Urban territorial space multi-element coupling intelligent optimization method and system

Abstract

The invention discloses an urban territorial space multi-element coupling intelligent optimization method and system, and the method comprises the steps: quantifying the functional partition transition probability of an ecological protection region, a basic farmland and an urban development region based on a Markov model, and simulating the spatial distribution of functional partitions through a future land utilization simulation model, constructing a first multi-target optimization model in combination with resistance surface analysis of the minimum cumulative resistance model; collecting urban POI data, screening residential, commercial and public service types as optimization objects, dividing the optimization objects by utilizing clustering analysis, extracting core points, and constructing a second multi-target optimization model; and constructing graph structure data based on the road network, defining node and edge features, training the graph structure of the road network by using a graph neural network, extracting the features, and constructing a third multi-target optimization model to make a multi-target optimization decision of the road network.

Description

Technical Field

[0001] The present invention relates to the field of optimization methods, and in particular to an intelligent optimization method and system for multi-factor coupling of urban land space. Background Art

[0002] Land resources are an important carrier of industrialization, urbanization and agricultural modernization. With the rapid development of urbanization, urban expansion has led to a reduction in arable land, ecological fragmentation, and increasingly obvious problems such as a high proportion of existing land but low utilization rate. Optimizing urban land space is an inevitable choice to cope with complex problems such as resource shortages, ecological degradation, and spatial conflicts in the process of urbanization.

[0003] In existing technologies, when optimizing national land space resources, land, ecology, economy and other factors are evaluated independently, and the coupling effect is ignored, resulting in ecological corridors being cut by roads, industrial layout and job-housing balance being unbalanced, and unable to meet the development needs of the urbanization process. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that when optimizing land space resources, factors such as land, ecology, and economy are evaluated independently, and the coupling effect is ignored, resulting in ecological corridors being cut by roads, industrial layout and job-housing balance being unbalanced, and unable to meet the development needs of the urbanization process. A method and system for intelligent optimization of multi-factor coupling of urban land space are provided.

[0005] The present invention solves the above-mentioned technical problems through the following technical solutions, which include:

[0006] Step 1: Quantify the functional zoning transfer probability of ecological protection areas, basic farmland, and urban development areas based on the Markov model. Use the future land use simulation model to simulate the spatial distribution of functional zones. Combined with the resistance surface analysis of the minimum cumulative resistance model, the first multi-objective optimization model is constructed. Ecological resilience, urban spatial resilience, and basic farmland protection and intensive development are determined as optimization objectives. Using geographic cellular automata and neural network algorithms, the optimization objective function and constraints are embedded. The evolution trend of urban spatial pattern is simulated, the dynamic change pattern is analyzed, and the spatial distribution of functional zones is optimized.

[0007] Step 2: Collect urban POI data and select residential, commercial, and public service types as optimization targets. Use cluster analysis to divide the optimization targets and extract core points. Build a second multi-objective optimization model to quantify the spatial relationships and traffic accessibility between nodes. Calculate and optimize the spatial density, network resilience, and compatibility of core points with future urban development areas. Use the NSGA-II algorithm to solve the optimization plan, analyze the spatial layout characteristics of core points, and optimize the spatial distribution of residential, commercial, and public service core points.

[0008] Step 3: Build graph structure data based on the road network, define node and edge features, use graph neural network to train the graph structure of the road network and extract features, build a third multi-objective optimization model, quantify the intensiveness and resilience of the road network, and embed the dynamically updated edge features of the third multi-objective optimization model into the message passing mechanism of the graph neural network to capture global topological characteristics to formulate multi-objective optimization decisions for the road network. Then return to step 1 for loop iterative optimization.

[0009] The decision includes at least one of adding new efficient connecting roads, expanding highly intensive roads, and deleting inefficient and redundant roads.

[0010] Optionally, the minimum cumulative resistance model includes:

[0011] C M =f∑D mn R m

[0012] Among them, C M It is a function of the minimum resistance of any grid center within the range and its distance to all sources; f is the negative correlation between the minimum cumulative resistance and its distance to all sources; D mn R is the distance from the source n to a grid in space through space m; m represents the resistance of species to cross the landscape space m;

[0013] The Markov model includes:

[0014] S (t+1) =P ab ·S (t)

[0015] Among them, S (t) ,S (t+1) is the land space type state matrix of the study area at time t and time t+1; P ab Represents the transition probability matrix from type a to type b;

[0016] The future land use simulation model simulates the spatial distribution of functional zones, including:

[0017] A neural network algorithm is used to obtain the suitability probability of each land use type within the study area from the base year national land spatial pattern data and multiple driving force factors including human and natural effects. The suitability probability is combined with the neighborhood factor, adaptive inertia coefficient, and conversion cost to obtain the overall conversion probability of each cell. The simulation results are obtained through a roulette competition mechanism.

[0018] The neural network algorithm includes:

[0019]

[0020] Where sp(p,i,t) is the suitability probability of type i space under grid p at time t; w j,i Represents the weight between the output layer and the hidden layer; sigmoid() is the excitation function from the hidden layer to the output layer; net j (p,t) represents the signal received by the j-th hidden layer grid p at time t.

[0021] Optionally, the ecological resilience includes the connectivity and compactness of the ecological protection zone. The spread and cohesion of the ecological protection zone are calculated to represent the ecological connectivity and compactness. The calculation method includes:

[0022]

[0023] Among them, CON is the spread index of the ecological protection zone, which represents the connectivity of the ecological protection zone; m is the total number of functional zoning types; g k The number of adjacent patches within the ecological protection area or between the ecological protection area and other types of functional zones;

[0024]

[0025] Among them, COH e is the cohesion index of the ecological reserve, which characterizes the compactness of the ecological reserve; P is the total area of the patches in the ecological reserve; p is the total boundary length of all patches in the ecological reserve; a is the total area of all patches in the ecological reserve; A is the total area of all functional zoning patches.

[0026] Optionally, the urban development zone is a source landscape type, and the ecological protection zone and basic farmland are sink landscape types. The urban spatial resilience is measured by calculating the source-sink landscape pattern index. The calculation process includes:

[0027]

[0028] Where: D is the average distance index of source landscape patches; d ij represents the distance between grid i in the source landscape patch and grid j in the sink landscape patch; m and n represent the number of grids in the source landscape patch and the number of grids in the sink landscape patch, respectively.

[0029] Optionally, the intensive development is characterized by calculating the cohesion of basic farmland and urban development areas, and the calculation method includes:

[0030]

[0031] Among them, COH c and COH uare the cohesion indexes of basic farmland and urban development zones, respectively, which characterize the degree of agglomeration of the two. P is the total area of the corresponding functional zone patches, p is the total boundary length of all patches of the corresponding functional zone, a is the total area of all patches of the corresponding functional zone, and A is the total area of all functional zone patches.

[0032] Optionally, the constraints include:

[0033]

[0034] Among them, A e 、A c 、A u They refer to the optimized ecological protection area, basic farmland and urban development area respectively. e,min 、A c,min 、A u,max The red line control limits for ecological protection zones, basic farmland and urban development zones in the national land space planning respectively.

[0035] Optionally, the calculation and optimization of the spatial density, network resilience, and compatibility of the core points with future urban development areas includes:

[0036]

[0037] in, is the average shortest geometric distance between all core points, which is used to measure spatial compactness; N is the total number of core points; d ij is the shortest geometric distance between the i-th and j-th core points.

[0038] Optionally, the using of the NSGA-II algorithm to solve the optimization solution includes:

[0039]

[0040] Population = {Position p ,Position c ,Position s}

[0041] Among them, Position p ,Position c ,Postition s are the spatial location matrices of residential core points, commercial core points, and public service core points, respectively. n,m is a specific spatial unit. If the core point is arranged in this unit, it is set to 1, otherwise it is set to 0. n and m respectively determine the spatial range where the core point can be arranged.

[0042] Optionally, the quantifying the compactness and resilience of the road network includes:

[0043]

[0044] Among them, S compact (e ij ) is edge e ij The intensiveness score of is used to indicate the intensiveness level of the edge; For edge e ij The average kernel density of the core points passing through the region represents the edge e ij The demand for traffic through the core point of the space; R(e ij ) represents edge e ij The geometric length of , which is used to represent the resource consumption of the edge;

[0045] The road network resilience is mainly characterized by maximizing the connectivity and adaptability of the road network in emergencies, which is measured by the connectivity of the network and the alternative paths of key nodes. The resilience score S of each edge is defined as resilience (e ij )for:

[0046]

[0047] Among them, C b (e ij ) represents edge e ij The betweenness centrality measures the importance of an edge in the global shortest path and reflects the traffic it carries in the network. For each pair of nodes (V i ,V j ) ij The betweenness centrality of is defined as:

[0048]

[0049] Among them, V is the set of all nodes; σ ij Passing through node V i To node V j The number of all shortest paths; σ ij (e ij ) are the paths passing through edge e ij the number of

[0050] P backup (e ij ) represents edge e ij The number of alternative paths reflects the replacement capacity when the edge fails. Specifically, if the edge e ij Failure, the ability to maintain connectivity between nodes through other paths, defined as P backup (e ij )for:

[0051]

[0052] Among them, P ij Passing through node V i To node V j The set of all shortest paths; O() is the indicator function, if the path P ij Contains edge e ij Otherwise, it is assigned to 1.

[0053] On the other hand, this application also provides an urban land space multi-factor coupling intelligent optimization system, including:

[0054] The urban land space feature optimization module is used to quantify the functional zoning transfer probability of ecological protection areas, basic farmland, and urban development zones based on the Markov model. The spatial distribution of functional zones is simulated using the future land use simulation model. Combined with the resistance surface analysis of the minimum cumulative resistance model, the first multi-objective optimization model is constructed. Ecological resilience, urban spatial resilience, and basic farmland protection and intensive development are determined as optimization targets. Geographic cellular automata and neural network algorithms are used to embed optimization objective functions and constraints, simulate the evolution trend of urban spatial patterns, analyze dynamic change patterns, and optimize the spatial distribution of functional zones.

[0055] The urban land space point element optimization module is used to collect urban POI data, select residential, commercial, and public service types as optimization targets, use cluster analysis to divide the optimization targets and extract core points, construct a second multi-objective optimization model, quantify the spatial relationship and traffic accessibility between nodes, calculate and optimize the spatial concentration, network resilience, and adaptability of core points to future urban development areas, use the NSGA-II algorithm to solve the optimization plan, analyze the spatial layout characteristics of core points, and optimize the spatial distribution of core points of residential, commercial, and public service types;

[0056] The urban land space line feature optimization module is used to construct graph structure data based on the road network, define node and edge features, use graph neural networks to train the graph structure of the road network and extract features, build a third multi-objective optimization model, quantify the intensiveness and resilience of the road network, and embed the dynamically updated edge features of the third multi-objective optimization model into the message passing mechanism of the graph neural network to capture global topological characteristics to formulate multi-objective optimization decisions for the road network, and then return to step one for loop iterative optimization;

[0057] The decision includes at least one of adding new efficient connecting roads, expanding highly intensive roads, and deleting inefficient and redundant roads.

[0058] Compared with the existing technology, the present invention has the following advantages: it integrates the optimization modules of urban land space points, lines and surfaces, optimizes step by step in the order of "surface-point-line", performs cyclic iterative analysis through the transmission of optimization results, and finally realizes intensive and resilience-oriented urban land space multi-element coupling intelligent optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is the overall flow chart of the present invention;

[0060] Figure 2 It is the NSGA-II algorithm operation logic of the present invention;

[0061] Figure 3 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0062] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0063] like Figure 1-3 As shown, this embodiment provides a technical solution: a method for intelligent optimization of multi-factor coupling of urban land space, including:

[0064] Step 1: Based on the Markov model, the functional zoning transfer probability of ecological protection areas, basic farmland, and urban development areas was quantified. The spatial distribution of functional zoning was simulated using the Future Land Use Simulation (FLUS) model. Combined with the resistance surface analysis of the Minimum Cumulative Resistance (MCR) model, the first multi-objective optimization model was constructed. Ecological resilience, urban spatial resilience, and basic farmland protection and intensive development were determined as optimization objectives. Geographic Cellular Automata (Geo-CA) and Artificial Neural Network (ANN) algorithms were used to embed optimization objective functions and constraints, simulate the evolution trend of urban spatial pattern, analyze the dynamic change law, and optimize the spatial distribution of functional zoning.

[0065] For example, the surface element optimization module is constructed based on the MCR-Markov-FLUS framework. The MCR-Markov-FLUS framework is an important tool in the optimization and simulation research of urban land space functional zoning. Among them, the MCR model optimizes the functional zoning configuration through resistance surface analysis to minimize the cumulative resistance of the system to disturbance. The Markov model quantifies and predicts the transition probability between functional zones based on time series data. The FLUS model uses neural network algorithms and geographic cellular automata to generate the spatial distribution of functional zones, emphasizing spatial heterogeneity and complexity. The analytical framework composed of the three can be applied to simulate the changing trend of functional zones and achieve spatial optimization configuration. The basic architecture of the framework is as follows:

[0066] (1) MCR model

[0067] The essence of the MCR model lies in the minimum resistance that needs to be overcome to reach the target location from the source, which reflects a kind of accessibility. The evolution of functional zoning is regarded as the result of the diffusion of ecological sources and construction sources to the surrounding areas. The minimum cumulative resistance surface of the ecological source and the minimum cumulative resistance surface of the construction source in the study area are constructed to represent the spatial span characteristics of functional zoning. The calculation formula is:

[0068] C M =f∑D mn R m

[0069] Among them, C M It is a function of the minimum resistance of any grid center within the range and its distance to all sources; f is the negative correlation between the minimum cumulative resistance and its distance to all sources; D mn R is the distance from the source n to a grid in space through space m; m represents the resistance of species to cross the landscape space m.

[0070] (2) Markov model

[0071] The Markov model simulates the spatial change of land by setting the state of a certain spatial type at time t+1 to be only related to time t. Its expression is as follows:

[0072] S (t+1) =P ab ·S (t)

[0073] Among them, S (t) ,S (t+1) is the land space type state matrix of the study area at time t and time t+1; P ab Represents the transition probability matrix from type a to type b.

[0074] (3) Flus model

[0075] The neural network algorithm (ANN) is used to obtain the suitability probability of each land use type within the study area from the national land spatial pattern data of the base year and a variety of driving factors including human and natural effects. The suitability probability is then combined with the neighborhood factor, adaptive inertia coefficient and conversion cost to obtain the overall conversion probability of each cell. The simulation results are finally obtained through a roulette competition mechanism.

[0076] ANN is divided into input layer, hidden layer and output layer. By normalizing the basic data of driving factors and sampling the land space data and driving factors using uniform sampling method, the suitability probability of different spaces is calculated by ANN. The specific calculation formula is:

[0077]

[0078] Where sp(p,i,t) is the suitability probability of type i space under grid p at time t; w j,i Represents the weight between the output layer and the hidden layer; sigmoid() is the excitation function from the hidden layer to the output layer; net j (p, t) represents the signal received by the jth hidden layer grid p at time t. The sum of the suitability probabilities of each spatial type obtained by the ANN is always 1. That is:

[0079]

[0080] The neighborhood development density is selected to measure the neighborhood effect, and the calculation formula is as follows:

[0081]

[0082] in, Indicates the total number of rasters of the i-th land class after the previous iteration on the N×N Moore neighborhood window, w i The neighborhood factor parameter range is [0,1]. The closer it is to 1, the stronger the expansion capacity of the functional zoning type.

[0083] The first multi-objective optimization model:

[0084] In urban spatial optimization, objective functions and constraints are core elements for achieving scientific planning. The objective function defines the direction of optimization, provides a quantitative basis for optimization, and ensures that the results are consistent with planning goals. Constraints limit the feasible scope of the optimization solution to avoid violating actual needs or policy regulations. In [1], the objective function for optimizing the surface elements of the urban land space pattern is set as follows:

[0085] (1) Ecological resilience

[0086] The connectivity and compactness of ecological protection areas are important indicators for measuring ecological resilience. The better the connectivity of an ecological protection area, the higher the degree of connection between different patches within the ecological protection area, which reduces the island effect and improves the stability and resilience of the ecosystem. The better the compactness of an ecological protection area, the closer the ecological patches are to each other and the more concentrated their distribution is, which helps to reduce the damage to the ecological core area caused by external interference. Based on the idea of landscape pattern index, the connectivity and compactness of the ecological protection area are measured respectively by calculating the spread and cohesion of the ecological protection area, and used as a measure of ecological resilience. The specific calculation formula is as follows:

[0087]

[0088] Among them, CON is the spread index of the ecological protection zone, which represents the connectivity of the ecological protection zone; m is the total number of functional zoning types; g k It refers to the number of adjacent patches within the ecological protection area or between the ecological protection area and other types of functional zones.

[0089]

[0090] Among them, COH e is the cohesion index of the ecological reserve, which characterizes the compactness of the ecological reserve; P is the total area of the patches in the ecological reserve; p is the total boundary length of all patches in the ecological reserve; a is the total area of all patches in the ecological reserve; A is the total area of all functional zoning patches.

[0091] (2) Urban spatial resilience

[0092] Urban spatial resilience focuses on the form of urban construction itself and the ecological pattern it affects, and on the interaction between the internal structure and the external environment. The integration of urban development space and ecological protection space is a landscape pattern that is conducive to eliminating adverse effects, while the decentralized layout of urban development space helps to improve the local self-organization and independence of the city and increase the resilience of the city. The source-sink landscape theory measures the morphological resilience characteristics of the city from the perspective of landscape ecology. The theory believes that heterogeneous landscapes can be divided into two types: source and sink landscapes. The source landscape refers to the landscape type that can promote the development of the process, and the sink landscape is the landscape type that can prevent or delay the development of the process. The negative effects generated by the source landscape can be absorbed and reduced by the sink landscape. In the , urban development space is defined as the source landscape, and the ecological protection area and basic farmland are defined as the sink landscape, and the urban spatial resilience is measured by calculating the source-sink landscape pattern index. The specific calculation formula is as follows:

[0093]

[0094] Where: D is the average distance index of source landscape patches; d ijrepresents the distance between grid i in the source landscape patch and grid j in the sink landscape patch; m and n represent the number of grids in the source landscape patch and the number of grids in the sink landscape patch, respectively. A smaller D value indicates a stronger coupling between the source and sink landscapes, and thus greater urban spatial resilience. Conversely, a larger D value indicates a significant spatial separation between the source and sink landscapes, making it difficult for the sink landscape to offset the negative effects of the source landscape, and thus poorer urban spatial resilience.

[0095] (3) Intensive development of basic farmland and urban space

[0096] The intensive development of basic farmland and urban space requires ensuring that basic farmland and urban development space are as concentrated and compact as possible to minimize spatial fragmentation and improve regional efficiency and sustainability through large-scale development and shared infrastructure. Urban development should focus on the efficient use of space, realize the development model of compact cities, reduce the scattered spread of cities into surrounding cultivated land, and reduce the fragmentation of originally large-scale farmland. In the paper, based on the idea of landscape pattern index, the compactness of basic farmland and urban development areas is calculated by calculating the cohesion of the two as a measure of intensive development. The specific calculation method is as follows:

[0097]

[0098] Among them, COH c and COH u are the cohesion indexes of basic farmland and urban development areas, respectively, characterizing the degree of agglomeration of the two; P is the total area of the corresponding functional zoning patches; p is the total boundary length of all patches of the corresponding functional zoning; a is the total area of all patches of the corresponding functional zoning; A is the total area of all functional zoning patches.

[0099] The constraints are mainly the "three zones and three lines" red line control restrictions in the national land space planning documents. The "three zones and three lines" include the ecological protection red line, the permanent basic farmland protection red line, and the urban development boundary red line. The ecological protection red line refers to the area that must be strictly protected to ensure national ecological security and maintain ecological functions; the permanent basic farmland protection red line refers to the foundation for ensuring national food security and stipulates the minimum protection area of basic farmland; the urban development boundary red line refers to the development boundary in urban planning and stipulates the maximum area of urban development. Specifically, the constraints can be expressed as:

[0100]

[0101] Among them, A e 、A c 、A u Refers to the optimized ecological protection area, basic farmland and urban development area respectively; A e,min 、A c,min 、A u,maxThe red line control limits for ecological protection areas, basic farmland and urban development areas in the national land space planning respectively.

[0102] Step 2: Collect urban POI data and select residential, commercial, and public service types as optimization targets. Use cluster analysis to divide the optimization targets and extract core points. Build a second multi-objective optimization model to quantify the spatial relationships and traffic accessibility between nodes. Calculate and optimize the spatial density, network resilience, and compatibility of core points with future urban development areas. Use the NSGA-II algorithm to solve the optimization plan, analyze the spatial layout characteristics of core points, and optimize the spatial distribution of core points of residential, commercial, and public service types.

[0103] It is understandable that in order to better optimize the land space resources, when collecting data, multi-source data can be collected, such as Figure 2 As shown in the figure, for example: meteorological data, terrain data, land use data, soil data, administrative boundary data, POI data and road network data, among which POI data (Points of Interest) is an important data type in the Geographic Information System (GIS), representing discrete point locations with specific functions or meanings in the real world, usually containing attribute information such as name, category, coordinates, and address.

[0104] Considering future spatial changes in urban development, optimizing the layout of existing core points and selecting new ones becomes crucial. Based on complex network theory, we optimize the layout of existing core points while rationally locating new ones based on future spatial changes. The goal of this optimization is to improve spatial efficiency, enhance network resilience, and ensure that new core points can be effectively integrated into future urban development plans.

[0105] The nodes in the second multi-objective optimization model are the core points obtained through identification, and the edges represent the connection relationship between the nodes. The weight of the edge is determined by the spatial distance and traffic accessibility between these nodes. The weight of the edge is specifically expressed as:

[0106]

[0107] Among them, w i,j is the weight of the edge between node i and node j, reflecting the connection strength between the nodes; d i,j is the geometric space distance between node i and node j; k i,j is the number of directly connected roads between node i and node j.

[0108] The objective function for optimizing the spatial layout of urban core points mainly includes three aspects: spatial density, network resilience, and adaptability to future urban development space. Spatial density requires that the layout of urban functional points be as compact as possible to reduce space waste. In the optimization of core point layout, the density goal is expressed as minimizing the average geometric distance between core points, achieving efficient sharing of resources and services, and improving spatial utilization efficiency and resource allocation efficiency. The objective function of intensive optimization is expressed as:

[0109]

[0110] in, is the average shortest geometric distance between all core points, which is used to measure spatial compactness; N is the total number of core points; d ij is the shortest geometric distance between the i-th and j-th core points.

[0111] The resilience of an urban network refers to the ability of an urban system to quickly recover and maintain its basic functions when subjected to disturbances or shocks (such as traffic accidents, natural disasters, etc.). From a complex network perspective, network resilience is usually enhanced by increasing redundant connections between nodes. The objective function of resilience optimization is to maximize the redundant paths and network connectivity between urban functional points, ensuring that the system can recover through other paths when certain nodes fail. Therefore, resilience is measured by the betweenness centrality of the network. The specific calculation method is expressed as:

[0112]

[0113] Among them, C b is the overall resilience of the network; σ ij is the number of shortest paths from node i to node j, σ ij (k) is the number of shortest paths passing through node k.

[0114] The results of optimizing urban land area elements define new urban development areas. Core points need to be distributed appropriately to accommodate the needs of the new urban development areas. The adaptability objective function measures the coverage of the new urban development areas by core points. The objective function is as follows:

[0115]

[0116] Among them, C s is the core point coverage, which is used to characterize the adaptability; A p The area of the urban development area covered by the service area of the core point is obtained by overlaying the area within 15 minutes of driving distance of each core point based on the road network; A u is the total area of urban development areas.

[0117] In summary, the objective function of urban core point spatial layout optimization can be expressed as:

[0118]

[0119] Among them, F is the objective function set for optimizing the spatial layout of urban core points; the remaining parameters are consistent with the above statements.

[0120] The constraints for optimizing the spatial layout of urban core points define the distribution logic of core points, mainly including urban development space constraints, urban core area constraints, and core quantity constraints. Among them, the urban development space constraint requires that residential core points can only be distributed within the urban development space, which can be expressed as:

[0121]

[0122] Among them, x i is the core point; {Urban} is the collection of urban development spaces.

[0123] In addition, the layout of existing urban centers and sub-centers should not change drastically during the optimization process, and the optimization process should ensure that the distribution pattern of these core areas remains stable. Therefore, the following constraints are set:

[0124] |d ij -d ij,0 |≤∈

[0125] Among them, d ij is the optimized core point x i and x j The geometric distance between ij,0 is the optimized core point x i and x j The geometric distance between them; ∈ is the tolerance, which is used to control the consistency between the optimization results and the existing urban structure pattern.

[0126] At the same time, the number of core points should be consistent with urban development and future planning. Therefore, the following constraints are set:

[0127] N∈[N plan ]

[0128] Where N is the number of core points; [N plan ] is the planned layout quantity range of commercial core points (or public service core points) determined based on urban development conditions and planning documents.

[0129] The NSGA-II algorithm in the genetic algorithm is used to solve the optimization scheme and obtain the optimized spatial layout. The NSGA-II algorithm is a multi-objective optimization algorithm based on the principles of inheritance, mutation and evolution in biology. The NSGA-II algorithm combines the non-dominated sorting and congestion operators through the elite strategy of population evolution. In addition, the algorithm uses a fast non-dominated sorting method based on the concepts of Pareto dominance and optimality to distinguish and solve. The spatial positions of residential core points, commercial core points and public service core points are respectively characterized into a matrix. Each combination of the spatial position matrices of the three types of core points is regarded as an individual, and the collection of individuals constitutes a population. The details are as follows:

[0130]

[0131] Population = {Positon p ,Position c ,Position s}

[0132] Among them, Position p ,Position c ,Position s are the spatial location matrices of residential core points, commercial core points, and public service core points respectively; x n,m is a specific spatial unit. If the core point is arranged in this unit, it is set to 1, otherwise it is set to 0. n and m respectively determine the spatial range where the core point can be arranged.

[0133] (1) Basic operating logic of NSGA-II algorithm

[0134] like Figure 2 As shown in the figure, the NSGA-II algorithm, based on the constraints set above, determines the permissible composition of each individual in the population, including its parents and all subsequent offspring. These constraints constitute the external environment that determines the survival of the population. Individuals unable to adapt to the external environment face "natural selection" and are unable to advance to the next iteration. The objective function in the optimization model reflects the degree of adaptation of an individual to the external environment and determines its hierarchical position within the population. Individuals with greater adaptability have a survival advantage, and their adaptive "genes" can be passed on to future generations. To prevent homogenization within the population, the algorithm calculates the next generation's values through multi-point crossover and Gaussian mutation. This process ensures that offspring can efficiently inherit "favorable genes" from their parents while introducing a certain amount of "genetic mutations," ensuring individual diversity and preventing convergence of optimization results. In the example, the population size is set to 100, the mutation range is set to [0, 1], the mutation probability is set to 10%, and the number of iterations is set to 1000.

[0135] (2) Fast non-dominated sorting and congestion operator

[0136] The fast non-dominated sort in the NSGA-II algorithm constitutes a population hierarchical structure based on the level of individual non-inferior solutions. This stratification effectively directs the search to the Pareto optimal solution set. The fast non-dominated sort represents a cyclic process that adapts to the value hierarchy. The analysis logic is: find the overall non-dominated solution set, called the first non-dominated layer F1. Remove all individuals in F1 from the population. Then continue to find the non-dominated solution set of the remaining population, called the second non-dominated layer F2, and continue to remove individuals in F2 from the remaining population. Continue this analysis logic until the entire population is stratified and all F rank Individuals in the same stratum have the same non-dominated ordering. Fast non-dominated sorting sorts all individuals in the population, but it does not establish an ordering between individuals in the same non-dominated ordering layer. If the selection of parents for the next generation exceeds the predetermined constraints after incorporating individuals that are not adapted to the environment, it is necessary to determine which individuals are eligible to become parents of the next generation. Therefore, the crowding operator is used to sort individuals in the same non-dominated layer of the non-dominated ordering to achieve parent selection.

[0137] The crowding operator refers to the average function distance C between two individuals in the same non-dominated layer calculated based on the objective function value. i The average function distance is expressed as the perimeter of a rectangle, with an individual and its nearest neighbor constituting the vertices. Points on the rectangle boundary are set to infinite crowding. For all other intermediate individuals, the crowding is the sum of the normalized absolute differences of all objective function values between two adjacent individuals.

[0138] (3) Natural selection process

[0139] Non-dominated layer level F based on non-dominated sorting rank and congestion operator C i Constructing a natural selection sequence [F rank ,C i Based on this sequence, all individuals are ranked. The natural selection process prioritizes the non-dominated stratum level, followed by the crowding operator ranking within the same non-dominated stratum. The next generation of parents is selected for each iteration based on the natural selection process and the set population size.

[0140] Step 3: Build graph structure data based on the road network, define node and edge features, use graph neural network to train the graph structure of the road network and extract features, build a third multi-objective optimization model, quantify the intensiveness and resilience of the road network, and embed the dynamically updated edge features of the third multi-objective optimization model into the message passing mechanism of the graph neural network to capture global topological characteristics to formulate multi-objective optimization decisions for the road network. Then return to step 1 for loop iterative optimization.

[0141] The decision includes at least one of adding new efficient connecting roads, expanding highly intensive roads, and deleting inefficient and redundant roads.

[0142] Representing a road network as a graph structure (Graph) data G = (V, E) is a cutting-edge modeling method currently studied in related fields, where V represents a node set and E represents an edge set. Node v∈V represents a road intersection, and edge e ij ∈E represents a road segment connecting two nodes. Each node and edge is attached with specific features to reflect its geographical attributes or traffic attributes.

[0143] (1) Node construction and node feature setting

[0144] Each intersection of the road network vector data is obtained through topological analysis as a node. Each node v i The eigenvector of ∈V is represented as X i ∈R d , where d is the feature dimension. In i The position feature is defined as:

[0145] X i =[x i ,y i ,r i ,q i ]

[0146] Among them, X i For node v i The eigenvector of x i For node v i Longitude; y i For node v i Latitude; r i Represents node v i Whether it is located in the urban development area, if it is located in the urban development area, it is assigned to 1, otherwise it is assigned to 0; q i Represents node v i Affected by various surrounding core points, nodes with a high degree of influence have higher demands on traffic carrying capacity, which is measured by the core point kernel density value of the node location (calculated in the same way as Equation 11).

[0147] (2) Construction and feature setting

[0148] The edge set E represents the topological structure of the road segment. Each road segment between intersections is extracted from the road network as a connection between nodes. Each edge e ij ∈E represents node v i and v j The connection relationship, its characteristic is e ij ∈Rk It reflects the physical or attribute information of the road segment. The feature vector of each edge can be expressed as:

[0149]

[0150] Among them, l ij For edge e ij The length of r ij Characterizing edge e ij Whether it passes through the urban development area. If it does, it is assigned to 1, otherwise it is assigned to 0; Represents edge e ij Affected by various core points around, the edges with high degree of influence have higher requirements on traffic carrying capacity. ij It is measured by the average kernel density of the core points across the region.

[0151] (3) Adjacency matrix construction

[0152] Adjacency matrix A∈R n×n Used to describe the topological structure of a road network graph, where n is the number of nodes. i and v j If there is an edge between ij =1, otherwise A ij =0.

[0153] 2. Construction of the third multi-objective optimization model

[0154] (1) Objective function

[0155] The intensive objective in the objective function is mainly characterized by maximizing the service efficiency of the road network for traffic demand while reducing the waste of redundant resources. In the graph structure, the intensive objective is measured by the coverage of the traffic demand by the spatial layout of the edge to the core point. For each edge e ij , its intensiveness score S compact (e ij ) is expressed as the ratio of the traffic demand of the core points of the edge passing through the space to the length of the edge:

[0156]

[0157] Among them, S compact (e ij ) is edge e ij The intensiveness score of is used to indicate the intensiveness level of the edge; For edge e ij The average kernel density of the core points passing through the region represents the edge e ij The demand for traffic through the core point of the space; R(e ij ) represents edge e ijThe geometric length of the edge is used to represent the resource consumption of the edge.

[0158] The resilience goal in the objective function is to maximize the connectivity and adaptability of the road network in emergencies, which is measured by the connectivity of the network and the alternative paths of key nodes. Define the resilience score S of each edge resilience (e ij )for:

[0159]

[0160] Among them, C b (e ij ) represents edge e ij The betweenness centrality measures the importance of an edge in the global shortest path and reflects the traffic it carries in the network. For each pair of nodes (V i ,V j ) ij The betweenness centrality of is defined as:

[0161]

[0162] Among them, V is the set of all nodes; σ ij Passing through node V i To node V j The number of all shortest paths; σ ij (e ij ) are the paths passing through edge e ij The number of

[0163] P backup (e ij ) represents edge e ij The number of alternative paths reflects the replacement capability when the edge fails. Specifically, if the edge e ij Failure, the ability to maintain connectivity between nodes through other paths. Define P backup (e ij )for:

[0164]

[0165] Among them, P ij Passing through node V i To node V j The set of all shortest paths; O() is the indicator function, if the path P ij Contains edge e ij Otherwise, it is assigned to 1.

[0166] (2) Constraints

[0167] The constraints in the optimization model are mainly connectivity constraints, which means that the network must still maintain connectivity after optimization, that is, the spectral radius of the adjacency matrix A′ is greater than zero:

[0168] λ min (A′)>0

[0169] Among them, A′ is the optimized graph structure adjacency matrix; λ min is the smallest non-zero eigenvalue of A′.

[0170] In the embodiment of the present application, urban land space point, line and surface element optimization modules are integrated, and the optimization is gradually carried out in the order of "surface-point-line". The optimization results are transferred through cyclic iterative analysis, and finally the intensive and resilience-oriented urban land space multi-element coupling intelligent optimization is achieved.

[0171] In another embodiment of the present application, Figure 3 As shown, an urban land space multi-factor coupling intelligent optimization system is provided, including:

[0172] The urban land space feature optimization module is used to quantify the functional zoning transfer probability of ecological protection areas, basic farmland, and urban development zones based on the Markov model. The spatial distribution of functional zones is simulated using the future land use simulation model. Combined with the resistance surface analysis of the minimum cumulative resistance model, the first multi-objective optimization model is constructed. Ecological resilience, urban spatial resilience, and basic farmland protection and intensive development are determined as optimization targets. Geographic cellular automata and neural network algorithms are used to embed optimization objective functions and constraints, simulate the evolution trend of urban spatial patterns, analyze dynamic change patterns, and optimize the spatial distribution of functional zones.

[0173] The urban land space point element optimization module is used to collect urban POI data, select residential, commercial, and public service types as optimization targets, use cluster analysis to divide the optimization targets and extract core points, construct a second multi-objective optimization model, quantify the spatial relationship and traffic accessibility between nodes, calculate and optimize the spatial concentration, network resilience, and adaptability of core points to future urban development areas, use the NSGA-II algorithm to solve the optimization plan, analyze the spatial layout characteristics of core points, and optimize the spatial distribution of core points of residential, commercial, and public service types;

[0174] The urban land space line feature optimization module is used to construct graph structure data based on the road network, define node and edge features, use graph neural networks to train the graph structure of the road network and extract features, build a third multi-objective optimization model, quantify the intensiveness and resilience of the road network, and embed the dynamically updated edge features of the third multi-objective optimization model into the message passing mechanism of the graph neural network to capture global topological characteristics to formulate multi-objective optimization decisions for the road network, and then return to step one for loop iterative optimization;

[0175] The decision includes at least one of adding new efficient connecting roads, expanding highly intensive roads, and deleting inefficient and redundant roads.

[0176] In the embodiment of the present application, urban land space point, line and surface element optimization modules are integrated, and the optimization is gradually carried out in the order of "surface-point-line". The optimization results are transferred through cyclic iterative analysis, and finally the intensive and resilience-oriented urban land space multi-element coupling intelligent optimization is achieved.

[0177] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0178] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0179] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A multi-factor coupled intelligent optimization method for urban land space, characterized by: include: Step 1: Quantify the functional zoning transfer probability of ecological protection areas, basic farmland, and urban development areas based on the Markov model. Use the future land use simulation model to simulate the spatial distribution of functional zones. Combined with the resistance surface analysis of the minimum cumulative resistance model, the first multi-objective optimization model is constructed. Ecological resilience, urban spatial resilience, and basic farmland protection and intensive development are determined as optimization objectives. Using geographic cellular automata and neural network algorithms, the optimization objective function and constraints are embedded. The evolution trend of urban spatial pattern is simulated, the dynamic change pattern is analyzed, and the spatial distribution of functional zones is optimized. Step 2: Collect urban POI data and select residential, commercial, and public service types as optimization targets. Use cluster analysis to divide the optimization targets and extract core points. Build a second multi-objective optimization model to quantify the spatial relationships and traffic accessibility between nodes. Calculate and optimize the spatial density, network resilience, and compatibility of core points with future urban development areas. Use the NSGA-II algorithm to solve the optimization plan, analyze the spatial layout characteristics of core points, and optimize the spatial distribution of residential, commercial, and public service core points. Step 3: Build graph structure data based on the road network, define node and edge features, use graph neural network to train the graph structure of the road network and extract features, build a third multi-objective optimization model, quantify the intensiveness and resilience of the road network, and embed the dynamically updated edge features of the third multi-objective optimization model into the message passing mechanism of the graph neural network to capture global topological characteristics to formulate multi-objective optimization decisions for the road network. Then return to step 1 for loop iterative optimization. The decision includes at least one of adding new efficient connecting roads, expanding highly intensive roads, and deleting inefficient and redundant roads.

2. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized by: The minimum cumulative resistance model includes: C M =f∑D mn R m Among them, C M It is a function of the minimum resistance of any grid center within the range and its distance to all sources; f is the negative correlation between the minimum cumulative resistance and its distance to all sources; D mn R is the distance from the source n to a grid in space through space m; m represents the resistance of species to cross the landscape space m; The Markov model includes: S (t+1) =P ab ·S (t) Among them, S (t) ,S (t+1) is the land space type state matrix of the study area at time t and time t+1; P ab Represents the transition probability matrix from type a to type b; The future land use simulation model simulates the spatial distribution of functional zones, including: A neural network algorithm is used to obtain the suitability probability of each land use type within the study area from the base year national land spatial pattern data and multiple driving force factors including human and natural effects. The suitability probability is combined with the neighborhood factor, adaptive inertia coefficient, and conversion cost to obtain the overall conversion probability of each cell. The simulation results are obtained through a roulette competition mechanism. The neural network algorithm includes: Where sp(p,i,t) is the suitability probability of type i space under grid p at time t; w j,i Represents the weight between the output layer and the hidden layer; sigmoid() is the excitation function from the hidden layer to the output layer; net j (p,t) represents the signal received by the j-th hidden layer grid p at time t.

3. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized by: The ecological resilience includes the connectivity and compactness of ecological protection areas. The spread and cohesion of ecological protection areas are calculated to represent the ecological connectivity and compactness. The calculation methods include: Among them, CON is the spread index of the ecological protection zone, which represents the connectivity of the ecological protection zone; m is the total number of functional zoning types; g k The number of adjacent patches within the ecological protection area or between the ecological protection area and other types of functional zones; Among them, COH e is the cohesion index of the ecological reserve, which characterizes the compactness of the ecological reserve; P is the total area of the patches in the ecological reserve; p is the total boundary length of all patches in the ecological reserve; a is the total area of all patches in the ecological reserve; A is the total area of all functional zoning patches.

4. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized by: The urban development zone is a source landscape type, and the ecological protection zone and basic farmland are sink landscape types. The urban spatial resilience is measured by calculating the source-sink landscape pattern index. The calculation process includes: Where: D is the average distance index of source landscape patches; d ij represents the distance between grid i in the source landscape patch and grid j in the sink landscape patch; m and n represent the number of grids in the source landscape patch and the number of grids in the sink landscape patch, respectively.

5. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized by: The intensive development is characterized by calculating the cohesion of basic farmland and urban development areas, and the calculation method includes: Among them, COH c and COH u are the cohesion indexes of basic farmland and urban development zones, respectively, which characterize the degree of agglomeration of the two. P is the total area of the corresponding functional zone patches, p is the total boundary length of all patches of the corresponding functional zone, a is the total area of all patches of the corresponding functional zone, and A is the total area of all functional zone patches.

6. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized by: The constraints include: Among them, A e 、A c 、A u They refer to the optimized ecological protection area, basic farmland and urban development area respectively. e,min 、A c,min 、A u,max The red line control limits for ecological protection zones, basic farmland and urban development zones in the national land space planning respectively.

7. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized by: The calculation and optimization of the spatial density, network resilience, and adaptability of core points to future urban development areas include: in, is the average shortest geometric distance between all core points, which is used to measure spatial compactness; N is the total number of core points; d ij is the shortest geometric distance between the i-th and j-th core points.

8. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized by: The NSGA-II algorithm is used to solve the optimization solution, including: Population={Position p ,Position c ,Position s } Among them, Position p ,Position c ,Position s are the spatial location matrices of residential core points, commercial core points, and public service core points, respectively. n,m is a specific spatial unit. If the core point is arranged in this unit, it is set to 1, otherwise it is set to 0. n and m respectively determine the spatial range where the core point can be arranged.

9. The method for intelligent optimization of urban land space multi-factor coupling according to claim 1 is characterized in that: The quantification of the density and resilience of the road network includes: Among them, S compact (e ij ) is edge e ij The intensiveness score of is used to indicate the intensiveness level of the edge; For edge e ij The average kernel density of the core points passing through the region represents the edge e ij The demand for traffic through the core point of the space; R(e ij ) represents edge e ij The geometric length of , which is used to represent the resource consumption of the edge; The road network resilience is mainly characterized by maximizing the connectivity and adaptability of the road network in emergencies, which is measured by the connectivity of the network and the alternative paths of key nodes. The resilience score S of each edge is defined as resilience (e ij )for: Among them, C b (e ij ) represents edge e ij The betweenness centrality measures the importance of an edge in the global shortest path and reflects the traffic it carries in the network. For each pair of nodes (V i ,V j ) ij The betweenness centrality of is defined as: Among them, V is the set of all nodes; σ ij Passing through node V i To node V j The number of all shortest paths; σ ij (e ij ) are the paths passing through edge e ij the number of P backup (e ij ) represents edge e ij The number of alternative paths reflects the replacement capacity when the edge fails. Specifically, if the edge e ij Failure, the ability to maintain connectivity between nodes through other paths, defined as P backup (e ij )for: Among them, P ij Passing through node V i To node V j The set of all shortest paths; O() is the indicator function, if the path P ij Contains edge e ij Otherwise, it is assigned to 1.

10. An urban land space multi-factor coupling intelligent optimization system, characterized by: include: The urban land space feature optimization module is used to quantify the functional zoning transfer probability of ecological protection areas, basic farmland, and urban development zones based on the Markov model. The spatial distribution of functional zones is simulated using the future land use simulation model. Combined with the resistance surface analysis of the minimum cumulative resistance model, the first multi-objective optimization model is constructed. Ecological resilience, urban spatial resilience, and basic farmland protection and intensive development are determined as optimization targets. Geographic cellular automata and neural network algorithms are used to embed optimization objective functions and constraints, simulate the evolution trend of urban spatial patterns, analyze dynamic change patterns, and optimize the spatial distribution of functional zones. The urban land space point element optimization module is used to collect urban POI data, select residential, commercial, and public service types as optimization targets, use cluster analysis to divide the optimization targets and extract core points, construct a second multi-objective optimization model, quantify the spatial relationship and traffic accessibility between nodes, calculate and optimize the spatial concentration, network resilience, and adaptability of core points to future urban development areas, use the NSGA-II algorithm to solve the optimization plan, analyze the spatial layout characteristics of core points, and optimize the spatial distribution of core points of residential, commercial, and public service types; The urban land space line feature optimization module is used to construct graph structure data based on the road network, define node and edge features, use graph neural networks to train the graph structure of the road network and extract features, build a third multi-objective optimization model, quantify the intensiveness and resilience of the road network, and embed the dynamically updated edge features of the third multi-objective optimization model into the message passing mechanism of the graph neural network to capture global topological characteristics to formulate multi-objective optimization decisions for the road network, and then return to step one for loop iterative optimization; The decision includes at least one of adding new efficient connecting roads, expanding highly intensive roads, and deleting inefficient and redundant roads.

Images