Community built environment regulation scenario generation method and system for multi-objective optimization

Abstract

This invention discloses a method and system for generating community built environment regulation scenarios oriented towards multi-objective optimization. By constructing a spatial regression model at the community scale between the built environment and residents' travel carbon emissions, travel time, and travel costs, the method quantifies the regulatory relationship between built environment variables and multi-dimensional travel characteristics. A multi-objective optimization mathematical model of "carbon emission reduction - time reduction - cost reduction" is established, and an improved NSGA-II algorithm based on response priority is used to solve for the community, regulation variables, and regulation intensity of built environment regulation. This achieves efficient generation of refined community-scale built environment regulation scenarios and multi-objective decision support for coordinated carbon reduction and cost reduction in travel.

Description

Technical Field

[0001] This invention belongs to the field of urban planning and design technology, specifically relating to a method and system for generating urban community built environment control scenarios oriented towards multi-objective optimization, which is particularly suitable for generating refined control schemes in scenarios such as digital urban and rural planning and design, and human settlement environment design services. Background Technology

[0002] Currently, in the field of urban planning and design services, especially in human settlement design services and digital urban and rural planning, some publicly available technical solutions attempt to assess the impact of the built environment on travel behavior through modeling and analysis. Spatial regression models and machine learning algorithms are commonly used to establish macro-level carbon emission prediction models for cities, providing a reference for planning decisions. However, the inventors' research found that, based on the results of three spatial regression models at the fine spatial scale of the community level—daily road segment travel carbon emissions, average travel time for residents, and average travel cost—different communities exhibit significant differences in their response to the regulation of these three built environment variables. Existing technologies still have the following limitations: 1) The research scale is relatively macroscopic, lacking refined regulatory mechanisms at the community level. Existing studies mostly focus on the macro-scale of cities or regions, making it difficult to capture the heterogeneity within the basic living unit of the community. Due to differences in location, population structure, and land use patterns, different communities may respond significantly differently to the same built environment variable. Macro-scale modeling often masks these differences, resulting in a lack of targeted regulatory solutions.

[0003] 2) Single-objective focus makes it difficult to handle conflicting relationships between multiple objectives. Existing technical solutions often focus on optimizing a single objective (such as considering only carbon emission minimization) or use simple linear weighted sum methods to handle multi-objective problems. However, the regulation of the built environment actually involves multiple interrelated and even conflicting objectives. For example, increasing the density of community road networks has a significant negative effect on the travel time and cost of residents in some communities (i.e., reducing time and cost), but a significant positive effect on carbon emissions from travel on community road sections (i.e., increasing carbon emissions). This phenomenon of "asynchronous regulation of carbon emissions, time consumption, and cost" indicates that single-objective optimization may lead to the deterioration of other objectives. Currently used multi-objective optimization methods often fail to objectively obtain the weight coefficients of each objective, usually requiring subjective methods such as expert scoring, and cannot reveal the complex trade-offs between multiple objectives.

[0004] 3) Low efficiency when dealing with large-scale decision variables. In practical urban planning applications, the number of communities to be regulated may reach hundreds, with dozens of regulation variables and multiple levels of regulation intensity. This large-scale combinatorial optimization problem places extremely high demands on the algorithm's solution efficiency. Traditional multi-objective optimization algorithms (such as the conventional NSGA-II) use random initial population generation, which results in slow convergence speed when facing such a large search space, making it difficult to generate a high-quality Pareto solution set within a reasonable time.

[0005] 4) Lack of systematic decision support for multi-objective scenarios. Existing technologies typically output only one or more optimal solutions, but fail to provide differentiated control scenarios for different macroeconomic policy objectives (such as focusing solely on carbon reduction, carbon reduction while reducing time consumption, carbon reduction while reducing costs, etc.). Planners and decision-makers find it difficult to select suitable control solutions from the massive set of solutions based on actual policy priorities.

[0006] In summary, there is an urgent need for a technical solution to generate built environment control scenarios that can achieve refined regulation at the community level, effectively handle multi-objective conflict relationships, have efficient solution capabilities, and provide multi-scenario decision support, in order to solve the above problems. Summary of the Invention

[0007] To address the aforementioned problems in existing technologies, this invention provides a method and system for generating community built environment regulation scenarios oriented towards multi-objective optimization. By constructing a spatial regression model at the community scale between the built environment and residents' travel carbon emissions, travel time, and travel costs, the method quantifies the regulatory relationship between built environment variables and multi-dimensional travel characteristics. A multi-objective optimization mathematical model of "carbon emission reduction - time reduction - cost reduction" is established, and an improved NSGA-II algorithm based on response priority is used to solve for the community, regulation variables, and regulation intensity of built environment regulation. This achieves efficient generation of refined community-scale built environment regulation scenarios and multi-objective decision support for coordinated carbon reduction and cost reduction in travel.

[0008] Multi-objective optimization is often used to address potential contradictions during the convergence of multiple objectives. Currently, commonly used multi-objective optimization methods include evaluation function methods and intelligent algorithms. Evaluation function methods, represented by the linear weighted sum method, involve first assigning weights to each objective function, with the sum of the weights being 1. Then, the weighted objective functions are summed to obtain the overall objective function. Finally, the solution is obtained by solving the objective function under constraints. Intelligent algorithms, represented by genetic algorithms, mimic the genetics and natural selection processes in biology, using a global search strategy to find the optimal solution in the global space. Common types include Non-Dominated Sorting Genetic Algorithm (NSGA) and Non-Dominated Sorting Genetic Algorithm with Elite Preservation Strategy (NSGA-II). NSGA, lacking an elite strategy, suffers from high computational complexity. NSGA-II, by introducing a crowding distance operator and combining it with non-dominated sorting, can quickly achieve the objective while maintaining diversity, and optimizes each objective as much as possible. Of the two representative methods, the linear weighted sum method is simple and can often find multiple solutions with different meanings under multi-objective optimization by using different evaluation functions. However, the weight coefficients of the evaluation function are difficult to obtain, and subjective methods such as expert scoring are usually required. Genetic algorithms, on the other hand, have strong search capabilities and can generate Pareto sets of optimal solutions for multiple objectives, but in specific recommendations, human decision-making based on actual conditions is still necessary.

[0009] This invention constructs a spatial regression model of travel carbon emissions, travel time, and travel costs at the community scale, establishes a multi-objective optimization mathematical model of "reducing carbon emissions, reducing travel time, and reducing costs", and uses an improved NSGA-II algorithm based on response priority to solve the model, thereby achieving efficient generation of community-scale built environment regulation scenarios and multi-objective decision support.

[0010] The technical solution of this invention is: a method for generating community built environment regulation scenarios oriented towards multi-objective optimization, comprising the following steps: Step 1: Construct a community-scale multidimensional spatial regression model of travel behavior. For all communities in the target city, establish spatial regression models for the three dependent variables: built environment independent variable and daily road segment carbon emissions, average travel time and average travel cost. Obtain the linear fitting equations for the regulation of the three dependent variables for different built environment variables in each community. Step 2: Calculate the response rate of a single-variable unit regulation intensity for a single community in the built environment—Based on the spatial regression model obtained in Step 1, calculate the change rates of urban-level travel carbon emissions, average travel time, and average travel cost for each community under a single-variable unit regulation intensity, and obtain the carbon reduction contribution rate e corresponding to each community i under a unit regulation intensity of variable j. ij Contribution rate of time reduction t ij , contribution rate of reduced expenses c ij and unit control cost qij ; Step 3: Construct a multi-objective optimization mathematical model for reducing carbon emissions and costs in travel—Establish a multi-objective optimization model with the objectives of maximizing the total contribution rate of carbon reduction at the city level, maximizing the total contribution rate of time reduction, maximizing the total contribution rate of cost reduction, and minimizing the total cost of regulation. Step 4: Classification of regulatory variables and regulatory communities based on decision tree — Using the CART decision tree algorithm, with the response rate calculated in Step 2 as the feature, the regulatory variables and regulatory communities are classified into high, medium and low response rates based on the principle of minimizing the GINI coefficient. Step 5: Solving using the NSGA-II algorithm based on responsiveness priority—First, based on the contribution rate of a single community and a single variable to the total carbon emissions, average travel time, and average travel cost of urban travel at the unit control intensity, a decision tree algorithm is used to divide high-response variables and high-response communities; then, different responsiveness is used to set population priorities for communities and variables, and logistic mapping is used to generate an initial population based on chaotic sequences in the intersection set of high-response variables and high-response communities, and Pareto optimal solution set is obtained through non-dominated sorting, crowding calculation, and crossover mutation operations. Step 6: Multi-objective scenario decision analysis – Set different combinations of macro-objectives and solve them using the algorithm in Step 5 to obtain the control scenarios under each combination of objectives. Then, extract typical control schemes according to three criteria: minimum number of control communities, minimum number of control variables, and minimum control intensity.

[0011] Preferably, the multi-objective optimization mathematical model in step 3 is: ; The constraints are: ; in, This represents the sum of carbon reduction contributions under the selected variables and communities. This represents the sum of the time-saving contribution rates under the selected variables and the community. The sum of the cost reduction contribution rates under the selected variables and communities; i represents the i-th community, m represents the total number of communities in the city. j represents the j-th variable, and n represents the total number of variables considered; Indicates community In variables The corresponding carbon reduction contribution rate under the unit control intensity Indicates community In variables The contribution rate of time reduction under the unit control intensity, Indicates community In variables The contribution rate of cost reduction under the unit's regulatory intensity; Indicates community variables The intensity of regulation, Indicates community variables The unit control cost; This indicates the carbon emission reduction target set at the city level. This represents the time reduction target set at the city level. This indicates the cost reduction target set at the city level.

[0012] Preferably, step 4, the classification of the response rate of the control variables, specifically includes: summing the city-level contribution rates of all communities corresponding to each control variable; using the change rate of total travel carbon emissions as the first feature, the change rate of average travel time as the second feature, and the change rate of average travel cost as the third feature, and classifying them using the CART decision tree algorithm; the formula for calculating the GINI coefficient used for classification is as follows:

[0013] In the formula, K is the number of label categories in the decision tree training set. In this paper, both variables and communities are divided into three categories. For the variable or community sample size at this node, For this node belonging to The number of variables or community samples in the class.

[0014] Preferably, step 4, which involves classifying the community response level, specifically includes: statistically analyzing the average rate of change of high and medium response variables for a single community, using the rate of change of total travel carbon emissions as the first feature, the rate of change of average travel time as the second feature, and the rate of change of average travel cost as the third feature; simultaneously, using the proportion of high-carbon spaces in the community as the fourth feature and the proportion of high-carbon individuals among residents as the fifth feature, and classifying them using the CART decision tree algorithm, with the high and medium response variables representing the response level obtained from the classification.

[0015] Preferably, step 5 specifically includes: Step 5.1: Establish the cross subset of all variables and all community regulation change rates under different regulation intensities, and apply the regulation intensity to each individual in the subset. The following site selection With change rate and cost Assigning a value; Step 5.2: Generate an initial chaotic solution. Based on the set target value, calculate the rate of change of travel carbon emissions, travel time, and travel costs for the high-response community under the changes of all high-response variables, and select the control intensity range below and above the target. , The logistic mapping is used to generate an initial population based on chaotic sequences at the intersection of high-response variables and high-response communities. Step 5.3: Perform individual ranking based on non-dominated ranking and crowding distance; Step 5.4: Perform crossover and mutation operations according to the individual order, and merge the current parent generation and offspring generation into a single population; Step 5.5: Determine whether the set target constraints are met. , , If the maximum number of iterations has been reached, the algorithm terminates and outputs the optimal solution set. If the condition is not met, the algorithm returns to step 5.3 to continue sorting the merged new population.

[0016] Preferably, the combination of macroscopic objectives set in step 6 includes: Target Combination A: A 1% reduction in total carbon emissions from travel, with no increase in travel time or cost; Target combination B: Total carbon emissions from travel decrease by 1%, travel time decreases by 1%, and travel costs do not increase; Target combination C: Total carbon emissions from travel decrease by 1%, travel time does not increase, and travel costs decrease by 1%; Target combination D: Total carbon emissions from travel, travel time, and travel costs all decrease by 1%.

[0017] Preferably, the unit control intensity is set to 10%.

[0018] This invention also provides a multi-objective optimization-oriented urban community built environment regulation scenario generation system, comprising: The data acquisition module is used to collect land use data, transportation facility data, transportation policy and operating environment data of various communities in the target city.

[0019] The spatial regression model building module and the multi-objective optimization model building module are used to construct multi-objective optimization mathematical models for carbon emission reduction, time consumption reduction, and cost reduction.

[0020] The response rate calculation module is used to calculate the response rate of a single community under a single variable unit of control intensity.

[0021] The multi-objective optimization model building module is used to construct a multi-objective optimization mathematical model for reducing carbon emissions, time consumption, and costs.

[0022] The response classification module is used to classify the response of regulatory variables and regulatory communities based on the decision tree algorithm.

[0023] The optimization module is used to solve multi-objective optimization models based on the NSGA-II algorithm with a response-first approach. Specifically, it includes: first, classifying high-response variables and high-response communities based on the contribution rate of a single community's single variable to the total carbon emissions, average travel time, and average travel cost at the city level under a unit control intensity; then, prioritizing communities and variables based on different response intensities, setting the unit control intensity to 10%, and generating an initial population based on a chaotic sequence using logistic mapping in the intersection of high-response variables and high-response communities; finally, obtaining the Pareto optimal solution set through non-dominated sorting, crowding calculation, and crossover / mutation operations.

[0024] The scenario decision-making module is used to generate control scenarios for different combinations of macroeconomic objectives and extract typical control schemes according to three criteria: minimum number of control communities, minimum number of control variables, and minimum control intensity.

[0025] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for generating urban community built environment control scenarios for multi-objective optimization.

[0026] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for generating urban community built environment control scenarios oriented towards multi-objective optimization.

[0027] The technical solution provided in this application has at least the following technical effects or advantages: 1. Achieving refined regulation at the community level. This invention, by constructing a spatial regression model of travel carbon emissions, travel time, and travel costs at the community level, reveals the differences in the responses of different communities to built environment variables. It can formulate precise regulation plans based on the differentiated characteristics of different communities, overcoming the limitations of macro-level "one-size-fits-all" regulation.

[0028] 2. Effectively addressing the issue of asynchronous regulation among multiple objectives. This invention constructs a multi-objective optimization mathematical model of "carbon emission reduction - time consumption reduction - cost reduction," which can handle potential contradictions during the convergence of multiple objectives. By solving the model, the synergistic and conflicting relationships between different variables in carbon reduction, time consumption reduction, and cost reduction can be identified, providing a basis for scientific decision-making.

[0029] 3. Improve the efficiency of solving large-scale optimization problems. This invention proposes the NSGA-II algorithm with a response priority. It pre-divides high, medium, and low response variables and communities through decision tree classification, sets population priorities for communities and variables based on different response degrees, and uses chaotic sequences to generate the initial population in the intersection of high response variables and high response communities. This significantly improves the search efficiency when facing a large number of communities, multiple variables, and multiple levels of regulation.

[0030] 4. Provides comprehensive decision support for multi-objective scenarios. This invention solves four different combinations of macro-objectives and systematically analyzes the differences in control strategies under different objective preferences. The results show that when the objective of reducing travel costs is added, the number of feasible solutions decreases significantly, confirming the contradiction between carbon reduction and cost reduction for some variables, providing important technical support for planning decisions. Furthermore, typical control scenarios are extracted according to three criteria: minimum number of controlled communities, minimum number of controlled variables, and minimum control intensity, enhancing the operability of the solutions.

[0031] 5. It has good scalability. The method of this invention can be extended to the control of the built environment at the community scale in other cities, and can also be extended to multi-objective optimization problems of other travel modes (such as slow traffic and public transportation), showing good universality and application prospects. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the decision tree classification process for the control variables in this invention.

[0033] Figure 2 This is a schematic diagram of the decision tree classification process for regulating communities in this invention.

[0034] Figure 3 The figure shows the solution results for different carbon reduction and cost reduction objectives under the control scenarios of this invention. Detailed Implementation

[0035] To better understand the above technical solutions, the technical solutions of the present invention will be fully described below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited to these embodiments.

[0036] Example: This example uses 904 communities in Nanjing as the research object, and fully realizes the generation of urban community built environment regulation scenarios for multi-objective optimization. In this example, different levels of regulation are applied to different variables and different communities, resulting in different regulation effects. Based on the foregoing analysis, three spatial regression models at the community scale—travel carbon emissions, resident travel time, and resident travel costs—generated linear fitting equations for the three dependent variables for all communities. Further testing with single-variable, single-community unit regulation intensity allows for the calculation of the contribution rate to total urban travel carbon emissions, average travel time, and average travel costs. Among the 22 regulation variables, except for community aging and the proportion of residents aged 14 and under, which are used to simulate changes in the natural population structure, the remaining regulation variables will generate different regulation costs under different regulation intensities and for different communities. Specifically, land use environmental variables mainly represent land development costs for different land use types; transportation facility environmental variables mainly represent the construction costs of corresponding facilities; and transportation policy and operational environmental variables mainly represent operation and management costs.

[0037] Therefore, the essence of the multi-objective solution process of this technical solution is to solve the combination of contribution rates of the control community, control variables, and control intensity under the condition of minimizing the control cost, in order to meet the goals of carbon reduction, time consumption reduction, and cost reduction at the city level.

[0038] First, the study area and data preparation were carried out, and independent variables of the built environment were collected, including land use data, transportation facility data, transportation policy data, and operational environment data. (1) Land use data: including the proportion of aging population in the community (U1), the proportion of people aged 14 and under in the community (U2), the density of jobs in the community (U4), the mixed use of land in the community (U5), the distance from commercial and office land in the community to the district-level administrative center (U7), and the distance from residential land in the community to the high-end commercial center (U8), etc.

[0039] (2) Traffic facility data: including community road network density (F1), community intersection density (F2), community bus network density (F4), community rail station density (F5), community bus stop density (F6), distance from residential land to bus stop within the community (F8), distance from commercial and office land to subway station within the community (F9), community shared bicycle density (F11), etc.

[0040] (3) Traffic policy and operating environment data: including community public parking space density (M1), area of ​​community land within a 30-minute bus / subway circle (M3), area of ​​community commercial and office land within a 15-minute walking circle (M5), average speed of community buses (M6), average bus passenger load factor in the community (M7), average rail passenger load factor in the community (M8), coefficient of variation of community bus operating environment (M9), coefficient of variation of community bus riding environment (M10), etc.

[0041] The specific steps of the method in this embodiment are as follows: The first step is to construct a spatial regression model of community-scale travel behavior. This embodiment focuses on 904 communities in Nanjing City, establishing multi-scale, geographically weighted spatial regression models for all-day road segment carbon emissions, average resident travel time, and average travel cost at a fine spatial scale at the community level. Through spatial regression analysis, linear fitting equations for the three dependent variables are obtained for each community. The models pass the spatial autocorrelation test and can be used for subsequent response rate calculations. The expression for the multi-scale, geographically weighted spatial regression model is as follows: ; In the formula, Let be the dependent variable value of the i-th spatial region; n is the total number of built environment variables; It is the j-th explanatory variable of the i-th spatial region; Let i be the centroid coordinates of the i-th spatial region; For the intercept term; Represents the first Bandwidth used for regression coefficients of individual variables For the i-th spatial region The regression coefficients of each variable; This represents the error between the dependent variable value and the predicted value for the i-th spatial region.

[0042] According to the definition of the multi-scale geographical weighted regression model, this model uses a linear function for local weighted regression. Therefore, it can specifically combine the calculation results of the travel carbon emission spatial regression model, the average travel time spatial regression model, and the average travel time spatial regression model to output the fitting equations for the total travel carbon emissions, average travel time, and average travel cost of a specific community.

[0043] If the significant variables in the spatial regression model of total travel carbon emissions in Zijin Xiaoying Community, Xuanwu District are U1, U4, F1, F4, F8, F9, M3, M5, and M6, then the fitting equation for its total travel carbon emissions is: ; In the formula: The total carbon emissions from travel in Zijin Xiaoying Community, Xuanwu District, are represented by U1, U4, F1, F4, F8, F9, M3, M5, and M6, which are the corresponding built environment variables mentioned above.

[0044] The second step is to calculate the response rate of a single-community, single-variable unit of regulatory intensity. Based on the spatial regression model obtained in the previous step, and combined with the spatial regression fitting equations for total travel carbon emissions, average travel time, and average daily travel cost in different communities, the city-level change rate under a 10% univariate adjustment was calculated for each specific community. For example, a 10% increase in land use mix in the Moling Road community would result in changes of -0.00095%, -0.00275%, and -0.00001% in total travel carbon emissions, average travel time, and average travel cost at the city level, respectively. This indicates that under the U5 variable adjustment for this community, the response strength of the change in average travel time is higher than that of the changes in total carbon emissions and travel costs. Calculations were performed for all variables in all communities, resulting in a summary table (Table 1) of the unit-strength adjustment response of the univariate for each community. Table 1. Changes in "Carbon Emissions - Time Consumption - Costs" under Unit Control Intensity in a Single Community and with a Single Variable

[0045] Similar calculations were performed on all variables across all communities to obtain a summary table of unit-level regulatory response rates for each community and its single variable. Taking the land use mix (U5) variable as an example, the response rates for some communities are shown in Table 2: Table 2. Changes in "Carbon Emissions - Time Consumption - Costs" under Unit Control Intensity in All Communities (Example of U5 Variables)

[0046] The third step is to construct a multi-objective optimization mathematical model for reducing carbon emissions and costs in travel. Based on the response rate data calculated in the second step, the following multi-objective optimization mathematical model is established: ; ; in, This represents the sum of carbon reduction contributions under the selected variables and communities. This represents the sum of the time-saving contribution rates under the selected variables and the community. The sum of the cost reduction contribution rates under the selected variables and communities; i represents the i-th community, m represents the total number of communities in the city. j represents the j-th variable, and n represents the total number of variables considered; Indicates community In variables The corresponding carbon reduction contribution rate under the unit control intensity Indicates community In variables The contribution rate of time reduction under the unit control intensity, Indicates community In variables The contribution rate of cost reduction under the unit's regulatory intensity; Indicates community variables The intensity of regulation, Indicates community variables The unit control cost; This indicates the carbon emission reduction target set at the city level. This represents the time reduction target set at the city level. This indicates the cost reduction target set at the city level.

[0047] Step 4: Decision tree-based classification of regulatory variables and community response. Existing technologies widely used data classification methods include K-means algorithm, decision tree algorithm, logistic regression, and Bayesian classifier. Among them, the decision tree algorithm is widely used in machine learning and data analysis due to its high classification efficiency. Based on the characteristics of the response dataset, this embodiment specifically selects the CART decision tree algorithm for classifying the control variables and the high, medium, and low response levels of the control community. The formula for calculating the GINI coefficient is as follows: ; In the formula, K is the number of label categories in the decision tree training set. In this paper, both variables and communities are divided into three categories. For the variable or community sample size at this node, For this node belonging to The number of variables or community samples in the class.

[0048] Step 4.1: Division of the response degree of the regulation variable When using the CART decision tree algorithm to classify control variables, the sum of the city-level contribution rates of all communities corresponding to each control variable is calculated: the change rate of total travel carbon emissions is denoted as feature1; the change rate of average travel time is denoted as feature2; and the change rate of average travel cost is denoted as feature3. Ten representative variables are selected, and the above data is used as the training set to train the variable classification tree model. The classification process is as follows: Figure 1 As shown in Table 3: Table 3. Results of Response Determination of Regulation Variables

[0049] The results show that, with a 10% control level, seven variables with high response will cause changes of -2.97%, -2.83%, and -1.71% in total carbon emissions, average travel time, and average travel cost at the city level, respectively. Six variables with medium response will cause changes of -1.06%, -0.98%, and 0.23% in total carbon emissions, average travel time, and average travel cost at the city level, respectively. Seven variables with low response will cause changes of 3.85%, 0.11%, and 1.14% in total carbon emissions, average travel time, and average travel cost at the city level, respectively, at a 10% control level. It is evident that the inclusion of medium-response variables will decrease total carbon emissions and average travel time while increasing average travel cost. Conversely, the inclusion of low-response variables will increase all three factors.

[0050] Step 4.2: Regulating Community Response Levels When using the CART decision tree algorithm to classify communities, the mean change rates for high-response and medium-response variables in a single community are statistically analyzed: the change rate of total travel carbon emissions is denoted as feature1; the change rate of average travel time is denoted as feature2; and the change rate of average travel cost is denoted as feature3. Simultaneously, the proportion of high-carbon spaces in a community (the ratio of high-carbon grid area to community area) is denoted as feature4, and the proportion of high-carbon individuals among residents (the ratio of high-carbon users to total users) is denoted as feature5. One hundred typical communities are selected as the training set to train the community classification tree model. The classification process is as follows: Figure 2 As shown in Table 4: Table 4. Results of Community Response Level Classification

[0051] The results showed that 241 communities were highly responsive to high- and medium-response regulation, with approximately 50% of their residents having high-carbon spaces or high-carbon individuals. These communities exhibited high responses to total travel carbon emissions, average travel time, and average travel costs. 334 communities had medium responses, with an average high-carbon space ratio of 97.80%, while the proportion of high-carbon individuals among residents was only 29.24%. This indicates that high-carbon spaces in these communities were primarily due to work-related travel. Under individual regulation of high- and medium-response variables, medium-response communities also contributed significantly to reducing total travel carbon emissions, but their contribution to reducing residents' travel time and costs was lower than that of high-response communities. 329 communities had low responses, primarily consisting of low-carbon spaces, but 39.07% of their residents were high-carbon individuals. This suggests that these residents typically traveled long distances and their travel routes were generally outside their own communities. Therefore, although low-response communities contributed less to total travel carbon emissions, their sensitivity to regulation of residents' travel time and costs was higher than that of medium-response communities.

[0052] Step 5: Solve using the NSGA-II algorithm based on response priority. The traditional NSGA-II algorithm uses a randomly generated initial population, and the crossover and mutation processes are also performed randomly, resulting in a low convergence speed when facing large-scale search sets. This embodiment, considering traffic policy and operational indicators, introduces variable responsiveness and community responsiveness labels into NSGA-II, and uses chaotic sequences to generate the initial population from high-response variables and high-response communities for iterative solution.

[0053] This embodiment employs the NSGA-II algorithm, based on response priority, to solve multiple objectives under a search scope of 904 regulatory communities, 20 regulatory variables, and 10 levels of regulatory intensity. The algorithm parameters in this embodiment are set as follows: population size 200, maximum number of iterations 500, crossover probability 0.9, and mutation probability 0.1.

[0054] The specific steps are as follows: Step 5.1: Establish the regulation subset Establish a subset of the cross-variables and the rate of change of regulation in all communities under different levels of regulation, and select the location for each individual in the subset under regulation level k. With change rate and cost The assignment.

[0055] Step 5.2: Generate the initial solution for chaos Based on the set target values, the rates of change in travel carbon emissions, travel time, and travel costs for high-response communities under changes in all high-response variables were calculated. Under a 10% control level for all high-response variables, the total carbon emissions change rate for all high-response communities was -0.89%, the travel time change rate was -0.78%, and the cost change rate was -0.48%. Therefore, a control level interval of [10%, 20%] was selected, and an initial population based on a chaotic sequence was generated using a logistic mapping at the intersection of high-response variables and high-response communities. The individual's characteristics under the control level of [10%, 20%] were then considered. The attributes are passed to the objective function and constraints in the third step.

[0056] Step 5.3: Non-dominated sorting and crowding calculation Individual ranking is performed based on non-dominated ranking and crowding distance. For example, in this embodiment, under a 10% control level, individual X... 12 The resulting overall cost Y is less than X 22 And the carbon emission change rate E sum Travel time variation rate T sum Travel cost change rate C sum All are not greater than X 22 Individual X is called X. 12 Dominant individual X 22 If individual X 12 If an individual is not dominated by other individuals, it becomes a non-dominated individual and its crowding level is calculated.

[0057] Step 5.4: Crossover Mutation and Population Merging Perform crossover and mutation operations according to the individual order, merge the current parent and offspring into a population, and continue to perform step 5.3 on the merged population.

[0058] Step 5.5: Iteration Termination Judgment Determine whether the target constraints are met. , , If the maximum number of iterations has been reached (500 generations in this embodiment), the algorithm ends and the optimal solution set is output. If the condition is not met, the algorithm returns to step 5.3 and continues to sort the merged new population.

[0059] Step 6: Multi-objective scenario decision analysis To comprehensively interpret the built environment regulation scenario decision-making mechanism for optimizing the multi-dimensional travel distribution of "carbon emissions-time consumption-cost," this embodiment sets and solves the targets for calculating total carbon emissions, travel time consumption, and travel costs at the city level using different units. Specifically, it sets four types of target combinations for solution: Target Combination A: A 1% reduction in total carbon emissions from travel, with no increase in travel time and cost; Target combination B: Total carbon emissions from travel decrease by 1%, travel time decreases by 1%, and travel costs do not increase; Target combination C: Total carbon emissions from travel decrease by 1%, travel time does not increase, and travel costs decrease by 1%; Target combination D: Total carbon emissions from travel, travel time, and travel costs all decrease by 1%.

[0060] The solution results are as follows Figure 3 As shown, typical control scenarios are further extracted from the solution set according to three criteria: minimum number of control communities, minimum number of control variables, and minimum control intensity. Among them: Figure 3 (a) is the solution result of the objective combination A [-1%, -0%, -0%], that is, the Pareto optimal solution set distribution under the objective of reducing the total carbon emissions of travel by 1% and not increasing travel time and cost. Figure 3 (b) is the solution result of the objective combination B [-1%, -1%, -0%], that is, the Pareto optimal solution set distribution under the objectives of reducing total carbon emissions from travel by 1%, reducing travel time by 1%, and not increasing travel costs. Figure 3 (c) represents the solution result of the objective combination C [-1%, -0%, -1%], which is the Pareto optimal solution set distribution under the objectives of reducing total carbon emissions from travel by 1%, not increasing travel time, and reducing travel costs by 1%. Figure 3 (d) shows the solution results for the objective combination D [-1%, -1%, -1%], i.e., the Pareto optimal solution set distribution under the objective of reducing total travel carbon emissions, travel time, and travel costs by 1%. Each subplot illustrates the distribution characteristics of the Pareto optimal solution set under different objective constraints, reflecting the influence of different objective combinations on the number of feasible solutions and the distribution of the solution set. The specific solution results for the four objective combinations are as follows: (1) Result of solving the objective combination A [-1%, -0%, -0%]: The optimal solution set obtained by the model solution consists of 37 non-dominated solutions, and the typical solutions are shown in Table 5: Table 5

[0061] (2) Result of solving the objective combination B [-1%, -1%, -0%]: The optimal solution set obtained by the model solution consists of 11 non-dominated solutions, and typical solutions are shown in Table 6: Table 6

[0062] (3) Result of solving the target combination C [-1%, -0%, -1%] The optimal solution set obtained from the model solution consists of four non-dominated solutions, and typical solutions are shown in Table 7: Table 7

[0063] (4)Solution results of the target combination D [-1%, -1%, -1%]: The optimal solution set obtained by model solving consists of 9 non-dominated solutions. The typical solutions are shown in Table 8: Table 8

[0064] It can be seen from the solution results that: When only the target of reducing travel carbon emissions (Target A) is constrained at the macro level, the optimal solution set obtained by model solving consists of 37 non-dominated solutions. It can be further divided into three categories: the fewest regulated communities, the fewest regulated variables, and the smallest regulation intensity. The typical solutions are shown in Table 5. It can be seen that the solution with the fewest regulated communities (136 communities) generally includes a combination of high and medium response variables and two levels of regulation intensity, and the regulated communities are limited to high-response communities; the solution with the fewest regulated variables (2 - 3 variables) is generally limited to high-response communities and high-response variables, but the regulation intensity is relatively high (20%); while for the solution with the smallest regulation intensity (10%), some medium-response variables or medium-response communities need to be added to achieve the target.

[0065] When both the target of reducing travel carbon emissions and travel time consumption (Target B) are constrained, the solution results are reduced to 11. Compared with Target A, the number of regulated communities has increased, indicating that a wider regulation range is required to achieve both carbon reduction and time reduction simultaneously. Since the total travel carbon emissions and the changes in travel time consumption are relatively consistent, the solution results of this target have certain similarities with those of Target A. In particular, the solutions with fewer regulated communities and the fewest regulated variables are basically the same, only different in the solutions with smaller regulation intensity. The typical solutions are shown in Table 6. It can be seen that when both the total travel carbon emissions and the average travel time consumption decrease by 1% simultaneously, the number of regulated communities that need to be included further increases.

[0066] When the reduction of travel cost is also added to the target constraints (Target C), the number of the optimal solution set is significantly reduced to 4, which verifies that the regulation response direction of travel cost distribution to some built environment variables is inconsistent with carbon emissions and travel time consumption. From the solution processes of the first two types of targets, it can be observed that the travel cost changes less compared with the total carbon emissions and travel time consumption. Especially in the solutions with the fewest regulated variables, it has inconsistent change characteristics with carbon emissions and travel costs. Therefore, there are certain differences between the solution results of Target C and the first two types. Similarly, the typical solutions are shown in Table 7 according to the fewest regulated variable communities, the fewest regulated variables, and the smallest regulation intensity. It can be seen that when the target of reducing travel cost by 1% is added, the overall number of regulated variables becomes smaller, while the number of regulated communities increases.

[0067] When all three indicators are required to decrease by 1% (target D), the optimal solution set obtained by the model consists of 9 non-dominated solutions. In the solution with the fewest control communities, compared with target A which only considers carbon emission reduction, the number of control variables decreases, the number of variables with a control intensity of 20% increases, and the total number of control communities increases, confirming that some control variables have certain contradictions regarding travel carbon emissions, travel time, and travel costs. As shown in Table 8, in the solution with the fewest control variables and the solution with the smallest control intensity, the number of control variables and control communities does not change significantly under the four targets. This indicates that comprehensive measures to control key variables of the built environment can effectively achieve various set targets.

[0068] The results from different objectives show that when only the macro-level constraint of reducing travel carbon emissions is applied, a large number of solutions are obtained. Secondly, considering both reducing travel carbon emissions and travel time simultaneously further illustrates that if only the macro-level built environment regulation is considered to optimize the total amount and spatial distribution of travel carbon emissions, it may lead to an increase in the average travel time for some community residents. Therefore, these regulation scenarios cannot effectively transmit the guidance function of multi-modal travel combinations to individuals at the micro-level. When reducing travel costs is also included as an objective constraint, the number of optimal solutions decreases significantly, confirming that the direction of the regulatory response of travel cost distribution to some built environment variables is inconsistent with that of travel carbon emissions and travel costs.

[0069] The results of this embodiment show that the technical solution can effectively address the problem of asynchronous regulation among multiple objectives of "carbon emissions-time consumption-cost". It can efficiently solve the problem in a large-scale search space through the NSGA-II algorithm with response priority, and provide differentiated regulation scenarios for different combinations of macro objectives, thus providing scientific and operable technical support for urban planning decisions.

[0070] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A method for generating community built environment regulation scenarios oriented towards multi-objective optimization, characterized in that, Includes the following steps: Step 1: Construct a community-scale multidimensional spatial regression model of travel behavior. For all communities in the target city, establish spatial regression models for the three dependent variables: built environment independent variable and daily road segment carbon emissions, average travel time and average travel cost. Obtain the linear fitting equations for the regulation of the three dependent variables for different built environment variables in each community. Step 2: Calculate the response rate of a single-variable unit regulation intensity for a single community in the built environment—Based on the spatial regression model obtained in Step 1, calculate the change rates of urban-level travel carbon emissions, average travel time, and average travel cost for each community under a single-variable unit regulation intensity, and obtain the carbon reduction contribution rate e corresponding to each community i under a unit regulation intensity of variable j. ij Contribution rate of time reduction t ij , contribution rate of reduced expenses c ij and unit control cost q ij ; Step 3: Construct a multi-objective optimization mathematical model for reducing carbon emissions and costs in travel—Establish a multi-objective optimization model with the objectives of maximizing the total contribution rate of carbon reduction at the city level, maximizing the total contribution rate of time reduction, maximizing the total contribution rate of cost reduction, and minimizing the total cost of regulation. Step 4: Classification of regulatory variables and regulatory communities based on decision tree — Using the CART decision tree algorithm, the regulatory variables and regulatory communities are classified into high, medium and low responsiveness based on the response rate calculated in Step 2. Step 5: Solving using the NSGA-II algorithm based on responsiveness priority—First, based on the contribution rate of a single community and a single variable to the total carbon emissions, average travel time, and average travel cost of urban travel at the unit control intensity, a decision tree algorithm is used to divide high-response variables and high-response communities; then, different responsiveness is used to set population priorities for communities and variables, and logistic mapping is used to generate an initial population based on chaotic sequences in the intersection set of high-response variables and high-response communities, and Pareto optimal solution set is obtained through non-dominated sorting, crowding calculation, and crossover mutation operations. Step 6: Multi-objective scenario decision analysis – Set different combinations of macro-objectives and solve them using the algorithm in Step 5 to obtain the built environment regulation scenarios under each combination of objectives. Then, extract typical regulation schemes according to three criteria: minimum number of regulated communities, minimum number of regulated variables, and minimum regulation intensity.

2. The method for generating community built environment regulation scenarios oriented towards multi-objective optimization according to claim 1, characterized in that, The multi-objective optimization mathematical model in step 3 is as follows: ； The constraints are: ； in, This represents the sum of carbon reduction contributions under the selected variables and communities. This represents the sum of the time-saving contribution rates under the selected variables and the community. The sum of the cost reduction contribution rates under the selected variables and communities; i represents the i-th community, m represents the total number of urban communities, j represents the j-th variable, and n represents the total number of variables considered; Indicates community In variables The corresponding carbon reduction contribution rate under the unit control intensity Indicates community In variables The contribution rate of time reduction under the unit control intensity, Indicates community In variables The contribution rate of cost reduction under the unit's regulatory intensity; Indicates community variables The intensity of regulation, Indicates community variables The unit control cost; This indicates the carbon emission reduction target set at the city level. This represents the time reduction target set at the city level. This indicates the cost reduction target set at the city level.

3. The method for generating community built environment regulation scenarios oriented towards multi-objective optimization according to claim 1, characterized in that, Step 4, the classification of the response rate of the control variables, specifically includes: summing the city-level contribution rates of all communities corresponding to each control variable; using the change rate of total travel carbon emissions as the first feature, the change rate of average travel time as the second feature, and the change rate of average travel cost as the third feature; and classifying them using the CART decision tree algorithm. The formula for calculating the GINI coefficient used for classification is as follows: ； In the formula, K is the number of label categories in the decision tree training set. In this paper, both variables and communities are divided into three categories. For the variable or community sample size at this node, For this node belonging to The number of variables or community samples in the class.

4. The method for generating community built environment regulation scenarios oriented towards multi-objective optimization according to claim 1, characterized in that, Step 4, which involves classifying the community responsiveness, specifically includes: statistically analyzing the average rate of change of high and medium response variables for a single community; using the rate of change of total travel carbon emissions as the first feature, the rate of change of average travel time as the second feature, and the rate of change of average travel cost as the third feature; simultaneously, using the proportion of high-carbon spaces in the community as the fourth feature and the proportion of high-carbon individuals among residents as the fifth feature; and using the CART decision tree algorithm for classification. The high and medium response variables are the responsiveness levels obtained from the classification.

5. The method for generating community built environment regulation scenarios oriented towards multi-objective optimization according to claim 1, characterized in that, Step 5 specifically includes: Step 5.1: Establish the cross subset of all variables and all community regulation change rates under different regulation intensities, and apply the regulation intensity to each individual in the subset. The following site selection With change rate and cost Assigning a value; Step 5.2: Generate an initial chaotic solution. Based on the set target value, calculate the rate of change of travel carbon emissions, travel time, and travel costs for the high-response community under the changes of all high-response variables, and select the control intensity range below and above the target. , The logistic mapping is used to generate an initial population based on chaotic sequences at the intersection of high-response variables and high-response communities. Step 5.3: Perform individual ranking based on non-dominated ranking and crowding distance; Step 5.4: Perform crossover and mutation operations according to the individual order, and merge the current parent generation and offspring generation into a single population; Step 5.5: Determine whether the set target constraints are met. , , If the maximum number of iterations has been reached, the algorithm terminates and outputs the optimal solution set. If the condition is not met, the algorithm returns to step 5.3 to continue sorting the merged new population.

6. The method for generating community built environment regulation scenarios oriented towards multi-objective optimization according to claim 1, characterized in that, The combination of macro-level objectives set in step 6 includes: Target Combination A: A 1% reduction in total carbon emissions from travel, with no increase in travel time or cost; Target combination B: Total carbon emissions from travel decrease by 1%, travel time decreases by 1%, and travel costs do not increase; Target combination C: Total carbon emissions from travel decrease by 1%, travel time does not increase, and travel costs decrease by 1%; Target combination D: Total carbon emissions from travel, travel time, and travel costs all decrease by 1%.

7. The method for generating community built environment regulation scenarios oriented towards multi-objective optimization according to claim 1 or 2, characterized in that, The unit control intensity is set to 10%.

8. A scenario generation system for regulating the built environment of urban communities oriented towards multi-objective optimization, characterized in that, include: The data acquisition module is used to collect land use data, transportation facility data, transportation policy and operating environment data of various communities in the target city; The spatial regression model building module is used to establish a spatial regression model between the built environment independent variable and three dependent variables: daily road segment carbon emissions, average travel time for residents, and average travel cost. The response rate calculation module is used to calculate the response rate of a single community under a single variable unit of control intensity. The multi-objective optimization model building module is used to construct a multi-objective optimization mathematical model for carbon emission reduction, time consumption reduction, and cost reduction. The response classification module is used to classify the response of regulatory variables and regulatory communities based on the decision tree algorithm. The optimization solution module is used to solve multi-objective optimization models based on the response-first NSGA-II algorithm; The scenario decision-making module is used to generate control scenarios for different combinations of macroeconomic objectives and extract typical control schemes according to three criteria: minimum number of control communities, minimum number of control variables, and minimum control intensity.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the method for generating urban community built environment regulation scenarios for multi-objective optimization as described in any one of claims 1 to 7.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method for generating urban community built environment regulation scenarios oriented towards multi-objective optimization as described in any one of claims 1 to 7.

Images