio of the total length of the outer contour line of the area to the area of the region, reflects whether the shape of the region edge is too complex, without units. The condition for retention is A j > Amin (the threshold of the minimum layable area, such as 0.05㎡); D j > Dmin (the threshold of the minimum connected distance, such as 150mm); B j < Bmax (the threshold of the maximum allowable complexity, such as 1.8).

[0106] Preferably, the structural drive spreading strategy simulation is specifically as follows:

[0107] Step S21: Construct a structural constraint graph for the three-dimensional model of the solar wing to obtain structural constraint graph data;

[0108] In an embodiment, the key structural points in the three-dimensional substrate model are defined as the nodes of the graph, mainly including grid points (i.e., the vertices of the surface triangular grid or polygon grid); region seam points (the connection edge points of different structural materials or modules); the end points of the rotating shaft or the moving joint points. Each node v i contains coordinates and structural attributes (such as "fixed point", "edge point", "rotatable point"). The construction edges between adjacent structural patches establish graph edges e ij , and each edge represents the structural constraint path between nodes. The types of structural edges are divided into rigid edges (non-spreadable areas); hinged edges (with angular constraints); transition edges (allowing laying changes); and the associated parameters of each edge are the maximum spreading angle θ max , the allowable deformation amount δ max , and the connection type.

[0109] Step S22: Schedule the laying behavior graph for the structural constraint graph data to obtain laying behavior graph data, where the laying behavior graph scheduling includes a horizontal filling strategy, a vertical filling strategy, a central symmetry filling strategy, and an S-shaped filling strategy;

[0110] In one embodiment, each node represents a "laying unit"; the node contains attributes such as current status (not laid / laying in progress / completed) and activation conditions (such as "adjacent laid areas reach 3 sides"). The lateral filling strategy (advancing along the X-axis) involves laying operations sequentially along the long side of the solar wing (X-axis), starting from one wingtip and progressing to the other; the activation condition is that the unit is activated after the adjacent unit to its left has completed laying. The longitudinal filling strategy (advancing along the Y-axis) involves laying operations along the short side of the solar wing (Y-axis), progressing from the wing root (closer to the body) to the wingtip (outer edge); the activation condition is that the adjacent unit on the inner side (closer to the wing root) has completed laying. The centrally symmetrical filling strategy involves simultaneously unfolding the laying path symmetrically along the main axis to both sides, starting from the geometric center of the solar wing, ensuring a symmetrical distribution of the laying load; the requirement is that the target node has structural symmetry, and its symmetrical nodes simultaneously meet the laying activation conditions. The S-shaped filling strategy (serpentine path) lays the paper along the intended printhead's round-trip path, alternating between left-to-right and right-to-left directions. It is suitable for laying large strip-shaped areas and effectively reduces the reversing frequency and execution latency of the motion system. The dynamic execution flow of the laying behavior graph is represented using a finite state machine or a directed acyclic graph model. Each node's state transition is controlled by activation conditions; state transition events include operations such as "layout activation," "state transition," and "path completion." In each simulation round, the system traverses the graph structure based on the current laying graph state, identifies the set of currently activatable nodes, updates their states, and records the laying behavior path, providing input for control execution.

[0111] Step S23: Verify the legality of the laying behavior diagram data to obtain fixed simulation data.

[0112] In one embodiment, after each laying step, the splicing error Δx between the laid surfaces is checked to be less than ε (e.g., 0.5 mm); iterative Boolean operations are used to verify whether there are intersecting, suspended, or broken areas between the laid surfaces. Structural simulation is performed on the current laying state, such as setting nodal boundary conditions (fixed / free / elastic); applying loads (gravity, solar pressure); and calculating the maximum stress σ. max With allowable stress σ allow σ is required max <0.9·σ allow If the conditions are not met, the strategy is rolled back or the laying order is reset. It is determined whether the laying path traverses all target regions; it is checked whether the policy dependency chain is closed-loop (to avoid logical deadlock); if path islands or non-convergent behavior exist, it is marked as "non-convergent simulation".

[0113] Preferably, the generation of the perturbation layout scheme specifically involves:

[0114] Step S24: Perform local graph partitioning on the structural constraint graph data to obtain local graph data;

[0115] In one embodiment, the overall structural diagram is divided according to spatial proximity or functional partition attributes, specifically including the following two typical strategies: (1) Taking the key node in the structural diagram as the center, expand outward to form a set of k-order adjacent nodes to form a local subgraph. k is an integer, representing the maximum connection depth between nodes in the diagram (for example, k=2 means that starting from the core node, the adjacent nodes are retained after a maximum of two layers of connection). Construct a spatially compact and highly connected subgraph region for local disturbance testing. (2) Based on the functional area division of the solar wing structure, the nodes in the diagram are partitioned and mapped according to their structural functions, such as the main wing section, auxiliary deployment mechanism, end structure area, etc. Each functional segment is mapped to an independent local diagram. On the basis of functional consistency, strategy optimization and simulation deduction are performed to improve the local accuracy of structural function simulation. Regardless of which partitioning strategy is adopted, the system ensures that the generated local diagram maintains connectivity in the structural topology and has the characteristics of being independently disturbed and analyzed in logic, which facilitates the generation of diversity and local stability screening in subsequent layout schemes.

[0116] Step S25: Perform perturbation selection on the local map data to obtain perturbation map data;

[0117] In one embodiment, the system quantifies and scores each local graph based on three perturbation sensitivity indices: structural degrees of freedom (DOF), i.e., the total number of unconstrained nodes in the local graph, reflecting the potential deformation capacity of the region. Higher DOF indicates a greater likelihood of significant response from external forces or spreading perturbations; node density, i.e., the ratio of the number of nodes to their area in the principal projection direction. This density index represents the complexity and tightness of the local structure; and structural stress concentration, obtained based on historical simulation data or structural static analysis results, reflecting the maximum stress distribution trend of the local region under stress, and is an important basis for judging the structural response sensitivity. Based on these three indices, the system constructs a weighted scoring function, assigning a set of perturbation score values ​​to the sensitivity of each local graph. By setting the weighting coefficients for each index (e.g., configuring the emphasis on nonlinear deformation or structural strength response according to the actual simulation strategy, such as adding randomness with weights of 0.5, 0.25, and 0.25), the system scores and ranks all local graphs. The system can use one of the following strategies to select perturbation map data: (1) select the top-k local maps with the highest scores as perturbation candidate regions; (2) randomly select several local maps whose scores exceed the set threshold; (3) execute the full coverage strategy to perform perturbation simulation processing on all local maps.

[0118] Step S26: Select disturbance space from the laying behavior map data based on the disturbance map data to obtain disturbance space data;

[0119] In one embodiment, the system maps the set of structural nodes contained in the perturbation graph to the corresponding set of laying units in the laying behavior graph. If the activation condition of a laying unit depends on the local structural response path, i.e., there is a behavioral chain dependency structure, then the unit is determined to be a target area with perturbation potential and marked as a perturbable space. The perturbation space can be divided into the following types according to the laying behavior characteristics and perturbation mode it undertakes in the behavior graph: path perturbation type, which realizes the perturbation rearrangement of the chain behavior path by adjusting the execution order of local laying units in the behavior graph; delay perturbation type, which introduces a control delay in the activation process of local behavioral units to simulate the intervention of response lag or structural feedback mechanism on the laying process; strategy replacement type, which partially replaces the original laying method, such as changing the sequential or S-shaped laying method to a symmetrical or alternating layout strategy, thereby constructing a multi-strategy perturbation space.

[0120] Step S27: Perform perturbation processing based on the perturbation spatial data to obtain perturbation result data;

[0121] In one embodiment, the system applies various disturbances to the corresponding units of the laying behavior graph, thereby constructing multiple perturbed layout candidate paths and control schemes, and evaluating their stability and robustness during the structural response process. Specific disturbance types and operations include, but are not limited to, the following: sequence disturbances, which introduce changes in the local path sequence by exchanging the execution order of adjacent laying units to explore the impact of the laying logic order on the overall structural response; angle disturbances, which introduce directional offsets within a certain range at the structural node level to simulate angular errors or structural installation deviations during node placement; positional disturbances, which apply small spatial displacement disturbances to local nodes in the laying behavior graph to simulate geometric position changes caused by physical deviations during placement; and strategy disturbances, which replace the original laying strategy path, for example, replacing the "S-shaped placement method" with a "bidirectional symmetrical strategy," to evaluate the structural energy efficiency, placement path length, and response robustness under different strategy forms. During the disturbance processing, the system progressively advances the simulation calculation process based on the perturbed behavior graph. The simulation module will update the state of the structural model in real time, dynamically record the structural displacement field distribution, the evolution process of the laying state diagram, and the trend of energy function changes under each disturbance step, and generate complete structural evolution trajectory data.

[0122] Step S28: Verify the validity of the disturbance result data to obtain random simulation data.

[0123] In one embodiment, the system performs legality judgment from two dimensions: geometric integrity and mechanical stability. Geometric legality verification specifically involves the system performing a spatial topological consistency check on the disturbed structural model, mainly including surface overlap detection, whether the layout units are suspended or disconnected, and whether the structural surfaces are damaged or broken. Specifically, the system calculates the minimum distance between all adjacent structural surfaces. If the minimum surface spacing is greater than 0.5 mm, the structure is considered geometrically legal; otherwise, the sample will be discarded. Mechanical legality verification specifically involves the system using finite element simulation tools to perform physical response analysis on the disturbed structure, focusing on two parameters: (1) the maximum stress value of the structure under loading conditions; and (2) the maximum response displacement of the structure. The system evaluates the difference between the maximum stress value in the disturbed sample and the stress reference value under the undisturbed state. If the absolute value of the stress change is less than a set threshold (e.g., 15 MPa) and the maximum displacement does not exceed the upper limit allowed by the system (e.g., 5 mm), the mechanical response is deemed reasonable. The perturbation results that satisfy the above geometric and mechanical conditions will be identified as valid perturbation samples, constituting a stochastic simulation dataset for the system to be used for strategy optimization, diverse layout generation and robustness evaluation.

[0124] Preferably, the regional structure domain analysis specifically includes:

[0125] Step S31: Construct the structural domain response tensor from the fixed simulation data and the random simulation data to obtain the structural domain response tensor data;

[0126] In one embodiment, the system extracts key structural response parameters from each laying unit in both fixed and random simulation data. Each laying unit can be a triangular facet, a rectangular tile, or other meshable surface unit. For each unit, the system records the following response fields: stress value σ, displacement δ, laying status marker s (laid / not laid), laying angle θ, and local energy dissipation index E. The system divides the entire simulation area into a two-dimensional spatial grid. For each grid unit, the system integrates its structural response fields in a fixed order to form a structural response vector. If the laying behavior occurs during a multi-step simulation process, and the response at each time step exhibits an evolutionary trend, the system can superimpose the response tensors at each time step into a third-order tensor structure. The dimension of this tensor can be represented as... m represents the spatial dimension, n represents the temporal dimension, and d represents the number of feature dimensions (e.g., 5 or 6).

[0127] Step S32: Calculate the deployable region coverage based on the structural domain response tensor data to obtain the deployable region coverage data;

[0128] In one embodiment, a binary field representing the element placement state is extracted from the structural domain response tensor, denoted as s. i,jWhere i and j represent the position indices in the two-dimensional spatial grid, and s i,j A value of 1 indicates that the unit has been deployed, while a value of 0 indicates that the unit has not been deployed. The system divides the entire deployment area into multiple sub-regions of fixed size, denoted as R. k Each sub-region contains several adjacent grid cells. Preferably, each sub-region can be set as a 5x5 grid block (5 rows and 5 columns) for local area statistics. In each sub-region R k Within the region, the system counts the number of deployed cells and calculates the ratio of this number to the total number of cells in the area, yielding the deployable coverage rate of the region, denoted as $\rho_k$. The calculation method is as follows: [The text abruptly ends here, likely due to an incomplete sentence or missing information.] k The coverage rate ρ is calculated by summing the number of all cells with a coverage state of 1 and dividing by the total number of cells in that area. k It equals the ratio of the number of units already deployed in that area to the total number of units. If ρ k ≥0.75, marked as "high coverage area"; if 0.4≤ρ k If the value is less than 0.75, it is marked as "medium coverage area"; otherwise, it is marked as "low coverage area".

[0129] Step S33: Calculate the local symmetry offset based on the structural domain response tensor data to obtain the local symmetry offset data;

[0130] In one embodiment, a reference axis of symmetry is uniformly defined across the entire laying surface or solar array structure. Preferably, the geometric center line or the X-axis of the structural coordinate system can be selected as this reference axis of symmetry, denoted as axis of symmetry x = x0. This axis corresponds to the solar array deployment direction or the main line of symmetry in the design drawing. Subsequently, for each local layout region R... k The system calculates the two-dimensional geometric centroid coordinates of this region as (x c ,y c ), where x c This indicates the position of the centroid of the region in the X direction, and y... c This represents the centroid position in the Y direction. The centroid is calculated by weighted averaging of the position coordinates of all laid units within the region. After obtaining the centroid position, the system uses x = x0 as the symmetry reference axis and defines the first-order symmetry offset of the region as the absolute difference Δ between the centroid of the region and the symmetry axis in the X direction. sym (R k )=|x cThe closer this value is to zero, the more symmetrical the layout of the area, theoretically resulting in more balanced stress, making it suitable for battery modules with high structural symmetry requirements. In this embodiment, a second-order symmetry offset metric is further defined to evaluate the overall difference between the shape of a region and its mirror image about the axis of symmetry. The region's outline or layout shape is mirrored with respect to the axis of symmetry, and the difference in spatial distribution between the original and mirrored regions is calculated using the mean square error as a mirror offset metric. The second-order symmetry offset can be defined as the mean square error between the original region and its mirrored region, reflecting the consistency of the overall shape symmetry. The smaller this value, the more symmetrical the region's shape. Through the joint calculation of the first-order and second-order symmetry offsets, the system can quantify the symmetry performance of each local region in multiple dimensions.

[0131] Step S34: Perform region label inference based on the deployable area coverage data and local symmetry offset data to obtain region label data;

[0132] In one embodiment, the system pre-defines a set of region labels to characterize the structural adaptability level of a region. This label set includes three categories: Preferred regions indicate regions with both high deployment density and good structural symmetry, suitable for priority battery placement; Available regions indicate regions with moderate deployment density and symmetry, suitable for selection when resources are limited or deployment requirements are complex; Risk regions indicate regions with significant defects, such as low coverage or poor symmetry, and are not recommended as preferred deployment targets. The reasoning process for these labels is based on two key indicators: coverage, representing the ratio of already deployed units to deployable units within the region, reflecting deployment density; and symmetry offset, representing the degree of centroid offset of the region relative to the axis of symmetry, reflecting structural stability. The system uses the following classification rules for label reasoning: If the coverage of an area is not less than 75% and the symmetry offset does not exceed 5 mm, the area is marked as a preferred area, indicating that it performs well in terms of both laying efficiency and structural symmetry; if the coverage of an area is between 40% and 75% and the symmetry offset does not exceed 15 mm, the area is marked as a usable area, indicating that it has basic deployment value; if an area does not meet any of the above conditions, i.e., the coverage is less than 40% or the symmetry offset exceeds 15 mm, the area is marked as a risk area and is not recommended for priority use.

[0133] Step S35: Construct candidate maps for regional structure by dividing and sorting fixed simulation data and random simulation data according to regional label data, and obtain regional structure data.

[0134] In one embodiment, the system uses each structural region as a map node and archives the corresponding multidimensional structural attributes and response characteristics of that region as node attributes, forming the basic unit of the regional candidate map. Each regional node contains the following feature fields: the deployable area coverage value, reflecting the deployment density of the region; symmetry offset, indicating the degree of deviation of the region layout from the symmetry axis; the region average stress value, representing the structural load response intensity of the region during the simulation process (which can be calculated based on historical experience or obtained by finite element calculation based on simulation data obtained from preset parameters and strategies); the region label type, such as "preferred area", "available area" or "risk area", representing the structural adaptability level of the region; and the recommended laying direction, such as along the X direction, Y direction or diagonal direction, reflecting the suitable laying path planning strategy. While constructing nodes, the system further establishes the connection relationship (i.e., the edge structure) between regional nodes in the map to express the spatial adjacency or physical response similarity between regions. The system determines whether two regions are connected based on any of the following conditions: spatial connectivity (if the Euclidean distance between the geometric centers of the two regions is less than a set spatial threshold, they are considered spatially adjacent and a connection is established); and response gradient continuity (if the main indices (such as stress or displacement values) of the two regions in the simulation response show strong gradient continuity, they are considered to have a response coupling relationship and a connection can be established). The system organizes all candidate regions, connection relationships, and label information into complete regional structure domain map data.

[0135] Preferably, the preferred layout order is as follows:

[0136] Step S36: Construct regional weight tensor data by fusing regional structural domain data;

[0137] In one embodiment, multi-dimensional attributes of the region (such as coverage, stress, symmetry, etc.) are integrated to form a weighted region tensor, which serves as the basis for layout ranking. For each region R... k Extract the following feature indicators, such as the deployable coverage rate ρ k Symmetry offset Δ sym,k Mean stress Number of units N k The score for the continuity of the laying direction is C. k (Extract path vectors; for each pair of adjacent path vectors, calculate the cosine of the angle between the unit vectors; based on the cosine of the angle between the unit vectors, cosθ...) i Calculate the average directional consistency, such as If all directions are consistent, C k →1 represents strong continuity; if the direction changes drastically, C k →0 represents poor coherence), label confidence P kAll metrics are min-max standardized to maintain the range [0,1], and a tensor is constructed such that each region corresponds to a multidimensional weight vector. All regions form a three-dimensional tensor structure. Where n is the number of regions and d is the feature dimension.

[0138] Step S37: Perform nonlinear mapping sorting on the region weight tensor data to obtain preliminary layout sorting data;

[0139] In one embodiment, a nonlinear scoring mapping strategy is used to weight the vectors of each region. (Refer to the previous step for parameter names and meanings) Sort the parameters to form a preliminary priority result. Use a weighted nonlinear mapping function, such as Sigmoid, ReLU, or an exponential function, to enhance response sensitivity. Ranking and scoring function. Where x i For the region weight tensor data, f(x) i ) represents a nonlinear transformation (e.g., f(x) = log(1 + αx), where α can be set to 5); weight w i This can be set according to user preferences or system goals, such as coverage priority → w1 = 0.4, symmetry priority → w2 = 0.2, lower stress is better → w3 = 0.2, and other balancing terms → w4 + w5 + w6 = 0.2, for all R... k According to S k Sort from largest to smallest; retain index, region ID, and sort score.

[0140] Step S38: Perform semantic annotation enhancement on the preliminary layout ranking data to obtain the layout ranking data.

[0141] In one embodiment, semantic descriptions are added to each candidate region based on the ranking to assist in manual review or system interpretability. The system infers the structural stability level of a region during actual deployment based on its deployable coverage and average stress index. If a region exhibits a combination of "high coverage + low average stress" (judged based on an empirically set threshold), it is labeled a "stable region"; if a region exhibits "low coverage + high stress" (judged based on an empirically set threshold), it is labeled a "high-risk region"; and intermediate cases are labeled "acceptable risk." Geometric morphology identification is performed on the boundary shapes of each region, based on convex hull analysis, aspect ratio calculation, and boundary connectivity identification. For example, convex regions have virtually no concavity at the boundary, suitable for symmetrical deployment; strip-shaped regions have a long side much larger than the short side, suitable for S-shaped deployment strategies; and concentrated block-shaped regions have relatively compact boundaries, suitable for deployment expanding from the center. The system combines the structural characteristics exhibited by the region in the ranking tensor to recommend the most suitable deployment path type for that region. If a region exhibits a "strip-like shape with high continuity," then "suitable for S-shaped paving paths" is recommended; if a region has a centrally symmetrical shape and stable structure, then "suitable for centrally symmetrical paving" is recommended; if there are no obvious features or the boundaries are broken, "locally adaptive paving" can be suggested. The system extracts the key factors that have the greatest impact on the score from the ranking scoring function to generate highly readable recommendation explanations. If the score of a region is dominated by high coverage and low symmetry offset, the system can automatically generate the statement: "Due to the high coverage and good symmetry of this region, paving is recommended as a priority"; if it is mainly controlled by stress, the system generates: "This region has a low stress level and good structural bearing capacity."

[0142] Preferably, step S4 specifically includes:

[0143] Step S41: Perform 3D layout projection based on the layout sorting data to obtain 3D layout projection data;

[0144] In one embodiment, if the region information in the sorted data directly corresponds to a predefined local grid segment on the surface of the solar array, its grid index and node information can be read directly without coordinate transformation.

[0145] In one embodiment, if the layout region is given by two-dimensional planar analysis or external coordinate system results, it needs to be mapped to the surface of the three-dimensional model through the following steps: using the nearest point projection method, the centroid coordinates and boundary point set of the layout region are projected onto the mesh of the three-dimensional model surface to determine its mapping target patch set. During the projection process, ensure the normal direction of the layout region (denoted as...) ) and the direction of the normal to the surface of the solar array (denoted as ) The orientations are consistent, satisfying the following orientation consistency conditions. If this condition is not met, the region's attitude is flipped or reflected to ensure surface fit. The laying plane in the external reference frame is mapped to the model's local coordinate system.

[0146] Step S42: Unfold the facets based on the 3D layout projection data to obtain the unfolded facet data;

[0147] In one embodiment, the layout area patches in three-dimensional space are geometrically unfolded into a two-dimensional planar structure suitable for drawing representation. A geometric unfolding algorithm (such as LSCM) is used to maintain angular integrity; for complex curved surface patches, partitioned unfolding or physical simulation unfolding (spring model) can be used; after unfolding, each patch F... i Includes 2D vertex coordinates, edge lengths, and original 3D reference information. Unfolded patches are arranged in layout order, maintaining relative spatial logic; add reference direction marker arrows (such as the direction of solar incidence, the starting edge of the tile, area numbering, and strategy labeling).

[0148] Step S43: Generate drawing elements based on the unfolded patch data to obtain drawing element data;

[0149] In one embodiment, the system uses each unfolded laying unit (such as a triangle, rectangle, or other polygonal patch) as the basic drawing object, extracts its two-dimensional boundary contour, and supplements it with engineering information related to the laying process. This generates the following types of drawing elements: geometric boundary line elements, used to represent the spatial contour of each laying unit in the drawing, in the form of a closed polyline or closed polygonal structure; laying start mark elements, used to identify the laying start point of each laying block, often presented as a starting dot, starting number arrow, etc.; unit number and battery model label elements, generating a unique number for each laying unit and attaching the battery type information selected for that unit, such as "B-158×158-72," used to guide battery component configuration; spatial reference information elements, marking the mapping relationship of the laying unit in the original three-dimensional structure, including the original structure ID, laying grid position number, etc.; and auxiliary layer elements, used to display additional attributes of the laying area, such as simulated stress distribution contour maps, occlusion area markers, and preserved area contours. The system introduces a layer management mechanism to organize and manage different categories of drawing elements across multiple layers. Preferably, the following layer structure is set: Layer 0: stores structural outline elements as a basic framework reference; Layer 1: stores laying path information and numbering markers to guide layout and installation operations; Layer 2: used to add text annotations, technical specifications, and other textual labels; Layer 3: reserved for special areas used in process planning, such as wiring space and buffer areas. The system encapsulates each complete drawing unit (composed of boundary lines, numbering, type labels, etc.) into a drawing element block. Each block is saved as an independent object, supporting operations such as copying, adjusting, and renumbering, facilitating drawing version iteration or modular application.

[0150] Step S44: Export the drawing format based on the drawing element data to obtain the solar panel battery selection data.

[0151] In one embodiment, the system supports exporting drawings in various common CAD formats, including .DXF, .DWG, .SVG, and .PDF. Graphic drawing and drawing generation can be accomplished through CAD development interfaces, such as the AutoCAD.NET API or OpenCascade development tools. During the export process, the system can automatically set the view scale, print area, text annotation size, and drawing frame structure of the drawing, and generate a metadata column below the drawing containing information such as drawing number, drafter, reviewer, and version number to meet the specifications of standard engineering drawings. Simultaneously, the system generates structured battery selection information based on the geometric dimensions and layout density of each battery placement area recorded in the drawing elements. This includes fields such as: area identifier (area ID), used to identify independent placement areas on the solar panel; number of units, indicating the total number of battery units that can be accommodated in the area; matching battery model, automatically recommending a matching standard battery model based on the area shape and unit specifications; power estimate, calculating the estimated total power output of the area based on the rated power and quantity of each battery; and electrical connection method suggestion, suggesting series, parallel, or mixed connection methods based on the layout topology for reference during electrical system design.

[0152] Preferably, this application also provides a solar panel battery selection system based on a three-dimensional layout, used to perform the solar panel battery selection method based on a three-dimensional layout as described above. The three-dimensional layout solar panel battery selection system includes:

[0153] The 3D substrate modeling module is used to acquire 3D data of the solar array and construct a 3D substrate model based on the 3D data of the solar array to obtain the 3D model of the solar array.

[0154] The layout simulation generation module is used to perform structure-driven deployment strategy simulation on the 3D model of the solar array to obtain fixed simulation data, and to perform perturbation-type layout scheme generation on the 3D model of the solar array using fixed strategy simulation to obtain random simulation data.

[0155] The structural region analysis and optimization sorting module is used to perform regional structural domain analysis on fixed simulation data and random simulation data to obtain regional structural domain data, and to perform optimization layout sorting on the regional structural domain data to obtain layout sorting data.

[0156] The drawing generation and configuration output module is used to export two-dimensional drawings from the layout sorting data to obtain solar panel battery configuration data.

[0157] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended application documents rather than the foregoing description. Thus, it is intended that all changes falling within the meaning and scope of the equivalents of the application documents be incorporated into the invention.

[0158] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for solar array cell sizing based on three-dimensional layout, characterized in that, The method comprises the following steps: Step S1: obtaining solar wing three-dimensional data, and constructing a three-dimensional substrate model according to the solar wing three-dimensional data to obtain a solar wing three-dimensional model; Step S2: performing structural driving spreading strategy simulation on the solar wing three-dimensional model to obtain fixed simulation data, and generating a perturbation type layout scheme for the solar wing three-dimensional model to obtain random simulation data; Step S3: performing regional structure domain analysis on the fixed simulation data and the random simulation data to obtain regional structure domain data, and performing optimal layout sorting on the regional structure domain data to obtain layout sorting data; Step S4: performing two-dimensional drawing export on the layout sorting data to obtain solar wing cell matching data; The regional structure domain analysis specifically comprises: performing structural domain response tensor construction on the fixed simulation data and the random simulation data to obtain structural domain response tensor data, wherein the structural domain response tensor data comprises key structural response parameters of each laying unit in the system extracted fixed simulation data and random simulation data, the entire simulation region is divided into a two-dimensional space grid, and for each grid unit, a structural response field is integrated in a fixed order to form a structural response vector; performing a layout area coverage rate calculation according to the structural domain response tensor data to obtain layout area coverage rate data; performing local symmetry offset calculation according to the structural domain response tensor data to obtain local symmetry offset data; performing regional label reasoning according to the layout area coverage rate data and the local symmetry offset data to obtain regional label data; performing domain sorting candidate graph construction on the fixed simulation data and the random simulation data according to the regional label data to obtain the regional structure domain data; performing regional weight tensor fusion construction on the regional structure domain data to obtain regional weight tensor data; performing nonlinear mapping sorting on the regional weight tensor data to obtain preliminary layout sorting data; performing candidate scheme semantic annotation enhancement on the preliminary layout sorting data to obtain the layout sorting data.

2. The method of claim 1, wherein, Step S1 specifically comprises: obtaining solar wing three-dimensional data; performing heterogeneous three-dimensional structure perception according to the solar wing three-dimensional data to obtain heterogeneous three-dimensional structure data; performing laying adaptation domain perception on the heterogeneous three-dimensional structure data to obtain laying adaptation domain data; generating a substrate model according to the laying adaptation domain data to obtain a solar wing three-dimensional model; The heterogeneous three-dimensional structure perception specifically comprises: performing heterogeneous data analysis according to the solar wing three-dimensional data to obtain heterogeneous data; performing structure semantic mapping on the heterogeneous data to obtain structure semantic data; generating a three-dimensional structure expression body according to the structure semantic data to obtain heterogeneous three-dimensional structure data; The laying adaptation domain perception specifically comprises: performing surface geometry paving ability calculation according to the heterogeneous three-dimensional structure data to obtain preliminary geometry paving domain data; performing structure function area elimination on the preliminary geometry paving domain data to obtain function mask data; performing semantic mask fusion according to the function mask data and the heterogeneous three-dimensional structure data to obtain semantic enhanced laying domain data; performing regional connectivity screening on the semantic enhanced laying domain data to obtain the laying adaptation domain data.

3. The method of claim 1, wherein, The structural driving spreading strategy simulation specifically comprises: The structure constraint graph data is subjected to a laying behavior graph scheduling to obtain laying behavior graph data, wherein the laying behavior graph scheduling comprises a horizontal filling strategy, a vertical filling strategy, a center symmetry filling strategy and an S-shaped filling strategy. The structure constraint graph data is subjected to a laying behavior graph scheduling to obtain laying behavior graph data, wherein the laying behavior graph scheduling comprises a horizontal filling strategy, a vertical filling strategy, a center symmetry filling strategy and an S-shaped filling strategy. The structure constraint graph data is subjected to a laying behavior graph scheduling to obtain laying behavior graph data, wherein the laying behavior graph scheduling comprises a horizontal filling strategy, a vertical filling strategy, a center symmetry filling strategy and an S-shaped filling strategy.

4. The method of claim 1, wherein, The structure constraint graph data is subjected to a local graph division to obtain local graph data. The local graph data is subjected to a perturbation graph selection to obtain perturbation graph data. The perturbation space data is subjected to a perturbation processing to obtain perturbation result data. The perturbation result data is subjected to a legality verification to obtain random simulation data. The step S4 specifically comprises: The three-dimensional layout projection data is subjected to a sheet unfolding to obtain sheet unfolding data.

5. The method of claim 1, wherein, The sheet unfolding data is subjected to a drawing element generation to obtain drawing element data. The drawing element data is subjected to a drawing format export to obtain solar wing cell matching data. The solar wing cell matching system for performing the three-dimensional layout-based solar wing cell matching method as claimed in claim 1 comprises: A three-dimensional substrate modeling module is configured to obtain solar wing three-dimensional data and construct a three-dimensional substrate model based on the solar wing three-dimensional data to obtain a solar wing three-dimensional model. A layout simulation generation module is configured to perform a structure-driven spreading strategy simulation on the solar wing three-dimensional model to obtain fixed simulation data and perform a perturbation layout scheme generation on the solar wing three-dimensional model to obtain random simulation data.

6. A three-dimensional layout based solar wing cell sizing system, characterized in that, A structure region analysis and optimal sorting module is configured to perform a region structure domain analysis on the fixed simulation data and the random simulation data to obtain region structure domain data and perform an optimal layout sorting on the region structure domain data to obtain layout sorting data. A drawing generation and matching output module is configured to perform a two-dimensional drawing export on the layout sorting data to obtain solar wing cell matching data. ​ ​ ​

Images