BIM-based substation steel structure life cycle carbon reduction optimization design system

Abstract

The application discloses a BIM-based substation steel structure full-life-cycle carbon reduction optimization design system and relates to the technical field of substation steel structure design. The system comprises a carbon pre-diagnosis module, which acquires a BIM initial model of a substation steel structure and performs carbon emission prediction, and activates a model fracture recombination process when the predicted carbon emission exceeds a preset threshold; a fracture disassembly module, which disassembles the BIM initial model into multiple independent component units according to steel structure connecting nodes, and records geometric parameters and material properties; a graph network clustering module, which calculates carbon transmission dependency values among components through a graph neural network, and generates candidate component combinations according to the clustering of the independent component units; a topological reconstruction module, which performs topological reconstruction on each candidate component combination to generate an optimized component submodel; and a carbon fusion module, which performs carbon correlation fusion on all the optimized component submodels to generate a carbon reduction optimized BIM model. The system realizes automatic carbon reduction design of the full-life-cycle of a substation steel structure.

Description

Technical Field

[0001] This invention relates to the field of substation steel structure design technology, specifically a BIM-based full life-cycle carbon reduction optimization design system for substation steel structures. Background Technology

[0002] Lifecycle-wide carbon emission control of substation steel structures is a crucial issue in green building. Current designs commonly employ BIM models to extract component volume and material information, then estimate overall carbon emissions based on carbon emission factor databases. When predicted carbon emissions exceed design specifications, engineers rely on experience to manually adjust component cross-sectional dimensions, change steel grades, or alter anti-corrosion coating types, thereby recalculating carbon emissions. This approach remains at the level of material replacement and sizing, lacking a deep understanding of the carbon emission impact of the steel structure connection system itself.

[0003] Existing technical solutions have significant drawbacks. First, they fail to identify the interrelationships in carbon emissions between components due to spatial proximity and force transmission paths, treating the carbon reduction potential of each component in isolation. This results in limited overall carbon reduction effects and easily leads to situations where repeated modifications to individual components do not prevent overall carbon emissions from exceeding limits. Second, the optimization process does not involve the structural reorganization of connection nodes. The welds, bolts, and additional node plates in the node area not only generate additional material usage and processing energy consumption, but an excessive number of nodes also amplifies the implicit carbon emissions of the entire steel structure. Simply adjusting the component parameters while maintaining the original connection topology cannot fundamentally reduce the carbon emission load brought by the nodes.

[0004] The problem addressed by this invention is how to automatically establish a quantitative description of carbon transfer dependence between components during the BIM design process of substation steel structures. Based on this description, the originally fixed overall structure is decomposed and re-clustered according to carbon dependence to form candidate combinations with carbon reduction potential. Furthermore, the invention addresses how to generate optimized sub-models with significantly reduced number of connecting nodes by rearranging the topology and simplifying the nodes of these combinations, and finally integrate them into an overall carbon reduction model to achieve a systematic reduction in carbon emissions throughout the entire life cycle. Summary of the Invention

[0005] The purpose of this invention is to provide a BIM-based full life-cycle carbon reduction optimization design system for substation steel structures, in order to solve the problems in the existing technology that cannot perform structural fracture and reorganization based on the carbon dependence relationship between components, and cannot reduce carbon emissions of connection nodes through topology reconstruction system.

[0006] The objective of this invention can be achieved through the following technical solutions: This invention provides a BIM-based full life-cycle carbon reduction optimization design system for substation steel structures. The system includes a carbon pre-diagnosis module, a fracture analysis module, a graph network clustering module, a topology reconstruction module, and a carbon fusion module. The carbon pre-diagnosis module acquires the initial BIM model of the substation steel structure and performs carbon emission prediction based on the initial BIM model. When the predicted carbon emissions exceed a preset threshold, the model fracture reconstruction process is activated, thereby initiating proactive intervention in high-carbon emission design schemes and avoiding subsequent detailed design under unfavorable conditions.

[0007] In the model fracture and reconstruction process, the fracture decomposition module decomposes the initial BIM model into multiple independent component units according to the steel structure connection nodes, and records the geometric parameters and material properties of each independent component unit. By performing component-level decomposition on the overall model, the basic information of each structural unit can be obtained, providing data support for subsequent refined carbon transport analysis and component reconstruction, and enabling the optimization process to be applied to the smallest design unit.

[0008] The graph clustering module calculates the carbon transport dependency between components based on the geometric parameters and material properties of each independent component unit using a graph neural network. It then clusters multiple independent component units based on this carbon transport dependency, generating multiple candidate component assemblies. By calculating the carbon emission impact weights between components using a graph network approach, the clustering results reflect the correlation strength between components at the carbon emission level. The resulting candidate component assemblies are more conducive to achieving significant carbon emission reduction effects when adjusting local structures.

[0009] The topology reconstruction module performs topology reconstruction on each candidate component assembly, generating multiple optimized component sub-models. By reorganizing the components at the topological level within the assembly, the number of unnecessary connection nodes is reduced, resulting in more compact and material-efficient component geometric entities. This reduces material usage and the resulting carbon emissions while maintaining mechanical functionality.

[0010] The carbon fusion module performs carbon correlation fusion on all optimized component sub-models to generate a carbon-reduced optimized BIM model. The sub-models are fused and assembled through docking surface matching detection, and the carbon emission reduction coefficient of each node is attached as a label to the fused component. The final carbon-reduced optimized BIM model maintains the original spatial layout, reduces the overall number of connection nodes, lowers the carbon emission intensity of materials, and significantly improves the traceability of the model.

[0011] As a further preferred embodiment of the present invention, the carbon pre-diagnosis module extracts the spatial layout data and material list of the steel structure from the substation construction drawings, imports them into BIM software to generate an initial BIM model including component geometry, steel grade, connection method, and anti-corrosion coating type; then, it extracts the volume and surface data of each component from the initial BIM model, obtains the carbon emission factor per unit mass based on the steel grade, and obtains the coating carbon emission factor based on the anti-corrosion coating type. The volume data is multiplied by the steel density to obtain the component mass, and then multiplied by the carbon emission factor per unit mass to obtain the component substrate carbon emission. Simultaneously, the surface data is multiplied by the coating carbon emission factor to obtain the component coating carbon emission. These are then summed to obtain the carbon emission prediction result of the initial BIM model. This method comprehensively considers the carbon emission contributions of the steel structure substrate and the anti-corrosion coating, resulting in a more accurate carbon emission prediction result that truly reflects the carbon intensity during the substation steel structure design phase.

[0012] As a further preferred embodiment of the present invention, when the carbon pre-diagnosis module compares the predicted carbon emissions with a preset threshold, the preset carbon emission threshold is obtained by multiplying the average carbon emissions of steel structures of similar substations in the design specifications by a safety factor. When the predicted carbon emissions exceed the threshold, an activation signal is sent to the control interface of the BIM software to trigger the model fragmentation and reorganization process; when the predicted carbon emissions are less than or equal to the threshold, the carbon reduction optimization process is terminated, and the initial BIM model is output as the final design model. This setting ensures that the optimization system intervenes only when carbon emissions exceed the limit, avoiding invalid modifications to schemes that already meet low-carbon design requirements.

[0013] As a further preferred embodiment of the present invention, when disassembling the initial BIM model, the fracture disassembly module traverses the topological relationships of all steel structure connection nodes in the model, identifies the beam, column, and support identifiers associated with each connection node, extracts the corresponding component geometric entities, separates them from the connection nodes to form independent component units, assigns a unique number to each independent component unit, and records its original spatial coordinates and connection node index table in the initial BIM model. The resulting independent component units fully retain the geometric and topological mapping relationship with the original model, which is beneficial for subsequent component reassembly and final model reconstruction.

[0014] Furthermore, when recording geometric parameters and material properties, the fracture analysis module extracts cross-sectional shape, cross-sectional dimensions, length data, and volume data from the geometric entity of each independent component unit, summarizing them as geometric parameters. It also queries the attribute database of the initial BIM model for the steel grade, yield strength, anti-corrosion coating type, and fireproof coating type corresponding to each independent component unit, summarizing them as material properties. A mapping relationship is established between the unique identifier and the geometric parameters and material properties, and stored in the extended data table of the BIM model. Through attribute mapping and extended storage, the physical and mechanical characteristics of different component units are recorded in a standardized manner, providing structured input for the feature encoding of the graph neural network.

[0015] As a further preferred embodiment of the present invention, when constructing graph data, the graph clustering module uses each independent component unit as a node in the graph neural network and the connection relationship between adjacent independent component units as edges; it encodes geometric parameters and material properties into feature vectors and assigns them to the corresponding nodes; it runs the message passing mechanism of the graph neural network, enabling each node to aggregate the feature vectors of its neighboring nodes, and calculates the carbon transfer dependency value between the node and its neighboring nodes based on the aggregation result. This carbon transfer dependency value reflects the influence weight of the carbon emissions of a component unit on the carbon emissions of its neighboring component units, making the carbon dependency relationship between components quantifiable into a comparable value, thus providing a basis for clustering.

[0016] Furthermore, during clustering, the graph clustering module uses each independent component unit as the initial cluster center and calculates its carbon transfer dependency value with all neighboring independent component units. This carbon transfer dependency value is compared with a preset clustering threshold. If the carbon transfer dependency value is greater than the threshold, the neighboring independent component unit is included in the candidate component assembly containing the current cluster center. After the clustering assignment of all independent component units is completed, the independent component units in each candidate component assembly are merged, and the total predicted carbon emission value of the candidate component assembly is calculated and used as the clustering label for that assembly. The clustering process driven by carbon transfer dependency values ​​can automatically identify component groups with close carbon emission correlations, forming candidate assemblies with high internal dependencies, thus defining an effective range for local topology reconstruction. The preset clustering threshold is preferably dynamically calculated based on the sum of the mean and standard deviation of all carbon transfer dependency values, thereby adapting to the actual distribution characteristics of carbon transfer dependencies under different design schemes and ensuring the adaptability of the clustering threshold and the rationality of the clustering results.

[0017] As a further preferred embodiment of the present invention, the topology reconstruction module extracts the spatial coordinates and connection node index table of all independent component units from the candidate component assembly, constructs an initial topology matrix of the candidate component assembly, performs singular value decomposition on the initial topology matrix, extracts the principal singular vector, and rearranges the connection order of the independent component units in the component assembly according to the principal singular vector; and recombines the independent component units into new component geometric entities according to the rearranged connection order, forming an optimized component sub-model, and records the reduction in the number of connection nodes of the optimized component sub-model. By obtaining the principal direction of the topology structure through singular value decomposition and re-optimizing the component connection sequence, the number of unnecessary connection components and nodes can be effectively reduced, thereby reducing the additional material consumption and carbon emissions caused by connection construction. Preferably, the number of singular values ​​retained during the singular value decomposition process is determined by the square root of the number of independent component units in the candidate component assembly, thus achieving a balance between retaining topological feature information and suppressing redundant dimensions, ensuring the stability of the reconstruction result.

[0018] As a further preferred embodiment of the present invention, when the carbon fusion module fuses and optimizes the component sub-models, it obtains the reduction value of the number of connection nodes of each optimized component sub-model and calculates the node carbon reduction coefficient of each optimized component sub-model based on this value; it performs matching detection on the docking surfaces between adjacent optimized component sub-models, and when the geometry and size of two docking surfaces are consistent, the two optimized component sub-models are fused through sharing the docking surface to form a fused component group; all fused component groups are reassembled with the optimized component sub-models that did not participate in the fusion according to the spatial coordinates of the original BIM model to generate a carbon reduction optimized BIM model, and a node carbon reduction coefficient label for each fused component group is attached to the carbon reduction optimized BIM model. The node carbon reduction coefficient label intuitively quantifies the contribution of each fused component to the overall carbon reduction, which is beneficial for designers to quickly identify key carbon reduction parts and conduct targeted acceptance.

[0019] As a further preferred embodiment of the present invention, after generating the carbon reduction optimized BIM model, the process further includes an iterative verification process for the carbon reduction optimized BIM model. The carbon reduction optimized BIM model is used as a new initial BIM model, and carbon emission prediction is re-executed to obtain a secondary carbon emission prediction result. The difference between the secondary carbon emission prediction result and a preset carbon emission threshold is calculated. When the difference is less than a preset convergence tolerance, the iterative process is terminated, and the current carbon reduction optimized BIM model is output as the final design model. When the difference is greater than or equal to the preset convergence tolerance, the model fragmentation and reconstruction process is reactivated, and a new round of fragmentation, reconstruction, and fusion operations are performed on the current carbon reduction optimized BIM model until the difference is less than the preset convergence tolerance or the preset maximum number of iterations is reached. Through iterative verification and closed-loop optimization, the system can gradually reduce carbon emissions in multiple rounds of structural reconstruction, ensuring that the final output model stably meets carbon emission design requirements, avoiding the problem of incomplete carbon reduction that may exist in a single optimization, and improving the engineering reliability and low-carbon performance of the design scheme.

[0020] The beneficial effects of this invention are: Carbon transfer dependencies between components are calculated using a graph neural network. Geometric parameters and material properties are encoded as node feature vectors, and a graph data structure is constructed using the connections between components as edges. The message passing mechanism of the graph neural network enables each node to aggregate the features of its neighboring nodes and output quantified dependency weights, representing the degree of mutual influence of carbon emissions from adjacent components. Clustering using these carbon transfer dependencies no longer relies solely on spatial distance or cross-sectional similarity, but instead includes neighboring components with carbon impact weights higher than the dynamic clustering threshold into the same candidate component ensemble. This mechanism breaks the fixed component division of the original design, allowing components closely related at the carbon emission level to be grouped together, forming ensemble units with carbon impact relationships as boundaries. This provides a precise scope for subsequent reorganization aimed at carbon reduction goals, avoiding blind global searches and ensuring that each step of the reorganization operation directly targets the component group that contributes the most to carbon emission reduction.

[0021] Topology reconstruction based on singular value decomposition (SVD) is performed on candidate component assemblies. The initial topology matrix between components within the assembly is extracted and subjected to SVD; the number of retained singular values ​​is determined by the square root of the number of independent component elements within the assembly. The connection order of components is rearranged using the principal singular vectors, thereby generating new component geometric entities with fewer connection nodes. This topology reconstruction does not rely on preset empirical rules but instead discovers connection methods that maintain structural functionality while reducing the number of nodes through matrix feature extraction. Reducing connection nodes directly leads to material omissions and reduced processing energy consumption for gusset plates, welds, bolts, and coatings. Furthermore, the reduction in the number of nodes further reduces carbon emissions during the fabrication and installation phases. Finally, when optimizing the fusion of component sub-models, the node carbon reduction coefficient is calculated based on the node reduction value and added as a label to the model. This ensures that the final carbon reduction optimized BIM model not only provides the overall carbon reduction but also visually indicates the node reduction contribution of each fused component group. Attached Figure Description

[0022] The invention will now be further described with reference to the accompanying drawings.

[0023] Figure 1 This is a flowchart of a BIM-based substation steel structure full life-cycle carbon reduction optimization design system; Figure 2 This is a flowchart of carbon prediagnosis for substation steel structures based on BIM; Figure 3 This is a flowchart of BIM model fracture disassembly and component attribute mapping; Figure 4 This is a flowchart of component clustering based on graph attention networks; Figure 5 It is a flowchart of carbon fusion and iterative carbon reduction optimization. Detailed Implementation

[0024] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] See Figure 1This invention provides a BIM-based full life-cycle carbon reduction optimization design system for substation steel structures, including a carbon pre-diagnosis module, a fracture decomposition module, a graph network clustering module, a topology reconstruction module, and a carbon fusion module. The carbon pre-diagnosis module acquires an initial BIM model of the substation steel structure and performs carbon emission prediction based on the initial BIM model. When the predicted carbon emissions exceed a preset threshold, a model fracture reconstruction process is activated. During the model fracture reconstruction process, the fracture decomposition module decomposes the initial BIM model into multiple independent component units according to the steel structure connection nodes, and records the geometric parameters and material properties of each independent component unit. The graph network clustering module calculates the carbon transfer dependency value between components using a graph neural network based on the geometric parameters and material properties of each independent component unit, and clusters multiple independent component units based on this carbon transfer dependency value to generate multiple candidate component assemblies. The topology reconstruction module performs topology reconstruction on each candidate component assembly to generate multiple optimized component sub-models. The carbon fusion module performs carbon correlation fusion on all optimized component sub-models to generate a carbon reduction optimized BIM model.

[0026] In practice, the BIM-based substation steel structure full life-cycle carbon reduction optimization design system performs a carbon pre-diagnosis process. (See also...) Figure 2 Spatial layout data and a bill of materials for the steel structure were extracted from the substation's construction drawings. The spatial layout data included the component's positioning axes, elevations, and span information, while the bill of materials included component numbers, steel grades, cross-sectional specifications, and anti-corrosion coating types. The spatial layout data and bill of materials were then imported into BIM software. The BIM software generated an initial BIM model based on the imported data. This initial BIM model contained the geometric dimensions, steel grade, connection method, and anti-corrosion coating type for each component. The geometric dimensions were determined by the cross-sectional specifications and length information, and the connection method was identified as welded or bolted connections from the node details in the construction drawings.

[0027] Volume and surface data for each component are extracted from the initial BIM model. In the BIM software, volume data is obtained through volumetric analysis, and surface data is obtained through surface area calculation; the surface data represents the unfolded area of ​​the component's outer surface. Based on the steel grade of each component, a pre-defined carbon emission factor per unit mass database is consulted to obtain the corresponding carbon emission factor. This database stores carbon emission factors for different steel grades at various production stages, sourced from a life cycle assessment database. Based on the anti-corrosion coating type of each component, a pre-defined coating carbon emission factor database is consulted to obtain the corresponding coating carbon emission factor. This database stores carbon emissions per unit area for various anti-corrosion coatings; the coating carbon emission factor value is related to the coating thickness, which is obtained from the bill of materials.

[0028] When calculating the carbon emission prediction results of the initial BIM model, the volume data of each component is multiplied by the steel density to obtain the component mass. The steel density is determined according to the steel grade and ranges from 7850 kg / m³ to 8050 kg / m³. The component mass is then multiplied by the carbon emission factor per unit mass to obtain the carbon emission of the component's substrate. Simultaneously, the surface data of each component is multiplied by the coating carbon emission factor to obtain the carbon emission of the component's coating. The carbon emission of the substrate and the carbon emission of the coating of all components are then summed to obtain the carbon emission prediction results of the initial BIM model. The calculation of the carbon emission prediction results is expressed by the following formula:

[0029] in, The carbon emission prediction results of the initial BIM model are in kilograms of carbon dioxide equivalent. This represents the total number of components in the initial BIM model. For the first Volume data of each component, in cubic meters; For the first The steel density of each component, in kilograms per cubic meter; For the first The carbon emission factor per unit mass of each component is expressed as kilogram carbon dioxide equivalent per kilogram. For the first Surface data of each component, in square meters; For the first The carbon emission factor of the coating corresponding to each component is expressed in kilograms of carbon dioxide equivalent per square meter. In specific implementation, The method for obtaining the carbon emission factor per unit mass is as follows: in the unit mass carbon emission factor database, a matching query is performed using the steel grade as the index, and the database returns the unit mass carbon emission factor value corresponding to the steel grade. The method for obtaining the data is as follows: In the coating carbon emission factor database, a matching query is performed using the anti-corrosion coating type and coating thickness as a joint index, and the database returns the coating carbon emission factor value.

[0030] After obtaining the carbon emission prediction results from the initial BIM model, these results are numerically compared with a preset carbon emission threshold. The preset carbon emission threshold is obtained by multiplying the average carbon emission of steel structures of similar substations in the design specifications by a safety factor. The design specifications adopt the average life-cycle carbon emission of steel structures of similar substations statistically analyzed in the current substation steel structure design standards. The safety factor is set to 1.1, based on the uncertainty of carbon emissions caused by material performance fluctuations and construction process deviations, with a 10% redundancy. When the carbon emission prediction result of the initial BIM model is greater than the preset carbon emission threshold, an activation signal is sent to the control interface of the BIM software. The control interface is the application programming interface provided by the BIM software, and the activation signal is a predefined control command that triggers the start of the model fragmentation and reconstruction process. When the carbon emission prediction result of the initial BIM model is less than or equal to the preset carbon emission threshold, the carbon reduction optimization process is terminated, and the initial BIM model is output as the final design model.

[0031] In specific implementation, please refer to Figure 3 After the model fracture reconstruction process is initiated, the fracture decomposition module decomposes the initial BIM model into multiple independent component units based on the steel structure connection nodes. When traversing the topological relationships of all steel structure connection nodes in the initial BIM model, it obtains the set of connection node objects built into the initial BIM model through the component association query interface provided by the BIM software. For each connection node object in the set, it reads the list of connected component identifiers stored in the connection node object, identifying the component identifier associated with each connection node. Component identifiers include beam identifiers, column identifiers, and brace identifiers. Beam identifiers consist of the prefix character "B" followed by a number, column identifiers consist of the prefix character "C" followed by a number, and brace identifiers consist of the prefix character "BR" followed by a number.

[0032] For each component identifier read, the fracture disassembly module extracts the corresponding component geometry from the component library of the initial BIM model using the component identifier. The extraction process calls the BIM software's data access interface, using the component identifier as the query key, to obtain the 3D boundary representation model data of the component geometry. When separating the component geometry from the connection node, the topology modification interface is called to delete the parent-child relationship between the component geometry and the connection node object, making the component geometry an independent geometric object.

[0033] Each independent geometric object separated from a connection node forms an independent component unit. A unique number is assigned to each independent component unit, in the format "CU-" followed by four sequential digits, incrementing from "CU-0001". The fracture disassembly module records the original spatial coordinates and connection node index table of each independent component unit in the initial BIM model. The original spatial coordinates are obtained by reading the coordinates of the positioning base point and direction vector of the component geometric entity in the world coordinate system of the initial BIM model. The positioning base point coordinates include X-axis, Y-axis, and Z-axis coordinate values, and the direction vector includes X-axis, Y-axis, and Z-axis direction components. The connection node index table stores a list of node identifiers for all connection node objects associated with this independent component unit before separation; each node identifier is a globally unique identifier for the connection node object.

[0034] When recording the geometric parameters and material properties of each independent component unit, geometric parameters are extracted from the component's geometric entity. These geometric parameters include cross-sectional shape, cross-sectional dimensions, length data, and volume data. The cross-sectional shape is obtained by analyzing the cross-sectional profile of the component's geometric entity; the cross-sectional shape can be one of an H-section, box section, circular tube section, or channel section. Cross-sectional dimensions are extracted based on the cross-sectional shape. When the cross-sectional shape is H-shaped, the dimensions include the cross-sectional height, flange width, web thickness, and flange thickness; when the cross-sectional shape is box section, the dimensions include the cross-sectional height, cross-sectional width, and wall thickness; when the cross-sectional shape is circular tube section, the dimensions include the outer diameter and wall thickness. Length data is obtained by calculating the distance between the center points of the two ends of the component's geometric entity. Volume data is obtained by calling the volume calculation function of the BIM software to calculate the component's geometric entity.

[0035] Material properties are retrieved from the attribute database of the initial BIM model. The attribute database stores material information for each component using its component identifier as the primary key. The fracture analysis module uses the original component identifier corresponding to each independent component element before assigning a unique number as the query condition to retrieve the values ​​of fields such as steel grade, yield strength, anti-corrosion coating type, and fireproof coating type from the attribute database. The retrieved steel grade is a string corresponding to the material standard, the yield strength is a numerical value in megapascals (MPa), the anti-corrosion coating type is a string representing the coating system name, and the fireproof coating type is a string representing the fireproof coating model. The extracted cross-sectional shape, cross-sectional dimensions, length data, and volume data for each independent component element are summarized into the geometric parameters of the independent component element, and the retrieved steel grade, yield strength, anti-corrosion coating type, and fireproof coating type are summarized into the material properties of the independent component element.

[0036] The fracture analysis module establishes a mapping relationship between the unique identifier of each independent component element and the summarized geometric parameters and material properties. This mapping relationship uses a key-value pair data structure, where the key is the unique identifier and the value is a data object containing geometric parameter and material property fields. All key-value pairs in this mapping relationship are then batch-written into the extended data table of the BIM model. The extended data table of the BIM model is a custom extended attribute storage area provided by the BIM software, appended to the initial BIM model file in table form, and supports retrieval by unique identifier.

[0037] In specific implementation, please refer to Figure 4 The graph clustering module receives the set of independent component units output by the fracture disassembly module, along with the geometric parameters and material properties of each independent component unit. The module first constructs a graph data structure, using each independent component unit as a node in the graph neural network and the connections between adjacent independent component units as edges. The connections between adjacent independent component units are determined by reading the connection node index table of each independent component unit: when the connection node index tables of one independent component unit and another independent component unit have the same connection node identifier, a connection is established between these two independent component units, and an undirected edge is created between the corresponding nodes of these two independent component units in the graph data structure.

[0038] The graph neural network architecture employs a graph attention network structure, comprising an input feature transformation layer and two stacked graph attention layers. The input feature transformation layer is a linear transformation matrix used to map the original features of each independent component unit to an initial node feature vector of a predetermined dimension. The first graph attention layer receives the initial node feature vector and outputs intermediate node feature vectors. The second graph attention layer receives the intermediate node feature vectors and outputs the final node feature vector, while simultaneously calculating the attention weights between nodes as carbon transfer dependencies.

[0039] The process of encoding the geometric parameters and material properties of each independent component unit into a feature vector is as follows: For each independent component unit, the cross-sectional shape, cross-sectional dimensions, length data, and volume data are extracted from the geometric parameters, and the steel grade, yield strength, anti-corrosion coating type, and fireproof coating type are extracted from the material properties. The cross-sectional shape is converted into a four-bit binary vector using one-hot encoding. The steel grade is mapped to an integer value according to a preset steel grade index table, and the anti-corrosion coating type and fireproof coating type are also mapped to integer values ​​through a preset type index table. The cross-sectional dimensions are composed of numerical vectors from the individual dimension values; for non-numerical cross-sectional shape codes, the numerical vector is not added again. The length and volume data retain their original values. The yield strength retains its original value. After concatenating all numerical data, it is then concatenated with the one-hot encoded cross-sectional shape to form the original feature vector. The original feature vector has a dimension of 13. The input feature transformation layer converts the 13-dimensional original feature vector into a 64-dimensional initial node feature vector through a 13x64 linear transformation matrix.

[0040] In some implementations, the graph neural network takes a graph data structure and initial node feature vectors for all nodes as input, and outputs the carbon transfer dependency value for each edge. The computation process of the first graph attention layer is as follows: for node... Using a learnable weight matrix to assign weights to nodes Initial node feature vectors and neighbor nodes The initial node feature vectors are linearly transformed, the two transformed vectors are concatenated, and a dot product is performed with the learnable attention parameter vector. The LeakyReLU activation function is then applied to obtain the original attention coefficients. Finally, the Softmax function is applied to the nodes... Normalize over all neighboring nodes to obtain the node With nodes The first layer of attention weights is calculated by summing these first layer attention weights with the linearly transformed feature vectors of the neighboring nodes, resulting in the node's weighted sum. The intermediate node feature vectors are used as inputs. The calculation process for the second graph attention layer is the same as the first graph attention layer. The input is the intermediate node feature vectors, while the learnable weight matrix and attention parameter vectors are set independently. The output is the final node feature vectors and node... with neighboring nodes The second layer of attention weights is used directly from the second graph attention layer. The carbon transport dependency value is calculated using the second layer of attention weights, with a value ranging from 0 to 1. The calculation process for the carbon transport dependency value is expressed by the following formula:

[0041] in, Independent component element node With independent component element nodes Carbon transfer dependence between them; The attention parameter vector for the second graph attention layer has a dimension of 128. Its values ​​are randomly initialized and loaded during the initialization of the graph clustering module, and are not updated during training. The weight matrix for the second graph attention layer has a dimension of 64 rows and 128 columns, and is also randomly initialized and fixed. For nodes The intermediate node feature vector obtained after passing through the first graph attention layer has a dimension of 64. For nodes The feature vector of the intermediate node; This represents a vector concatenation operation; Represents a node The set of neighboring nodes; the negative slope parameter of the LeakyReLU activation function is set to 0.2. Attention parameter vector. and weight matrix The Xavier uniform distribution initialization method is used to generate specific values. The parameters of the Xavier uniform distribution are calculated using the sum of the input and output dimensions as the denominator, and the parameters remain unchanged after initialization. This setting is based on the fact that carbon transport dependence values ​​can be determined solely by the geometry and inherent material properties of the component elements and their topological connectivity, without the need for fitting through training data. A graph attention network with fixed parameters can be used to extract the structured dependencies between components.

[0042] After the graph neural network completes its operation, the graph clustering module obtains the carbon transfer dependency value on each edge and enters the clustering phase. In the clustering phase, each independent component unit serves as the initial cluster center, and initially, each independent component unit forms its own candidate component combination. For each independent component unit that serves as a cluster center, all neighboring independent component units in the graph data structure are queried, and the carbon transfer dependency value between this independent component unit and each of its neighboring independent component units is obtained. The carbon transfer dependency value is then compared with a preset clustering threshold. The preset clustering threshold is dynamically calculated based on the sum of the mean and standard deviation of all carbon transfer dependency values. Specifically, the calculation method is as follows: collect the carbon transfer dependency values ​​on all edges in the graph data structure to form a numerical set, and calculate the mean of the numerical set. and standard deviation Clustering threshold Set as . The calculation method is to sum all carbon transport dependency values ​​and then divide by the total number of edges. The calculation method is to use each carbon transport dependency value and The square root of the sum of the squares of the differences.

[0043] When the carbon transport dependency between an independent component and a neighboring independent component is greater than the clustering threshold... When a neighboring independent component unit is included in the candidate component assembly containing the current cluster center, it is considered an independent cluster center. After inclusion, the neighboring independent component unit no longer serves as an independent cluster center, but the candidate component assembly continues to expand outward from the original cluster center. The neighbors of newly included neighboring independent component units are checked, and the comparison and clustering steps are repeated until the carbon transport dependency value of all neighboring independent component units is less than or equal to the clustering threshold. Until then, after the clustering and assignment of all independent component units are completed, the independent component units in each candidate component assembly are merged. Merging means marking multiple independent component units belonging to the same candidate component assembly with the same assembly identifier. When calculating the total carbon emission prediction value of each candidate component assembly, the carbon emission of the component substrate and the carbon emission of the component coating of each independent component unit in the candidate component assembly are obtained. The carbon emission of the component substrate and the carbon emission of the component coating are reused from the values ​​calculated for each component in the carbon prediagnosis module. The carbon emission of the component substrate and the carbon emission of the component coating of all independent component units in the candidate component assembly are added together to obtain the total carbon emission prediction value of the candidate component assembly. This total carbon emission prediction value is used as the clustering label of the candidate component assembly and stored in the data record of the candidate component assembly.

[0044] In practice, the topology reconstruction module receives multiple candidate component assemblies output by the graph clustering module and performs topology reconstruction on each candidate component assembly. For a given candidate component assembly, the topology reconstruction module extracts the spatial coordinates and connection node index table of all independent component units from that assembly. The spatial coordinates are obtained from the original spatial coordinate records of the independent component units, including the coordinates of the positioning base point and the direction vector of each independent component unit. The connection node index table is read from the extended data table records corresponding to each independent component unit.

[0045] When constructing the initial topology matrix of the candidate component assembly, the matrix dimension is the number of independent component units contained in the candidate component assembly. Let the candidate component assembly contain... If there are 1 independent component unit, then the initial topology matrix is... for OK A square matrix of columns. The first column in the initial topological matrix. Line number Column elements Indicates the relationship between the i-th independent component unit and the i-th... The connection relationship between individual component units, where and The range of values ​​is from arrive Integers. When the first... The connection node index table of each independent component unit and the first When the connection node index table of an independent component element contains at least one identical connection node identifier, the element Set the value to This indicates a connection between two independent component elements. When no two connection node identifiers exist in the connection node index table, the element... Set the value to This indicates that there is no connection between two independent component elements. (Matrix diagonal elements) The values ​​are all set to .

[0046] The initial topology matrix has been constructed. Then, the topology reconstruction module refactors the initial topology matrix. Perform singular value decomposition. Singular value decomposition will initialize the topological matrix. It can be decomposed into the product of three matrices, expressed by the following formula:

[0047] in, Let be the initial topological matrix with dimension . OK List; It is a left singular vector matrix with dimension . OK List; It is a singular value diagonal matrix with dimension . OK The columns and diagonal elements form the initial topological matrix. The singular values ​​are arranged in descending order; It is a right singular vector matrix with dimension . OK List; Represents a right singular vector matrix The transpose of .

[0048] The number of singular values ​​retained during singular value decomposition is determined by the number of independent component elements in the candidate component assembly. The square root determines the number of singular values ​​retained. Specifically, it determines the number of singular values ​​retained. Set as ,in Indicates rounding down. Sets the number of singular values ​​to retain. Number of independent component units The basis for rounding down the square root is that in the low-rank approximation of a matrix, the relationship between the number of singular values ​​retained and the matrix size follows the square root scaling rule. This can reduce the matrix dimension while retaining the main connection pattern information in the initial topological matrix, avoiding the introduction of noise components by retaining too many singular values, and also avoiding the loss of effective connection structure information by retaining too few singular values.

[0049] In specific implementation, when When the value is 16, The calculation result is 4, which represents the number of singular values ​​retained. It is 4. When When the value is 20, The calculated result is approximately 4.47, and the number of singular values ​​retained after rounding down is... The singular value diagonal matrix is ​​4. Only the former is retained in the middle. Find the largest singular value and set the rest of the singular values ​​to zero to obtain the dimension-reduced singular value diagonal matrix. Correspondingly, from the right singular vector matrix Extracting the first The three right singular vectors form the dimension-reduced right singular vector matrix. , dimension OK List.

[0050] When extracting the principal singular vectors of the initial topological matrix, start from the right singular vector matrix. Extract the first right singular vector corresponding to the largest singular value from the set, and use it as the principal singular vector. The principal singular vector is a vector of length 1. A column vector, where each element corresponds to the weight value of an independent component unit in the candidate component assembly. This principal singular vector... The weight values ​​are sorted in descending order of numerical value, with independent component units having larger weight values ​​appearing earlier in the connection sequence. Based on the descending sorting result, a new sequence index is assigned to each independent component unit in the candidate component assembly, starting from... It starts to increase, corresponding to the positions before and after in the arrangement.

[0051] Based on the rearranged connection order, the independent component elements in the candidate component assembly are recombined into new component geometric entities. During recombination, the elements are ordered from the first element to the next element. arrive Process them sequentially. The first one processed is the one with the sequential index of... The independent component element is used as the starting component for a new component geometry. The sequence index is... The independent component elements and their sequential indices are The independent component units are connected, with the connection position determined by the common connection node identifier matched in the connection node index table of the two independent component units. The two component geometries are joined at the connection node through coordinate alignment. Subsequent independent component units are joined to the already combined geometries one by one in the same way until all independent component units are combined to form a complete new component geometry. In the new component geometry, the multiple connection nodes that originally existed between the independent component units are merged and reduced, resulting in a decrease in the number of connection nodes compared to the total number of connection nodes for each independent component unit in the original candidate component assembly.

[0052] The new component geometry is output as an optimized component sub-model. The topology reconstruction module records the reduction in the number of connection nodes in this optimized component sub-model. This reduction is calculated by subtracting the actual number of connection nodes retained in the optimized component sub-model from the total number of connection nodes in the original independent component units of the candidate component assembly. The total number of connection nodes in the original independent component units is obtained by performing a union operation on the connection node index tables of all independent component units in the candidate component assembly, calculating the total number of duplicate connection node identifiers. The actual number of connection nodes retained in the optimized component sub-model is the actual number of connection nodes present in the new component geometry after assembly.

[0053] In specific implementation, please refer to Figure 5 The carbon fusion module receives all optimized component sub-models output by the topology reconstruction module, along with the reduction value of the number of connection nodes corresponding to each optimized component sub-model. When obtaining the reduction value of the number of connection nodes for each optimized component sub-model, it reads from the data recorded in the topology reconstruction module. The reduction value is the total number of connection nodes of the original independent component units in the candidate component assembly that generated the optimized component sub-model, minus the actual number of connection nodes retained in the optimized component sub-model. The total number of connection nodes of the original independent component units is the number of duplicate connection node identifiers obtained by performing a union operation on the connection node index tables of all independent component units in the candidate component assembly. The actual number of connection nodes retained in the optimized component sub-model is the number of connection points existing after the geometric entities of the internal components of the optimized component sub-model are joined.

[0054] The nodal carbon emission reduction coefficient for each optimized component sub-model is calculated based on the reduction in the number of connected nodes for each sub-model. The nodal carbon emission reduction coefficient is calculated using the following formula:

[0055] in, For the first The nodal carbon reduction coefficients of each optimized component sub-model The value of is a real number between 0 and 1; For the first The reduction in the number of connection nodes in each optimized component sub-model. It is a non-negative integer; To generate the first The total number of connection nodes of the original independent component units in the candidate component assembly corresponding to each optimized component sub-model. It is a positive integer. The method for obtaining this is as follows: During the candidate component assembly processing stage of the topology reconstruction module, record the size of the union of the connection node index tables of all independent component units in the candidate component assembly. When it is 0, A value of 0 indicates that the optimized component sub-model did not reduce the number of connected nodes.

[0056] After calculating the carbon emission reduction coefficient at each node, the carbon fusion module performs matching detection on the mating surfaces between adjacent optimized component sub-models. For each optimized component sub-model, the boundary surface of the optimized component sub-model is extracted in the BIM environment. The boundary surface is the surface where the optimized component sub-model contacts or connects with other components. The mating surface is identified by analyzing the geometry of the boundary surface. The geometry types of the mating surface include planar, arc, and polygonal. The geometric parameters of the mating surface are read from the BIM data of the optimized component sub-model. For planar mating surfaces, the geometric parameters include the vertex coordinate sequence and area of ​​the planar contour; for arc mating surfaces, the geometric parameters include the arc radius, central angle, and arc height. The criterion for determining adjacent optimized component sub-models is that the distance between the original spatial coordinates of the two optimized component sub-models in the initial BIM model is less than a preset proximity threshold. The preset proximity threshold is set to 10 mm, and this value is determined according to the allowable deviation standard for steel structure installation.

[0057] When two adjacent optimized component sub-models are matched for mating surfaces, the geometric shape types of the two mating surfaces are compared. If the geometric shape types are consistent, the geometric dimensions of the mating surfaces are further compared. For planar mating surfaces, the Euclidean distance between corresponding vertices of the contour vertex coordinate sequences of the two mating surfaces after spatial alignment is compared. If the Euclidean distance between all corresponding vertices is less than the preset geometric matching tolerance, the geometric shape and dimensions of the two mating surfaces are determined to be consistent. The preset geometric matching tolerance is set to 2 mm, a value determined according to the steel structure manufacturing precision standard. For arc-shaped mating surfaces, the difference in arc radius and the difference in arc height of the two mating surfaces are compared. If the absolute value of the difference in arc radius is less than 2 mm and the absolute value of the difference in arc height is less than 2 mm, the geometric shape and dimensions of the two mating surfaces are determined to be consistent.

[0058] When the geometry and dimensions of two mating surfaces are identical, the carbon fusion module merges these two optimized component sub-models through a shared mating surface. The fusion operation calls the Boolean union function of the BIM software, merging the geometric entities of the two optimized component sub-models into a single geometric entity with the mating surface as the common face, forming a fused component group. During the fusion process, redundant geometric surfaces at the shared mating surface are removed, and the connection nodes within the two optimized component sub-models are merged into a single connection node at the mating surface. After forming the fused component group, the comprehensive node carbon emission reduction coefficient of the fused component group is calculated. The comprehensive node carbon emission reduction coefficient is the sum of the carbon emission reduction coefficients of all optimized component sub-models participating in the fusion within the fused component group. The average value is calculated by weighting the volume of the optimized component sub-model.

[0059] All integrated component groups and non-integrated optimized component sub-models are reassembled according to the spatial coordinates of the original BIM initial model. For each integrated component group and non-integrated optimized component sub-model, the positioning base point coordinates and direction vectors are obtained from the original spatial coordinate records of the independent component units contained within it. The assembly reference point of the combined overall geometric entity is calculated, and the assembly reference point is aligned with the corresponding coordinate position in the world coordinate system of the BIM initial model. After reassembly, a carbon reduction optimized BIM model is generated. In the carbon reduction optimized BIM model, each integrated component group is attached with a node carbon emission reduction coefficient label. The node carbon emission reduction coefficient label is stored as BIM attribute data, with the attribute field name being "node carbon emission reduction coefficient" and the attribute value being the comprehensive node carbon emission reduction coefficient value of the integrated component group. The non-integrated optimized component sub-models are also attached with node carbon emission reduction coefficient labels, with the attribute value being the optimized component sub-model itself. Numerical value.

[0060] After generating the carbon reduction optimized BIM model, iterative verification is performed on iterative verification. The first step of iterative verification is to use the carbon reduction optimized BIM model as a new initial BIM model and re-execute carbon emission prediction. When re-executing carbon emission prediction, the carbon pre-diagnosis module is invoked. According to the execution logic of the carbon pre-diagnosis module, the volume and surface data of each component are extracted from the carbon reduction optimized BIM model, the carbon emission factor per unit mass and the carbon emission factor of the coating are queried, the carbon emission of the component substrate and the carbon emission of the component coating are calculated and accumulated to obtain the secondary carbon emission prediction result. After obtaining the secondary carbon emission prediction result, the difference between the secondary carbon emission prediction result and the preset carbon emission threshold is calculated, and the absolute value of the difference is taken. This difference is compared with the preset convergence tolerance, which is set to 5% of the preset carbon emission threshold. The basis for setting the convergence tolerance to 5% of the carbon emission threshold is that when the carbon emission has dropped to a small range near the threshold, the marginal benefit of continuing optimization decreases, and stopping iteration can avoid over-optimization while retaining sufficient carbon reduction effect.

[0061] When the difference between the secondary carbon emission prediction result and the preset carbon emission threshold is less than the preset convergence tolerance, the iteration process terminates, and the current carbon reduction optimization BIM model is output as the final design model. When the difference between the secondary carbon emission prediction result and the preset carbon emission threshold is greater than or equal to the preset convergence tolerance, the model fragmentation and reconstruction process is reactivated. The activation method is the same as that in the carbon prediagnosis module, sending an activation signal to the control interface of the BIM software. After reactivation, the fragmentation and decomposition module performs a new round of decomposition on the current carbon reduction optimization BIM model. During decomposition, it traverses the connection nodes in the carbon reduction optimization BIM model, decomposing the model into a new set of independent component units. Then, the graph network clustering module, topology reconstruction module, and carbon fusion module are executed sequentially to generate a new round of carbon reduction optimization BIM model, and iterative verification is performed again. The iteration process continues until the difference between the secondary carbon emission prediction result and the preset carbon emission threshold is less than the preset convergence tolerance in a certain iteration, or the number of iterations reaches the preset maximum number of iterations. The maximum number of iterations is set to 5, based on the analysis of system convergence efficiency, and it is expected that the carbon emission reduction effect will tend to stabilize within 5 iterations.

[0062] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A BIM-based substation steel structure full life-cycle carbon reduction optimization design system, characterized in that, include: The carbon pre-diagnosis module acquires the initial BIM model of the substation steel structure and performs carbon emission prediction based on the initial BIM model. When the predicted carbon emissions exceed the preset threshold, the model fracture and reconstruction process is activated. The fracture dismantling module decomposes the initial BIM model into multiple independent component units according to the steel structure connection nodes during the model fracture reconstruction process, and records the geometric parameters and material properties of each independent component unit. The graph clustering module calculates the carbon transport dependency value between components based on the geometric parameters and material properties of each independent component unit using a graph neural network, and clusters multiple independent component units based on the carbon transport dependency value to generate multiple candidate component assemblies. The topology reconstruction module performs topology reconstruction on each candidate component assembly, generating multiple optimized component sub-models. The carbon fusion module performs carbon correlation fusion on all optimized component sub-models to generate a carbon reduction optimized BIM model.

2. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 1, characterized in that, The specific implementation method for the step of obtaining the initial BIM model of the substation steel structure and performing carbon emission prediction based on the initial BIM model is as follows: Extract the spatial layout data and material list of the steel structure from the construction drawings of the substation, import the spatial layout data and material list into BIM software, and generate an initial BIM model that includes the geometric dimensions of the components, steel grades, connection methods and anti-corrosion coating types. Extract volume and surface data of each component from the initial BIM model, and query the corresponding carbon emission factor per unit mass based on the steel grade, and query the corresponding coating carbon emission factor based on the anti-corrosion coating type. The volume data of each component is multiplied by the steel density to obtain the component mass. Then, the component mass is multiplied by the carbon emission factor per unit mass to obtain the carbon emission of the component substrate. At the same time, the surface data of each component is multiplied by the coating carbon emission factor to obtain the carbon emission of the component coating. The carbon emission of the substrate and the carbon emission of the coating of all components are added together to obtain the carbon emission prediction result of the BIM initial model.

3. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 1, characterized in that, The specific implementation method of the step of activating model fragmentation and reconstruction when the predicted carbon emissions exceed the preset threshold is as follows: The carbon emission prediction results of the initial BIM model are compared with the preset carbon emission threshold, which is obtained by multiplying the average carbon emission of steel structures of the same type of substation in the design code by a safety factor. When the carbon emission prediction result is greater than the preset carbon emission threshold, an activation signal is sent to the control interface of the BIM software, which triggers the start of the model fragmentation and reconstruction process. When the carbon emission prediction result is less than or equal to the preset carbon emission threshold, the carbon reduction optimization process is terminated, and the BIM initial model is output as the final design model.

4. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 1, characterized in that, In the process of model fracture and reconstruction, the step of decomposing the initial BIM model into multiple independent component units according to the steel structure connection nodes is specifically implemented as follows: Traverse the topological relationships of all steel structure connection nodes in the initial BIM model and identify the component identifier associated with each connection node, which includes beam identifier, column identifier and brace identifier; For each component identifier, extract the component geometric entity corresponding to the component identifier, and separate the component geometric entity from the connection node to form an independent component unit; Each independent component unit is assigned a unique number, and its original spatial coordinates and connection node index table in the initial BIM model are recorded.

5. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 4, characterized in that, The specific implementation method for recording the geometric parameters and material properties of each independent component unit is as follows: Extract the cross-sectional shape, cross-sectional dimensions, length data, and volume data of each independent component element from its component geometry entity, and summarize these data into the geometric parameters of that independent component element. Query the steel grade, yield strength, anti-corrosion coating type and fireproof coating type corresponding to each independent component unit from the attribute database of the initial BIM model, and summarize these data into the material properties of the independent component unit; A mapping relationship is established between the unique number of each independent component element and the geometric parameters and material properties of that independent component element, and stored in the extended data table of the BIM model.

6. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 1, characterized in that, The specific implementation method of the step of calculating the carbon transport dependence value between components using a graph neural network based on the geometric parameters and material properties of each independent component unit is as follows: Each independent component unit is used as a node in the graph neural network, and the connection relationship between adjacent independent component units is used as the edge of the graph neural network to construct a graph data structure; The geometric parameters and material properties of each independent component element are encoded into a feature vector, and this feature vector is assigned to the corresponding node in the graph neural network. The message passing mechanism of the graph neural network enables each node to aggregate the feature vectors of its neighboring nodes, and calculates the carbon transfer dependency value between each node and its neighboring nodes based on the aggregation result. This carbon transfer dependency value represents the influence weight of the carbon emissions of a component unit on the carbon emissions of its neighboring component units.

7. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 6, characterized in that, The specific implementation method of the step of clustering multiple independent component units based on the carbon transport dependency value to generate multiple candidate component assemblies is as follows: Using each independent component unit as the initial cluster center, calculate the carbon transfer dependency between that independent component unit and all its neighboring independent component units; The carbon transport dependency value is compared with a preset clustering threshold. When the carbon transport dependency value is greater than the clustering threshold, the neighboring independent component unit is included in the candidate component combination where the current cluster center is located. After all independent component units have been clustered and assigned, the independent component units in each candidate component assembly are merged, and the total carbon emission prediction value of the candidate component assembly is calculated. The total carbon emission prediction value is used as the cluster label of the candidate component assembly.

8. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 7, characterized in that, The preset clustering threshold is dynamically calculated based on the sum of the mean and standard deviation of all carbon transport dependencies.

9. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 1, characterized in that, The specific implementation method of performing topological reconstruction on each candidate component assembly to generate multiple optimized component sub-models is as follows: Extract the spatial coordinates and connection node index table of all independent component units from each candidate component assembly, and construct the initial topology matrix of the candidate component assembly; Perform singular value decomposition on the initial topology matrix, extract the principal singular vector of the initial topology matrix, and rearrange the connection order of the independent component units in the candidate component assembly according to the principal singular vector; Based on the rearranged connection order, the independent component units in the candidate component assembly are recombined into new component geometric entities. These new component geometric entities serve as optimized component sub-models, and the reduction in the number of connection nodes in these optimized component sub-models is recorded.

10. The BIM-based substation steel structure full life-cycle carbon reduction optimization design system according to claim 1, characterized in that, The specific implementation method of fusing all optimized component sub-models through carbon correlation to generate a carbon reduction optimized BIM model is as follows: Obtain the reduction value of the number of connected nodes in each optimized component sub-model, and calculate the node carbon emission reduction coefficient of each optimized component sub-model based on the reduction value of the number of connected nodes; Matching detection is performed on the docking surfaces between adjacent optimized component sub-models. When the geometry and size of the two docking surfaces are consistent, the two optimized component sub-models are merged by sharing the docking surface to form a fused component group. All integrated component groups and optimized component sub-models that did not participate in the integration are reassembled according to the spatial coordinates of the original BIM initial model to generate a carbon reduction optimized BIM model, and node carbon emission reduction coefficient labels for each integrated component group are added to the carbon reduction optimized BIM model.

Images