A multi-objective optimization processing method and system for low-carbon updating of urban communities

Abstract

The application discloses a kind of urban community low-carbon renewal multi-objective optimization processing method and system, belong to big data technical field.It includes the following steps: obtaining data and processing data, the data obtained includes geographic space data and population data, constructs multi-objective optimization agent database and space database;According to the agent database constructed, create multi-objective optimization simulation model;According to simulation model and combine external influence parameter, design update process mechanism, formulate update scheme selection scene and condition and agent interaction, generate optimal update strategy;Extract and visualize multi-objective optimization result of update strategy.Compared with prior art, the beneficial aspects of the present application are that multi-objective optimization based on multi-agent interaction and spatiotemporal visualization of update process are realized, the interaction between and within update subjects such as residents and government is simulated, and multi-objective optimization such as minimum cost, maximum carbon reduction and maximum resident satisfaction is completed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data, and more particularly to a multi-objective optimization processing method and system for urban community low-carbon renewal. BACKGROUND

[0002] As a basic component of urban structure, the carbon emissions of a community account for a certain proportion in the overall carbon emissions of a city. Community renewal, also known as community reconstruction, is an improvement and update of various problems or needs in a community, aiming to improve the quality of living environment, promote social harmony and development, and meet the needs and expectations of residents.

[0003] However, current community renewal actions are mostly limited to an economic perspective, or only focus on a single dimension such as society, ecology, and culture, lacking comprehensive consideration in multi-objective optimization methods. The multi-objective optimization method of community low-carbon renewal is relatively insufficient, and the measurement is mostly based on a single target, ignoring the complex adaptive system characteristics of urban renewal based on different combinations of update subjects, funds, and elements. Therefore, how to coordinate multiple subjects, balance the cost of renewal and the long-term carbon reduction goals and tasks, and scientifically complete planning decisions through public participation has become an important problem in current community low-carbon renewal. With the continuous evolution of information technology, geographic information systems and artificial intelligence provide new ideas and analysis methods for planning decisions, and multi-objective optimization algorithms can balance multiple conflicting goals to find a relatively optimal solution, and are widely used in planning decision research.

[0004] Currently, multi-objective optimization based on genetic algorithms has been gradually applied in multiple fields, but most of them are based on programming software platforms such as MATLAB, Python, and Grasshopper, or although they can achieve chart visualization of multi-objective optimization result data based on static data, they cannot interact with resident agents and dynamically display resident's update scheme selection in space and time. Although Anylogic software can display the update status of residents under different parameters through multi-agent simulation in real time, its internal optimization module only supports single-objective optimization and cannot achieve effective integration of multi-objective optimization and complex interactive dynamic simulation of agents. SUMMARY

[0005] 1. Technical problems to be solved

[0006] In view of the problems of single current community low-carbon updating measurement target and insufficient dynamic updating consideration of implementation subject demand in the prior art, the present application provides a city community low-carbon updating multi-target optimization processing method and system, which can realize spatial digitalization construction and dynamic demand analysis of multiple updating subjects, simulate the interaction of intelligent agents such as residents and government, maximize the carbon reduction effect, minimize the updating investment cost and maximize the resident satisfaction as the updating target, and perform multi-target iterative optimization on the quantitative indexes in the updating strategy, thereby supporting the planning and decision-making of adaptive low-carbon oriented updating.

[0007] 2. Technical solution

[0008] The object of the present application is achieved by the following technical solution.

[0009] The content part of the present application is used to introduce the concepts in a brief form, which will be described in detail in the specific embodiment part. The content part of the present application is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0010] Some embodiments of the present application propose a city community low-carbon updating multi-target optimization processing method and system to solve the technical problems mentioned in the background part.

[0011] As a first aspect of the present application, some embodiments of the present application provide a city community low-carbon updating multi-target optimization processing method, comprising the following steps: acquiring data and processing the data, the acquired data including geographic space data and population data, and constructing a multi-target optimization intelligent agent database and a space database; creating a simulation model capable of multi-target optimization according to the constructed intelligent agent database; generating an optimal updating strategy by designing an updating flow mechanism, formulating updating scheme selection scenarios and conditions and intelligent agent interaction according to the simulation model and in combination with external influence parameters; and extracting and visualizing the multi-target optimization results of the optimal updating strategy.

[0012] Further, the data processing includes acquiring building original carbon emission data, acquiring preset carbon emission reduction data, creating an updating scheme and constructing an intelligent agent database; the building original carbon emission data includes residential building original carbon emission data and non-residential building original carbon emission data.

[0013] Further, the geographic space data includes street terrain data, plot boundary data, building data, green land data, water system data and road network data; the building data includes building contour data, building latitude and longitude data, building plot number data and building attribute data.

[0014] Further, the spatial database is composed of block terrain data, plot boundary data, building contour data, green land data, water system data and road network data.

[0015] Further, in the process of constructing the agent database, the building and the corresponding residents of the building are regarded as a sample, and a community full sample data set is constructed; the samples in the community full sample data set are sampled to obtain a sampling sample set, and the resident data of the sampling sample in the sampling sample set is collected; the resident data of all samples in the community full sample data set is obtained through data augmentation according to the building attribute data.

[0016] Further, a first data table is established according to the building attribute data in the community full sample data set, a second data table is established according to the building attribute data of the sampling data set, and the most similar row in the first data table is found in the second data table, and the sample most close to the unsampled sample in building attribute is found in the sampling sample.

[0017] Further, in the process of searching, the Euclidean distance is used as a similarity measurement standard to obtain the resident data of all samples in the community full sample data set; it is assumed that the feature of each sample is represented by a vector, the feature vector of the unsampled sample i is represented as X i =(z i1 ,z i2 ,…,z in ),the building attribute data of the unsampled sample i is represented as z ik ; the feature vector of the sampled sample j is represented as X j =(z j1 ,z j2 ,…,z jn ),the building attribute data of the sampled sample j is represented as z jk ; k is an index used to traverse the dimension of each vector, k is a natural number, and the expression for calculating the Euclidean distance is as follows:

[0018]

[0019] In the formula, d represents the Euclidean distance between the unsampled sample i and the sampled sample j, and n represents the dimension of the vector, that is, the number of building attribute variables.

[0020] Further, the process of creating a simulation model includes: adding agents according to the agent database, and embedding NSGA-Ⅱ; the agents include a first agent group, a second agent group and a NSGA-Ⅱ single agent.

[0021] Further, the implementation process of the NSGA-II includes: initializing a population, generating random individuals; performing non-dominated sorting and calculating the crowded distance; generating a child population through selection, crossover and mutation; merging the parent and child populations, performing non-dominated sorting and crowded distance calculation again to generate a new population; through the iteration cycle of selecting the parent, performing crossover and mutation and updating the population, the Pareto optimal solution set is obtained.

[0022] As a second aspect of the present application, some embodiments of the present application provide a system based on the above-mentioned urban community low-carbon update multi-objective optimization processing method, comprising a data acquisition module: acquiring data and processing the data, the acquired data including geographic spatial data and population data, constructing a multi-objective optimization agent database and a spatial database; a model construction module: creating a simulation model capable of multi-objective optimization according to the constructed agent database; a strategy optimization module: according to the simulation model and combining external influence parameters, generating an optimal update strategy by designing an update process mechanism, selecting scenarios and conditions for an update scheme, and intelligent agent interaction; a result output module: extracting and visualizing the multi-objective optimization results of the optimal update strategy.

[0023] 3. Beneficial effects

[0024] Compared with the prior art, the advantages of the present application are that: the present scheme can realize multi-objective optimization based on multi-agent interaction and spatio-temporal visualization of the update process, simulate the interaction between and within the update-related subjects such as residents and government, complete multi-objective optimization of minimum cost, maximum carbon reduction, and maximum resident satisfaction, and realize the visualization of long-time sequence community low-carbon update progress on the basis of spatial digitalization construction, providing scientific and quantitative index suggestions for the next update action, assisting the update subjects in evaluating investment cost, carbon reduction effect and update timing, and supporting urban community low-carbon adaptive update. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 Flowchart of the urban community low-carbon update multi-objective optimization processing method in an embodiment of the present application;

[0026] Figure 2 Flowchart of the urban community low-carbon update multi-objective optimization processing method in an embodiment of the present application;

[0027] Figure 3 Flowchart of the urban community low-carbon update multi-objective optimization processing method in an embodiment of the present application;

[0028] Figure 4 Flowchart of the urban community low-carbon update multi-objective optimization processing method in an embodiment of the present application;

[0029] Figure 5A time line graph of positive influence parameters of a community low-carbon renewal scheme in an embodiment of the present application;

[0030] Figure 6 A time line graph of the implementation proportion of a renewal scheme of community low-carbon renewal in an embodiment of the present application;

[0031] Figure 7 A time line graph of the current annual renewal input value of community low-carbon renewal in an embodiment of the present application;

[0032] Figure 8 A time line graph of the cumulative renewal carbon reduction amount of community low-carbon renewal in an embodiment of the present application;

[0033] Figure 9 A time line graph of feedback parameters of community low-carbon renewal in an embodiment of the present application;

[0034] Figure 10 A time sequence diagram of the update selection transition of an update object in an embodiment of the present application. DETAILED DESCRIPTION

[0035] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0037] According to Figures 1 to 10 As shown in the figure, a multi-objective optimization processing method for urban community low-carbon renewal of the present application includes the following steps: obtaining data and processing the data, constructing a multi-objective optimization intelligent agent database and a spatial database; creating a simulation model capable of multi-objective optimization according to the constructed intelligent agent database; deepening the simulation model design, generating an optimal renewal strategy through the simulation model combined with external influence parameters; extracting multi-objective optimization results, visualizing the community low-carbon renewal process based on the spatial database.

[0038] As shown in the figure, in a specific embodiment, the specific steps of the multi-objective optimization processing method for urban community low-carbon renewal include: Figure 1

[0039] S1, constructing a database

[0040] Obtain data and process the data, construct a multi-objective optimization intelligent agent database and a spatial database.

[0041] Specifically, the obtained data includes geographic spatial data and population data.

[0042] The geographic spatial data includes street terrain data, plot boundary data, building data, green space data, water system data, and road network data.​

[0043] Building data includes building outline data, building latitude and longitude data, building plot number data, and building attribute data. Building attribute data includes building ownership, architectural style, building quality, architectural heritage protection strategy type, building function, number of floors, number of residents, building perimeter, and building area.

[0044] Specifically, the spatial database consists of block terrain data, plot boundary data, building outline data, green space data, water system data and road network data.

[0045] In a specific embodiment, the collected data is processed to obtain carbon emission data and construct an agent database, and the process is as follows:

[0046] (1) Obtaining raw building carbon emission data

[0047] Based on the geographic space data, the building data within the corresponding range of the community is locked, and the original carbon emission data of the building is calculated according to the building data and the population data corresponding to the building data.

[0048] Specifically, buildings include residential buildings and non-residential buildings, and the raw carbon emission data of buildings include the raw carbon emission data of residential buildings and the raw carbon emission data of non-residential buildings.

[0049] The raw carbon emission data of residential buildings includes transportation carbon emission data and living carbon emission data, and the raw carbon emission data of residential buildings is calculated based on population data. The expression for obtaining transportation carbon emission data is:

[0050] C t =∑D u *M u

[0051] Where C t is the total carbon emissions from community transportation, in kg / year; u is the type of transportation, D u represents the annual travel distance under transportation mode u, M u is the carbon emission coefficient corresponding to different modes of transportation.

[0052] In this embodiment, the carbon emission coefficient values ​​corresponding to different modes of transportation are shown in Table 1.

[0053] Table 1 Carbon emission coefficients of transportation carbon emission data

[0054] Mode of transport Carbon emission factor Walking 0 Bicycle 0 Electric vehicle 0.0696 Motorcycle 0.1136 Private car 0.1786 Taxi 0.1786 Bus 0.0738 Company shuttle 0.0738

[0055] The life carbon emission data includes carbon emission data of water consumption, waste water, electricity consumption, gas consumption and garbage disposal based on population data, and the building and the corresponding residents of the building are regarded as a sample i, and the expression for obtaining the life carbon emission data is:

[0056] C r =∑[R i *(W i *M W +Wi i *P w *M N +E i *M E +Q i *M Q +F i *M F )];

[0057] In the formula, C r is the total life carbon emission of the community, and the unit is Kg / year; R i represents the total number of residents obtained according to the population data, W i represents the water consumption per capita of the residential building, and the unit is Kg / (person·year); M W is the carbon emission coefficient of water consumption, P W is the pollution coefficient of water consumption, M N is the carbon emission coefficient of waste water, E i represents the electricity consumption per capita of the residential building, and the unit is kilowatt-hour / (person·year); M E represents the carbon emission coefficient of electricity consumption, Q i represents the gas consumption per capita of the residential building, and the unit is m 3 / (person·year); M Q represents the carbon emission coefficient of gas consumption, F i represents the garbage production per capita of the residential building, and the unit is Kg / (person·year); M F represents the carbon emission coefficient of garbage disposal.

[0058] Based on the obtained traffic carbon emission data and life carbon emission data, the total community traffic carbon emission and the total community life carbon emission are added to obtain the total residential building original carbon emission, and the expression for calculating the residential building original carbon emission data based on the total residential building original carbon emission is:

[0059]

[0060] In the formula, CarbonInitial r represents the residential building original carbon emission data, and the unit is Kg / year; S a represents the building area; S b represents the per capita living area.

[0061] Specifically, for non-residential buildings, the expression for obtaining the original carbon emission data of non-residential buildings is:

[0062] CarbonInitial o = S a * (W S * (M w + P W * M N ) + E s * M E + F S * M F

[0063] In the formula, CarbonInitial o represents the original carbon emission data of non-residential buildings, with the unit of Kg / year; W S represents the water consumption of non-residential buildings, with the unit of Kg / (m2·year); E s represents the electricity consumption of non-residential buildings, with the unit of kilowatt-hour / (m2·year); F S represents the waste emission of non-residential buildings, with the unit of Kg / (m2·year); M W is the carbon emission coefficient of water consumption, P W represents the pollution coefficient of water consumption, M N represents the carbon emission coefficient of waste water, M E represents the carbon emission coefficient of electricity consumption, M Q represents the carbon emission coefficient of gas consumption, and M F represents the carbon emission coefficient of garbage disposal.

[0064] In this embodiment, the carbon emission coefficients of water consumption, waste water, electricity consumption, gas consumption, and garbage disposal are as shown in Table 2.

[0065] Table 2 Carbon emission coefficients of life carbon emission data

[0066] Carbon emission factor Value M W ]]> 0.0003 P W ]]> 0.8 M n ]]> 0.00025 M E ]]> 0.9044 M Q ]]> 2.9538 M F ]]> 0.3

[0067] Through the above steps, the original carbon emission data of buildings is obtained.

[0068] (2) Creating an update plan

[0069] According to the analysis of carbon emission data, geographic data, and building attribute data, an update plan is created, and the carbon reduction amount and input value of the update plan are obtained.

[0070] ​Specifically, the updating scheme includes a low-amplitude updating scheme, a medium-amplitude updating scheme, and a high-amplitude updating scheme. Let l represent the updating scheme type, l=1 represents selection of the low-amplitude updating scheme; l=2 represents the medium-amplitude updating scheme; l=3 represents the high-amplitude updating scheme; let p l For the updating strategy, p1 represents a positive impact parameter of the low-amplitude updating scheme; p2 represents a positive impact parameter of the medium-amplitude updating scheme; and p3 represents a positive impact parameter of the high-amplitude carbon reduction updating scheme.

[0071] The specific process of obtaining the carbon reduction amount and the input value of the updating scheme is as follows:

[0072]

[0073] In the formula, CarbonReduce l represents the annual carbon reduction amount of a family when a resident selects an l-type updating scheme, h represents a carbon reduction measure for measuring the carbon reduction effect in units of people, and CMeasure h represents the carbon reduction amount per person of the carbon reduction measure h, and v represents a carbon reduction measure for measuring the carbon reduction effect in units of area, and CMeasure v represents the carbon reduction amount per unit area of the carbon reduction measure v, and S a represents the building area; S b represents the per capita living area.

[0074]

[0075] In the formula, Investment l represents the updating and reconstruction input when a resident selects an l-type updating scheme, e represents a carbon reduction measure for measuring the carbon reduction input in units of people, and IMeasure e represents the input value per person of the carbon reduction measure h, and f represents a carbon reduction measure for measuring the carbon reduction input in units of area, and IMeasure f represents the input value per unit area of the carbon reduction measure f, and S a represents the building area; S b represents the per capita living area.

[0076] (3) Constructing an agent database

[0077] Based on the geospatial data, building original carbon emission data and population data, the building and the residents corresponding to the building are regarded as a sample, a community full sample data set is constructed, the samples in the community full sample data set are sampled to obtain a sampling sample set, and the resident data of the sampling sample in the sampling sample set is collected, and the resident data of all samples in the community full sample data set is obtained according to the building attribute data. According to the building original carbon emission data, the building latitude and longitude data, the building block number data, the carbon reduction amount and the input value data of the updating scheme, and the resident data, an agent database is constructed.

[0078] Specifically, the samples in the community full sample data set include building attribute data, building latitude and longitude data, building block number data, building original carbon emission data, carbon reduction amount and input value of updating scheme, and the data collected from the sampling sample is the resident data of the sampling sample. Data enhancement refers to a technology for expanding training data by using algorithms, and a method for expanding a training data set by generating more similar generated data using a small amount of data prior knowledge. Through the process of data enhancement of the sampling sample, the resident data of all samples in the community full sample data set is obtained. This step can effectively expand the data set, propose a remedy scheme for the scene where full sample research cannot be realized in the updating practice process, and at the same time, minimize the data error.

[0079] In the embodiment, the data enhancement adopts a propensity score matching method (PSM). The PSM method is a non-experimental attribution analysis method, which uses observable variables and logistic regression to find a group of comparable control groups (control groups). By calculating the propensity value, multiple confounding variables are integrated, and the matching difficulty is greatly reduced. The PSM method uses a propensity score function to compress the information of a multi-dimensional vector into one dimension, and then matches according to the propensity score. In this way, under the given observable characteristic variables, the sample of the treatment group and the sample of the control group are made as similar as possible. The basic idea of PSM includes: finding a sample of the control group (control group) to make it as similar as possible (similar) to the sample of the treatment group (experimental group) in the value of the observable variable. In the embodiment, the propensity score matching is realized based on the MatchIt package and the dplyr package in R language.

[0080] Specifically, a first data table is established according to the building attribute data in the community full sample dataset, a second data table is established according to the building attribute data in the sampling dataset, and the sample most similar to the unsampled sample in building attribute is found in the sampling sample by searching the second data table. More specifically, the Euclidean distance is used as a similarity measure in the search process. The smaller the Euclidean distance, the higher the similarity of the two samples in building attribute, effectively screening out the sampling sample most similar to the unsampled sample in building attribute, thereby obtaining the resident data of all samples in the community full sample dataset.

[0081] In a specific embodiment, it is assumed that the characteristics of each sample are represented by a vector, and the characteristic vector of the unsampled sample i is represented as X i i1 i2 in , and the building attribute data of the unsampled sample i is represented as z ik ; the characteristic vector of the sampled sample j is represented as X j j1 j2 jn , and the building attribute data of the sampled sample j is represented as z jk ; k is an index used to traverse the dimension of each vector, and k is a natural number. The expression for calculating the Euclidean distance is as follows:

[0082]

[0083] In the formula, d represents the Euclidean distance between the unsampled sample i and the sampled sample j, and n represents the dimension of the vector, i.e., the number of building attribute variables.

[0084] Traverse the unsampled sample dataset: take unsampled sample 1 as an example, calculate the Euclidean distance between this sample and each sampled sample, select the sampled sample with the smallest Euclidean distance, extract its serial number, and match the resident data corresponding to the sampled sample; unsampled samples 2, 3, 4… are the same. According to the serial number, find the serial number of the resident data and the corresponding resident data, export and save as csv format, which is the resident data of all samples in the community full sample dataset.

[0085] According to the building latitude and longitude data, the building plot number data, the building original carbon emission data, the carbon reduction amount and the input value of the update scheme, and the resident data, an agent database is constructed.

[0086] S2, create a simulation model

[0087] According to the constructed agent database, a simulation model capable of multi-objective optimization is created. ​​​​​​

[0088] Specifically, the problem to be optimized is modeled, and then a simulation model is created, agents are added according to the agent database constructed in step S1, a multi-objective optimization genetic algorithm is embedded, the simulation of the actions and decisions of the multi-agent in the updating process is realized, and the updating strategy iteration optimization experiment is updated. The optimal updating strategy is generated based on the influence factor and the updating scheme. In this embodiment, the influence factor can be the degree of inclination of the updating object to the updating scheme, and the degree of inclination to the updating scheme includes inclination and non-inclination. The specific process is as follows:

[0089] (1) Establishing a problem model

[0090] Establishing a problem model includes defining genes, objective functions, and constraint conditions. Among them, the gene refers to the specific object to be optimized, that is, the index of the updating strategy: positive impact parameter.

[0091] First, the problem to be solved is modeled as an optimizable problem. In this embodiment, the problem to be solved is: how to balance the feedback parameter, the input value required by the updating scheme, and the carbon reduction amount to achieve low-carbon updating based on the data of the agent database? More precisely, by locking the optimal updating object in different time periods and generating the current optimal carbon reduction amount at the same time, the optimal updating strategy is formed.

[0092] Specifically, the feedback parameter can be a value representing the degree of satisfaction of the updating object to the updating scheme.

[0093] Set fitness as a variable for verifying the optimization result; the input value of the current year is fitness1; and the cumulative updating carbon reduction amount is fitness2.

[0094] Specifically, the defined objective function includes minimizing the input value of the current year and maximizing the cumulative updating carbon reduction amount, and the expression of the objective function is as follows:

[0095]

[0096] In the formula, cost il represents the input value under the scenario that the updating object i selects scheme l, and carbonreduce l represents the carbon reduction amount generated by the updating object i selecting scheme l.

[0097] Set the carbon reduction basic amount carbonTask as a constraint condition, and use the carbon reduction basic amount to represent the minimum value of the updating carbon reduction amount fitness2, and the expression of the carbon reduction basic amount is as follows:

[0098] carbonTask = carbonTotal * n + carbonreduce;

[0099] In the formula, carbonTotal represents the total carbon emission of the community since the last update, n represents the proportion of the minimum carbon emission reduction generated by the current update relative to the total carbon emission since the last update, and carbonreduce represents the cumulative carbon reduction of the community since the last update.

[0100] (2) Embedding a multi-objective optimization genetic algorithm

[0101] NSGA-Ⅱ is used as the embedded multi-objective optimization genetic algorithm. NSGA-Ⅱ (Non-dominated Sorting Genetic Algorithm Ⅱ) is a multi-objective optimization genetic algorithm used to solve optimization problems with multiple conflicting objective functions. In this embodiment, since the greater the carbon reduction, the higher the carbon reduction investment value, in order to balance the carbon reduction and the investment value, NSGA-Ⅱ is selected. The process of NSGA-Ⅱ includes: initializing the population, generating individual encoding and evaluating the objective function value of the individual; performing non-dominated sorting and crowded distance calculation on the individuals in the initial population, generating a child population through selection, crossover and mutation operations; merging the parent population and the child population, and performing non-dominated sorting and crowded distance calculation on the individuals in the merged population again to generate a new population; repeating the above process until the preset termination condition is met. Through such a process, a set of optimal balanced solutions between multiple conflicting objectives can be found, i.e. the Pareto optimal solution set.

[0102] Specifically, the preset termination condition can be reaching the maximum number of iterations.

[0103] The update strategy set according to the update scheme (including the low-amplitude update scheme, the medium-amplitude update scheme and the high-amplitude carbon reduction update scheme) is taken as an individual, and the encoding data of each individual includes: genes, fitness, non-dominated level, dominated number, crowded distance, dominated set and dominated number of times.

[0104] NSGA-Ⅱ is embedded in the simulation model, and the specific steps are as follows:

[0105] First, a simulation model is built in AnyLogic, and the model database, resources, environmental base map and other related parameters of the simulation model are configured. The other related parameters include space type, time unit, simulation memory, model time execution mode, model time execution ratio and running configuration memory.

[0106] In a specific embodiment, the space type is defined as GIS, the city block and building shape file are added as the environmental base map, the time unit is set to month, the simulation memory is 50000 Mb, the model time execution ratio is 1, and the running configuration memory is 16384 Mb.

[0107] Secondly, the agent is constructed including the first agent group, the second agent group and the NSGA-Ⅱ single agent, the parameter setting of the first agent group is completed through the mapping agent database, and the encoding data of the updating strategy is established in the second agent group.

[0108] In a specific embodiment, the first agent group is the resident agent group, and the second agent group is the updating strategy agent group. The process of constructing the first agent group is as follows: right-clicking the Main agent in the engineering list, newly building the agent and naming it as Resident, taking the table recording the updating feature data as the basic database table, selecting all columns in the table as the parameters of the first agent group Resident, performing the parameter mapping in the Resident, and completing the creation of the first agent. The variables choice, dyy, zyy, gyy and satisfaction are newly built in the Resident, which are respectively used for recording the selection of the updating object to the updating scheme, the tendency degree of selecting the low-amplitude updating scheme, the tendency degree of selecting the medium-amplitude updating scheme, the tendency degree of selecting the high-amplitude updating scheme and the feedback parameter. The feedback parameter is the value reflecting the satisfaction degree of the updating object to the updating scheme.

[0109] The process of constructing the second agent is as follows: right-clicking the Main agent in the engineering list, newly building the agent and naming it as Zc; double-clicking the second agent group Zc in the engineering, adding the agent components through the panel, copying the first agent group Resident and pasting it into the second agent group Zc except the encoding information, which is used for recording the different feedback of the updating object in different positive influence parameter combinations. Since the positive influence parameter p l is point data, the positive influence parameter interested by the updating object is interval data, therefore, the conversion variable Type l (l=1, 2, 3) is additionally added in the second agent group Zc for the convenience of subsequent comparison. In the embodiment, the specific values of the conversion variable Type l and the corresponding positive influence parameters are shown in Table 3.

[0110] Table 3 the specific values of the conversion variable Type l and the corresponding positive influence parameters

[0111] Update policy <![CDATA[Conversion variable Type l <!-- 8 -->]]> p l = 0 1 0.01 < p l ≤ 0.25 2 0.26 < p l ≤ 0.50 3 0.51 < p l ≤ 0.75 4 0.76 < p l ≤ 1 5

[0112] After the first and second agent groups are constructed, a new NSGA-Ⅱ single agent is created, and the NSGA-Ⅱ is divided into different "states" based on the state chart module in the AnyLogic agent panel. Different "states" are connected through "transitions". The specific process is realized through internal entry action code or calling specified functions to determine whether to enter the next state. The state is the specific situation or mode of the agent at a certain time. In the NSGA-Ⅱ single agent, each state can represent a stage or step of the algorithm, such as initial population initialization, non-dominated sorting, crowding distance calculation, selection, crossover, mutation, etc. Transition refers to the process or event of the agent moving from the current state to the next state, indicating the conversion of the algorithm from one stage to another, such as the conversion from the initialization state to the fast non-dominated sorting state.

[0113] As shown in Figure 2 , the implementation process of NSGA-Ⅱ in the simulation model is as follows:

[0114] 1) Initialize the population and generate random individuals:

[0115] Set the population size, generate random combinations as individuals and calculate the fitness value, select individuals that meet the basic carbon reduction amount carbonTask and generate the initial population, set the size of the target Pareto optimal solution set and the maximum number of iterations, etc. The update strategy p l is a random number with two decimal places and a value between 0 and 1.

[0116] 2) Non-dominated sorting and crowding distance calculation:

[0117] Calculate the domination relationship and the number of dominated individuals for each individual; group the individuals according to the domination relationship; repeatedly extract non-dominated solutions from the ungrouped individuals until all individuals are grouped.

[0118] Specifically, the core of NSGA-Ⅱ is that it uses the fast non-dominated sorting method to sort the individuals in the population. This sorting method is based on the domination relationship between individuals. In optimization problems, if one individual is not worse than another individual in all objective functions and at least one objective function is better than the other individual, it is said that the individual dominates the other individual.

[0119] In one specific embodiment, custom dominates() and nonDominatedSorting() functions are used for non-dominated sorting. The dominates() function is used to determine the dominance relationship between two individuals, and returns true if individual p dominates individual q; otherwise, it returns false. The nonDominatedSorting() function then sorts the individuals according to the dominance relationship and adds them to different fronts. The smaller the front value, the higher the rank of the individuals in the front in the population (i.e., they are non-dominated solutions or less dominated by other solutions).

[0120] The calculation of the crowding distance requires that the individuals are first sorted according to the objective function values, where the crowding distance of the boundary individuals is infinite, and the formula for the calculation of the crowding distance of the intermediate individuals is as follows:

[0121]

[0122] where i d denotes the crowding distance of individual i in the population, denotes the function value of the gth objective function of the (i+1)th individual, i and g are natural numbers, denotes the function value of the gth objective function of the (i-1)th individual, maxf g denotes the maximum function value of the gth objective function of individual i, minf g denotes the minimum function value of the gth objective function of individual i, m denotes the total number of objective functions.

[0123] In this embodiment, the calculation of the crowding distance of the individuals in the population is achieved by the calculateCrowdingDistance() function, and the steps include: for each individual in the population, the crowding distance is initialized to 0; for each objective function, the crowding distances of the best (i.e., the smallest value) and the worst (i.e., the largest value) individuals in the population are set to infinity; for the non-boundary individuals in the population, the crowding distances of each objective function are calculated; the differences are added to obtain the crowding distance of the individual in the given objective space; finally, the calculated crowding distance is returned to each individual.

[0124] 3) Generate a child population by selection, crossover and mutation:

[0125] First, a specified number of parent individuals are selected by using the binary tournament method. Two individuals are randomly selected from the original population, and the non-dominated rank of the two individuals is compared. The individual with the smaller rank value is selected as the parent individual. When the rank values of the two individuals are the same, i.e., they are in the same front, the individual with the larger crowding distance is selected as the parent individual.

[0126] In one specific embodiment, a specified number of parent individuals are co-screened to join the parent population. The original population size is 100, the number of parent individuals is 15, the parent population storing the parent individuals is the set parents, and the set type is ArrayList.

[0127] Secondly, the pairing, crossover and mutation operations are performed. In the NSGA-II algorithm, commonly used crossover operation methods include simulated binary crossover, single-point crossover, multi-point crossover and uniform crossover, and mutation operation methods include polynomial mutation, Gaussian mutation, non-uniform mutation and uniform mutation.

[0128] Specifically, the gene sequence in the present embodiment only involves a single variable type, so uniform crossover and uniform mutation methods are selected, the gene mutation probability is set to 0.1, and the target offspring population size is 100. The specific process of performing the pairing, crossover and mutation operations is as follows: randomly generating two-by-two pairings of a specified number of individuals from the selected parent population, in the present embodiment, 15 pairs of pairings are generated and stored in the set pairs; performing uniform crossover on the genes in the parent individuals in the pairings to generate a corresponding number of offspring individuals; performing uniform mutation on each offspring individual according to the mutation probability, screening offspring individuals that meet the basic amount of carbon reduction to generate an offspring population of a specified size, and storing the offspring population in the set children, and the set type is ArrayList.

[0129] 4) Merging the parent and offspring populations, performing non-dominated sorting and crowded distance calculation again to generate a new population:

[0130] Performing a merging operation on the initial population and the offspring population to generate a merged new population, the size of the merged new population being the sum of the number of individuals in the initial population and the offspring population; calculating the fitness of the merged new population, and performing non-dominated sorting and calculating crowded distance to ensure the integrity of the encoding information of the individuals in the merged population; finally, according to the Pareto level of the individuals and the internal sorting of the level, a specified number of front individuals are selected to form a new population.

[0131] In one specific embodiment, the function merge() is used to merge the population, the merged population is stored in the set merged_population, the set type is ArrayList, and the population size is 200; the method used to select the specified number of individuals in the population is the custom function regenerateChoose(targetNum), and targetNum represents the target number of individuals to be selected, and the data type of targetNum is int. The first 100 individuals in the set merged_population are selected to generate a new population through this step.

[0132] 5) Iterative loop to obtain the Pareto optimal solution set:

[0133] By constantly selecting parents, crossing and mutation, updating the population, until the maximum number of iterations is reached. After the end of the iteration, according to the Pareto level of the individuals in the final population and the internal sorting, a specified number of front individuals are selected as the Pareto optimal solution set.

[0134] In a specific embodiment, the number of iterations is set to 10, and the individual ranked first in the first layer of the front is selected as the Pareto optimal solution.

[0135] (3) Run the simulation model

[0136] After embedding the multi-objective optimization genetic algorithm, the first agent group, the second agent group and the NSGA-II single agent are set up, including the selection of agent time mode, the first update time or recurrence time, and whether to create a dataset for dynamic variables; After the setting is completed, the final simulation model is obtained.

[0137] By running the simulation model, the console output information is verified to obtain the running result of the simulation model.

[0138] In a specific embodiment, the running result of the simulation model is verified by the traceln() function, and the running result is evaluated after being verified without error, the evaluation process includes judging whether the number of individuals in the Pareto optimal solution set and the related coding value reach the preset value, and the traceln() function can be used to call and view the real-time value of the specified variable during the evaluation process. Finally, the optimal update strategy is obtained through the constructed simulation model.

[0139] S3, generating an optimal update strategy

[0140] According to the constructed simulation model, and combining the influence of the external influence parameter on the updated object during the update process, an optimal update strategy is generated.

[0141] Specifically, the external influence parameter can be a parameter that reflects the degree of influence of the updated object on the updated object. The specific process of generating the optimal update strategy is as follows:

[0142] (1) Design the update flow mechanism

[0143] Generate an initial update scheme; execute NSGA-II to generate an optimal strategy. Due to the influence of the positive influence parameter, the influence factor will change for the first time, and due to the influence of the external influence parameter, the influence factor will change for the second time. Therefore, according to the change of the influence factor, the multi-objective optimization strategy needs to be updated again, through the cyclic iteration of the above process, until the target node.

[0144] In this embodiment, the impact factor is the degree of inclination of the update object to the update scheme; the external influence parameter is a parameter reflecting the degree of influence of the updated update object on the update object that has not been updated; and the positive influence parameter reflects the incentive of the update strategy to implement the update scheme.

[0145] As shown in Figure 3 , the flow of the update mechanism. In this embodiment, the update simulation model is run from October 2024, based on the impact factor, an initial update scheme is generated, and the optimal update strategy is obtained; after 6 months, due to the influence of the positive influence parameter, the impact factor changes for the first time, the inclination of the update object to the update scheme changes, and the optimal update strategy of this round changes accordingly; after 12 months, due to the influence of the external influence parameter, the impact factor changes for the second time, the inclination of the update object to the update scheme changes again, and the update scheme changes again, based on this situation, the final update strategy is obtained, and the iteration cycle is repeated according to this rule until 2030 and the update simulation model stops running.

[0146] (2) Develop update scheme selection scenarios and conditions

[0147] The update scheme selection scenario includes the scenario where the update strategy parameter interval is greater than the expected parameter interval of the update object, the update strategy parameter interval is equal to the expected parameter interval of the update object, or is affected by the external influence parameter. When the inclination of the update object to multiple update schemes is consistent and is inclined, a higher amplitude update scheme is selected. Based on the external influence parameter, the calculation formula of the probability of the change of the inclination of the update object to the update scheme is as follows:

[0148]

[0149] Where exist is used to indicate whether there is an update object that updates with this update scheme, exist = true, indicating that there is an update object that updates with this update scheme, exist = false, indicating that there is no update object that updates with this update scheme. TQ indicates the degree of influence of the update object on the external influence parameter, maxQS indicates the degree of influence of the update object with the maximum external influence parameter value, and the values of TQ and maxQS are both integers between 1 and 5.

[0150] Based on geographic spatial data, building data and population data, since there are cases where the update object tends to update but does not update in the actual update process, therefore, in this embodiment, a watch mechanism is established, which is: the more update objects that implement the update scheme, the higher the probability of the impact factor converting to update selection, thereby improving the random probability setting of the inclination of the impact factor.

[0151] (3) Agent interaction

[0152] The process of interaction of the agent is mainly achieved through direct message passing, shared variables, using environment objects as intermediaries, and event-triggered interaction. In the embodiment, the agent interaction includes realizing agent communication based on the message sending and receiving mechanism and realizing agent interaction based on shared variables.

[0153] Specifically, the agent communication is realized based on the message sending and receiving mechanism, and the function send() is used to send a message to a single agent. The subject of the function is as follows: public void send (java.lang.object msg, Agent dest). Wherein, msg refers to a message, and dest refers to the target agent that accepts the message.

[0154] Message receiving is realized through the state diagram inside the agent. The trigger of transition is set as a message through a mode. When the trigger transition type is a specified message and the message is consistent with the information content of the sent message, the link (connection) between the agents can receive the message, trigger the transition, and complete the communication between the agents.

[0155] Specifically, the agent interaction is realized based on shared variables, that is, multiple agents can access and modify shared variables to interact. In the embodiment, the second agent group Zc can generate an update strategy according to the influence factor and the selection of the update scheme, and the update object adjusts the influence factor and the selection of the update scheme according to the update strategy. The influence factor and the selection of the update scheme are shared variables that affect the update object of the feedback update strategy, and finally the optimal update strategy is generated.

[0156] S4, multi-objective optimization result extraction and update process visualization

[0157] After the construction of the update simulation model is completed, in order to better show the multi-objective optimization result under the optimal update strategy, the environment base map setting of the simulation model is improved based on the spatial database, the optimal update object in different time periods is locked, the update simulation model running result is extracted first, that is, the index data is obtained; secondly, through the agent animation and adding analysis charts and other ways, the update process is spatio-temporal dynamic visualized.

[0158] Specifically, the index data includes the positive influence parameter of each update scheme, the update proportion of each update scheme, the input value of the current year, the cumulative updated carbon reduction amount, and the feedback parameter.

[0159] The update of the index data is synchronized with the change of the update strategy adjustment. In the embodiment, the time line chart is used for chart visualization, and the time window of the time line chart is set to 6 years. For example, Figures 5 to 9The time line chart of the relevant index data is shown.

[0160] The update rate rate of each update scheme l The calculation formula is as follows:

[0161]

[0162] Wherein, count l The total number of update objects in the community that select the update scheme l for updating, R i Indicates the total number of residents.

[0163] In this embodiment, the update object is the resident corresponding to the building, and the calculation formula of the feedback parameter satisfaction is as follows:

[0164]

[0165] Wherein, level is the grade value set according to the currently selected update scheme, level = 0, 1, 2, 3;

[0166] When level = 0, it corresponds to no update of the current selection;

[0167] When level = 1, it corresponds to a low-amplitude update scheme of the current selection;

[0168] When level = 2, it corresponds to a medium-amplitude update scheme of the current selection;

[0169] When level = 3, it corresponds to a high-amplitude update scheme of the current selection;

[0170] Wherein, m is the matching degree of the update strategy interval relative to the expected parameter interval of the update object, m = 1, 2, 3;

[0171] When m = 1, it means that the update strategy interval is lower than the expected parameter interval of the update object;

[0172] When m = 2, it means that the update strategy interval is equal to the expected parameter interval of the update object;

[0173] When m = 3, it means that the update strategy interval is higher than the expected parameter interval of the update object;R i Indicates the total number of residents.

[0174] The space visualization is based on the selection of the update scheme of the update object at different time periods, the animation effect of the first intelligent agent changes with the update of the update object, and the function setFillColor() of the AnyLogic simulation model is used to define the graphic animation fill color, and the function body is: Public void setFillColor (java.awt.Color fillColor).

[0175] As shown in Figure 4 , the implementation procedure of the resident intelligent agent update state visualization is shown. Figure 10 As shown in

[0176] As shown in Table 4, the intelligent agent animation fill color corresponding to the selection of the update scheme of the update object in the embodiment is shown.

[0177] Table 4 Intelligent agent animation fill color corresponding to resident update selection

[0178] Selection of update scheme Agent animation fill color Unselected initial state white No update gainsboro Low amplitude update scheme yellowGreen Medium amplitude update scheme green High amplitude update scheme darkGreen

[0179] In one specific embodiment, a city community low-carbon update multi-objective optimization processing system includes a data acquisition module: collecting data and processing the data, constructing a multi-objective optimization intelligent agent database and a space database; a model construction module: creating a simulation model capable of multi-objective optimization according to the constructed intelligent agent database; a strategy optimization module: deepening the design of the simulation model, generating an optimal update strategy through the simulation model combined with external influence parameters; a result output module: extracting multi-objective optimization results, visualizing the community low-carbon update process based on the space database.

[0180] The city community low-carbon update multi-objective optimization processing method and system of the present application, through the construction based on spatial digitalization and the dynamic demand analysis of multiple update subjects, simulates the interaction of intelligent agents such as residents and governments, takes the maximum carbon reduction effect, the minimum update investment cost and the highest resident satisfaction as the update target, realizes the multi-objective iterative optimization of the quantitative indexes in the update strategy to support the planning and decision-making of adaptive update with low-carbon orientation.

[0181] The above description of the application and its embodiments is illustrative and not restrictive, and the application can be practiced in other specific forms without departing from the spirit or essential character thereof. The drawings are intended to be illustrative, and not limiting, and the appended claims should not be limited to the drawings. Thus, if a person of ordinary skill in the art is inspired to design a similar structure and embodiment to the technical solution without departing from the spirit of the invention, it should be within the scope of protection of the patent. In addition, the word "comprising" does not exclude other elements or steps, and the word "one" before an element does not exclude the inclusion of "multiple" such elements. Multiple elements stated in a product claim can also be implemented by one element through software or hardware. The words "first", "second", etc. are used to indicate names, not any specific order.

Claims

1. A method for multi-objective optimization of low-carbon renewal of urban community, comprising the following steps: obtaining and processing data, the obtained data including geographic spatial data and population data, and constructing a multi-objective optimization agent database and a spatial database; the agent database construction process includes: taking a sample unit composed of a building and its corresponding residents as an agent, constructing a community full-sample dataset; sampling the community full-sample dataset to obtain a sampling sample set; collecting resident data based on the sampling sample set, and performing data augmentation according to building attribute data to obtain resident data of all samples in the community full-sample dataset; the data augmentation process includes: establishing a first data table according to the building attribute data in the community full-sample dataset, establishing a second data table according to the building attribute data of the sampling dataset, and using the Euclidean distance to find the sample in the second data table closest to the first data table, the expression being: where d is the Euclidean distance between the unsampled sample i and the sampled sample j, n is the dimension of the vector, X i is the feature vector of the unsampled sample, X i = (z i1 , z i2 , …, z in ); X j is the feature vector of the sampled sample, X j = (z j1 , z j2 , …, z jn ); z ik is the building attribute data of the unsampled sample, z jk is the building attribute data of the sampled sample, k is the index and is a natural number; according to the constructed agent database, a simulation model embedded with the multi-objective optimization algorithm NSGA-II is created; generating an optimal renewal strategy by designing a renewal process mechanism, formulating a renewal scheme selection scenario and conditions, and agent interaction according to the simulation model and combining external influence parameters; extracting and visualizing the multi-objective optimization results of the renewal strategy. 2.The method for multi-objective optimization of low-carbon renewal of urban community according to claim 1, wherein: processing data includes obtaining building original carbon emission data, obtaining preset carbon emission reduction data, creating a renewal scheme, and constructing an agent database; the building original carbon emission data includes residential building original carbon emission data and non-residential building original carbon emission data. 3.The method for multi-objective optimization of low-carbon renewal of urban community according to claim 1, wherein: geographic spatial data includes block terrain data, plot boundary data, building data, green space data, water system data, and road network data; building data includes building contour data, building latitude and longitude data, building plot number data, and building attribute data. 4.The method for multi-objective optimization of low-carbon renewal of urban community according to claim 3, wherein: the spatial database is composed of block terrain data, plot boundary data, building contour data, green space data, water system data, and road network data. 5.The method for multi-objective optimization of low-carbon renewal of urban community according to claim 1, wherein: the process of creating a simulation model includes: adding agents according to the agent database, and embedding NSGA-Ⅱ; the agents include a first agent group, a second agent group, and an NSGA-Ⅱ single agent; the first agent group is a resident agent group, and the second agent group is a renewal strategy agent group. 6.The method for multi-objective optimization of low-carbon renewal of urban community according to claim 1, wherein: the implementation process of NSGA-Ⅱ includes: initializing a population and generating random individuals; performing non-dominated sorting and calculating the crowding distance; generating a child population through selection, crossover, and mutation; merging the parent and child populations, and performing non-dominated sorting and crowding distance calculation again to generate a new population; through the iteration loop of selecting the parent, performing crossover and mutation, and updating the population, the Pareto optimal solution set is obtained.

7. The system for the process of multi-objective optimization of low-carbon renewal of urban communities according to any of claims 1 to 6, characterized by the fact that, The data acquisition module includes acquiring and processing data, the acquired data including geospatial data and population data, constructing a multi-objective optimization agent database and a spatial database; The agent database construction process includes: taking a sample unit composed of a building and its corresponding residents as an agent, constructing a community full-sample dataset; sampling the community full-sample dataset to obtain a sampling sample set; collecting resident data based on the sampling sample set, and performing data augmentation according to the building attribute data to obtain resident data of all samples in the community full-sample dataset; The data augmentation process includes: establishing a first data table according to the building attribute data in the community full-sample dataset, establishing a second data table according to the building attribute data of the sampling dataset, and using the Euclidean distance to find the closest sample in the second data table to the first data table, the expression being: where d is the Euclidean distance between the unsampled sample i and the sampled sample j, n is the dimension of the vector, X i is the feature vector of the unsampled sample, X i i1 i2 in ; X j is the feature vector of the sampled sample, X j j1 j2 jn ; z ik is the architectural attribute data of the unsampled sample, z jk is the architectural attribute data of the sampled sample, k is an index and is a natural number;​​​​​​ The model construction module creates a simulation model capable of multi-objective optimization based on the constructed agent database; The strategy optimization module generates an optimal update strategy by designing an update process mechanism, selecting scenarios and conditions for an update scheme, and interacting with agents based on the simulation model and external influence parameters; The result output module extracts and visualizes the multi-objective optimization results of the update strategy.

Images