BIM-based photovoltaic construction comprehensive management method and system
By constructing a three-dimensional voxel mesh system and a constraint propagation correction process, the computational complexity and electrical path planning problems at photovoltaic construction sites were solved, enabling real-time deviation detection and standardized path generation on mobile terminals.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHONGCHENG ENERGY GRP CO LTD
- Filing Date
- 2026-03-31
- Publication Date
- 2026-05-29
Smart Images
Figure CN121936086B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent management, and in particular to a BIM-based integrated management method and system for photovoltaic construction. Background Technology
[0002] As a core component of clean energy, photovoltaic (PV) power generation has seen rapid growth in installed capacity in recent years, with an increasing number of large-scale ground-mounted PV power plants and distributed rooftop PV systems. Consequently, the complexity of construction management has also increased significantly. Against this backdrop, Building Information Modeling (BIM) technology, with its powerful 3D visualization capabilities and full lifecycle information management capabilities, has been widely adopted in the design and construction management of PV power plants. However, in the actual construction process of PV power plants, spatial deviations inevitably exist between the ideal state described by the BIM design model and the actual state of the construction site. These deviations originate from various sources: differential settlement due to uneven foundation bearing capacity can cause the support pile foundation to deviate from the design elevation; inherent limitations in the operational precision of construction personnel can lead to deviations in component installation positions from the design coordinates; elastic deformation and creep displacement of metal supports and PV modules under temperature fluctuations and wind loads can occur; and differences exist between the microscopic undulations of the construction site terrain and the geological exploration report.
[0003] There is an urgent need for an automated photovoltaic construction management method that can unify deviation detection, cascade adjustment of related components, electrical path replanning, shading verification, and cost control within the same technical framework. This method should have deterministic computational logic to ensure project safety, while having sufficiently low computational complexity to support real-time operation on mobile terminals at the construction site. Summary of the Invention
[0004] One of the objectives of this invention is to provide a BIM-based integrated management method and system for photovoltaic construction, in order to solve the problems of high computational complexity, uncontrollable correction results, and electrical path planning that do not conform to engineering specifications in the existing technology for photovoltaic construction sites.
[0005] This invention is achieved through the following technical solution: a BIM-based integrated management method for photovoltaic construction, comprising the following steps: acquiring initial building information model data of the photovoltaic construction site and measured point cloud data collected on site; spatially discretizing the initial building information model data to construct a three-dimensional voxel mesh system, wherein each voxel unit in the three-dimensional voxel mesh system is marked as one of entity attributes, spatial attributes, or connectivity attributes; mapping the measured point cloud data to the three-dimensional voxel mesh system, identifying abnormal voxel units with spatial positional deviations between the design position and the actual position through Boolean difference operations, and generating a deviation vector set; based on the deviation vector set, initiating a constraint propagation correction process, starting from the abnormal voxel unit, propagating adjustment signals layer by layer to surrounding voxel units according to preset topological adjacency rules until a preset geometric convergence condition is met, generating a corrected local voxel layout; in the corrected local voxel layout, based on the voxel units marked as connectivity attributes, recalculating the electrical connection paths between photovoltaic modules using deterministic path search logic, and generating and outputting construction instructions containing updated construction coordinates and a bill of materials.
[0006] Furthermore, after outputting the construction instructions containing the updated construction coordinates and bill of materials, the process may further include: comparing the output bill of materials with the preset bill of quantities item by item, calculating the difference in the quantity of work caused by the location correction and the corresponding cost change; determining whether the cost change exceeds a preset cost threshold; if it exceeds the preset cost threshold, triggering an approval lock mechanism, suspending the output of subsequent construction instructions to the site, and generating an anomaly report containing change details requesting manual intervention.
[0007] Furthermore, a three-dimensional voxel mesh system is constructed, including: establishing a global Cartesian coordinate system covering the photovoltaic construction site; calculating the offset of each spatial point in the initial building information model data relative to the origin of the global Cartesian coordinate system, dividing the offset by a preset voxel resolution and rounding down to obtain the three-dimensional integer index of the voxel unit mapped by each spatial point; establishing a voxel state matrix based on the three-dimensional integer index, and assigning a state value to each voxel unit using an incrementing integer code to identify the entity attribute, the spatial attribute, and the connectivity attribute, respectively.
[0008] Further, the step of mapping the measured point cloud data to the three-dimensional voxel mesh system, identifying abnormal voxel units where there is a spatial deviation between the design position and the actual position, and generating a deviation vector set includes: converting the coordinates of each point in the measured point cloud data into a mesh index in the global Cartesian coordinate system, and counting the number of points falling into each voxel unit; when the number of points exceeds a preset density threshold, determining the corresponding voxel unit as a measured occupied voxel; comparing the position of the measured occupied voxel with the position of the design voxel marked as the entity attribute in the initial building information model data one by one; if the position of the measured occupied voxel does not coincide with the position of the design voxel, determining the measured occupied voxel as the abnormal voxel unit, and calculating the spatial vector difference between the geometric centers of the measured occupied voxel and the corresponding design voxel as a deviation vector in the deviation vector set.
[0009] Further, based on the set of deviation vectors, a constraint propagation correction process is initiated, including:
[0010] Extract the target deviation vector from the set of deviation vectors and identify its corresponding source voxel node in the initial building information model data;
[0011] Query the parent and child nodes of the source voxel node, and construct a locally dependent directed graph centered on the source voxel node;
[0012] Determine whether the position of the source voxel node after displacement according to the target deviation vector spatially overlaps with the voxel unit marked as the spatial attribute;
[0013] If no spatial overlap occurs, the displacement signal is transmitted to the child node based on the local dependency directed graph.
[0014] Furthermore, the constraint propagation correction process also includes: if spatial overlap occurs, the source voxel node is displaced along the opposite direction of the target deviation vector by a preset step size, and the position after the displacement attempt is recursively checked to see if it satisfies the non-overlap constraint; if it satisfies the non-overlap constraint, the new position of the source voxel node is locked, and the secondary deviation generated by the displacement is passed as a new deviation vector to the child node in the locally dependent directed graph.
[0015] Furthermore, when transmitting displacement signals to the child nodes, the displacement response of each node in the next iteration is calculated by comprehensively considering the following three factors: the product of the corrected displacement received by the current node from the upstream node and the preset propagation attenuation factor; the topology transfer coefficient determined according to the physical connection type between the current node and the upstream node; and the obstacle avoidance correction amount generated based on local spatial state information when the expected displacement enters the region marked as the spatial attribute. The combined effect of the product of the propagation attenuation factor, the topology transfer coefficient, and the obstacle avoidance correction amount is used as the displacement response amount for the next iteration.
[0016] Furthermore, the geometric convergence condition is determined as follows: the absolute magnitude of the displacement change of all nodes in the locally dependent directed graph between two adjacent iterations is calculated; when the sum of the displacement changes of all nodes is less than a preset convergence threshold, the geometric convergence condition is satisfied, the iteration is terminated, and the corrected local voxel layout is output; the method also includes a deadlock handling mechanism: a recursion depth counter is set to record the number of constraint propagation iterations; when the number of iterations exceeds a preset maximum depth threshold and the geometric convergence condition is still not satisfied, a deadlock state is determined, the automatic correction process is paused, and a manual intervention request signal is output.
[0017] Furthermore, the electrical connection path between photovoltaic modules is recalculated using deterministic path search logic, including: in the corrected local voxel layout, determining the starting voxel corresponding to the output end of the photovoltaic module and the ending voxel corresponding to the input end of the combiner device; in the three-dimensional voxel mesh system, removing all voxel units marked with the entity attribute and the spatial attribute, and retaining only voxel units marked with the connection attribute to form a walkable space; within the walkable space, executing a pathfinding algorithm based on orthogonal constraints, restricting the path extension direction to only along the three orthogonal axes of the coordinate axes, prohibiting diagonal crossings, thereby generating the electrical connection path composed of straight line segments parallel to the coordinate axes and orthogonal bends.
[0018] Furthermore, the orthogonal constraint pathfinding algorithm determines the optimal electrical connection path by calculating the total cost of the path nodes. The calculation process for the total cost is as follows: the basic cost of a single-step movement is added to the state weight penalty of the voxel cell where the current node is located to obtain the actual cumulative cost from the starting point to the current node; the sum of the absolute values of the voxel coordinates of the current node and the voxel coordinates of the termination voxel in each coordinate axis direction component is calculated as the heuristic estimated cost; it is determined whether the movement direction from the parent node to the current node is orthogonal to the movement direction from the grandparent node to the parent node. If so, a deflection penalty cost caused by the change in path direction is applied; the actual cumulative cost, the heuristic estimated cost, and the deflection penalty cost are summed to obtain the total cost.
[0019] Furthermore, a construction instruction containing updated construction coordinates and a bill of materials is generated and output, including: summing the absolute distances of all straight segments in the electrical connection path to obtain the total length of the foundation straight line; calculating the additional cable length required for a circular bend to replace a theoretical right-angle bend at each orthogonal bend in the electrical connection path based on the minimum allowable bending radius of the cable, as a slack length compensation value; adding the total length of the foundation straight line, the slack length compensation values at all orthogonal bends, and the wiring allowance reserved at both ends to obtain the final cable cutting length and incorporating it into the bill of materials.
[0020] Furthermore, the recalculation of the electrical connection path between photovoltaic modules using deterministic path search logic also includes a collision self-checking step based on bounding boxes: on the generated electrical connection path, an axis-aligned bounding box is defined for each path voxel unit according to the actual physical cross-sectional dimensions of the cable tray; it is detected whether any voxel unit marked as the spatial attribute is included among all voxel units covered by the bounding box; if so, a backtracking mechanism is triggered, forcibly setting the weight value of the collided path voxel unit in the path search to infinity, and re-executing the orthogonal constraint pathfinding algorithm until a collision-free electrical connection path is found.
[0021] Furthermore, after generating the corrected local voxel layout, a deterministic shading detection step based on the sun's position is also included: obtaining the sun's direction vector at a specific time node based on the geographical coordinates of the photovoltaic construction site and the current construction date; for each voxel unit marked as a photovoltaic module in the corrected local voxel layout, starting from its voxel center, ray tracing is performed in the three-dimensional voxel mesh system along the opposite direction of the sun's direction vector with a preset voxel resolution step size; in each step, the state of the voxel unit at the current mapping position is queried; if, before reaching the mesh space boundary, the voxel unit along the ray path is marked as the entity attribute or as an insurmountable obstacle in the spatial attribute, then the corresponding photovoltaic module is determined to be in a shading state, and a feedback instruction containing a displacement suggestion is generated.
[0022] Another aspect of the present invention provides a BIM-based photovoltaic construction integrated management system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements any of the BIM-based photovoltaic construction integrated management methods described above.
[0023] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0024] 1. This invention reduces the computational dimensionality by discretizing the voxel space, enabling the system to run in real time on mobile terminals with limited computing power. Under this framework, complex floating-point geometric operations that traditionally require triangular facets, such as face-to-face intersection in collision detection and nearest point search in point cloud registration, are transformed into arithmetic operations based on integer indices and simple state value comparison operations. This dimensionality reduction transformation from continuous spatial operations to discrete index operations significantly reduces the computational complexity of the system, preventing it from growing exponentially with the improvement of model geometric accuracy. As a result, the entire process of deviation detection, collision judgment, and path search can run smoothly on mobile terminals such as tablets at the construction site, realizing real-time on-site construction management.
[0025] 2. This invention achieves automated conversion from noisy measured data to a high-confidence deviation vector field by using point cloud noise filtering based on density thresholds and voxel-level Boolean difference operations. By statistically analyzing the number of points falling into each voxel unit and comparing it with a preset density threshold, and by performing Boolean difference operations on the measured voxel set and the designed voxel set, it efficiently identifies abnormal voxel units with deviations. Furthermore, it generates a set of deviation vectors with clear physical meaning by calculating the difference in the geometric centers of matching voxel pairs. The entire process requires no manual intervention and is fully automated from the input of the original point cloud data to the output of the deviation vector field. Moreover, the direction and magnitude of the deviation vector directly correspond to the spatial offset of the component from the designed position to the actual position, providing clear physical interpretability.
[0026] 3. This invention employs an improved three-dimensional A* search algorithm for electrical connection path planning. Its core improvements lie in two aspects: First, it uses Manhattan distance instead of Euclidean distance as the heuristic function, forcing the path search to expand only along the positive and negative directions of the coordinate axes, prohibiting diagonal crossings. This results in generated paths naturally consisting of straight lines parallel to the coordinate axes and 90-degree bends, conforming to the standard requirements for laying cables horizontally and vertically along cable trays or supports in actual engineering. Second, it introduces a turning penalty term into the total node cost, causing the algorithm to automatically favor path schemes with fewer bends while ensuring path reachability. This reduces construction difficulty and the risk of damage to cable insulation due to bending. Furthermore, a weighted penalty mechanism based on voxel state values is introduced into the path cost function. This means that although reserved voxels (such as maintenance channels) may be physically empty, they are given a high passage cost in the path search, which effectively prevents cable paths from encroaching on functional areas that are required to be reserved by engineering specifications. Compared with the simplified estimation method based on Euclidean straight distance in the prior art, the orthogonal path generated by this invention truly reflects the actual laying trajectory of the cable in three-dimensional space, fundamentally eliminating the systematic underestimation bias of the straight distance estimation method. Attached Figure Description
[0027] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:
[0028] Figure 1 The above is a flowchart of the overall method provided in Embodiment 1 of the present invention.
[0029] Figure 2 This is a diagram illustrating the control effect of the propagation attenuation factor on the influence range provided in Embodiment 1 of the present invention.
[0030] Figure 3 This is a diagram showing the influence of the topology transfer coefficient on the displacement transfer attenuation characteristics provided in Embodiment 1 of the present invention.
[0031] Figure 4 This is a schematic diagram illustrating the difference between normal convergence and deadlock oscillation provided in Embodiment 1 of the present invention.
[0032] Figure 5 This is a schematic diagram of the control effect provided in Embodiment 1 of the present invention.
[0033] Figure 6 This is a schematic diagram showing the comparison of detection efficiency provided in Embodiment 1 of the present invention. Detailed Implementation
[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0035] Example 1
[0036] This embodiment discloses a BIM-based integrated management method for photovoltaic construction. Figure 1 The overall method flowchart of this embodiment is shown. As can be seen from the figure, this embodiment includes the following steps:
[0037] Step 1: Obtain the initial BIM design model data of the photovoltaic construction site and the measured point cloud data collected on site.
[0038] The initial BIM design model data refers to the three-dimensional digital model data containing all design information of the photovoltaic power station, generated by the design unit based on Building Information Modeling (BIM) technology before the official commencement of the photovoltaic construction project. This initial BIM design model data not only includes the three-dimensional geometric shape and spatial location information of various components such as photovoltaic modules, brackets, combiner boxes, and cable trays, but also the topological connection relationships between components (e.g., which components share the same purlin, which components are connected to the same combiner box, etc.), the material property information of the components, and the preset Bill of Materials (BOM).
[0039] Understandably, the initial BIM design model data serves as the benchmark for the entire construction management process. All subsequent deviation detection, location correction, and material calculations are compared and evaluated using this model data as the design target value.
[0040] Measured point cloud data refers to the actual spatial information of a photovoltaic construction site, expressed as a set of three-dimensional coordinate points, collected through on-site measurements using laser scanners (such as ground-based 3D laser scanners, UAV-borne LiDAR, etc.), structured light sensors, or photogrammetry. Each point cloud data point contains its coordinate values in three-dimensional space, used to describe the true physical location of various components and terrain features at the construction site.
[0041] It is understandable that, due to unavoidable factors such as foundation settlement, installation errors, and material deformation at construction sites, there is often a certain degree of spatial deviation between the actual site conditions reflected by the measured point cloud data and the initial BIM design model data. Accurately obtaining the measured point cloud data is a prerequisite for subsequent deviation detection and automated correction.
[0042] In this embodiment, the initial BIM design model data can be obtained from the project design management platform or the BIM collaboration server, and its format can be IFC standard format, Revit native format or other common BIM data exchange formats.
[0043] In this embodiment, the measured point cloud data can be collected in real time by laser scanning equipment deployed at the construction site, or by aerial scanning of the construction area using a drone equipped with a LiDAR module. Before being input into the system, the collected raw point cloud data can undergo preprocessing steps, including but not limited to outlier removal, coordinate system alignment (i.e., aligning the coordinate system of the point cloud data with the coordinate system of the BIM design model to the same global reference frame), and data downsampling.
[0044] Step 2: Spatial discretization is performed on the initial BIM design model data to construct a three-dimensional voxel mesh system; each voxel in the three-dimensional voxel mesh system is labeled as one of the entity attributes, spatial attributes, or connectivity attributes.
[0045] Spatial discretization refers to the process of dividing the continuous three-dimensional space occupied by the photovoltaic construction site into a large number of uniform and regular cubic units according to a preset physical resolution. Each such cubic unit is called a voxel, which is a natural extension of the pixel concept in three-dimensional space to the volume dimension.
[0046] A three-dimensional voxel mesh system refers to a regular array of cubes covering the entire construction area, formed after the aforementioned spatial discretization process. This array constitutes a unified spatial indexing framework. Within this framework, any spatial coordinate point can be uniquely mapped to a specific voxel unit through simple integer operations. This transforms the complex and costly problem of continuous spatial geometric collision detection into an efficient and extremely low-cost problem of discrete index lookup and integer comparison, fundamentally reducing the computational dimensionality and enabling the system to run efficiently on devices with limited computing power, such as mobile devices.
[0047] Entity attributes, spatial attributes, and connectivity attributes are classification labels for the physical occupancy state of voxel units. Entity attributes correspond to the occupied state of photovoltaic brackets or modules; spatial attributes correspond to the reserved state of functional areas such as reserved maintenance passages and safety distances, as well as the blocked state with immovable obstacles; connectivity attributes correspond to the free state where wiring or displacement is possible.
[0048] In this embodiment, to ensure that the three-dimensional voxel mesh system can accurately represent the spatial information of the construction site, the steps for constructing the three-dimensional voxel mesh system specifically include:
[0049] Establish a global Cartesian coordinate system covering the entire construction area. This global Cartesian coordinate system refers to a unified three-dimensional rectangular coordinate reference system established for the entire photovoltaic construction area. Its origin and coordinate axis directions are determined at the initial stage of the project and remain unchanged throughout the entire construction management cycle. The origin of this coordinate system can be set at a corner point or feature control point of the construction site, and the coordinate axis directions are usually aligned with the main boundary directions of the construction site. The purpose of establishing a global Cartesian coordinate system is to provide a unified spatial reference framework for BIM design model data and measured point cloud data, enabling accurate spatial comparison and calculation between the two within the same coordinate system.
[0050] The construction area is divided into continuous cubic units according to a preset physical resolution. The preset physical resolution, or voxel resolution, refers to the side length of each cubic voxel unit. The value of this parameter directly determines the accuracy of spatial discretization—a smaller voxel resolution results in a more accurate spatial description, but also increases the required storage space and computational load; conversely, a larger voxel resolution results in less accurate spatial description. In practical engineering, the value of the voxel resolution needs to be set in a balance between the minimum feature dimensions of the photovoltaic module (e.g., module frame width, bolt hole diameter, etc.) and the system's computing power constraints.
[0051] Specifically, for any spatial coordinate point within the construction site, the 3D integer index of the unique voxel to which the point belongs can be obtained by calculating the offset of the point relative to the origin of the global Cartesian coordinate system, dividing by the voxel resolution, and rounding down. This process achieves a constant-time transformation from continuous coordinates to discrete indices.
[0052] For example, in this embodiment, the three-dimensional integer index of any spatial coordinate point can be calculated using the following formula:
[0053]
[0054] in, As a floor function, it ensures that each spatial coordinate point is uniquely mapped to a specific voxel; The global Cartesian coordinate system origin coordinates preset for the construction site serve as the reference datum for the entire voxel grid coordinate system. The three-dimensional coordinates of the spatial coordinate point to be mapped; Voxel resolution, which is the side length of each cubic voxel unit.
[0055] It is understandable that the aforementioned voxel index calculation method is essentially a spatial quantization operation, quantizing continuous three-dimensional spatial coordinates into discrete integer index triples. Through this quantization operation, all subsequent spatial position determination, collision detection, and neighborhood query operations can be completed through integer index arithmetic operations, avoiding the high computational complexity and floating-point error accumulation problems caused by performing complex floating-point geometric operations in continuous space.
[0056] A voxel state matrix is established, and a state value is assigned to each cube element. The voxel state matrix is a 3D state tensor covering the entire voxel mesh, where each element is an integer representing the physical property category of the corresponding voxel element. This integer state labeling simplifies subsequent collision detection from complex triangular facet intersection operations to simple integer comparison operations, significantly improving the algorithm's efficiency.
[0057] For example, in this embodiment, the state tensor Each element The definition can be shown in the following formula:
[0058]
[0059] in, , , These represent the number of voxels in the three dimensions of length, width, and height of the voxel mesh, which are determined by the spatial extent of the construction area and the voxel resolution. The free state indicates that the space region where the voxel is located is empty and can be freely planned and used. In other words, it is a connection attribute and is a walkable space for subsequent electrical connection path planning. This indicates that the area is occupied by physical components such as photovoltaic brackets and modules, which is also a physical attribute. This indicates a reserved state, meaning that the area is designated as a functional space such as a maintenance passage or safety distance. Although it may be physically empty, it should not be occupied during the planning process. In other words, it is a type of spatial attribute. This indicates a blocked state, meaning that the area contains bedrock or permanent, non-removable obstacles that cannot be traversed or occupied under any circumstances; it is another type of spatial attribute.
[0060] Understandably, the state definitions described above use an incrementing integer encoding method, which allows many judgment logics to be implemented through simple numerical comparisons. For example, to determine whether a voxel is impassable, it is only necessary to check whether its state value is greater than zero; to determine whether a voxel is an absolute obstacle, it is only necessary to check whether its state value is infinity. This design greatly simplifies the implementation logic of subsequent algorithms and effectively reduces computational complexity.
[0061] In this embodiment, voxel resolution The value ranges from 0.05 meters to 0.5 meters. When the minimum feature size of the photovoltaic module is small (e.g., the module spacing is only a few centimeters), Choose the smaller value to ensure spatial description accuracy; when the construction site area is large and computing resources are limited. A larger value is chosen to control the total number of voxels and computational overhead.
[0062] In this embodiment, the three-dimensional geometric information of each component in the initial BIM design model data is read one by one, mapped to the voxel mesh using the aforementioned voxel index calculation method, and the state value of the corresponding voxel is assigned according to the component type. For example, the voxels occupied by photovoltaic modules and supports are marked as solid occupied state (state value 1), the voxels of the maintenance passage area marked in the design drawings are marked as reserved state (state value 2), the voxels occupied by bedrock or non-removable structures marked in the geological exploration data are marked as blocked state (state value infinity), and the remaining voxels not covered by any component or constrained area are kept in free state by default (state value 0).
[0063] Step 3: Map the measured point cloud data to a 3D voxel mesh system, identify anomalous voxel units with spatial positional deviations through Boolean difference operations, and generate a set of deviation vectors.
[0064] Mapping refers to converting each three-dimensional coordinate point in the measured point cloud data from continuous three-dimensional coordinates to a discrete grid index in the global Cartesian coordinate system using the voxel index calculation method defined in step 2, thereby determining the voxel unit to which each point cloud data point belongs.
[0065] Boolean difference operation refers to comparing the actual occupied voxel set determined based on measured point cloud data with the design voxel set marked as occupied in the initial BIM design model one by one to find the parts that do not overlap between the two—that is, voxel units where the design position deviates from the actual position.
[0066] Abnormal voxel units refer to voxel units whose designed positions do not match their actual positions, as identified after the above Boolean difference operation.
[0067] The deviation vector set refers to the ordered set of deviation vectors calculated from all abnormal voxel elements. Each deviation vector describes the spatial offset direction and offset distance of the corresponding component from the design position to the actual position.
[0068] In this embodiment, since the original point cloud data inevitably contains noise (such as flying spots and reflection artifacts in laser scanning), treating all point cloud data as valid data indiscriminately will severely interfere with the deviation calculation results. Therefore, the steps for generating the deviation vector set specifically include:
[0069] The coordinates of each point in the measured point cloud data are converted into a grid index in the global Cartesian coordinate system. This conversion process uses the voxel index calculation method defined in step 2, which calculates the offset of each point coordinate relative to the origin of the global coordinate system, divides it by the voxel resolution, and rounds down to obtain the three-dimensional integer index of the voxel to which the point belongs.
[0070] The number of point clouds falling into each voxel unit is counted. When the number exceeds the preset density threshold, the voxel unit is determined to be a measured occupied voxel.
[0071] Density threshold is a key parameter in noise filtering. Its core idea is that a voxel is only considered occupied by a real physical entity when it has accumulated a sufficient amount of point cloud data. A small number of noise points that accidentally fall into a voxel are automatically filtered out because their number is insufficient to reach the set density threshold. Through this statistical filtering mechanism based on point cloud density, measured occupancy information reflecting the location of real physical components can be reliably extracted from noisy point cloud data.
[0072] For example, in this embodiment, the measured occupancy indicator function of a voxel can be calculated using the following formula:
[0073]
[0074] in, It is a collection of measured point clouds, containing all three-dimensional coordinate points collected by sensors; This is an indicator function; it takes the value 1 when the condition inside the parentheses is true, and 0 otherwise. This is a spatial mapping function, specifically the voxel index calculation method defined in step 2, used to map point cloud coordinates. Mapped to its corresponding voxel index; The density threshold is used for noise filtering. The indicator function outputs 1 only when the number of point clouds falling within the voxel exceeds this threshold, indicating that the voxel is considered to be occupied by a physical entity; otherwise, it outputs 0, indicating that the point cloud within the voxel is considered to be noise.
[0075] Understandable, The introduction of this technology effectively eliminates flying point noise and reflection artifacts in laser scanning, ensuring that the subsequently generated deviation vector is based on real physical entities rather than sensor errors. The specific value needs to be calibrated based on the measurement accuracy and scanning density of the scanning equipment used.
[0076] The locations of the measured occupied voxels are compared with the locations of voxels marked as occupied in the initial BIM design model. After the density filtering described above, two sets of voxels are obtained: one set is the set of design voxels marked as occupied in the initial BIM design model. The other set is the actual occupied voxel set obtained based on cloud computing of measured points. Boolean difference operations involve comparing these two sets one by one to identify components whose designed positions deviate from their actual positions.
[0077] If the two positions do not coincide, the spatial vector difference between their geometric centers is calculated, defined as the deviation vector, and stored in the deviation queue. The direction of the deviation vector indicates the direction of the component's offset, and its magnitude indicates the distance of the offset. This transforms the disordered point cloud differences into an ordered vector field, clearly describing the physical process of the component's offset from its design position to its actual position.
[0078] For example, in this embodiment, for the source voxels in the design model Its deviation vector can be calculated using the following formula:
[0079]
[0080]
[0081] in, This is a function for calculating the coordinates of the geometric center of a voxel, i.e., taking the three-dimensional coordinates of the center point of the voxel cube; These are the source voxels to be inspected in the design model, belonging to the design voxel set. ; The target voxel in the measured data belongs to the actual occupied voxel set. ; This is a local nearest neighbor matching operator used to find the most likely actual position voxel corresponding to each design voxel in the measured voxel set. Constraints ensure that matching only occurs between the design set and the measured set, guaranteeing that the physical meaning of the deviation vector is clear—that is, the vector accurately describes the direction and amount of offset of a specific component from its design position to its actual position.
[0082] A deviation queue is a first-in-first-out (FIFO) data structure used to store all calculated deviation vectors in an orderly manner. Subsequent constraint propagation and correction processes will extract deviation vectors from this queue for processing in sequence.
[0083] Understandably, constructing this deviation vector field essentially reduces the complex 3D point cloud registration problem to a voxel-level discrete correspondence lookup problem. Voxelization significantly reduces computational complexity and naturally smooths out small random deviations—because multiple point cloud points within the same voxel are unified into a single voxel-level position representation. This deviation vector field serves as the direct input for automated constraint propagation and adjustment in subsequent step 4.
[0084] Step 4: Based on the deviation vector set, initiate the constraint propagation correction process: with the abnormal voxel element as the wave source center, propagate the adjustment signal layer by layer to the surrounding voxel elements according to the preset topological adjacency rules until the preset geometric convergence condition is met, thereby generating the corrected local voxel layout.
[0085] Among them, the constraint transmission correction process refers to the process of deterministically transmitting the source deviation along the topological connection relationship between components by constructing a local dependency graph and designing a wavefront propagation algorithm based on the deviation vector set generated in step 3, thereby realizing the automated collaborative adjustment of related components.
[0086] Topological adjacency rules refer to a set of rules that define the physical connection relationships between voxels of components in a photovoltaic array (such as purlin connections, shared supports, etc.). These rules determine which adjacent and related components a deviation signal will propagate from from one component to.
[0087] Geometric convergence condition refers to the mathematical criterion for determining whether the wavefront propagation process has reached a steady state. When the sum of the displacement changes of all affected nodes between two adjacent iterations is less than the preset convergence threshold, the system is considered to have reached a steady state, and the propagation process terminates.
[0088] The corrected local voxel layout refers to the final adjusted set of positions of all affected component voxels within the deviation area after the constraint transmission correction process.
[0089] It is understandable that the components in photovoltaic construction are not independent of each other, but rather form a complex topological connection through structural components such as supports and purlins. A positional deviation in one component will inevitably be transmitted to its adjacent related components through physical connections—the so-called "one small change affects the whole system" problem. This step solves this problem through a deterministic wavefront propagation mechanism, enabling precise and predictable calculation of the adjustment amount for related components without relying on uncontrollable AI prediction models.
[0090] In this embodiment, the constraint propagation correction process can specifically execute the following logic:
[0091] S4.1: Extract a deviation vector from the deviation queue and identify its corresponding source voxel node. The source voxel node refers to the voxel element marked as occupied in the initial BIM design model corresponding to the deviation vector. This node is the starting point of wavefront propagation, and its deviation vector will propagate outward as the initial excitation signal.
[0092] S4.2: Query the parent and child nodes of the source voxel node and construct a locally dependent directed graph. The parent node refers to the component node upstream of the source voxel node in the physical connection hierarchy of the photovoltaic array, such as the main beam voxel of the support structure installed by the source voxel node; the child node refers to the component node downstream of the source voxel node, such as the adjacent photovoltaic module voxel connected to the source voxel node via purlins.
[0093] A locally dependent directed graph is a graph theory structure centered on a source voxel node, containing its parent and child nodes and the directed edges connecting them. In this directed graph, the set of nodes represents the voxel representation of the photovoltaic module, the set of edges represents the physical connection constraints between modules, and the direction of the edges represents the propagation direction of the deviation signal.
[0094] Understandably, the construction of this locally dependent directed graph simulates the force transmission mechanism of rigid or semi-rigid connected structures in the physical world: in a rigidly connected structure, applying displacement to a node will cause the entire structure to move by an equal amount; while in a semi-rigid connection, the displacement transmission will be reduced.
[0095] S4.3: Determine whether the deviation vector causes spatial overlap between the source voxel node and a neighboring voxel marked as blocked. Spatial overlap refers to the situation where, after the source voxel node is displaced according to the direction and distance indicated by the deviation vector, the voxel at its new position is in a blocked state in the state tensor (i.e.,...). If the coordinates of the source voxel node are not found to be in the voxel index, then spatial overlap is determined. This determination process only requires converting the new position coordinates of the source voxel node into a grid index using the voxel index calculation method, and then querying the value of the corresponding element in the state tensor. The computational complexity is constant.
[0096] S4.4: If overlap occurs, the source voxel node is displaced along the opposite direction of the deviation vector at a preset step size, and the position after displacement is recursively checked to ensure that the non-overlap constraint is met. The preset step size is typically equal to the voxel resolution. Or an integer fraction thereof, used to control the minimum adjustment unit for each displacement attempt. The design intent of attempting displacement in the opposite direction of the deviation vector is that when a component encounters an obstacle in the deviation direction and cannot be positioned, the system automatically attempts to push the component away from the obstacle until a legal position is found that does not overlap with any obstructing voxels.
[0097] S4.5: If the displacement is successful, lock the new position of the source voxel node and pass the secondary deviation generated by the displacement to the child nodes in the locally dependent directed graph. This deviation is stored as a new deviation vector in the deviation queue until the queue is empty or the boundary conditions are met. The secondary deviation refers to the difference between the actual displacement of the source voxel node due to obstacle avoidance correction and the original deviation vector. This difference needs to be passed to downstream child nodes through the locally dependent directed graph to ensure that the positional relationships of related components remain consistent.
[0098] In this embodiment, during the wavefront propagation process described above, the transmission of displacement signals from the source node to neighboring nodes needs to take into account three key factors: first, the amount of displacement transmitted from the upstream node; second, the reduction effect of the connection type on displacement transmission; and third, whether there are spatial conflicts caused by obstacles at the target node.
[0099] For example, in this embodiment, when the source node is at the , After the corrected displacement is generated in the second iteration, its neighboring nodes are in the... The displacement response at the next iteration can be calculated using the following formula:
[0100]
[0101] in, The number of iterations for propagation indicates how many layers the signal has propagated from the source node; For nodes In the The correction displacement in the next iteration is a three-dimensional vector that describes the direction and distance of the node's position adjustment in this iteration; This is the topology transfer coefficient matrix, which describes the transfer coefficients from the nodes. To the node The influence of the connection type on displacement transfer.
[0102] It is understandable that when the node With nodes When the connection is rigid (such as welding or bolted connection), the topology transfer factor is 1, which means that the displacement is completely transferred - this corresponds to the equal displacement transfer characteristic of rigid structures in the physical world; when the connection is flexible (such as elastic clips or flexible connectors), the topology transfer factor is less than 1, which means that the displacement is partially absorbed - this corresponds to the reduction characteristic of displacement transfer in semi-rigid structures. The propagation attenuation factor is a constant between 0 and 1, used to control the spatial range of the deviation's influence. With each layer of propagation, the displacement signal is multiplied by this attenuation factor, gradually weakening it and preventing the deviation's influence from spreading infinitely throughout the entire photovoltaic array, ensuring that adjustments are made only within a local area. This is the core efficiency advantage of this algorithm compared to global recalculation methods—limiting the adjustment range to the neighborhood of the deviation source through the attenuation factor avoids recalculating the entire model. This is the obstacle avoidance potential energy function, used to handle spatial conflicts. When a node... When the expected displacement causes the node to enter a region marked as a blocked state in the state tensor, the function generates a reverse corrective thrust to push the node away from the obstacle, thereby ensuring that the adjusted layout does not violate spatial constraints. Represents a node Surrounding local spatial state information, i.e., based on nodes It is the set of state values of each voxel within a certain range around the center. Figure 2 This diagram illustrates the control effect of different propagation attenuation factor values on the influence range in this embodiment. Figure 3 A schematic diagram illustrating the effect of different topological transfer coefficients on displacement transfer attenuation characteristics in this embodiment is shown.
[0103] Understandably, the wavefront propagation equation described above is an iterative process. In each iteration, all affected nodes synchronously update their own displacements based on the displacements of their upstream nodes. This iterative process requires a definite termination condition to ensure the determinism and termination of the algorithm.
[0104] In this embodiment, a convergence criterion is defined to determine the timing of iteration termination: when the sum of the displacement changes of all nodes between two adjacent iterations is less than a preset convergence threshold, the system is considered to have reached a steady state and the propagation process terminates.
[0105] For example, in this embodiment, the convergence condition can be expressed as follows:
[0106]
[0107] in, is the Euclidean norm of the displacement vector, used to calculate the absolute magnitude of the displacement change of each node between two adjacent iterations; It is the set of all nodes in a locally dependent directed graph; The preset convergence threshold is typically much smaller than the voxel resolution. This is to ensure that the accuracy of the convergence results is sufficient to meet engineering requirements. Figure 4 This diagram illustrates the difference between normal convergence and deadlock oscillation under convergence conditions in this embodiment.
[0108] Understandably, when the above convergence condition is met, it means that the displacements of all affected nodes have stabilized, and further iteration will not produce meaningful positional changes; the system has reached a state of mechanical equilibrium. At this point, the final positions of each node output constitute the corrected local voxel layout.
[0109] In this embodiment, to address potential oscillations or deadlocks caused by certain special topologies (e.g., two interconnected nodes repeatedly pushing each other during obstacle avoidance correction), the constraint propagation correction process also includes a deadlock handling mechanism: a recursive depth counter is set to record the number of iterations in constraint propagation. This recursive depth counter is an integer variable that increments by 1 at the start of each iteration to track the number of iterations already performed in the current wavefront propagation. When the number of iterations exceeds a preset maximum depth threshold and the geometric convergence condition is still not met, a deadlock state is determined.
[0110] Among them, the maximum depth threshold This is the pre-defined maximum number of allowed iterations. This parameter ensures that the system will never fall into an infinite loop under any circumstances.
[0111] For example, in this embodiment, the termination state determination of the algorithm can be shown in the following formula:
[0112]
[0113] in, This represents the current number of iterations executed. The preset maximum number of iterations, This indicates successful convergence; This indicates a deadlock interruption; This decision logic, which serves as a marker for human intervention, ensures that the system does not fall into an infinite loop.
[0114] Terminate the current automatic correction process, highlight the boundary of the conflict area that caused the deadlock in the 3D voxel mesh system, and output a manual intervention request signal.
[0115] Understandably, this solution has a clear mathematical judgment mechanism for situations where there is no solution, ensuring the safety of engineering construction. In other words, the system will not output construction instructions when it cannot ensure correctness, but will promptly request experienced engineers to intervene manually, thereby avoiding engineering safety risks that may be caused by incorrect adjustment instructions. The design of highlighting the boundary of the conflict area is intended to help on-site engineers quickly locate the problem area and shorten the time for manual decision-making.
[0116] Step 5: In the corrected local voxel layout, the electrical connection paths between photovoltaic modules are recalculated using deterministic path search logic, and construction instructions including updated construction coordinates and bill of materials are generated and output.
[0117] Among them, the deterministic path search logic refers to the method of searching for the optimal electrical connection path from the output end of the photovoltaic module to the input end of the combiner box based on the improved three-dimensional A* pathfinding algorithm in the three-dimensional voxel mesh system established in step 2, using deterministic mathematical rules. This method does not rely on any randomization or AI prediction model and always produces the same output result for the same input.
[0118] Electrical connection paths refer to the cable laying routes between photovoltaic modules and between modules and combiner boxes. These routes must avoid all voxel cells marked as occupied or blocked in the 3D voxel mesh system, extending only within the walkable space formed by free voxel cells. Construction coordinates refer to the final spatial coordinates of each component voxel after correction, represented by 3D coordinate values in a global Cartesian coordinate system. The bill of materials is a summary list of quantities, including cable cutting lengths, cable types, and connector quantities, generated based on the recalculated electrical connection paths.
[0119] Understandably, at photovoltaic construction sites, cables must be laid horizontally and vertically along cable trays or supports, and cannot traverse space in a straight line as assumed by Euclidean distance. Traditional material estimation methods based on straight-line distance systematically underestimate the actual cable requirements, leading to significant errors between budget and actual consumption. This step employs an orthogonal pathfinding algorithm in voxel grid space to ensure that the planned cable paths naturally conform to the horizontal and vertical engineering wiring specifications.
[0120] In this embodiment, the step of recalculating the electrical connection paths between photovoltaic modules specifically includes:
[0121] The starting voxel (module output) and ending voxel (combiner box input) in the modified local voxel layout are determined. The starting voxel is the voxel cell corresponding to the location of the photovoltaic module's electrical output terminal in the modified local voxel layout; the ending voxel is the voxel cell corresponding to the location of the combiner box's electrical input terminal. These two voxels serve as the start and end points of the path search algorithm, respectively.
[0122] In a 3D voxel mesh system, all voxel cells marked as occupied or blocked are removed, leaving free-state voxel cells to form the walkable space. The walkable space refers to the 3D connected region comprised of all voxel cells with a state value of 0 (free state) in the state tensor. Within this space, electrical connection paths can extend freely. Voxel cells with state values greater than 0 are considered impassable regions, and the path search algorithm will not include these voxels in the candidate path nodes.
[0123] Understandably, by directly linking the definition of the walkable space to the value of the state tensor, the path search algorithm only needs to perform one integer comparison operation (i.e., check) to determine obstacle avoidance. Whether it is true or not, there is no need to perform complex geometric collision calculations, thus ensuring the computational efficiency of the pathfinding process.
[0124] Within the walkable space, an orthogonal pathfinding algorithm is executed, restricting the path extension direction to only along the X, Y, and Z axes of the coordinate system, prohibiting diagonal crossings. The orthogonal pathfinding algorithm is an improved 3D A* search algorithm based on Manhattan distance. Its core improvements lie in two aspects: First, Manhattan distance is used instead of the traditional Euclidean distance as the heuristic function, thus forcing the algorithm to prioritize searching paths parallel to the coordinate axes; second, a turning penalty term is introduced into the path cost function, making the algorithm inclined to choose paths with fewer bends, thereby reducing construction difficulty and protecting the cables.
[0125] The prohibition of diagonal crossings means that during the search process, each expansion step is only allowed along one of the positive or negative directions of the X, Y, and Z coordinate axes. In other words, each move can only reach one of the six orthogonal neighbor voxels sharing a face with the current voxel, and movement to diagonal neighbor voxels sharing edges or vertices is not permitted. This restriction ensures that the generated path consists entirely of straight lines parallel to the coordinate axes and 90-degree bends, fully complying with the horizontal and vertical alignment specifications of engineering wiring.
[0126] For example, in this embodiment, the total cost of a node during the search process can be calculated using the following formula:
[0127]
[0128] in, For nodes Given the total cost, the algorithm prioritizes expanding the node with the lowest total cost. From the starting point to the current node The actual known path cost; To start from the current node Heuristic cost estimation to reach the target endpoint; This refers to the penalty for turning caused by a change in path direction. Represents a node The parent node in the search tree.
[0129] For each component of the total cost function mentioned above, its construction logic needs to be explained separately: for the actual cost Its calculation adopts an additive method, starting from the cost of the parent node, adding the cost of one step of movement and the weight penalty generated by passing through different state voxels.
[0130] For example, in this embodiment, the actual cost It can be represented as follows:
[0131]
[0132] in, The cumulative path cost of the parent node; The basic cost of a single-step movement is typically equal to the voxel resolution. That is, the physical distance of moving one voxel; To be based on the current node The state of the voxel The applied weighted penalty.
[0133] Understandably, by introducing voxel state weights, the algorithm can automatically avoid regions unsuitable for wiring. For example, when a voxel is in a free state, No additional penalty is imposed; when the voxel is in a reserved state (maintenance channel), Take a very large positive value to effectively prevent the path from crossing protected areas such as maintenance access routes—areas that may be physically empty but are prohibited from being laid with cables according to engineering specifications.
[0134] For heuristic functions In this embodiment, Manhattan distance is used instead of traditional Euclidean distance.
[0135] For example, in this embodiment, the Manhattan heuristic function It can be represented as follows:
[0136]
[0137] in, For the current node voxel coordinates; The voxel coordinates of the target endpoint.
[0138] Understandably, Manhattan distance is calculated as the sum of the absolute values of the components along each coordinate axis. Geometrically, it is equivalent to the shortest distance from the current point to the target point in a grid where movement is only allowed along the coordinate axes. This choice has key engineering implications: it forces the search algorithm to prioritize exploring paths parallel to the coordinate axes, ensuring that the planned cable paths naturally conform to the horizontal and vertical engineering cabling specifications, avoiding non-standard paths that traverse space at an angle.
[0139] Penalty for turning Its design intent is to reduce unnecessary bends in the cable path.
[0140] For example, in this embodiment, the turning penalty item It can be represented as follows:
[0141]
[0142] in, From the parent node to the current node The direction vector of movement; This is the direction vector for movement from the grandparent node to the parent node; This indicates that the two direction vectors are parallel, meaning the path direction has not changed, and no penalty is applied in this case; This indicates that the two direction vectors are orthogonal, meaning the path has a 90-degree bend, at which point a bending constant is applied. As a punishment. Figure 5 The diagram illustrates the effect of the turning penalty on the number of bends and path length in this embodiment.
[0143] Understandably, in actual construction, each bend not only increases construction time and labor costs, but may also damage the cable insulation layer due to insufficient bending radius. The larger the value of , the more the algorithm tends to avoid bends, but this may lead to excessively long detours. Therefore, in actual engineering, a balance needs to be struck between construction costs and cable protection requirements.
[0144] Calculate the total length of the generated orthogonal path, and based on the bending radius constraint of the cable, add relaxation length compensation values to the voxel elements at the path corners to obtain the final cable cutting length.
[0145] The bending radius constraint refers to the minimum bending radius limit determined by the physical characteristics of the cable. At each 90-degree bend in the path, the actual cable will not bend sharply at a right angle, but will form a quarter-circle arc with the minimum bending radius.
[0146] The slack length compensation value refers to the additional cable length required at each bend in the path because the actual cable bends into arcs instead of theoretical right angles.
[0147] For example, in this embodiment, based on the generated path point sequence The final cut length of the cable can be calculated using the following formula:
[0148]
[0149] in, The L1 norm (i.e., Manhattan distance) between two adjacent path points represents the length of the straight path segment; Sum the lengths of all straight segments in the path to obtain the total length of the basic straight lines of the path; The minimum allowable bending radius of a cable is determined by the cable's physical properties and is a key parameter in the cable specification. For the bend correction term, at each 90-degree bend in the path, the arc length is... The length of the corresponding right-angled broken line is (i.e., the projection of the sum of the two radii onto the right-angled direction), the difference between the two This refers to the additional cable length required at each bend; The wiring allowance reserved at both ends of the cable is the redundant length required by the engineering construction specifications for end-cap fabrication and wiring operations.
[0150] Understandably, the above formula directly incorporates two physical factors that must be considered in actual construction—bending radius constraints and wiring allowances—into the mathematical model, thus elevating the quantity calculation results from theoretical estimates to accurate data that can be directly used for material procurement. This is the core advantage of this model compared to traditional straight-line estimation methods based on Euclidean distance.
[0151] In this embodiment, step 5 also includes a collision self-check step for the cable tray to ensure that the generated electrical connection path does not intrude into protected areas such as maintenance passages in the physical space.
[0152] On the generated orthogonal path, a bounding box is defined for each path voxel element. The bounding box is an axis-aligned bounding box defined with the path voxel element as the center, based on the actual physical cross-sectional dimensions of the cable tray. The size of this bounding box is greater than or equal to the physical cross-sectional dimensions of the cable tray, and it characterizes the actual space occupied by the cable tray at that path node.
[0153] Understandably, cables are abstracted in a voxel mesh as a one-dimensional path extending along the centerline of a path voxel. However, in the physical world, cables and their cable trays have a definite cross-sectional width and height. Therefore, simply checking whether the path voxel itself is located in a free state is insufficient—it is also necessary to check whether the spatial extent covered by the physical cross-section of the cable tray conflicts with the protected area. The introduction of bounding boxes is precisely to bridge the gap between the one-dimensional path representation and the three-dimensional physical cross-section.
[0154] The function checks whether the bounding box intersects with any voxel element marked as reserved. Intersection detection specifically determines whether all voxels covered by the bounding box of the path voxel element contain any voxels with a reserved state value. The voxel cells are axis-aligned, and the voxel mesh itself is also an axis-aligned regular structure. Therefore, intersection detection can be efficiently completed by traversing the voxel indices within the bounding box and querying the state tensor one by one. If an intersection is found, a backtracking mechanism is triggered, forcibly setting the weight value of the current path voxel cell to infinity, and re-executing the orthogonal pathfinding algorithm until a path that does not intersect with the preserved voxel is found.
[0155] The backtracking mechanism refers to setting the weight of the colliding voxel in the cost graph of the path search to infinity when a collision is detected between the path and a preserved voxel (even if it is equivalent to an impassable region in terms of cost), and then restarting the orthogonal pathfinding algorithm. Since the colliding voxel has been assigned an infinite weight, the algorithm will automatically bypass the voxel during the research, thus finding an alternative path that does not intersect with any preserved voxels.
[0156] Understandably, this collision self-checking and backtracking mechanism forms a deterministic self-healing closed loop: each time a collision is detected, the collision is eliminated by modifying the cost graph and searching again. Since each backtracking will mark at least one voxel as impassable, and the total number of voxels is finite, the process must terminate within a finite number of steps, either by finding a collision-free path or by proving that there is no feasible path under the current constraints, at which point the system should output an unreachable report.
[0157] Step 6: Deterministic Detection of Shading Based on Solar Elevation Angle. This step involves using a deterministic ray-stepping method within a 3D voxel mesh system to verify whether the corrected photovoltaic module layout is affected by shading from surrounding obstacles (such as buildings, tall supports, equipment enclosures, etc.) after component position adjustments and electrical connection path planning.
[0158] Understandably, in photovoltaic power plants, the power generation efficiency of a module directly depends on the amount of solar irradiance it receives. If the module is shaded by surrounding obstacles, its power generation will decrease significantly. Therefore, after adjusting the module's position, it is necessary to verify whether the new layout has any shading issues at critical time points (such as the winter solstice when the solar altitude angle is lowest). This step fully utilizes the fast indexing properties of the voxel mesh and state tensor established in step 2, employing a simple and efficient ray-stepping method to complete this verification.
[0159] In this embodiment, the deterministic detection step of shadow occlusion specifically includes:
[0160] Obtain the latitude and longitude of the construction site and the solar trajectory data for the current construction date. Latitude and longitude are the geographical coordinates of the photovoltaic construction site on the Earth's surface, serving as the fundamental input parameters for calculating the sun's position. Solar trajectory data refers to the azimuth and altitude angles of the sun at various times of the day, calculated using astronomical formulas based on the construction site's latitude and longitude and the specified date.
[0161] Understandably, the acquisition of solar trajectory data can be based on mature solar position algorithms (such as the SPA algorithm, PSA algorithm, etc.). These algorithms take the latitude and longitude, altitude, date and time of the construction site as input and output accurate solar azimuth and elevation angles. In shadow occlusion detection, it is usually necessary to select the time node with the most unfavorable occlusion (such as the winter solstice when the solar elevation angle is the lowest) for detection to ensure that the detection results have the greatest conservatism throughout the year.
[0162] For each voxel element labeled as a photovoltaic module, its ray projection vector relative to the sun's position is calculated. The ray projection vector is a three-dimensional direction vector pointing from the center of the photovoltaic module voxel towards the sun's position. This vector is calculated from the azimuth and elevation angles in the sun's trajectory data and describes the incident direction of sunlight reaching the module surface at a specific moment.
[0163] Ray stepping tracing is performed in the 3D voxel mesh system along the opposite direction of the ray projection vector. Ray stepping tracing refers to the detection process that starts from the center of the voxel to be detected and moves outwards voxel by voxel along the direction of the sunlight ray vector, with a step size equal to the voxel resolution. In each step, the state tensor is queried using the voxel index to determine the voxel state at the current position.
[0164] For example, in this embodiment, given time Sun direction vector For any voxel center of the component to be detected The ray stepping equation can be expressed as follows:
[0165]
[0166] in, Let be the coordinates of the voxel center of the component to be detected; This is the step factor, with a value of [value missing]. , indicating the number of steps advanced along the ray direction; In time The solar direction vector calculated by the solar position algorithm at any given moment; is the Euclidean norm of the solar direction vector, used to normalize the direction vector into a unit vector; The voxel resolution, i.e., the voxel side length defined in step 2, is used here as the step size for ray stepping; The maximum number of steps represents the upper limit of the number of steps required for the ray to reach the spatial boundary of the site (i.e., the sky bounding box). Figure 6 A schematic diagram comparing the detection efficiency of ray stepping tracking in this embodiment with that of the traditional three-dimensional projection method is shown.
[0167] Understandably, since voxel index lookup itself is a constant-time operation, the entire ray-stepping process is highly efficient. In each step, only the voxel index query needs to be performed... The coordinates are converted into grid indices using the voxel index calculation method in step 2, and then the value of the corresponding element in the state tensor can be queried to complete the judgment.
[0168] If the ray hits any voxel marked as occupied or blocked before reaching the sky bounding box, the current photovoltaic module voxel is determined to be in a shaded state, and a shift suggestion command is generated. The sky bounding box refers to the spatial boundary of the three-dimensional voxel grid system. When the ray steps beyond this boundary, it indicates that the ray has reached the open sky area without encountering any obstacles, and the module is not shaded.
[0169] Exemplarily, in this embodiment, the component voxel center The occlusion detection result can be determined by the following formula:
[0170]
[0171]
[0172] in, This is a Boolean value, which is True if and only if any occupied or blocked voxels exist along the ray path, indicating that the component is occluded; Indicates the ray at the 1st Steps to reach the location The state value of the corresponding voxel; This is an existential quantifier, indicating that if any step in the entire stepping process satisfies the condition, it is considered an occlusion; The logical OR operator indicates that if a ray encounters an entity in either an occupied state (such as a bracket or component) or a blocked state (such as a building), it is considered an occlusion.
[0173] The relocation suggestion instruction is an adjustment instruction automatically generated by the system when a shading problem is detected in a photovoltaic module, suggesting that the module be moved to an unshaded position. This instruction can be fed back as a new deviation vector to the constraint propagation correction process in step 4, triggering a new round of position adjustment iterations until all modules meet the unshading constraint or the conditions for manual intervention are triggered.
[0174] Understandably, the aforementioned shading detection method fully leverages the regular structure of voxel meshes and the fast indexing characteristics of state tensors, simplifying traditional complex 3D geometric projection calculations into a voxel-by-voxel state query operation along the ray direction. This enables the system to complete shading detection of a large number of components at multiple critical time points (such as each hour of the winter solstice) in a very short time, thereby ensuring that the adjusted component layout will not lead to a decrease in power generation efficiency due to shading issues.
[0175] Step 7: Based on the updated construction coordinates and cable lengths, calculate the difference in project quantity changes in real time and execute cost control logic.
[0176] The "Quantity Change Difference" refers to the difference between the updated Bill of Materials (BOM) generated in step 5 and the preset BOM in the initial BIM design model data. This difference reflects changes in material requirements caused by deviations and location corrections at the construction site.
[0177] In this embodiment, the initial BIM design model data obtained in step 1 includes a preset bill of quantities, and the method further includes:
[0178] Based on the updated construction coordinates and cable lengths generated in step 5, the difference in quantities is calculated in real time. Real-time calculation means that whenever step 5 outputs an updated bill of materials, the system automatically compares it item by item with the pre-set bill of quantities in the initial BIM design model, calculating the changes in each material (such as cable length, number of connectors, number of cable tray sections, etc.) and the corresponding cost changes.
[0179] Understandably, real-time calculation of quantity change differences provides project managers with a dynamic cost monitoring tool. By immediately feeding back the material changes caused by each deviation correction to management, potential cost overruns can be identified early on, avoiding the passive situation of discovering material shortages or severe budget overruns late in the construction process.
[0180] If the change difference exceeds a preset cost threshold, an approval lockout mechanism is triggered, suspending the output of construction instructions and generating an anomaly report. The preset cost threshold is the maximum permissible percentage or absolute amount of change in work quantity pre-set by project management personnel according to engineering budget management regulations. When the cost change corresponding to the change in work quantity exceeds this threshold, it indicates that the cost impact of the deviation correction has exceeded the scope of automatic decision-making authority.
[0181] The approval lock mechanism means that when cost changes exceed a threshold, the system automatically locks the output channel of construction instructions, suspending the sending of any new construction instructions to the construction site, and simultaneously generating a detailed anomaly report. This report includes the specific details of the changes (such as which components have shifted, how much the cable length has increased, how much the total cost has increased, etc.) and is sent to the project manager or cost engineer for approval. Only after approval will the system unlock and resume the output of construction instructions.
[0182] Understandably, the approval lock-in mechanism is a safety valve design—it ensures that within a controllable cost impact range, the system can automatically complete the entire process of deviation detection, position correction, and construction instruction generation; while when the cost impact exceeds the controllable range, the system will promptly pause and request manual decision-making to avoid significant economic losses caused by the blind execution of automated processes.
[0183] Example 2
[0184] This embodiment discloses a photovoltaic construction integrated management system based on BIM technology. Specifically, this system can be integrated into an electronic device, such as a terminal or server. The terminal can be a mobile phone, tablet, smart Bluetooth device, laptop, or personal computer equipped with geographic information software such as ArcGIS or QGIS; the server can be a single server or a server cluster consisting of multiple servers.
[0185] In this embodiment, the photovoltaic construction integrated management system based on BIM technology can also be integrated into multiple electronic devices, such as multiple servers, so that multiple servers can implement the photovoltaic construction integrated management method based on BIM technology in Embodiment 1 of this application.
[0186] In this embodiment, the BIM-based photovoltaic construction integrated management system can also be deployed on edge computing terminals or mobile devices at the construction site to realize real-time data collection, processing, and instant output of construction instructions at the photovoltaic construction site.
[0187] In this embodiment, the server can also be implemented in the form of a terminal.
[0188] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A BIM-based integrated management method for photovoltaic construction, characterized in that, The comprehensive management method includes: Acquire initial building information model data and measured point cloud data collected on-site at the photovoltaic construction site; The initial building information model data is spatially discretized to construct a three-dimensional voxel mesh system, wherein each voxel unit in the three-dimensional voxel mesh system is labeled as one of entity attributes, spatial attributes, or connectivity attributes. The measured point cloud data is mapped onto the three-dimensional voxel mesh system, and abnormal voxel units with spatial position deviations between the design position and the actual position are identified by Boolean difference operation, generating a set of deviation vectors; Based on the set of deviation vectors, the constraint propagation correction process is initiated. Starting from the abnormal voxel unit, the adjustment signal is propagated layer by layer to the surrounding voxel units according to the preset topological adjacency rules until the preset geometric convergence condition is met, and the corrected local voxel layout is generated. In the revised local voxel layout, based on the voxel units marked with the connection attribute, the electrical connection path between photovoltaic modules is recalculated using deterministic path search logic, and a construction instruction containing updated construction coordinates and bill of materials is generated and output. The construction of the three-dimensional voxel mesh system includes: Establish a global Cartesian coordinate system covering the photovoltaic construction site; By calculating the offset of each spatial point in the initial building information model data relative to the origin of the global Cartesian coordinate system, and dividing the offset by the preset voxel resolution and rounding down, the three-dimensional integer index of the voxel unit mapped by each spatial point is obtained. A voxel state matrix is established based on the three-dimensional integer index, and a state value is assigned to each voxel unit using an incrementing integer code to identify the entity attribute, the spatial attribute, and the connection attribute, respectively. The recalculation of electrical connection paths between photovoltaic modules using deterministic path search logic includes: In the revised local voxel layout, the starting voxel corresponding to the output end of the photovoltaic module and the ending voxel corresponding to the input end of the combiner device are determined. In the three-dimensional voxel mesh system, all voxel units marked as entity attributes and spatial attributes are removed, and only voxel units marked as connection attributes are retained to form a walkable space. Within the walkable space, a pathfinding algorithm based on orthogonal constraints is executed, restricting the path extension direction to only along the three orthogonal axes of the coordinate axes and prohibiting diagonal crossings, thereby generating the electrical connection path composed of straight line segments parallel to the coordinate axes and orthogonal bends; The orthogonal constraint-based pathfinding algorithm determines the optimal electrical connection path by calculating the total cost of the path nodes. The calculation process for the total cost is as follows: The basic cost of a single-step movement is added to the state weight penalty of the voxel cell where the current node is located to obtain the actual cumulative cost from the starting point to the current node; The sum of the absolute values of the voxel coordinates of the current node and the voxel coordinates of the termination voxel along each coordinate axis is calculated as the cost of heuristic estimation. Determine whether the movement direction from the parent node to the current node is orthogonal to the movement direction from the grandparent node to the parent node. If so, apply the deflection penalty cost caused by the change in path direction. The total cost is obtained by summing the actual cumulative cost, the heuristic estimated cost, and the bend penalty cost. The method of recalculating the electrical connection paths between photovoltaic modules using deterministic path search logic also includes a bounding box-based collision self-check step: On the generated electrical connection path, an axis-aligned bounding box is defined for each path voxel unit based on the actual physical cross-sectional dimensions of the cable tray; Detect whether any voxel units covered by the bounding box contain voxel units labeled with the spatial attribute; If a collision occurs, a backtracking mechanism is triggered, forcibly setting the weight of the collided path voxel unit in the path search to infinity, and re-executing the orthogonal constraint pathfinding algorithm until a collision-free electrical connection path is found.
2. The BIM-based integrated management method for photovoltaic construction according to claim 1, characterized in that, The step of initiating the constraint propagation correction process based on the set of deviation vectors includes: Extract the target deviation vector from the set of deviation vectors and identify its corresponding source voxel node in the initial building information model data; Query the parent and child nodes of the source voxel node, and construct a locally dependent directed graph centered on the source voxel node; Determine whether the position of the source voxel node after displacement according to the target deviation vector spatially overlaps with the voxel unit marked as the spatial attribute; If no spatial overlap occurs, the displacement signal is transmitted to the child node based on the local dependency directed graph.
3. The BIM-based integrated management method for photovoltaic construction according to claim 2, characterized in that, The constraint propagation correction process also includes: If spatial overlap occurs, the source voxel node is displaced along the opposite direction of the target deviation vector with a preset step size, and the position after the displacement attempt is recursively checked to see if it satisfies the non-overlap constraint. If the non-overlapping constraint is satisfied, the new position of the source voxel node is locked, and the secondary deviation generated by the displacement is passed as a new deviation vector to the child node in the local dependent directed graph.
4. The BIM-based integrated management method for photovoltaic construction according to claim 2, characterized in that, When transmitting displacement signals to the child nodes, the displacement response of each node in the next iteration is calculated by combining the following three factors: The product of the corrected displacement received by the current node from the upstream node and the preset propagation attenuation factor; The topology transfer coefficient is determined based on the physical connection type between the current node and the upstream node; as well as When the expected displacement enters the region marked with the spatial attribute, an obstacle avoidance correction amount is generated based on local spatial state information to push away the obstacle. The combined effect of the product of the propagation attenuation factor, the topology transfer coefficient, and the obstacle avoidance correction is used as the displacement response for the next iteration.
5. The BIM-based integrated management method for photovoltaic construction according to claim 1, characterized in that, After generating the corrected local voxel layout, the method further includes a deterministic detection step of shadow occlusion based on the sun's position: Based on the geographical coordinates of the photovoltaic construction site and the current construction date, obtain the solar direction vector at a specific time point; For each voxel unit marked as a photovoltaic module in the corrected local voxel layout, ray stepping is performed in the three-dimensional voxel mesh system from its voxel center along the opposite direction of the solar direction vector with a preset voxel resolution step size. At each step, query the state of the voxel unit at the current mapping position; If the voxel unit along the ray path is marked as the entity attribute or as an insurmountable obstacle in the spatial attribute before reaching the grid space boundary, the corresponding photovoltaic module is determined to be in a shaded state, and a feedback instruction containing a displacement suggestion is generated.
6. A BIM-based integrated photovoltaic construction management system, characterized in that, The integrated management system includes: processor; The memory stores a computer program, which, when executed by a processor, implements the BIM-based photovoltaic construction integrated management method as described in any one of claims 1 to 5.