Automated modeling and simulation method for urban multi-domain collaboration

Abstract

The present invention discloses an automated modeling and simulation method for urban multi-domain collaboration, which belongs to the field of model fusion technology. It solves the problem of poor accuracy of output results of the existing urban multi-domain collaborative modeling and simulation method due to the high modeling complexity; the present invention conducts research scenario analysis; determines the interactive collaborative objects, constructs a knowledge graph for polling retrieval, clusters and performs causal relationship inference on the interactive collaborative elements and their attributes that have domain interactive relationships with the modeling domain elements, and identifies high-dimensional covariates and the relationships between high-dimensional covariates; determines the modeling domain elements that need to be studied, selects their related variables, and obtains the relationships between the modeling domain elements based on the Pearson correlation coefficient between the related variables; constructs a topological model to realize multi-domain collaborative modeling and simulation. The present invention effectively improves the accuracy of the model results obtained by the cross-domain automated modeling method, and can be applied to simulate the operating status of urban multi-domain collaboration.

Description

Technical Field

[0001] The present invention relates to an automated modeling and simulation method, in particular to an automated modeling and simulation method for urban multi-domain collaboration, and belongs to the technical field of model fusion. Background Art

[0002] With the rapid development of urbanization, cities have evolved into highly complex complexes. Urban multi-domain collaborative modeling and simulation technology has emerged in response to this trend. Its core purpose is to integrate various types of information from multiple fields and to demonstrate the operating status of complex urban systems through simulation, thereby providing a scientific and reliable basis for urban planning, management and decision-making. Early urban modeling work focused on a single field, such as simulating traffic flow alone, or analyzing energy consumption only. However, with the increasing demand for urban development and the rapid advancement of simulation technology, multi-domain collaborative modeling and simulation has gradually become a popular research direction. It breaks the traditional boundaries of fields, realizes the deep integration and collaborative analysis of multi-field data, and can present the true face of the urban system in an all-round and dynamic manner.

[0003] In the existing technology, the advantages and disadvantages of some urban multi-domain collaborative modeling and simulation methods are as follows: (1) Multi-objective optimization mathematical modeling analysis method. In urban multi-domain collaborative modeling and simulation, multi-objective optimization mathematical modeling is a very valuable analysis method. First, the problem must be clearly defined, the boundaries of the urban multi-domain system must be clarified, the system input, output and key variables must be determined, and multiple interrelated and potentially conflicting goals must be determined. Then, decision variables must be selected. Different fields have different variable options. Then, the objective function must be constructed, and each goal must be quantified using mathematical expressions. At the same time, the constraints must be determined. In the solution stage, traditional algorithms or intelligent algorithms can be used to obtain non-inferior solutions. Finally, by analyzing the Pareto Frontier, sensitivity analysis and scenario analysis, in-depth understanding of system performance and changing trends; however, the model construction and solution of the mathematical modeling analysis method of multi-objective optimization are relatively complex, requiring a certain mathematical foundation and computing power. For large-scale multi-objective problems, the amount of calculation is large, the solution time is long, and there may be subjectivity and uncertainty when determining the objective function and constraints; (2) Hybrid modeling analysis method based on multiple rules. In urban multi-domain collaborative modeling and simulation, hybrid modeling based on multiple rules is a comprehensive and in-depth analysis method. First, it is necessary to have an in-depth understanding of the characteristics and operating mechanisms of various fields in the city, and transform the knowledge, experience and scientific principles of different fields into corresponding rules. Subsequently, these different types of rules are organically integrated to construct a hybrid model. In the hybrid model, different rules interact with each other to jointly simulate the operating status of the urban multi-domain system. During the analysis process, by inputting different initial conditions and parameter settings and observing the output results of the model, the impact of different strategies and plans on the urban multi-domain system can be evaluated, thereby providing scientific and reasonable solutions for urban planning and management. The decision-making basis of the rationale; however, the hybrid modeling analysis method based on multiple rules has many types of rules and high technical complexity. The system integration and data docking in different fields have a high technical threshold, and the accuracy of the hybrid model is limited by the data quality and the assumptions of the hybrid model rules; (3) Conventional system dynamics modeling and analysis methods. In urban multi-domain collaborative modeling and simulation, the use of system dynamics modeling and analysis is an effective means. First, it is necessary to sort out the complex causal relationships and feedback mechanisms in various fields of the city, determine the variables in the model, and then draw a flow chart to clearly show the flow and feedback paths between the variables and build a system dynamics model. After the system dynamics model is built, by setting different scenarios, adjusting the initial values and parameters of the variables, observing the dynamic changes of the system over time, and analyzing the changing trends of each variable, it is possible to provide forward-looking and feasible decision-making suggestions for the coordinated development of urban multi-domains; however, the modeling complexity of the system dynamics model is high, and the entire process relies on manual labor. It is difficult to quickly build a multi-system cross-domain fusion model, and the heuristic parameter setting is difficult to ensure the accuracy of the results;(4) Large model analysis. In urban multi-domain collaborative modeling and simulation, the use of large models for analysis can provide more in-depth and extensive suggestions for urban development. First, extensive urban multi-domain data is collected. Then, the multi-domain data is cleaned, labeled, and preprocessed to meet the input requirements of the large model. The large model is trained with massive data to allow the large model to learn the complex nonlinear relationships between various fields. In actual analysis, specific urban development scenario parameters are input. Finally, the output results of the large model are interpreted and verified. Combined with actual conditions, comprehensive and accurate suggestions are provided to urban planners and decision makers. However, the large model analysis method has high requirements for the input data of the large model, and training the large model requires a lot of computing resources and high costs. In addition, the internal structure of the model is complex and the interpretability is poor.

[0004] In summary, an automated modeling and simulation method for urban multi-domain collaboration is needed. Summary of the Invention

[0005] A brief overview of the present invention is provided below to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to identify key or important aspects of the present invention, nor is it intended to limit the scope of the present invention. Its purpose is simply to present certain concepts in a simplified form as a prelude to the more detailed description discussed later.

[0006] In view of this, in order to solve the problem of poor output accuracy of traditional urban multi-domain collaborative modeling and simulation methods in the prior art due to high modeling complexity, the present invention provides an automated modeling and simulation method for urban multi-domain collaboration.

[0007] The technical solution is as follows: An automated modeling and simulation method for urban multi-domain collaboration includes the following steps:

[0008] S1. Conduct research scenario analysis, i.e., generate a knowledge graph based on graph-structured retrieval enhancement and generation methods to perform retrieval and importance calculations, and identify the key elements and attributes of the research scenario.

[0009] S2. Based on the key elements and attributes of the research scenario, identify interactive and collaborative objects, construct a knowledge graph for polling retrieval, cluster and perform causal relationship inference on the interactive and collaborative elements and their attributes that have domain interaction relationships with the modeling domain elements, and identify high-dimensional covariates and the relationships between them.

[0010] S3. Determine the modeling domain elements to be studied, select their related variables, and obtain the relationships between the modeling domain elements based on the Pearson correlation coefficients between the related variables;

[0011] S4. Based on the identified high-dimensional covariates and the relationships between them, the modeling domain elements and the relationships between them, a topological model is constructed to achieve multi-domain collaborative modeling and simulation.

[0012] Furthermore, the step S1 includes the following steps:

[0013] S11. Identify the research scenario, use its relevant data as input data, and construct a knowledge graph based on the entities and relationships between entities extracted from the input data;

[0014] S12. Search the knowledge graph and calculate the importance of the retrieved entities;

[0015] S13. Determine the key elements of the research scenario, i.e., necessary nodes, based on the importance of the entity and pre-set importance criteria, and output the attributes of the necessary nodes;

[0016] In said S11, data related to the research scenario is collected from a public database, entities related to the research scenario are extracted therefrom, relationships between entities are established, relationships between entities are integrated, and a knowledge graph is constructed;

[0017] In S12, the key elements of the research scenario are used as queries to search within the knowledge graph to obtain relevant subgraphs, determine the entities corresponding to the subgraphs, i.e., nodes, and calculate the degree centrality, betweenness centrality, and closeness centrality of the nodes;

[0018] For a node i , its degree centrality The calculation formula is expressed as: ,in, For nodes i degree, that is, the degree of the node i The number of connected edges, n is the total number of nodes in the graph;

[0019] For a node i , its betweenness centrality The calculation formula is expressed as: ,in, For nodes s To Node t The number of shortest paths, For slave nodes s To Node t and passing through the node i The number of shortest paths;

[0020] For a node i , its proximity to centrality The calculation formula is expressed as: ,in, For nodes i To the node j The shortest path length;

[0021] In S13, the attributes of the necessary nodes are determined by the edges connected to the nodes and the related text descriptions.

[0022] Furthermore, the step S2 includes the following steps:

[0023] S21. Determine an interactive collaborative object, use its related data as input data, and construct a knowledge graph based on the interactive collaborative elements extracted from the input data and the relationships between the interactive collaborative elements;

[0024] S22. Using the modeling domain elements of the interactive collaborative object as a query, poll and search in the knowledge graph to obtain the interactive collaborative elements and their attributes that have a domain interaction relationship with the modeling domain elements;

[0025] S23. Clustering the interactive collaborative elements and their attributes that have domain interaction relationships with the modeling domain elements based on cosine similarity, clustering elements with similar relationships to form high-dimensional covariates;

[0026] S24. For interactive collaborative elements with logical relationships between attributes and domain interaction relationships with modeling domain elements, perform causal relationship inference based on Bayesian networks to determine the logical relationships between elements, cluster elements with causal logical relationships, and form high-dimensional covariates;

[0027] In said S21, data related to the interactive collaborative object is collected from a public database, interactive collaborative elements related to the interactive collaborative object are extracted therefrom, and the interactive collaborative elements and the relationships between the interactive collaborative elements are integrated to construct a knowledge graph;

[0028] In the step S23, the attribute vector of the interactive collaborative element having a domain interaction relationship with the modeling domain element is calculated, that is, the first attribute vector and the second attribute vector , perform numerical processing, further calculate the similarity of the results after numerical processing, and obtain the cosine similarity to complete the clustering. The calculation formula is expressed as ,in, is the element of the first attribute vector, is the element of the second attribute vector;

[0029] In the above S24, in the Bayesian network, for the interactive collaborative elements with logical relationships between attributes and domain interaction relationships between modeling domain elements, that is, the first variable X and the second variable Y, the joint probability distribution of the two is known to be P(X, Y), and the Bayesian result is , determine whether there is a causal relationship between the first variable X and the second variable Y. If the Bayesian result P(X|Y) is different from the first variable P(X) and meets the causal inference conditions, then it is considered that the second variable Y has a causal influence on the first variable X.

[0030] Furthermore, in S3, the modeling domain elements to be studied, i.e., the first modeling domain element and the second modeling domain element, are determined, and the related variables of the first modeling domain element and the second modeling domain element, i.e., the first variable X and the second variable Y, are selected, and the relationship between the modeling domain elements is obtained based on the Pearson correlation coefficient between the two.

[0031] Pearson correlation coefficient Expressed as:

[0032] ;

[0033] in, is the Pearson correlation coefficient, is the number of data points, is the first variable X observations, The second variable Y observations, is the mean of the first variable X, is the mean of the second variable Y;

[0034] The value range of the Pearson correlation coefficient is [-1, 1]. When , it means that the two variables are completely positively correlated. When , it means that the two variables are completely negatively correlated. , it means there is no linear correlation between the two variables.

[0035] Furthermore, in the above S4, in the process of constructing a topological model based on the modeling domain elements and the relationships between them, for the internal domain, the topological principle is applied, the modeling domain elements are taken as nodes, the relationships between the modeling domain elements are taken as edges, and the internal domain topological model is constructed. For cross-domain research scenarios, high-dimensional covariates are taken as nodes, and the multiple relationships associated with high-dimensional covariates are taken as edges. The networks of multiple domains are connected to form a complex cross-domain topological structure, and finally a topological model is obtained. , , where, for the topology model, is a collection of nodes, is a collection of edges, for the internal topology model of the domain, is a collection of modeling domain elements. It is a collection of feature relationships, targeting complex topological structures across domains. is a set of high-dimensional covariates, is a collection of multiple relationships between high-dimensional covariates.

[0036] The beneficial effects of the present invention are as follows: the present invention is not implemented by a single model, but a set of comprehensive methods formed by means of scene analysis, establishment of high-dimensional covariates, etc., and finally outputs a set of network topology simulation models, which can be applied to the field of collaborative simulation modeling of urban multi-domain systems; when analyzing key elements contained in the scene, element relationships or associations between fields, the present invention is no longer based on tedious manual operations, but is implemented through programs and algorithms, and humans only participate in correcting the output results. The whole process reduces the cost of manual participation and modeling deviations caused by subjective consciousness, and solves the problems of complex fusion model construction and high degree of manual dependence; when determining the research object, the present invention confirms the specific elements contained in the modeling research object through scene analysis, limits the boundaries of the research, and prevents excessive conjectures and assumptions from affecting the accuracy of the model. The problem of difficult collaborative integration of systems in different fields is solved; the present invention establishes high-dimensional covariates as the main mode of cross-domain collaboration, which can not only comprehensively analyze the various modes of cross-domain collaboration in multiple fields, but also clearly see the specific elements and degree of mutual influence that occur in cross-domain collaborative feedback, thereby improving the interpretability of the model; the present invention provides a full set of analytical processes: scenario analysis limits the method of solving the model and determines the elements included in the model, high-dimensional covariates identify and confirm the channels for collaboration in different fields, and multi-field model evolution model is jointly established to achieve fusion deduction; the present invention can realize automated modeling, which includes automated identification of model elements, high-dimensional covariates and logical interaction relationships (qualitative and quantitative) between elements; the present invention can perform cross-domain collaborative feedback through covariates, that is, connect different field models through multi-dimensional correlation relationships, and uniformly transmit the mutual influence between field models through covariates. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0038] Figure 1 Flowchart of the automated modeling and simulation method for urban multi-domain collaboration. DETAILED DESCRIPTION

[0039] To make the technical solutions and advantages of the embodiments of the present invention more clearly understood, exemplary embodiments of the present invention are further described in detail below with reference to the accompanying drawings. It should be noted that the embodiments described are only a portion of the embodiments of the present invention, and are not an exhaustive list of all embodiments. It should be noted that the embodiments of the present invention and the features thereof may be combined with each other unless they conflict.

[0040] refer to Figure 1 The present embodiment provides a detailed description of the automated modeling and simulation method for urban multi-domain collaboration, specifically including the following steps:

[0041] S1. Conduct research scenario analysis, i.e., generate a knowledge graph based on graph-structured retrieval enhancement and generation methods to perform retrieval and importance calculations, and identify the key elements and attributes of the research scenario.

[0042] S2. Based on the key elements and attributes of the research scenario, identify interactive and collaborative objects, construct a knowledge graph for polling retrieval, cluster and perform causal relationship inference on the interactive and collaborative elements and their attributes that have domain interaction relationships with the modeling domain elements, and identify high-dimensional covariates and the relationships between them.

[0043] S3. Determine the modeling domain elements to be studied, select their related variables, and obtain the relationships between the modeling domain elements based on the Pearson correlation coefficients between the related variables;

[0044] S4. Based on the identified high-dimensional covariates and the relationships between them, the modeling domain elements and the relationships between them, a topological model is constructed to achieve multi-domain collaborative modeling and simulation.

[0045] Furthermore, the step S1 includes the following steps:

[0046] S11. Identify the research scenario, use its relevant data as input data, and construct a knowledge graph based on the entities and relationships between entities extracted from the input data;

[0047] S12. Search the knowledge graph and calculate the importance of the retrieved entities;

[0048] S13. Determine the key elements of the research scenario, i.e., necessary nodes, based on the importance of the entity and pre-set importance criteria, and output the attributes of the necessary nodes;

[0049] In said S11, data related to the research scenario is collected from a public database, entities related to the research scenario are extracted therefrom, relationships between entities are established, relationships between entities are integrated, and a knowledge graph is constructed;

[0050] In S12, the key elements of the research scenario are used as queries to search within the knowledge graph to obtain relevant subgraphs, determine the entities corresponding to the subgraphs, i.e., nodes, and calculate the degree centrality, betweenness centrality, and closeness centrality of the nodes;

[0051] For a node i , its degree centrality The calculation formula is expressed as: ,in, For nodes i degree, that is, the degree of the node i The number of connected edges, n is the total number of nodes in the graph;

[0052] For a node i , its betweenness centrality The calculation formula is expressed as: ,in, For nodes s To Node t The number of shortest paths, For slave nodes s To Node t and passing through the node i The number of shortest paths;

[0053] For a node i , its proximity to centrality The calculation formula is expressed as: ,in, For nodes i To the node j The shortest path length;

[0054] In S13, the attributes of the necessary nodes are determined by the edges connected to the nodes and the related text descriptions.

[0055] Specifically, when the research scenario is determined, the urban areas that need to be analyzed are also identified. Using papers, research models, policy documents, and reports on this field as input, a graph-based retrieval enhancement generation method is used to perform a global query on the input data. Based on criteria such as importance, feasibility, security, and privacy, the method identifies the associated words that are closely related to the research content as necessary elements for the scenario to run, and simultaneously outputs the necessary attributes of these elements.

[0056] The process of determining the research scenario involves the graph-based retrieval enhancement generation method, associated word identification and element determination. The specific principles are as follows: (1) The principle of the graph-based retrieval enhancement generation method is to construct a knowledge graph, taking various entities (such as concepts, terms, objects, etc.) in the input data (i.e., papers, research models, policy documents, reports, etc. in public databases) as nodes, and the semantic relationships between entities (such as causal relationships, subordinate relationships, association relationships, etc.) as edges; through the above-mentioned graph structure, the intrinsic connections between data can be captured more comprehensively, compared with the traditional text matching-based retrieval method. (1) The knowledge graph can better understand semantics and improve the accuracy and relevance of retrieval. (2) For the identification of associated words and the determination of elements, the principle is to measure the importance of nodes (i.e. entities) in the retrieved related subgraphs according to indicators such as the degree of the node (the number of edges connected to the node) and the centrality of the node in the graph (such as betweenness centrality, closeness centrality, etc.). Nodes with higher degrees and stronger centrality are often more critical in the entire knowledge system and are more likely to be necessary elements for the operation of the scene. For each identified element, its attributes are determined by the type of edges connected to the node and the relevant text description.

[0057] The importance measurement of the node takes into account three indicators: degree centrality, betweenness centrality and closeness centrality. The specific measurement principles are as follows: (1) degree centrality, which reflects the activity of the node in the local network. The higher the degree, the more extensive the direct connection between the node and other nodes, and the more important it may be in the local scope; (2) betweenness centrality, which measures the bridge role of the node in the entire network. The higher the betweenness centrality of a node, the more frequently the node appears in the shortest path connecting other nodes, which has an important impact on the connectivity and information transmission of the network; (3) closeness centrality, which reflects the average distance between the node and all other nodes. The higher the closeness centrality, the shorter the average distance between the node and other nodes, which has advantages in information dissemination and other aspects, and also reflects its importance in the network;

[0058] In this embodiment, the research scenario is the bus priority development strategy in the urban intelligent transportation system. The specific scenario analysis process is as follows:

[0059] In S11, a series of papers, policy documents and reports on urban transportation, bus priority policy, intelligent transportation technology, etc. are collected from public databases. For example, one paper mentioned that "the intelligent bus dispatching system optimizes the departure frequency by real-time monitoring of bus locations and combining passenger flow data." From this, entities such as "intelligent bus dispatching system", "bus location monitoring", "passenger flow data", and "departure frequency optimization" are extracted, and relationships between entities are established. For example, "intelligent bus dispatching system" uses "bus location monitoring" and "passenger flow data" to achieve "departure frequency optimization". These entities and relationships are constructed into a knowledge graph.

[0060] In S12, the query "key elements of the bus priority development strategy" is used to search the knowledge graph to find relevant subgraphs. Assuming that this subgraph contains "bus lane node", "intelligent bus dispatch system node", "bus subsidy policy node", etc., the degree centrality, betweenness centrality and closeness centrality of the above nodes are calculated;

[0061] (1) Degree centrality calculation: Assume that the “bus lane node” is connected to the “traffic congestion relief node”, “bus operation efficiency improvement node”, and “bus priority right of way guarantee node”. The total number of nodes n in the graph is 10, then the degree of the “bus lane node” is 3, and its degree centrality is ;

[0062] (2) Calculation of betweenness centrality: Assume that there are five shortest paths from the "traffic congestion relief node" to the "urban sustainable development node", of which three pass through the "bus lane node", and there are four shortest paths from the "bus operation efficiency improvement node" to the "citizen travel satisfaction improvement node", of which two pass through the "bus lane node". At the same time, assume that there are other paths between node pairs. By calculating the ratio of the number of shortest paths between all node pairs passing through the "bus lane node" to the total number of shortest paths and summing them up, the betweenness centrality of the "bus lane node" is obtained. , after calculation, ;

[0063] (3) Calculation of closeness centrality: Assuming that the sum of the shortest path lengths from the “bus lane node” to the other nine nodes is 15 by calculating and accumulating the lengths of each shortest path in the knowledge graph, the closeness centrality of the “bus lane node” is obtained. , ;

[0064] The importance score of each node is calculated based on the weighted average of degree centrality, betweenness centrality, and closeness centrality. Assuming that the weights of degree centrality, betweenness centrality, and closeness centrality are 0.3, 0.4, and 0.3 respectively, the importance score of the "bus lane node" is obtained. , ;

[0065] In S13, by comparing the importance scores of each node, it is determined that the "bus lane node", "intelligent bus dispatching system node", "bus subsidy policy", etc. are necessary elements when the scenario is running. The "bus lane node" among the necessary elements is given as an example, and its attributes are determined by the connected edges and related text descriptions. For example, the edge from the "bus lane node" to the "traffic congestion relief node" indicates that it has the function of alleviating traffic congestion, and its construction standard attributes (such as width, setting location, etc.) can be known from the relevant public database.

[0066] Furthermore, the step S2 includes the following steps:

[0067] S21. Determine an interactive collaborative object, use its related data as input data, and construct a knowledge graph based on the interactive collaborative elements extracted from the input data and the relationships between the interactive collaborative elements;

[0068] S22. Using the modeling domain elements of the interactive collaborative object as a query, poll and search in the knowledge graph to obtain the interactive collaborative elements and their attributes that have a domain interaction relationship with the modeling domain elements;

[0069] S23. Clustering the interactive collaborative elements and their attributes that have domain interaction relationships with the modeling domain elements based on cosine similarity, clustering elements with similar relationships to form high-dimensional covariates;

[0070] S24. For interactive collaborative elements with logical relationships between attributes and domain interaction relationships with modeling domain elements, perform causal relationship inference based on Bayesian networks to determine the logical relationships between elements, cluster elements with causal logical relationships, and form high-dimensional covariates;

[0071] In said S21, data related to the interactive collaborative object is collected from a public database, interactive collaborative elements related to the interactive collaborative object are extracted therefrom, and the interactive collaborative elements and the relationships between the interactive collaborative elements are integrated to construct a knowledge graph;

[0072] In the step S23, the attribute vector of the interactive collaborative element having a domain interaction relationship with the modeling domain element is calculated, that is, the first attribute vector and the second attribute vector , perform numerical processing, further calculate the similarity of the results after numerical processing, and obtain the cosine similarity to complete the clustering. The calculation formula is expressed as ,in, is the element of the first attribute vector, The element vector is composed of various attribute values of the elements. By calculating the cosine similarity, the similarity between two elements in the attribute space can be measured. The higher the similarity, the more similar the two elements are, and the more likely they are to be clustered into the same class.

[0073] In the above S24, in the Bayesian network, for the interactive collaborative elements with logical relationships between attributes and domain interaction relationships between modeling domain elements, that is, the first variable X and the second variable Y, the joint probability distribution of the two is known to be P(X, Y), and the Bayesian result is , to determine whether there is a causal relationship between the first variable X and the second variable Y. If the Bayesian result P(X|Y) is significantly different from the first variable P(X) and meets the causal inference conditions (such as the causal Markov condition, etc.), then it is considered that the second variable Y has a causal influence on the first variable X.

[0074] Specifically, in the process of identifying high-dimensional covariates, we use papers and related models related to inter-domain interaction and collaboration in public databases as input, and use a graph-based retrieval enhancement generation method to poll all elements contained in the modeling domain obtained through scenario analysis, identify the elements required for domain interaction, and refine them to corresponding attributes; based on the element attributes, we cluster similar elements or elements with logically related attributes to form a high-dimensional covariate;

[0075] The identification of high-dimensional covariates mainly involves two methods, namely, a retrieval enhancement generation method based on a graph structure and a high-dimensional covariate formation by element clustering; wherein, the principle of the retrieval enhancement generation method based on a graph structure in step S2 is basically the same as that of step S1, but the retrieval process involves polling of the knowledge graph, and the principle of the process is to first pre-process the input data, extract the elements and relationships therein, and construct a knowledge graph, and then, for each element in the modeling domain determined by the scene analysis, perform a polling search in the knowledge graph, and by traversing the nodes and edges, find other elements that have domain interaction relationships with the element, and extract their relevant attributes; for element clustering to form high-dimensional covariates, the principle is to use the retrieved element attributes as the basis, and use a clustering algorithm to classify elements with similar attributes or logical relationships (such as causal relationships, complementary relationships, etc.) between attributes into one category, and each set of elements forms a high-dimensional covariate, the purpose of which is to integrate scattered and complex elements so as to more effectively analyze the collaborative relationship between domains;

[0076] In this embodiment, it is assumed that the interactive collaboration objects studied are urban transportation and energy fields;

[0077] In S21, papers, models, policy documents and other materials on urban transportation and energy are collected. For example, a paper mentions that "the popularization of electric vehicles depends on the improvement of charging pile infrastructure, and the use of electric vehicles will affect the distribution of urban traffic flow." From this sentence, interactive and collaborative elements such as "electric vehicles," "charging pile infrastructure," and "urban traffic flow distribution" are extracted, and the relationships between the interactive and collaborative elements are established. For example, there is a dependency relationship between "electric vehicles" and "charging pile infrastructure," and there is an influence relationship between "electric vehicles" and "urban traffic flow distribution," and a knowledge graph is constructed.

[0078] In said S22, it is assumed that the determined modeling domain elements include "transportation tools", "energy supply facilities", "traffic flow indicators", etc. Taking "transportation tools" as an example, polling and searching in the knowledge graph are performed to find elements that have a domain interaction relationship with "transportation tools", such as "electric vehicles", whose attributes include cruising range, charging time, energy consumption type, etc., "fuel vehicles", whose attributes include fuel type, fuel consumption per 100 kilometers, etc., "hybrid vehicles", whose attributes include attributes of "electric vehicles" and "fuel vehicles", so the attribute categories of transportation tools are cruising range, charging time, energy consumption type, fuel type, fuel consumption per 100 kilometers, and fuel tank capacity;

[0079] In S23, the attribute vectors of “electric vehicle” and “hybrid vehicle” are calculated. Assume that the attribute vector of “electric vehicle” is , which means that the cruising range is 400 kilometers, the charging time is 1 hour, the energy consumption type is electricity, and there are no fuel-related attributes (missing attributes are filled with zeros). The attribute vector of "hybrid electric vehicle" is , which represents a range of 500 kilometers, a refueling or charging time of 0.5 hours, an energy consumption type of hybrid electric, a fuel type of gasoline, a fuel consumption of 5 liters per 100 kilometers, and a fuel tank capacity of 40 liters. The attribute vector is digitized (for example, the energy consumption type is encoded as 1 for electricity, 2 for gasoline, and 3 for hybrid electric), and the cosine similarity of the digitized attribute vector is calculated;

[0080] Cosine similarity of the attribute vectors of "electric vehicle" and "hybrid vehicle" after numerical processing Expressed as:

[0081]

[0082]

[0083]

[0084] in, Elements of "electric vehicles", Elements of a “hybrid vehicle”;

[0085] Due to the high similarity, “electric vehicles” and “hybrid vehicles” can be clustered into the same category as a high-dimensional covariate about “new means of transportation”;

[0086] In S24, the relationship between “charging pile infrastructure” and “electric vehicle ownership” is analyzed using a Bayesian network, assuming that the joint probability distribution is obtained by collecting data. 、 、 、 ;

[0087] According to the Bayesian formula, the Bayesian result is calculated (In actual calculations, the probability value should be between 0 and 1. The data assumed here is only for illustration of the calculation process.) Bayesian results The probability of improving the charging pile infrastructure There are significant differences, indicating that there is a causal relationship between the two. "Charging pile infrastructure" and "electric vehicle ownership" can be clustered together to form a high-dimensional covariate on "electric vehicle energy supply and development".

[0088] Furthermore, in S3, the modeling domain elements to be studied, i.e., the first modeling domain element and the second modeling domain element, are determined, and the related variables of the first modeling domain element and the second modeling domain element, i.e., the first variable X and the second variable Y, are selected, and the relationship between the modeling domain elements is obtained based on the Pearson correlation coefficient between the two.

[0089] Pearson correlation coefficient Expressed as:

[0090] ;

[0091] in, is the Pearson correlation coefficient, is the number of data points, is the first variable X observations, The second variable Y observations, is the mean of the first variable X, is the mean of the second variable Y;

[0092] The value range of the Pearson correlation coefficient is [-1, 1]. When , it means that the two variables are completely positively correlated. When , it means that the two variables are completely negatively correlated. , it means there is no linear correlation between the two variables.

[0093] Specifically, scenario analysis and high-dimensional covariate identification can qualitatively determine the elements with logical causal relationships. In the process of identifying element relationships, the correlation analysis method in statistics is used to quantitatively calculate the correlation coefficients between different elements, and the linear relationship between the elements can be obtained.

[0094] Correlation analysis is a statistical method used to measure the strength and direction of the linear relationship between variables. It collects a large amount of data related to each factor and calculates the correlation coefficient to accurately represent the degree of association between the factors.

[0095] In this example, assuming that the relationship between urban traffic flow and air quality is studied, the daily traffic flow data (the number of vehicles passing through the selected area, unit: vehicle) of a certain area in the city over a period of time (e.g., one month, a total of 30 days) is selected as the first variable X, and the daily air quality index (AQI) is selected as the second variable Y. Based on the data related to each factor in Table 1, the Pearson correlation coefficient is obtained. ;

[0096]

[0097] Table 1

[0098] Pearson correlation coefficient The calculation process is expressed as:

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105] From this we can see that there is a certain positive correlation between traffic flow and the air quality index, but the correlation is not very strong, which means that as traffic flow increases, the air quality index has a certain degree of upward trend, but there are other factors that also affect air quality.

[0106] Furthermore, in the above S4, in the process of constructing the topological model based on the modeling domain elements and the relationships between them, for the internal domain, the topological principle is applied, the modeling domain elements are used as nodes, and the relationships between the modeling domain elements are used as edges to construct the internal domain topological model. For cross-domain research scenarios, high-dimensional covariates are used as nodes, and the multiple relationships (similar relationships and causal logical relationships) associated with high-dimensional covariates are used as edges to connect the networks of multiple domains to form a complex cross-domain topological structure, and finally a topological model is obtained. , , where, for the topology model, is a set of nodes (vertices), is a collection of edges, for the internal topology model of the domain, is a collection of modeling domain elements. It is a collection of feature relationships, targeting complex topological structures across domains. is a set of high-dimensional covariates, is a collection of multiple relationships between high-dimensional covariates.

[0107] Specifically, in step S4, the main task is to build a topology graph, and the final output is also a topology model composed of nodes and edges;

[0108] In this example, taking the urban transportation, energy, and environment fields as examples, the identified high-dimensional covariates are: "electric vehicle ownership" (related to the transportation and energy fields), "charging pile coverage" (energy field), "air quality index" (environment field), and "public transportation travel ratio" (transportation field);

[0109] The relationships between covariates include: there is a positive correlation between "electric vehicle ownership" and "charging pile coverage" (the higher the charging pile coverage, the more conducive it is to the ownership and use of electric vehicles); there is a negative correlation between "electric vehicle ownership" and "air quality index" (increased use of electric vehicles can reduce exhaust emissions and improve air quality); there is a certain substitution relationship between "proportion of public transportation travel" and "electric vehicle ownership" (when the proportion of public transportation travel is high, the demand for electric vehicles may be relatively reduced).

[0110] Build complex topologies across domains :

[0111]

[0112]

[0113] In addition to the above steps, the topology model can also be subjected to sensitivity analysis, correctness verification and other processing to further verify the performance of the topology model.

[0114] Although the present invention has been described with respect to a limited number of embodiments, it will be apparent to those skilled in the art, having benefit of the foregoing description, that other embodiments are contemplated within the scope of the invention thus described. Furthermore, it should be noted that the language used in this specification has been selected primarily for readability and didactic purposes, rather than for the purpose of explaining or limiting the subject matter of the present invention. Consequently, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the present invention is intended to be illustrative rather than restrictive of the scope of the invention, which is defined by the appended claims.

Claims

1. An automated modeling and simulation method for urban multi-domain collaboration, characterized by: The following steps are involved: S1. Conduct research scenario analysis, i.e., generate a knowledge graph based on graph-structured retrieval enhancement and generation methods to perform retrieval and importance calculations, and identify the key elements and attributes of the research scenario. S2. Based on the key elements and attributes of the research scenario, identify interactive and collaborative objects, construct a knowledge graph for polling retrieval, cluster and perform causal relationship inference on the interactive and collaborative elements and their attributes that have domain interaction relationships with the modeling domain elements, and identify high-dimensional covariates and the relationships between them. S3. Determine the modeling domain elements to be studied, select their related variables, and obtain the relationships between the modeling domain elements based on the Pearson correlation coefficients between the related variables; S4. Based on the identified high-dimensional covariates and the relationships between them, the modeling domain elements and the relationships between them, a topological model is constructed to achieve multi-domain collaborative modeling and simulation. Said S2 comprises the following steps: S21. Determine an interactive collaborative object, use its related data as input data, and construct a knowledge graph based on the interactive collaborative elements extracted from the input data and the relationships between the interactive collaborative elements; S22. Using the modeling domain elements of the interactive collaborative object as a query, poll and search in the knowledge graph to obtain the interactive collaborative elements and their attributes that have a domain interaction relationship with the modeling domain elements; S23. Clustering the interactive collaborative elements and their attributes that have domain interaction relationships with the modeling domain elements based on cosine similarity, clustering elements with similar relationships to form high-dimensional covariates; S24. For interactive collaborative elements with logical relationships between attributes and domain interaction relationships with modeling domain elements, perform causal relationship inference based on Bayesian networks to determine the logical relationships between elements, cluster elements with causal logical relationships, and form high-dimensional covariates; In said S21, data related to the interactive collaborative object is collected from a public database, interactive collaborative elements related to the interactive collaborative object are extracted therefrom, and the interactive collaborative elements and the relationships between the interactive collaborative elements are integrated to construct a knowledge graph; In the step S23, the attribute vector of the interactive collaborative element having a domain interaction relationship with the modeling domain element is calculated, that is, the first attribute vector and the second attribute vector Perform numerical processing, further calculate the similarity of the results after numerical processing, and obtain the cosine similarity to complete the clustering. The calculation formula is expressed as Among them, a i ′ is the element of the first attribute vector, b i ′ is the element of the second attribute vector; In S24, in the Bayesian network, for interactive collaborative elements that have a logical relationship between attributes and a domain interaction relationship with modeling domain elements, that is, a first variable X and a second variable Y, the joint probability distribution of the two is known, and the Bayesian result is used to determine whether there is a causal relationship between the first variable X and the second variable Y. If the Bayesian result is different from the first variable and meets the causal inference conditions, it is considered that the second variable Y has a causal influence on the first variable X.

2. The automated modeling and simulation method for urban multi-domain collaboration according to claim 1 is characterized in that: Said S1 comprises the following steps: S11. Identify the research scenario, use its relevant data as input data, and construct a knowledge graph based on the entities and relationships between entities extracted from the input data; S12. Search the knowledge graph and calculate the importance of the retrieved entities; S13. Determine the key elements of the research scenario, i.e., necessary nodes, based on the importance of the entity and pre-set importance criteria, and output the attributes of the necessary nodes; In said S11, data related to the research scenario is collected from a public database, entities related to the research scenario are extracted therefrom, relationships between entities are established, relationships between entities are integrated, and a knowledge graph is constructed; In S12, the key elements of the research scenario are used as queries to search within the knowledge graph to obtain relevant subgraphs, determine the entities corresponding to the subgraphs, i.e., nodes, and calculate the degree centrality, betweenness centrality, and closeness centrality of the nodes; For a node i, its degree centrality C D The calculation formula of (i) is expressed as: Where d(i) is the degree of node i, that is, the number of edges connected to node i, and n is the total number of nodes in the graph; For a node i, its betweenness centrality C B The calculation formula of (i) is expressed as: Among them, σ st is the number of shortest paths from node s to node t, σ st (i) is the number of shortest paths from node s to node t that passes through node i; For a node i, its proximity centrality C C The calculation formula of (i) is expressed as: Where d(i,j) is the shortest path length from node i to node j; In S13, the attributes of the necessary nodes are determined by the edges connected to the nodes and the related text descriptions.

3. The automated modeling and simulation method for urban multi-domain collaboration according to claim 2 is characterized in that: In S3, the modeling domain elements to be studied, i.e., the first modeling domain element and the second modeling domain element, are determined, and the related variables of the first modeling domain element and the second modeling domain element, i.e., the first variable X and the second variable Y, are selected, and the relationship between the modeling domain elements is obtained based on the Pearson correlation coefficient between the two. The Pearson correlation coefficient r is expressed as: Where r is the Pearson correlation coefficient, m is the number of data points, and x i is the i-th observation value of the first variable X, y i The i-th observation value of the second variable Y, is the mean of the first variable X, is the mean of the second variable Y; The value range of the Pearson correlation coefficient is [-1, 1]. When r = 1, it means that the two variables are completely positively correlated. When r = -1, it means that the two variables are completely negatively correlated. When r = 0, it means that there is no linear correlation between the two variables.

4. The automated modeling and simulation method for urban multi-domain collaboration according to claim 3 is characterized in that: In the above S4, in the process of constructing a topological model based on the modeling domain elements and the relationships between them, for the internal domain, the topological principles are applied, and the modeling domain elements are used as nodes and the relationships between the modeling domain elements are used as edges to construct an internal domain topological model. For cross-domain research scenarios, high-dimensional covariates are used as nodes and the multiple relationships associated with the high-dimensional covariates are used as edges. The networks of multiple domains are connected to form a complex cross-domain topological structure, and finally a topological model G is obtained, G = (V, E), where, for the topological model, V is a set of nodes and E is a set of edges. For the internal domain topological model, V is a set of modeling domain elements and E is a set of element relationships. For the complex cross-domain topological structure, V is a set of high-dimensional covariates and E is a set of multiple relationships between high-dimensional covariates.

Images