Industrial land intelligent site selection method based on spatial big data and artificial intelligence big model

Abstract

The invention discloses an industrial land intelligent site selection method based on spatial big data and an artificial intelligence big model. The method comprises the following steps: 1) obtaining a feature matrix containing spatial topology constraints and economic association rules; 2) extracting key analysis elements in user site selection requirements; 3) inputting the spatial key analysis elements into CNN convolutional neural network branches to obtain spatial features; the policy text type key analysis elements are input into a Transform branch, and policy features are obtained; inputting the industrial chain type key analysis elements into GNN graph neural network branches to obtain industrial chain features; 4) generating an industrial land analysis conclusion; and 5) generating a comparison report, a site selection proposal, a risk avoidance strategy and an emergency plan of each feasible scheme. According to the invention, through fusion of multi-source spatial data (industrial layout, traffic network and regional assessment) and an artificial intelligence large model, dynamic planning constraint matching, full-life-cycle risk assessment and multi-objective optimization site selection are realized.

Description

Technical Field

[0001] The present invention relates to the intersection of smart city planning, industrial upgrading and artificial intelligence technology, and specifically to an intelligent site selection method for industrial land based on spatial big data and artificial intelligence big models. Background Art

[0002] Industrial land site selection is the process of scientifically and rationally selecting a suitable land location for industrial project construction based on the specific needs of industrial production, taking into account multiple factors, including natural, economic, social, and environmental factors. It is a key component of industrial project planning and regional industrial layout, directly impacting a company's operating costs, production efficiency, and sustainable development capabilities.

[0003] However, traditional site planning methods have the following problems:

[0004] Data silos: It is difficult to efficiently integrate multi-source data such as industrial layout, transportation network, and regional assessment.

[0005] Static constraints: Traditional models cannot update dynamic constraints in planning in real time (such as planning adjustments and changes in planning constraints).

[0006] Insufficient risk assessment: The coverage of regional assessment indicators such as earthquake safety and environmental impact is incomplete, and there is a lack of full life cycle risk prediction capabilities.

[0007] Inefficiency: High reliance on manual labor makes it difficult to achieve multi-objective optimization (such as economy, safety, and sustainability).

[0008] Therefore, it is necessary to provide a dynamic planning and risk assessment system covering the entire process. Summary of the Invention

[0009] The purpose of the present invention is to provide an intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model, comprising the following steps:

[0010] 1) Obtain and process data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to obtain a feature matrix containing spatial topological constraints and economic association rules;

[0011] 2) Obtain user site selection requirements, perform semantic recognition, and extract key analysis elements in user site selection requirements;

[0012] 3) Build an industrial land location selection model, including a CNN convolutional neural network branch, a Transformer branch, a GNN graph neural network branch, and a multimodal fusion layer;

[0013] 4) Input the key analysis elements of spatial categories into the CNN convolutional neural network branch to obtain spatial features; input the key analysis elements of policy text categories into the Transformer branch to obtain policy features; input the key analysis elements of industrial chain categories into the GNN graph neural network branch to obtain industrial chain features;

[0014] Use the multimodal fusion layer to fuse spatial features, policy features, and industrial chain features to obtain multi-dimensional features;

[0015] 5) Input multi-dimensional features into the intelligent analysis engine, and generate industrial land analysis conclusions through knowledge graph-assisted reasoning and dynamic weight optimization algorithm;

[0016] 6) Build a dynamic constraint rule library;

[0017] 7) Combined with the dynamic constraint rule base and the industrial land analysis conclusions, a set of feasible solutions is screened out, and the risk assessment indicators of each feasible solution are calculated; if the risk assessment indicator exceeds the threshold, the parameter backtracking mechanism is automatically triggered to readjust the industrial land site selection model or constraint conditions; if the risk assessment indicator does not exceed the threshold, a comparison report, site selection proposal, risk avoidance strategy and emergency plan for each feasible solution are generated.

[0018] Furthermore, in step 1), the steps of processing the traffic network, industrial park boundaries, enterprise attributes, and regional economic assessment data include:

[0019] 1.1) Using natural language processing algorithms, we disambiguate entities in data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to identify the true, unique objects to which fuzzy entities refer.

[0020] 1.2) Establish a unified spatial coordinate system and time base, and through ontology mapping technology, associate the registered address of the enterprise with the GIS geographic coordinates to build a spatiotemporal index across data sources;

[0021] 1.3) Using spatial overlay analysis methods, calculate the topological accessibility of enterprise POIs and trunk roads to generate a "transportation-industry" coupling index;

[0022] 1.4) Data standardization;

[0023] 1.5) Use GeoHash to divide the spatial data into multiple grid cells and extract industrial park density and traffic flow values;

[0024] Convert non-spatial data into vectors through Embedding technology to extract enterprise types and regional assessment levels;

[0025] 1.6) The industrial park density, traffic flow value, enterprise type, and regional assessment level are weightedly integrated through the attention mechanism to obtain a feature matrix that contains spatial topological constraints and economic association rules.

[0026] Furthermore, in step 1.4), for numerical data, the data standardization method is Z-Score normalization; for spatial data, the data standardization method is UTM projection conversion.

[0027] Furthermore, in step 2), the method for performing semantic recognition includes semantic role labeling and intent recognition technology.

[0028] Furthermore, in step 6), the step of constructing a dynamic constraint rule base includes:

[0029] 6.1) Analyze local policies and obtain the results of local policy analysis, including the guiding requirements of local policies on industrial positioning, spatial distribution, and economic structure optimization;

[0030] 6.2) Based on the policy analysis results, analyze the mandatory requirements, regulatory requirements and special control conditions of the special plan;

[0031] 6.3) Delineate geological risk avoidance zones based on earthquake safety impact assessments; determine ecologically sensitive areas and pollutant emission thresholds based on environmental impact assessments; set safety production standards based on construction project safety reviews; determine extreme weather response measures based on climate feasibility studies; determine total water use and water conservation requirements based on water resource studies; and delineate cultural heritage protection areas based on cultural relic impact assessments.

[0032] 6.4) Combine steps 6.2) and 6.3) to build a dynamic constraint rule library.

[0033] Furthermore, mandatory planning requirements include functional zoning of land use and upper limits on floor area ratio;

[0034] Control requirements include building setbacks and green space ratio;

[0035] The special control conditions of the special plan include safety protection areas and historical and cultural area protection areas.

[0036] The system based on the intelligent site selection method for industrial land includes a data fusion layer, an intelligent analysis layer, a dynamic rule constraint layer, and a decision output layer;

[0037] The data fusion layer obtains and processes the traffic network, industrial park boundaries, enterprise attributes and regional economic assessment data to obtain a feature matrix containing spatial topological constraints and economic association rules;

[0038] The intelligent analysis layer obtains the user's site selection requirements, performs semantic recognition, extracts the key analysis elements in the user's site selection requirements, and then inputs the spatial key analysis elements into the CNN convolutional neural network branch to obtain spatial features, inputs the policy text key analysis elements into the Transformer branch to obtain policy features, and inputs the industrial chain key analysis elements into the GNN graph neural network branch to obtain industrial chain features;

[0039] The intelligent analysis layer uses the multimodal fusion layer to integrate spatial features, policy features, and industrial chain features to obtain multi-dimensional features, and generates industrial land analysis conclusions through knowledge graph-assisted reasoning and dynamic weight optimization algorithms;

[0040] The dynamic rule constraint layer constructs a dynamic constraint rule library;

[0041] The decision output layer combines the dynamic constraint rule base and the industrial land analysis conclusions to screen out a set of feasible solutions and calculate the risk assessment indicators of each feasible solution; if the risk assessment indicator exceeds the threshold, the parameter backtracking mechanism is automatically triggered to readjust the industrial land site selection model or constraint conditions; if the risk assessment indicator does not exceed the threshold, a comparison report, site selection proposal, risk avoidance strategy and emergency plan for each feasible solution are generated.

[0042] The technical effects of this invention are undeniable. By integrating multi-source spatial data (industrial layout, transportation network, regional assessment) with a large artificial intelligence model, this invention achieves dynamic planning constraint matching, full lifecycle risk assessment, and multi-objective optimization site selection. This method supports real-time updating of planning constraints and covers 11 regional assessment indicators, including earthquake safety impact, environmental impact, construction project safety, climate feasibility, water resource demonstration, cultural heritage impact, and overburdened mines. It solves the problems of low efficiency and data siloing in traditional site selection and is suitable for smart city and industrial upgrading scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Intelligent site selection system for industrial land;

[0044] Figure 2 This is the data preprocessing flow chart;

[0045] Figure 3 For intelligent analysis process;

[0046] Figure 4 Build processes for dynamic constraint rules;

[0047] Figure 5 Output process for site selection decision making. DETAILED DESCRIPTION

[0048] The present invention will be further described below with reference to the following examples, but it should not be understood that the scope of the present invention is limited to the following examples. Without departing from the above technical ideas of the present invention, various substitutions and modifications can be made according to common technical knowledge and customary means in the art, and all should be included in the scope of protection of the present invention.

[0049] Example 1:

[0050] See also Figures 1 to 5 ,The intelligent site selection method for industrial land based on spatial big data and artificial intelligence ,model includes the following steps:

[0051] 1) Obtain and process data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to obtain a feature matrix containing spatial topological constraints and economic association rules;

[0052] 2) Obtain user site selection requirements, perform semantic recognition, and extract key analysis elements in user site selection requirements;

[0053] 3) Build an industrial land location selection model, including a CNN convolutional neural network branch, a Transformer branch, a GNN graph neural network branch, and a multimodal fusion layer;

[0054] 4) Input the key analysis elements of spatial categories into the CNN convolutional neural network branch to obtain spatial features; input the key analysis elements of policy text categories into the Transformer branch to obtain policy features; input the key analysis elements of industrial chain categories into the GNN graph neural network branch to obtain industrial chain features;

[0055] Use the multimodal fusion layer to fuse spatial features, policy features, and industrial chain features to obtain multi-dimensional features;

[0056] 5) Input multi-dimensional features into the intelligent analysis engine, and generate industrial land analysis conclusions through knowledge graph-assisted reasoning and dynamic weight optimization algorithm;

[0057] 6) Build a dynamic constraint rule library;

[0058] 7) Combined with the dynamic constraint rule base and the industrial land analysis conclusions, a set of feasible solutions is screened out, and the risk assessment indicators of each feasible solution are calculated; if the risk assessment indicator exceeds the threshold, the parameter backtracking mechanism is automatically triggered to readjust the industrial land site selection model or constraint conditions; if the risk assessment indicator does not exceed the threshold, a comparison report, site selection proposal, risk avoidance strategy and emergency plan for each feasible solution are generated.

[0059] In step 1), the steps for processing the transportation network, industrial park boundaries, enterprise attributes, and regional economic assessment data include:

[0060] 1.1) Using natural language processing algorithms, we disambiguate entities in data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to identify the true, unique objects to which fuzzy entities refer.

[0061] 1.2) Establish a unified spatial coordinate system and time base, and through ontology mapping technology, associate the registered address of the enterprise with the GIS geographic coordinates to build a spatiotemporal index across data sources;

[0062] 1.3) Using spatial overlay analysis methods, calculate the topological accessibility of enterprise POIs and trunk roads to generate a "transportation-industry" coupling index;

[0063] 1.4) Data standardization;

[0064] 1.5) Use GeoHash to divide the spatial data into multiple grid cells and extract industrial park density and traffic flow values;

[0065] Convert non-spatial data into vectors through Embedding technology to extract enterprise types and regional assessment levels;

[0066] 1.6) The industrial park density, traffic flow value, enterprise type, and regional assessment level are weightedly integrated through the attention mechanism to obtain a feature matrix that contains spatial topological constraints and economic association rules.

[0067] In step 1.4), for numerical data, the data normalization method is Z-Score normalization; for spatial data, the data normalization method is UTM projection conversion.

[0068] In step 2), the method for semantic recognition includes semantic role labeling and intent recognition technology.

[0069] In step 6), the steps of building a dynamic constraint rule base include:

[0070] 6.1) Analyze local policies and obtain the results of local policy analysis, including the guiding requirements of local policies on industrial positioning, spatial distribution, and economic structure optimization;

[0071] 6.2) Based on the policy analysis results, analyze the mandatory requirements, regulatory requirements and special control conditions of the special plan;

[0072] 6.3) Delineate geological risk avoidance zones based on earthquake safety impact assessments; determine ecologically sensitive areas and pollutant emission thresholds based on environmental impact assessments; set safety production standards based on construction project safety reviews; determine extreme weather response measures based on climate feasibility studies; determine total water use and water conservation requirements based on water resource studies; and delineate cultural heritage protection areas based on cultural relic impact assessments.

[0073] 6.4) Combine steps 6.2) and 6.3) to build a dynamic constraint rule library.

[0074] Mandatory planning requirements include land use function zoning and upper limit of floor area ratio;

[0075] Control requirements include building setbacks and green space ratio;

[0076] Special control conditions of the special plan include ecological protection red lines and historical and cultural area protection areas.

[0077] Example 2:

[0078] The intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model includes the following steps:

[0079] 1) Obtain and process data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to obtain a feature matrix containing spatial topological constraints and economic association rules;

[0080] 2) Obtain user site selection requirements, perform semantic recognition, and extract key analysis elements in user site selection requirements;

[0081] 3) Build an industrial land location selection model, including a CNN convolutional neural network branch, a Transformer branch, a GNN graph neural network branch, and a multimodal fusion layer;

[0082] 4) Input the key analysis elements of spatial categories into the CNN convolutional neural network branch to obtain spatial features; input the key analysis elements of policy text categories into the Transformer branch to obtain policy features; input the key analysis elements of industrial chain categories into the GNN graph neural network branch to obtain industrial chain features;

[0083] Use the multimodal fusion layer to fuse spatial features, policy features, and industrial chain features to obtain multi-dimensional features;

[0084] 5) Input multi-dimensional features into the intelligent analysis engine, and generate industrial land analysis conclusions through knowledge graph-assisted reasoning and dynamic weight optimization algorithm;

[0085] 6) Build a dynamic constraint rule library;

[0086] 7) Combined with the dynamic constraint rule base and the industrial land analysis conclusions, a set of feasible solutions is screened out, and the risk assessment indicators of each feasible solution are calculated; if the risk assessment indicator exceeds the threshold, the parameter backtracking mechanism is automatically triggered to readjust the industrial land site selection model or constraint conditions; if the risk assessment indicator does not exceed the threshold, a comparison report, site selection proposal, risk avoidance strategy and emergency plan for each feasible solution are generated.

[0087] Example 3:

[0088] The intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model has the same technical content as Example 2. Furthermore, in step 1), the step of processing the traffic network, industrial park boundaries, enterprise attributes and regional economic evaluation data includes:

[0089] 1.1) Using natural language processing algorithms, we disambiguate entities in data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to identify the true, unique objects to which fuzzy entities refer.

[0090] 1.2) Establish a unified spatial coordinate system and time base, and through ontology mapping technology, associate the registered address of the enterprise with the GIS geographic coordinates to build a spatiotemporal index across data sources;

[0091] 1.3) Using spatial overlay analysis methods, calculate the topological accessibility of enterprise POIs and trunk roads to generate a "transportation-industry" coupling index;

[0092] 1.4) Data standardization;

[0093] 1.5) Use GeoHash to divide the spatial data into multiple grid cells and extract industrial park density and traffic flow values;

[0094] Convert non-spatial data into vectors through Embedding technology to extract enterprise types and regional assessment levels;

[0095] 1.6) The industrial park density, traffic flow value, enterprise type, and regional assessment level are weightedly integrated through the attention mechanism to obtain a feature matrix that contains spatial topological constraints and economic association rules.

[0096] Example 4:

[0097] The intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model has the same technical content as any one of Examples 2-3. Furthermore, in step 1.4), for numerical data, the data standardization method is Z-Score normalization; for spatial data, the data standardization method is UTM projection conversion.

[0098] Example 5:

[0099] The intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model has the same technical content as any one of Examples 2-4. Furthermore, in step 2), the method for semantic recognition includes semantic role labeling and intent recognition technology.

[0100] Example 6:

[0101] The intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model has the same technical content as any one of Examples 2-5. Furthermore, in step 6), the step of constructing a dynamic constraint rule library includes:

[0102] 6.1) Analyze local policies and obtain the results of local policy analysis, including the guiding requirements of local policies on industrial positioning, spatial distribution, and economic structure optimization;

[0103] 6.2) Based on the policy analysis results, analyze the mandatory requirements, regulatory requirements and special control conditions of the special plan;

[0104] 6.3) Delineate geological risk avoidance zones based on earthquake safety impact assessments; determine ecologically sensitive areas and pollutant emission thresholds based on environmental impact assessments; set safety production standards based on construction project safety reviews; determine extreme weather response measures based on climate feasibility studies; determine total water use and water conservation requirements based on water resource studies; and delineate cultural heritage protection areas based on cultural relic impact assessments.

[0105] 6.4) Combine steps 6.2) and 6.3) to build a dynamic constraint rule library.

[0106] Example 7:

[0107] An intelligent site selection method for industrial land based on spatial big data and artificial intelligence large models, with the same technical content as any one of Examples 2-6, and further, mandatory planning requirements including land function zoning and upper limit of floor area ratio;

[0108] Control requirements include building setbacks and green space ratio;

[0109] The special control conditions of special plans include airport clearance protection areas, geological disaster prevention and control zoning, safety protection areas, assessment of buried mineral areas, geological disaster risk assessment, and historical and scenic area protection areas.

[0110] Example 8:

[0111] A system based on the intelligent site selection method for industrial land according to any one of embodiments 1-7, comprising a data fusion layer, an intelligent analysis layer, a dynamic rule constraint layer, and a decision output layer;

[0112] The data fusion layer obtains and processes the traffic network, industrial park boundaries, enterprise attributes and regional economic assessment data to obtain a feature matrix containing spatial topological constraints and economic association rules;

[0113] The intelligent analysis layer obtains the user's site selection requirements, performs semantic recognition, extracts the key analysis elements in the user's site selection requirements, and then inputs the spatial key analysis elements into the CNN convolutional neural network branch to obtain spatial features, inputs the policy text key analysis elements into the Transformer branch to obtain policy features, and inputs the industrial chain key analysis elements into the GNN graph neural network branch to obtain industrial chain features;

[0114] The intelligent analysis layer uses the multimodal fusion layer to integrate spatial features, policy features, and industrial chain features to obtain multi-dimensional features, and generates industrial land analysis conclusions through knowledge graph-assisted reasoning and dynamic weight optimization algorithms;

[0115] The dynamic rule constraint layer constructs a dynamic constraint rule library;

[0116] The decision output layer combines the dynamic constraint rule base and the industrial land analysis conclusions to screen out a set of feasible solutions and calculate the risk assessment indicators of each feasible solution; if the risk assessment indicator exceeds the threshold, the parameter backtracking mechanism is automatically triggered to readjust the industrial land site selection model or constraint conditions; if the risk assessment indicator does not exceed the threshold, a comparison report, site selection proposal, risk avoidance strategy and emergency plan for each feasible solution are generated.

[0117] Example 9:

[0118] Intelligent site selection system for industrial land based on spatial big data and artificial intelligence big model,

[0119] Intelligent site selection is achieved through the following modules:

[0120] Data fusion layer: Integrates structured and unstructured data such as industrial layout, transportation network, and regional assessment, and uses distributed computing technology to achieve high-precision spatial analysis.

[0121] Intelligent Analysis Layer: By deeply integrating the complementary strengths of models such as CNN, Transformer, and GNN, an intelligent analysis and evaluation system integrating multimodal features has been constructed. During the feature extraction phase, multi-scale dilated convolutions are used to extract local spatial features, combined with a channel attention mechanism to enhance key region recognition. A heterogeneous graph relationship model is constructed, employing graph pooling techniques to generate high-order topological representations. The score generation module innovatively introduces interpretable components and a multi-task learning framework to achieve dual-path output for classification and regression. At the optimization strategy level, hybrid precision training and a distributed computing framework are used to accelerate model convergence for different modal characteristics, achieving multi-dimensional comprehensive scoring and improving site selection accuracy.

[0122] Dynamic Rule Constraint Layer: A dynamic rule constraint system is constructed by integrating intelligent policy text parsing, digital modeling of planning constraints, and multidimensional regional assessment analysis. This system uses content and spatial analysis methods to semantically analyze policy texts, such as industrial park development plans and management regulations, to establish a rule matrix encompassing mandatory indicators, guiding requirements, and special constraints. This system dynamically updates policy rules through comparison and threshold alarms. The dynamic rule base employs an intelligent matching algorithm to accurately map rules to plot characteristics. A conflict resolution mechanism ensures that superior laws take precedence and that new regulations override old ones. The assessment model incorporates a three-level screening process: an initial screening verifies land use compatibility, a secondary screening calculates planning indicators such as floor area ratio, and a final screening verifies special constraints such as aviation height restrictions and historical preservation. The system outputs a heat map of compliant plots, a list of constraints, and recommended adjustments. This system ensures site compliance through a technical path: policy semantic parsing → spatial constraint digitization → dynamic rule updating → intelligent plot matching.

[0123] Decision output layer: provides site selection recommendations, risk assessment, and decision support, including optimal site recommendations and risk avoidance suggestions.

[0124] Example 10:

[0125] The intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model includes the following steps:

[0126] Step 1: Data Preprocessing

[0127] The data preprocessing process, with planning data as its core framework, integrates four types of heterogeneous data: transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments. Multi-source data fusion is achieved through a three-stage process. The first stage involves dynamic semantic parsing and data alignment. The system invokes the natural language processing module to perform semantic disambiguation on entities such as "industrial park" and "transportation hub" in the dispersed data, establishing a unified spatial coordinate system (converted to CGCS2000) and time base (standardized to quarterly granularity). Furthermore, through ontology mapping technology, enterprise registration addresses are linked to GIS geographic coordinates to construct a spatiotemporal index across data sources. The second stage involves spatiotemporal data cleaning and reconstruction. Using spatial overlay analysis, the topological accessibility of enterprise POIs and main roads is calculated to generate a "transportation-industry" coupling indicator. Data standardization is also completed simultaneously during this stage, including Z-score normalization of numerical variables and UTM projection conversion of spatial data. The third stage involves cross-modal feature encoding and fusion. Spatial data is divided into 500-meter grid cells using GeoHash to extract industrial park density and traffic flow values. Embedding technology is used to convert non-spatial data into vectors, transforming enterprise types and regional assessment levels. These two types of features are weighted and fused using an attention mechanism to output a feature matrix that incorporates spatial topological constraints and economic association rules, supporting intelligent site selection analysis.

[0128] Step 2: Smart Analysis

[0129] Intelligent analysis utilizes a multimodal collaborative architecture to achieve full-chain intelligent processing, from demand input to output. The process begins with the user inputting a site selection request in natural language. The system uses a dynamic semantic parsing module to deeply deconstruct the request. Using semantic role labeling and intent recognition techniques, it automatically extracts key analytical elements and identifies data type characteristics. The parsing engine dynamically routes data based on data type characteristics: spatial data is transferred to the CNN convolutional neural network branch, where spatial features are extracted using multi-layer convolution kernels; policy text data enters the Transformer branch, where long-range semantic dependencies are captured using a self-attention mechanism; and industry chain-related data is assigned to the GNN graph neural network branch, which resolves complex topological structures through node embedding and message passing. The processing results of the three models are integrated at the feature level in the multimodal fusion layer, where a cross-modal attention mechanism is used to jointly learn representations of spatial, textual, and graph structures. The fused multidimensional features are fed into the intelligent analysis engine, where the final analysis conclusions are generated through knowledge graph-assisted reasoning and a dynamic weight optimization algorithm.

[0130] Step 3: Dynamic Constraint Rules

[0131] The dynamic constraint rule construction process is a systematic process centered on multi-level constraint analysis, gradually forming a quantifiable rule base. First, policy rule analysis focuses on dimensions such as industrial layout and industrial development. By sorting out the guiding requirements of local policies regarding industrial positioning, spatial distribution, and economic structure optimization, the policy boundaries and target framework for regional development are clarified, providing directional basis for subsequent constraint conditions. Second, planning constraint analysis is conducted. Based on the results of policy analysis, mandatory planning requirements (such as land use function zoning and floor area ratio limits) are analyzed. Next, control requirements (such as planning control indicators such as building setbacks and green space ratios) are extracted. Finally, special control conditions for special plans (such as ecological protection red lines and historical and cultural area protection areas) are integrated. Third, regional assessment constraints are integrated, establishing a checklist through cross-disciplinary special assessments: earthquake safety impact assessments define geological risk avoidance zones; environmental impact assessments identify ecologically sensitive areas and pollutant emission thresholds; construction project safety reviews set safety production standards; climate feasibility studies propose extreme weather response measures; water resource studies determine total water use and water conservation requirements; and cultural relic impact assessments define cultural heritage protection areas. Ultimately, through rule base integration and dynamic management, the above constraints are transformed into structured, digital rule entries, and a rule conflict verification mechanism and dynamic update interface are established to support the application scenario of intelligent site selection for industrial land.

[0132] Step 4: Site Selection Decision Output

[0133] The site selection decision-making process, centered around dynamic constraints and intelligent analysis, forms a scientific, closed-loop management system. First, a set of feasible solutions is identified through multi-dimensional intelligent analysis combined with dynamic constraints. Feasibility plans for multiple sites are then compared to produce a comparative report and decision-making support recommendations. If the initial site selection passes the risk assessment, an optimized site recommendation is generated. If risk indicators exceed thresholds, the process automatically triggers a parameter backtracking mechanism to readjust the analysis model or constraints. Finally, the site selection proposal, risk mitigation strategies, and contingency plans are integrated into a comprehensive decision report.

Claims

1. An intelligent site selection method for industrial land based on spatial big data and artificial intelligence big model, characterized by: The following steps are involved: 1) Obtain and process data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to obtain a feature matrix containing spatial topological constraints and economic association rules; 2) Obtain user site selection requirements, perform semantic recognition, and extract key analysis elements in user site selection requirements; 3) Build an industrial land location selection model, including a CNN convolutional neural network branch, a Transformer branch, a GNN graph neural network branch, and a multimodal fusion layer; 4) Input the key spatial analysis elements into the CNN convolutional neural network branch to obtain spatial features; Input the key analysis elements of the policy text into the Transformer branch to obtain policy features; Input the key analysis elements of the industrial chain into the GNN graph neural network branch to obtain the characteristics of the industrial chain; Use the multimodal fusion layer to fuse spatial features, policy features, and industrial chain features to obtain multi-dimensional features; 5) Input multi-dimensional features into the intelligent analysis engine, and generate industrial land analysis conclusions through knowledge graph-assisted reasoning and dynamic weight optimization algorithm; 6) Build a dynamic constraint rule library; 7) Combine the dynamic constraint rule base and the industrial land analysis conclusions to screen out a set of feasible solutions and calculate the risk assessment indicators of each feasible solution; If the risk assessment indicator exceeds the threshold, the parameter backtracking mechanism will be automatically triggered to readjust the industrial land site selection model or constraint conditions; if the risk assessment indicator does not exceed the threshold, a comparison report of each feasibility plan, a site selection proposal, a risk avoidance strategy and an emergency plan will be generated.

2. The method for intelligent site selection of industrial land based on spatial big data and artificial intelligence big model according to claim 1 is characterized in that: In step 1), the steps for processing the transportation network, industrial park boundaries, enterprise attributes, and regional economic assessment data include: 1.1) Using natural language processing algorithms, we disambiguate entities in data on transportation networks, industrial park boundaries, enterprise attributes, and regional economic assessments to identify the true, unique objects to which fuzzy entities refer. 1.2) Establish a unified spatial coordinate system and time base, and through ontology mapping technology, associate the registered address of the enterprise with the GIS geographic coordinates to build a spatiotemporal index across data sources; 1.3) Using spatial overlay analysis methods, calculate the topological accessibility of enterprise POIs and trunk roads to generate a "transportation-industry" coupling index; 1.4) Data standardization; 1.5) Use GeoHash to divide the spatial data into multiple grid cells and extract industrial park density and traffic flow values; Use Embedding technology to convert non-spatial data into vectors and extract enterprise types and regional assessment levels; 1.6) The industrial park density, traffic flow value, enterprise type, and regional assessment level are weightedly integrated through the attention mechanism to obtain a feature matrix that contains spatial topological constraints and economic association rules.

3. The method for intelligent site selection for industrial land based on spatial big data and artificial intelligence big model according to claim 2 is characterized in that: In step 1.4), for numerical data, the data normalization method is Z-Score normalization; for spatial data, the data normalization method is UTM projection conversion.

4. The method for intelligent site selection for industrial land based on spatial big data and artificial intelligence big model according to claim 1 is characterized in that: In step 2), the method for semantic recognition includes semantic role labeling and intent recognition technology.

5. The method for intelligent site selection for industrial land based on spatial big data and artificial intelligence big model according to claim 1 is characterized in that: In step 6), the steps of building a dynamic constraint rule base include: 6.1) Analyze local policies and obtain the results of local policy analysis, including the guiding requirements of local policies on industrial positioning, spatial distribution, and economic structure optimization; 6.2) Based on the policy analysis results, analyze the mandatory requirements, regulatory requirements and special control conditions of the special plan; 6.3) Delineate geological risk avoidance zones based on earthquake safety impact assessments; determine ecologically sensitive areas and pollutant emission thresholds based on environmental impact assessments; set safety production standards based on construction project safety reviews; determine extreme weather response measures based on climate feasibility studies; determine total water use and water conservation requirements based on water resource studies; and delineate cultural heritage protection areas based on cultural relic impact assessments. 6.4) Combine steps 6.2) and 6.3) to build a dynamic constraint rule library.

6. The method for intelligent site selection for industrial land based on spatial big data and artificial intelligence big model according to claim 5 is characterized in that: Mandatory planning requirements include land use function zoning and upper limit of floor area ratio; Control requirements include building setbacks and green space ratio; The special control conditions of the special plan include the ecological protection red line and the protection scope of the historical and cultural area.

7. A system based on the method for intelligent site selection for industrial land according to any one of claims 1 to 6, characterized in that: It includes data fusion layer, intelligent analysis layer, dynamic rule constraint layer, and decision output layer; The data fusion layer obtains and processes the traffic network, industrial park boundaries, enterprise attributes and regional economic assessment data to obtain a feature matrix containing spatial topological constraints and economic association rules; The intelligent analysis layer obtains the user's site selection requirements, performs semantic recognition, extracts the key analysis elements in the user's site selection requirements, and then inputs the spatial key analysis elements into the CNN convolutional neural network branch to obtain spatial features, inputs the policy text key analysis elements into the Transformer branch to obtain policy features, and inputs the industrial chain key analysis elements into the GNN graph neural network branch to obtain industrial chain features; The intelligent analysis layer uses the multimodal fusion layer to integrate spatial features, policy features, and industrial chain features to obtain multi-dimensional features, and generates industrial land analysis conclusions through knowledge graph-assisted reasoning and dynamic weight optimization algorithms; The dynamic rule constraint layer constructs a dynamic constraint rule library; The decision output layer combines the dynamic constraint rule base and the industrial land analysis conclusion to screen out a set of feasible solutions and calculate the risk assessment index of each feasible solution; If the risk assessment indicator exceeds the threshold, the parameter backtracking mechanism will be automatically triggered to readjust the industrial land site selection model or constraint conditions; if the risk assessment indicator does not exceed the threshold, a comparison report of each feasibility plan, a site selection proposal, a risk avoidance strategy and an emergency plan will be generated.

Images