An old cell inclusive low-carbon reconstruction method and system

Abstract

This invention discloses an inclusive low-carbon renovation method and system for old residential communities, relating to the field of data processing systems and methods. The method is based on a multi-dimensional digital twin model and employs a multi-index comprehensive evaluation method based on AHP-entropy weighting to score the age-friendliness, accessibility, public space accessibility, and social integration of the target community, forming four-dimensional scores and an inclusiveness baseline index. The total carbon emissions for the baseline year are calculated according to a unified accounting boundary and the method of "activity data × emission factor," establishing a carbon emission baseline. The inclusiveness scores, the inclusiveness baseline index, and the carbon emission baseline are used as state inputs and reward calculation benchmarks for multi-agent reinforcement learning, generating multiple sets of candidate low-carbon renovation schemes. An improved multi-objective optimization algorithm is used to solve the Pareto optimal frontier for the candidate schemes, outputting the optimal renovation scheme set, and combining residents' preferences to determine the implementation plan.

Description

Technical Field

[0001] This invention relates to the field of building energy conservation and urban renewal technology, and in particular to a method and system for inclusive low-carbon renovation of old residential communities. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Low-carbon renovation of old residential communities has become an important path for urban carbon emission reduction. However, existing renovation technologies for old residential communities have the following prominent problems: First, there is a lack of coordination between carbon reduction targets and inclusiveness targets. Existing renovation plans are mostly geared towards energy conservation and carbon reduction, prioritizing technical measures such as building envelope renovation and the installation of renewable energy systems, while neglecting inclusive needs such as age-friendly renovations and the construction of accessible environments.

[0004] Second, the generation of renovation plans relies heavily on expert experience and lacks sufficient intelligence. Current renovation plans primarily depend on the personal experience and judgment of architects and engineers, lacking systematic data analysis and optimization methods. Different communities exhibit significant differences in building structure, resident composition, and carbon emission characteristics. Experience-driven plans struggle to achieve site-specific, refined designs, easily leading to mismatches between renovation measures and actual needs. Summary of the Invention

[0005] In view of the shortcomings of existing technologies, the purpose of this invention is to provide an inclusive low-carbon renovation method for old residential communities, so as to achieve synergistic optimization of carbon emission reduction targets and inclusiveness improvement targets, improve the intelligence level of renovation schemes and residents' participation, and provide technical support for construction carbon emission control and post-renovation effect evaluation.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: An inclusive low-carbon renovation method for old residential communities includes the following: Data is collected from old residential communities, and multi-dimensional digital twin models are constructed using the collected data; Based on a multi-dimensional digital twin model, a multi-index comprehensive evaluation method based on AHP-entropy weight combination is adopted to score the age-friendliness, accessibility, public space accessibility and social integration of the target community, forming four-dimensional scores and calculating the inclusiveness baseline index; and establishing a carbon emission baseline. The inclusiveness scores, the inclusiveness baseline index, and the carbon emission baseline are used as the state input and reward calculation benchmark of the multi-agent reinforcement learning model to generate multiple sets of candidate low-carbon transformation schemes, which cover multiple dimensions. A multi-objective optimization model is established with the objectives of maximizing the carbon emission reduction rate and the comprehensive inclusive index improvement rate, and with the total renovation budget and the feasible construction period as constraints. An improved multi-objective optimization algorithm is used to solve the Pareto optimal front for candidate schemes, and the optimal renovation scheme set is output. The final implementation plan is determined using a multi-attribute decision-making method based on preference aggregation. Phased construction management and dynamic carbon footprint monitoring: Develop phased construction plans according to the final implementation plan, deploy IoT construction monitoring equipment to collect hourly construction carbon emission data in real time during the construction phase, and use digital twin models to analyze and warn of construction carbon emission deviations.

[0007] The inclusive low-carbon renovation method for old residential communities described above also includes the following: after the renovation is completed, the inclusive indicators are re-evaluated at multiple time periods, and the carbon emissions after the renovation are calculated according to the accounting boundary, activity data and corresponding emission factors consistent with the carbon emission baseline. The renovation effect is evaluated by comparing the baseline data, and the evaluation results are fed back to the multi-agent reinforcement learning model for policy iteration and update. The activity data includes the construction electricity consumption, centralized heating consumption, fuel consumption of category k, freight consumption, and daily maintenance material consumption of category q during the evaluation period.

[0008] The inclusive low-carbon renovation method for old residential communities described above, which involves constructing a multi-dimensional digital twin model based on collected data, includes the following: Building structure information collection: acquire point cloud data and orthophotos of building facades, process geometric parameters through point cloud data, and combine with urban construction archives to obtain building structure type, construction year, and existing insulation measures; Energy consumption data collection: collecting electricity consumption data, obtaining centralized heating heat data, and collecting energy consumption data of community public facilities; Traffic travel pattern data collection: Record the frequency and time distribution of entry and exit of motor vehicles, non-motor vehicles and pedestrians, and collect data on the distribution and utilization rate of parking facilities in the community; Data collection on resident demographic structure and preferences; Constructing a multi-dimensional digital twin model: The digital twin model adopts a three-layer architecture: the bottom layer is the BIM building information model layer, the middle layer is the GIS spatial geographic layer, and the top layer is the IoT real-time data layer.

[0009] The inclusive low-carbon renovation method for old residential communities described above employs a multi-index comprehensive evaluation method based on AHP-entropy weighting to score the target community's age-friendliness, accessibility, public space accessibility, and social integration, forming a four-dimensional score and an inclusive baseline index, including the following: Constructing an inclusive scoring system: Establishing a multi-level evaluation indicator system consisting of a target layer, a primary indicator layer, a secondary indicator layer, and a tertiary observation indicator layer; Composite weighting calculation: The AHP-entropy weighting method is used for combined weighting, and the subjective weight vector is obtained based on expert judgment. An objective weight vector is obtained based on the reference cell sample data. And a combined weight vector is obtained through game theory-based weighting. ; Using a game theory-based weighting model:

[0010] The objective is to minimize the total deviation between the combined weights and the subjective and objective weights, and then solve for the combined weight vector. ,in, This is the subjective weight vector of AHP. Let α be the objective weight vector of the entropy weight method, and let α be the combination coefficient of the principal and objective weights. , Denotes the L2 norm; the combined weights satisfy... ,according to We weighted the homogenized and dimensionless index values ​​to obtain scores for four dimensions: age-friendliness, accessibility, public space accessibility, and social integration, as well as an inclusive baseline index. , As a benchmark for comparison of the inclusive improvement in subsequent MAPPO reward functions and multi-objective optimization; The establishment of a carbon emission baseline includes the following: carbon emission baseline Carbon emissions corresponding to purchased electricity Carbon emissions corresponding to centralized heating Carbon emissions corresponding to natural gas Carbon emissions related to motor vehicle travel in the community And the carbon emissions associated with routine maintenance and repair materials constitute:

[0011] in, E represents the electricity consumption during the base period of construction. Electricity emission factors that are matched to the accounting year and region; H represents centralized heating. For heating emission factors; , For the k-th type of fuel consumption, Fuel emission factor; Calculated based on motor vehicle fuel consumption or vehicle driving activity data and freight emission factors; , This refers to the amount of materials used for routine maintenance and repair in category q. Let be the emission factor for material production, k and q be the category indices, and Σ represent the summation over all corresponding categories.

[0012] The above-described inclusive low-carbon renovation method for old residential communities includes intelligent generation of the low-carbon renovation plan, comprising the following: The MAPPO multi-agent reinforcement learning framework is adopted, with 6 agents, each corresponding to one of the six transformation action dimensions. The six transformation action dimensions are used to describe the types of transformation measures that can be selected, and together with the four inclusive evaluation dimensions in step S2, they play different roles as "decision variables" and "evaluation results".

[0013] To directly input the evaluation and calculation results of step S2 into the MAPPO model, the state space S is defined to include building thermal parameters, current energy consumption level, and carbon emission baseline. Inclusiveness four-dimensional score and inclusiveness baseline index The state space is defined by factors such as resident preference distribution, remaining budget, and construction constraints; each agent selects the corresponding modification action based on this state space. The global reward function R is:

[0014] Where R is the global reward value of the joint retrofit scheme; the digital twin model predicts the carbon emissions C(a) and the comprehensive inclusiveness index I(a) after the retrofit based on the joint actions, and calculates... , ; and The relative importance of carbon reduction targets and inclusiveness enhancement targets is controlled separately in the learning process to meet the following requirements. , through and Normalization is performed to avoid reward bias caused by differences in the dimensions and numerical scales of the two objectives; the larger R is, the better the overall performance of the joint modification scheme under the two types of objectives; C(a) represents the modification action a. i The predicted carbon emissions, I(a) represents the amount of retrofitting actions taken (a). i The post-projected comprehensive index of inclusiveness; ΔC and ΔI represent the carbon emission reduction and inclusiveness improvement relative to the baseline, respectively.

[0015] Each agent is based on the current global state s in the state space S ( ) Observe and independently output the transformation actions in their respective dimensions. i ( ), through MAPPO's centralized critic network The value of joint actions is evaluated to guide agents in learning cooperative strategies that benefit the global objective. Here, C represents the joint action value function output by the centralized critic network, and s represents the current global state. to Together they constitute a joint action.

[0016] As described above, in an inclusive low-carbon renovation method for old residential communities, the first intelligent agent is responsible for energy-saving renovation of the building envelope, the second intelligent agent is responsible for the deployment of renewable energy systems, the third intelligent agent is responsible for the design of green transportation micro-circulation systems, the fourth intelligent agent is responsible for rainwater harvesting and sponge city renovation, the fifth intelligent agent is responsible for the configuration of age-friendly facilities, and the sixth intelligent agent is responsible for the construction of an accessible environment.

[0017] The inclusive low-carbon renovation method for old residential communities described above, which establishes a multi-objective optimization model with the objectives of maximizing carbon emission reduction rate and maximizing the comprehensive inclusive index improvement rate, and with the total renovation budget and feasible construction period as constraints, includes the following: An improved multi-objective optimization algorithm is used for multi-objective optimization, and the objective function is defined as:

[0018] in, The objective function is carbon emission reduction. This represents the carbon emission reduction of the retrofit plan relative to the baseline. The baseline value for carbon emissions is used; by taking the negative value of the reduction rate, the problem of "maximizing the carbon emission reduction rate" is transformed into a minimization problem.

[0019] in, To improve the inclusiveness of the objective function, The transformation plan relative to the inclusive baseline index The increase in the comprehensive inclusiveness index; through the analysis of / Taking negative values ​​transforms the problem of "maximizing the rate of improvement of the inclusive comprehensive index" into a minimization problem.

[0020] Where st represents the constraint condition. Let Σ be the cost of the i-th modification measure, where i is the modification measure number. The total cost of all renovation measures, To meet the budget ceiling for renovation;

[0021] in, The estimated construction period for the renovation plan, The maximum allowable construction period; The improved multi-objective optimization algorithm is used to solve the Pareto optimal frontier of candidate schemes and output the optimal modification scheme set. It includes the following: Based on the standard NSGA-III, an adaptive reference point dynamic adjustment mechanism is introduced: After each generation of evolution, the distribution density of the current Pareto frontier in the two objective dimensions is analyzed. The reference point density is reduced in areas with excessively dense distribution and increased in areas with sparse distribution to ensure uniform coverage of the solution set on the frontier. At the same time, an elite retention strategy is introduced, and the best 10% of individuals in each generation are directly introduced into the next generation. All solutions on the Pareto optimal frontier constitute the optimal modification scheme set.

[0022] The inclusive low-carbon renovation method for old residential communities described above, wherein the final implementation plan is determined using a multi-attribute decision-making method based on preference aggregation, includes the following: The resident-participatory scheme confirmation and interaction platform pushes the best renovation scheme set to residents of the target community through a visual simulation effect. The functional architecture of the resident-participatory scheme confirmation and interaction platform includes: a 3D visualization engine layer to realize real-time 3D rendering and virtual roaming of renovation schemes; an age-friendly interactive interface layer; a scheme display and comparison layer; an opinion collection and statistical analysis layer; and a preference aggregation decision layer that combines residents' preference weights to rank the schemes. The platform is launched through both WeChat mini-program and APP, allowing residents to view renovation plans and provide feedback anytime, anywhere. After the collection period, the resident support rate for each plan is calculated, and the final implementation plan is determined by combining the preferences and aggregated decision-making results.

[0023] As described above, the inclusive low-carbon renovation method for old residential communities reduces the carbon emissions per hour during construction. The calculation is as follows:

[0024] in, Let t represent the electricity consumption for construction within the t-th hour; This represents the fuel consumption for the k-th type of construction within the t-th hour. This refers to the amount of routine maintenance materials for category m within hour t. and These refer to the corresponding material transportation weight and transportation distance; , , and These are emission factors from electricity, fuel, materials production, and freight transportation, respectively.

[0025] Secondly, the present invention also provides an inclusive low-carbon renovation system for old residential communities, comprising the following modules: Data acquisition and modeling module: integrates UAV data acquisition subsystem, IoT sensor network management subsystem and BIM / GIS data fusion subsystem, and constructs multi-dimensional digital twin model based on the acquired data; Inclusivity Assessment and Carbon Accounting Module: Employing a multi-indicator comprehensive evaluation method based on AHP-entropy weighting, the module scores the age-friendliness, accessibility, public space accessibility, and social integration of the target community, generating four-dimensional scores and an inclusive baseline index; and establishing a carbon emission baseline. Intelligent solution generation module: It uses the scores of each dimension of inclusiveness, the inclusiveness baseline index and the carbon emission baseline as the state input and reward calculation benchmark of the multi-agent reinforcement learning model to generate multiple sets of candidate low-carbon transformation solutions; Multi-objective optimization module: It establishes a multi-objective optimization model with the objectives of maximizing carbon emission reduction rate and maximizing the comprehensive inclusive index improvement rate, and with the total renovation budget and construction feasible period as constraints. It uses an improved multi-objective optimization algorithm to solve the Pareto optimal front for candidate schemes and outputs the optimal renovation scheme set. Resident participation module: Provides mobile application and web-based management backend, offering feedback data statistical analysis and preference aggregation decision-making functions; Construction monitoring module: This module integrates IoT construction monitoring equipment management, real-time carbon footprint calculation, and early warning information push functions. It supports linkage with digital twin models, mapping construction progress and hourly construction carbon emission data into the three-dimensional model in real time, thereby realizing visualized control of the construction process. The effect evaluation module provides evaluation task management, data comparison and analysis, and evaluation report generation functions. It supports automatic feedback of evaluation results to the multi-agent reinforcement learning model of the intelligent solution generation module to realize iterative updates of strategies. This module provides trend analysis charts to intuitively show the changing trend of the transformation effect over time.

[0026] The beneficial effects of the present invention are as follows: This invention provides a method for assessing inclusiveness needs and calculating carbon emission baselines, thereby achieving synergistic optimization of carbon emission reduction targets and inclusiveness improvement targets. This avoids the problem of energy-saving renovation and social inclusiveness solutions being separated in existing solutions, ensuring that the renovated old communities significantly improve their friendliness to elderly residents and people with disabilities while reducing carbon emissions. The scores of each dimension of inclusiveness, the inclusiveness baseline index, and the carbon emission baseline are used as the state input and reward calculation benchmark for multi-agent reinforcement learning. Multiple sets of candidate low-carbon transformation schemes are generated using multi-agent reinforcement learning algorithms. A multi-objective optimization model is established with the objectives of maximizing the carbon emission reduction rate and the comprehensive inclusive index improvement rate, and with the total renovation budget and the feasible construction period as constraints. An improved multi-objective optimization algorithm is used to solve the Pareto optimal front for candidate schemes and output the optimal renovation scheme set. While ensuring convergence, the uniformity of the distribution of the optimal solution set on the Pareto front is improved. Attached Figure Description

[0027] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0028] Figure 1 This is a flowchart of an inclusive low-carbon renovation method for old residential communities according to one or more embodiments of the present invention. Detailed Implementation

[0029] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0030] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless otherwise expressly indicated by the invention, the singular form is also intended to include the plural form. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. As introduced in the background section, there is a lack of coordination between carbon emission reduction targets and inclusiveness targets in the existing technology. In order to solve the above-mentioned technical problems, this invention proposes an inclusive low-carbon renovation method for old residential communities.

[0031] Example 1 In a typical embodiment of the present invention, reference is made to Figure 1 As shown, an inclusive low-carbon renovation method for old residential communities includes the following: Step S1: Data Acquisition and Basic Information Modeling. Through multi-source remote sensing imagery, drone aerial photography, IoT sensor networks, and on-site surveys, acquire information on the target old residential community's building structure, energy consumption, transportation patterns, resident population structure and preferences, and the distribution of green spaces and public spaces to construct a multi-dimensional digital twin model.

[0032] The specific implementation method of step S1 is as follows: (1) Building structure information collection: Aerial photography of the target old residential area was carried out using a drone equipped with a high-resolution RGB (color mode that mixes red, green and blue primary colors) camera and LiDAR to obtain point cloud data and orthophotos of the building facade. Geometric parameters such as building outline, number of floors, roof area, and wall area were extracted through point cloud data processing. Information such as building structure type (brick-concrete structure, frame structure, etc.), construction year, and existing insulation measures were obtained by combining urban construction archives.

[0033] (2) Energy consumption data collection: Smart meters are installed at the main distribution boxes of each building in the target community to collect electricity consumption data at 15-minute sampling intervals, with a continuous collection period of 12 months or more. For winter heating energy consumption, the heat consumption data of centralized heating is obtained through heat meters. At the same time, energy consumption data of public facilities such as community public area lighting, elevator operation, and water supply and drainage pumping stations are also collected.

[0034] (3) Traffic travel pattern data collection: Traffic flow detection equipment is deployed at the main entrances and exits of the community to record the frequency and time distribution of motor vehicles, non-motor vehicles and pedestrians entering and exiting. Data such as commuting mode, commuting distance and travel frequency are obtained through resident travel questionnaires, and data on the distribution and utilization rate of parking facilities in the community are collected.

[0035] (4) Data collection on residents’ population structure and preferences: Through the assistance of the community neighborhood committee, door-to-door surveys were conducted to collect data on residents’ age structure, family composition, health status (with a focus on the proportion of people with mobility impairments and disabled elderly), income level, and preference ranking of renovation measures. The survey coverage rate was greater than or equal to 80% of the total number of households in the community.

[0036] (5) Constructing a multi-dimensional digital twin model: The digital twin model adopts a three-layer architecture: The bottom layer is the BIM building information model layer, which contains detailed three-dimensional geometric models, material thermal parameters, and structural information of each building. The middle layer is the GIS spatial geographic layer, which integrates spatial data such as community topography, road network, pipeline layout, and green space distribution. The top layer is the IoT real-time data layer, which connects to real-time data streams from smart meters, environmental sensors (temperature, humidity, PM2.5, CO2 concentration), video surveillance, etc. The three layers are unified to the CGCS2000 (China Geodetic Coordinate System 2000, the latest national geodetic coordinate system in China) coordinate reference framework and unified time benchmark through a spatiotemporal data fusion algorithm. The IoT monitoring equipment also includes smart heat meters, smart gas meters, smart water meters, power distribution circuit energy metering terminals, human infrared or millimeter wave passenger flow counters, access control / vehicle identification terminals, parking space geomagnetic sensors, illuminance sensors, noise sensors, and LoRa, NB-IoT, or 4G / 5G IoT gateways for data aggregation and transmission; each device uploads timestamped data according to a preset sampling period.

[0037] Step S2: Inclusivity Needs Assessment and Carbon Emission Baseline Calculation. Based on the constructed multi-dimensional digital twin model, a multi-indicator comprehensive evaluation method based on AHP-entropy weighting is used to score the age-friendliness, accessibility, public space accessibility, and social integration of the target community, forming an inclusive baseline index. Simultaneously, the total carbon emissions for the baseline year are calculated according to the unified accounting boundary and the method of "activity data × emission factor". Establish a carbon emission baseline.

[0038] The specific implementation method of step S2 is as follows: (1) Constructing an inclusiveness scoring system: Establish a four-layer evaluation structure, including a target layer, a primary indicator layer, a secondary indicator layer, and a tertiary observation indicator layer. The target layer is the comprehensive inclusiveness index I; the primary indicators include age-friendliness (A1), barrier-free accessibility (A2), accessibility of public spaces (A3), and social integration (A4); each primary indicator has several secondary indicators, which are further refined into tertiary observation indicators based on data availability. For example, age-friendliness may include indicators such as elevator configuration rate, handrail installation rate, non-slip floor coverage rate, emergency call device coverage rate, and age-friendly lighting compliance rate.

[0039] (2) Composite weighting calculation: A combination of AHP (Analytic Hierarchy Process) and entropy weighting was used for weighting. More than 10 experts in urban planning, architectural design, and accessibility design were invited to make pairwise comparisons of the importance of the indicators, construct a judgment matrix, and calculate the subjective weight vector. Simultaneously, reference community samples with similar size, construction year, and climate zone to the target community were selected. The data for each indicator were homogenized and dimensionless. Information entropy and difference coefficients were calculated based on the sample dispersion to determine the objective weight vector. .

[0040] Using a game theory-based comprehensive weighting model

[0041] The objective is to minimize the total deviation between the combined weights and the subjective and objective weights, and then solve for the combined weight vector. .in, This is the subjective weight vector of AHP. The objective weight vector of the entropy weight method. For the combined weight vector, Denotes the L2 norm; the combined weights satisfy... , The combination coefficients of primary and objective weights and The obtained As the final scoring weight, it is used to convert the relevant data collected in step S1 into inclusive evaluation results that can be directly accessed in step S3. α ranges from 0 to 1, with a larger α indicating a higher proportion of AHP subjective weights in the combined weights.

[0042] (3) Inclusivity score calculation: For positive indicators, the following is adopted: Standardize and adopt negative indicators Standardize the data so that the standardized values ​​z of each indicator are obtained. ij All values ​​are between 0 and 1, with higher values ​​indicating better inclusiveness.

[0043] in, This represents the original value of the j-th indicator for the i-th evaluation object; Let j be the standardized value of the j-th indicator for the i-th evaluation object; and , i and j are the minimum and maximum values ​​of the j-th indicator in the sample, respectively; i is the index of the evaluation object, and j is the index of the indicator;

[0044] in, Let j be the standardized value of the j-th indicator of the target community. The corresponding combined weights are used; the inclusiveness baseline index I0 ranges from 0 to 100. For any first-level indicator Ak, its dimensional scores are normalized according to the combined weights of the indicators within that dimension, then weighted and summed, and converted to a score of 0 to 100. This yields the scores for the four dimensions A1 to A4 and the inclusiveness baseline index. This serves as input for subsequent MAPPO state space, reward function, and multi-objective optimization. In this context, the subscript 0 indicates the baseline state of the target community, Σ indicates the summation of all evaluation indicators, and Ak indicates the k-th first-level inclusive dimension, where k = 1, 2, 3, 4.

[0045] (4) Establish a carbon emission baseline: Select the activity data from the 12 consecutive months in step S1 as the baseline period, and ensure that the accounting boundaries remain consistent in subsequent assessments. Carbon emission baseline Carbon emissions corresponding to purchased electricity Carbon emissions corresponding to centralized heating Carbon emissions corresponding to natural gas Carbon emissions related to motor vehicle travel in the community And the carbon emissions associated with routine maintenance and repair materials constitute:

[0046] in, E represents the electricity consumption during the base period of construction. Electricity emission factors that are matched to the accounting year and region; H represents centralized heating. For heating emission factors; , For the k-th type of fuel consumption, Fuel emission factor; Calculated based on motor vehicle fuel consumption or vehicle driving activity data and freight emission factors; , This refers to the amount of materials used for routine maintenance and repair in category q. Let be the emission factor for material production, k and q be the category indices, and Σ represent the summation over all corresponding categories.

[0047] Electricity and heating are counted as indirect emissions; on-site fuel and fuel-powered travel related to the community are counted as direct emissions; and maintenance and repair materials are counted as implicit emissions. Emission factors should preferentially use data from greenhouse gas emission factor databases and data published by the competent authorities that match the accounting year, avoiding the mixing of factors from different years.

[0048] Step S3: Intelligent Generation of Low-Carbon Retrofit Schemes. The scores of the four inclusive dimensions (A1-A4), the inclusive baseline index I0, and the carbon emission baseline C0 obtained in Step S2 are input together with the building, energy consumption, resident preferences, budget, and construction condition data generated in Step S1 into the MAPPO multi-agent reinforcement learning model to generate multiple sets of candidate low-carbon retrofit schemes. The candidate schemes include six retrofit action dimensions: energy-saving retrofit of building envelope, deployment of renewable energy systems, design of green transportation micro-circulation systems, rainwater harvesting and sponge city transformation, configuration of age-friendly facilities, and construction of barrier-free environments.

[0049] The specific implementation method of step S3 is as follows: The MAPPO (Multi-Agent Proximal Policy Optimization) reinforcement learning framework was adopted, with six agents corresponding to the six transformation action dimensions mentioned above: Agent 1 was responsible for energy-saving renovation of the building envelope; Agent 2 was responsible for the deployment of renewable energy systems; Agent 3 was responsible for the design of green transportation micro-circulation systems; Agent 4 was responsible for rainwater harvesting and sponge city transformation; Agent 5 was responsible for the configuration of age-friendly facilities; and Agent 6 was responsible for the construction of an accessible environment. Each agent's action space consisted of corresponding transformation parameters, such as the insulation material and thickness of the building envelope, photovoltaic installation capacity, permeable pavement ratio, elevator and handrail configuration, and ramp parameters. All six dimensions belonged to the "transformation action space," while A1 to A4 in step S2 belonged to the "evaluation results," and a one-to-one correspondence between the two was not required.

[0050] To ensure that the evaluation results of step S2 and the carbon emission baseline are directly involved in policy learning, the state space S is defined to include building thermal parameters, current energy consumption level, and carbon emission baseline. Scores on the four inclusion dimensions (A1-A4) and the inclusion baseline index. Resident preference distribution, remaining budget, and construction constraints. Each agent is in a certain state. The output is a joint action 'a', and the corresponding carbon emission index C(a) and inclusiveness index I(a) are predicted by the digital twin model. The global reward function R is:

[0051] Where R is the global reward value of the joint transformation scheme; , . and Used to adjust the relative importance of carbon reduction targets and inclusiveness enhancement targets in training, to meet... ; Baseline carbon emissions in the denominator and Inclusivity Baseline Index This is used to normalize the improvement magnitude of the two objectives, reducing the impact of differences in dimensions and numerical scales on the reward. A larger R indicates a better overall performance of the joint modification scheme. Where a... i =(a1,a2,…,a6) represents the joint modification action of six agents; C(a) represents the modification action a taken. i The predicted carbon emissions, I(a) represents the amount of retrofitting actions taken (a). i The post-projected comprehensive index of inclusiveness; ΔC and ΔI represent the carbon emission reduction and inclusiveness improvement relative to the baseline, respectively.

[0052] During MAPPO training, the centralized critic network evaluates the value of the scheme using the complete state S and the joint actions of the six agents, and the six agent networks output their respective modification actions; therefore, step S2... , A1 to A4 are not isolated results, but rather influence the direction of candidate solution generation simultaneously through state input and reward calculation.

[0053] Each agent is based on the current global state s in the state space S ( ) Observe and independently output the transformation actions in their respective dimensions. i ( ), through MAPPO's centralized critic network The value of joint actions is evaluated to guide agents in learning cooperative strategies that benefit the overall goal. Multiple candidate low-carbon transformation schemes are formed after several rounds of interaction. The training process can employ a curriculum learning strategy, first pre-training with small-scale virtual community data, and then fine-tuning with real community data. To avoid confusion with the carbon emission symbol C, the centralized commentator output is denoted as Vc(s,a1,…,a6); Vc represents the estimated value under state s and joint actions, and i is the agent number.

[0054] Step S4: Multi-objective collaborative optimization. Using the candidate schemes generated in Step S3 as the search objects, a multi-objective optimization model is established with the objectives of maximizing the carbon emission reduction rate and maximizing the comprehensive inclusive index improvement rate, and with the total renovation budget and construction feasibility period as constraints. The improved NSGA-III algorithm is used to solve the Pareto optimal front for the candidate schemes, and the optimal renovation scheme set is output.

[0055] The specific implementation method of step S4 is as follows: The improved NSGA-III algorithm (multi-objective optimization algorithm) is used for multi-objective optimization, and the objective function is defined as:

[0056] in, The objective function is carbon emission reduction. This represents the carbon emission reduction of the retrofit plan relative to the baseline. = , = , The baseline value for carbon emissions is used; by taking the negative value of the reduction rate, the problem of "maximizing the carbon emission reduction rate" is transformed into a minimization problem.

[0057] in, To improve the inclusiveness of the objective function, , Both are objective functions to be minimized. The transformation plan relative to the inclusive baseline index The increase in the comprehensive inclusiveness index; through the analysis of / Taking negative values ​​transforms the problem of maximizing the improvement rate of the inclusive comprehensive index into a minimization problem. = , = ;

[0058] Where st represents the constraint condition. Let Σ be the cost of the i-th modification measure, where i is the modification measure number. The total cost of all renovation measures, To meet the budget ceiling for renovation;

[0059] in, The estimated construction period for the renovation plan, The maximum allowable construction period; and Use the same time unit.

[0060] Building upon the standard NSGA-III, an adaptive reference point dynamic adjustment mechanism is introduced: after each generation of evolution, the distribution density of the current Pareto front in both target dimensions is analyzed. Reference point density is reduced in overly dense regions and increased in sparse regions to ensure uniform coverage of the solution set on the front. Simultaneously, an elite retention strategy is introduced, directly transferring the top 10% of individuals from each generation to the next, outputting all solutions on the optimal Pareto front to form the optimal modification scheme set.

[0061] Step S5: Resident Participatory Solution Confirmation: Through a mobile interactive platform, a visual simulation of the optimal renovation solution set is pushed to residents of the target community. Residents' votes and feedback are collected, and a multi-attribute decision-making method based on preference aggregation is used to determine the final implementation plan.

[0062] The specific implementation method of step S5 is as follows: Considering that the renovation of old residential areas involves the interests of all residents, the current process for collecting residents' opinions is mostly based on formal questionnaires. There is a lack of effective interactive means to allow residents to intuitively understand the effects of the renovation plan. Residents' preferences and needs have not been fully integrated into the decision-making process, which has affected the social acceptance and implementation effectiveness of the renovation plan.

[0063] Therefore, in this embodiment, the optimal renovation plan set is pushed to residents of the target community through a resident-participatory plan confirmation and interaction platform, which includes a visual simulation of the solution. The functional architecture of the resident-participatory plan confirmation and interaction platform includes: a 3D visualization engine layer (real-time 3D rendering and virtual roaming of renovation plans based on WebGL (3D drawing protocol)), an age-friendly interactive interface layer (providing large font mode, high contrast mode, voice input mode and simplified gesture operation mode), a plan display and comparison layer (supporting side-by-side comparison of multiple plans, displaying the carbon emission reduction prediction value, renovation period, cost details and inclusive improvement indicators of each plan), an opinion collection and statistical analysis layer (supporting multiple feedback forms such as voting, rating, text opinions and voice opinions), and a preference aggregation decision layer (using the TOPSIS (top-lowest solution distance method) method combined with resident preference weights to rank each plan).

[0064] The platform is launched via both WeChat mini-program and mobile app, allowing residents to view renovation plans and provide feedback anytime, anywhere. For elderly residents with mobility issues, community volunteers provide tablets to assist with participation. After the data collection period, the system automatically calculates resident support rates for each plan and combines these with preferences to determine the final implementation plan.

[0065] Step S6: Phased Construction Management and Dynamic Monitoring of Carbon Footprint. Develop a phased construction plan according to the final implementation scheme, deploy IoT-based construction monitoring equipment to collect hourly construction carbon emission data in real time during each construction phase, and utilize a digital twin model for construction carbon emission deviation analysis and early warning.

[0066] The specific implementation method of step S6 is as follows: Given the lack of carbon emission control during the construction phase, existing research and practice often focus on the carbon emission reduction effect after the renovation is completed, while neglecting the carbon emissions during the construction process itself. The production and transportation of building materials, fuel consumption of construction machinery, and electricity use on the construction site all generate significant carbon emissions. The lack of real-time monitoring and control measures makes it difficult to assess and optimize the carbon footprint during the construction phase.

[0067] Therefore, the final implementation plan is broken down into several construction phases, with a clear carbon emission budget set for each phase. Smart meters are deployed in each construction area to monitor construction electricity consumption, fuel flow meters to monitor fuel consumption of construction machinery, and GPS positioning modules to record the trajectory and distance of material transport vehicles. These are connected to electronic ledgers or electronic weighbridges for material arrival to record material type, quantity, and transport quality. Video surveillance is used for verification of construction activities. The IoT construction monitoring equipment includes at least smart meters or power distribution circuit energy metering terminals, fuel flow meters or oil level sensors for construction machinery, GPS / BeiDou positioning terminals for material transport vehicles, vehicle load sensors or electronic weighbridges, QR code / RFID identification terminals for material arrival, terminals for monitoring the running time and idling speed of construction machinery, environmental temperature, humidity, and particulate matter sensors, video surveillance equipment, and LoRa, NB-IoT, or 4G / 5G IoT gateways. Data collected by each device is accompanied by device number and timestamp and uploaded to the construction monitoring module.

[0068] The carbon footprint calculation model uses hours (t) as the time granularity, uniformly converting four categories of emissions—electricity consumption at the construction site, fuel for construction machinery, production of modification materials, and transportation of materials—into carbon dioxide equivalents. (Single-hour construction carbon emissions) The calculation is as follows:

[0069] in, Let t represent the electricity consumption for construction within the t-th hour; This represents the fuel consumption for the k-th type of construction within the t-th hour. This refers to the amount of routine maintenance materials for category m within hour t. and These refer to the corresponding material transportation weight and transportation distance; , , and These are emission factors for electricity, fuel, materials production, and freight transportation, respectively. Actual carbon emissions at any construction phase. .Will Compared with the carbon emission budget value for this stage When comparing, The system triggers an early warning at specific times, with a warning threshold coefficient of 1.15. It identifies major sources of deviation, such as material waste, idling machinery, or unreasonable transportation routes, based on the contribution of each emission category, and sends rectification suggestions to construction management personnel. Here, t is the hourly index, k is the fuel category index, and m is the material category index.

[0070] Step S7: Evaluation and Continuous Optimization of Post-Transformation Effects. Inclusivity indicators and carbon emission measurements are conducted at 6, 12, and 24 months after the transformation. The transformation effects are evaluated by comparing the results with baseline data, and the evaluation results are fed back to the multi-agent reinforcement learning algorithm model for policy iteration and updates.

[0071] The specific implementation method of step S7 is as follows: It should be noted that there is a lack of a long-term mechanism for evaluating the effects of the retrofit: most existing retrofit projects end after completion and acceptance, lacking a systematic follow-up evaluation mechanism. Whether the actual carbon emission reduction effect and inclusive improvement effect of the retrofit measures have met the design expectations needs to be continuously monitored and evaluated over a longer period of time, and data support should be provided for the optimization of subsequent retrofit projects.

[0072] Therefore, in this embodiment, three evaluation nodes are set after the renovation is completed: T1 = 6 months after completion (short-term evaluation), T2 = 12 months after completion (mid-term evaluation), and T3 = 24 months after completion (long-term evaluation). The evaluation content of each node includes: (1) Inclusivity index reassessment: Using the same evaluation index system and scoring method as step S2, recalculate the scores of the four dimensions of age-friendliness, accessibility, public space accessibility and social integration, and compare them with the baseline data.

[0073] (2) Actual carbon emission measurement: Using the accounting boundary and emission factor that are completely consistent with the carbon emission baseline in step S2, the construction electricity consumption during the evaluation period p was collected. Central heating Natural gas and other fuel consumption Data on motor vehicle travel activities related to the community and the consumption of maintenance materials will be used to calculate the carbon emissions after the renovation. :

[0074]

[0075] in, This refers to the carbon emissions within the baseline period corresponding to the evaluation period p, using the same caliber. When the evaluation period length differs from the baseline year, a comparison with the same period or conversion to an annualized equivalent value is used before comparison. Environmental parameters such as temperature and humidity are used for energy consumption verification and climate condition correction, and are not directly added to energy consumption data to form carbon emissions. To evaluate the actual carbon emission reduction rate in period p, To evaluate the carbon emissions generated by motor vehicle travel within period p. To evaluate the carbon emissions generated by maintenance materials during the period p.

[0076] (3) Resident satisfaction survey: Collect residents’ satisfaction ratings and specific opinions on the renovation effect through questionnaire survey.

[0077] The measured data (actual carbon emission reduction rate, actual degree of inclusiveness improvement) of each assessment node are used as feedback signals to input into the reward function of the MAPPO model, and the model strategy is iteratively updated to make the subsequent project solutions more in line with the actual results.

[0078] The renovation method provided in this embodiment enables residents to participate immersively in the decision-making process through a participatory scheme confirmation and interaction platform. Virtual reality technology allows residents to intuitively understand the renovation effects, and an age-friendly interactive interface design ensures equal participation for residents of all ages, significantly improving the social acceptance of the renovation scheme. A real-time carbon footprint monitoring and early warning mechanism is established during the construction phase, filling the gap in existing technologies for carbon emission control during construction. The linkage between IoT devices and digital twin models enables automatic identification and early warning of carbon emission deviations, effectively controlling the construction carbon footprint. A multi-time-node effect evaluation and model iteration feedback mechanism is constructed after the renovation, providing data-driven scheme optimization for subsequent renovation projects through continuous tracking and evaluation, forming a complete closed loop from planning and design to implementation evaluation.

[0079] Example 2 This embodiment provides an inclusive low-carbon renovation system for old residential communities, including the following modules: The data acquisition and modeling module integrates a UAV data acquisition subsystem, an IoT sensor network management subsystem, and a BIM / GIS data fusion subsystem, enabling fully automated processing from raw data acquisition to digital twin model construction. This module includes a data quality control subsystem for outlier detection, missing value imputation, and consistency verification of the acquired data. The IoT sensor network management subsystem handles device registration, data acquisition, time synchronization, anomaly detection, and communication status management. Its connected devices include smart meters, smart heat meters, smart gas meters, environmental sensors, passenger and vehicle identification terminals, parking space geomagnetic sensors, and video surveillance equipment.

[0080] The Inclusivity Assessment and Carbon Accounting Module includes a built-in AHP-entropy weighting engine, a multi-indicator comprehensive scoring engine, and a carbon emission calculation engine. It receives data from the digital twin model and automatically calculates indicator standardization, combined weight calculation, A1-A4 four-dimensional scoring, the inclusiveness baseline index I0, and the carbon emission baseline C0. This module provides standardized assessment report generation capabilities and supports export in PDF and Word formats.

[0081] Intelligent solution generation module: Built-in MAPPO multi-agent reinforcement learning engine and transformation knowledge graph, it automatically generates multiple sets of candidate low-carbon transformation solutions based on evaluation results. The transformation knowledge graph includes professional knowledge such as building energy-saving technology library, renewable energy product library, and barrier-free design standard library, providing constraints and guidance for the agent's action space.

[0082] Multi-objective optimization module: Features a built-in improved NSGA-III solver, supports user-defined objective function weights and constraints, and outputs a set of Pareto optimal solutions. This module provides an interactive Pareto frontier visualization interface to help decision-makers understand the trade-offs between different solutions in terms of carbon reduction and inclusion.

[0083] Resident Participation Module: Offers a mobile application and a web-based management backend. The mobile application, designed for residents, provides functions such as browsing the plan, virtual tours, voting feedback, and submitting opinions. The web-based management backend, designed for project managers, provides feedback data statistical analysis and preference aggregation decision-making functions.

[0084] Construction Monitoring Module: This module integrates IoT construction monitoring equipment management, real-time carbon footprint calculation, and early warning information push functions. It supports linkage with a digital twin model, mapping construction progress and hourly carbon emission data into a 3D model in real time, enabling visualized control of the construction process. IoT construction monitoring equipment includes at least smart meters or power distribution circuit metering terminals, fuel flow meters or oil level sensors for construction machinery, GPS / BeiDou positioning terminals for material transport vehicles, vehicle load sensors or electronic weighbridges, QR code / RFID (Radio Frequency Identification) terminals for material entry, terminals for monitoring the running time and idling speed of construction machinery, environmental temperature, humidity, and particulate matter sensors, video surveillance equipment, and LoRa, NB-IoT, or 4G / 5G IoT gateways. Data collected by each device is accompanied by the device number and timestamp and uploaded to the construction monitoring module.

[0085] The effectiveness evaluation module provides evaluation task management, data comparison and analysis, and evaluation report generation functions. It supports automatic feedback of evaluation results to the MAPPO model in the intelligent solution generation module, enabling iterative strategy updates. This module provides trend analysis charts to intuitively display the changing trends of the transformation effects over time. Please refer to Example 1 for details.

[0086] Example 3 This embodiment provides an inclusive low-carbon renovation device for old residential communities, including at least one processor (CPU / GPU) and a memory (RAM / ROM / SSD) communicatively connected to the processor. The memory stores a computer program executable by the processor. When the processor executes the computer program, it implements all the steps of the method described in Embodiment 1. Specifically, the processor can be an Intel Xeon or AMD EPYC series high-performance server processor, and the GPU can be an NVIDIA A100 or RTX 4090 for training, inference, and 3D visualization rendering of the MAPPO model. The memory uses DDR4 RAM combined with an NVMe SSD to ensure fast read / write performance under large data volumes.

[0087] This embodiment also provides a computer-readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (such as SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc., which stores a computer program. When the computer program is executed by a processor, it implements the steps in the above method embodiment.

[0088] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for inclusive low-carbon renovation of old residential communities, characterized in that, Includes the following: Data is collected from old residential communities, and multi-dimensional digital twin models are constructed using the collected data; Based on a multi-dimensional digital twin model, a multi-index comprehensive evaluation method based on AHP-entropy weight combination is adopted to score the age-friendliness, accessibility, public space accessibility and social integration of the target community, forming four-dimensional scores and calculating the inclusiveness baseline index; and establishing a carbon emission baseline. The inclusiveness scores, the inclusiveness baseline index, and the carbon emission baseline are used as the state input and reward calculation benchmark of the multi-agent reinforcement learning model to generate multiple sets of candidate low-carbon transformation schemes, which cover multiple dimensions. A multi-objective optimization model is established with the objectives of maximizing the carbon emission reduction rate and the comprehensive inclusive index improvement rate, and with the total renovation budget and the feasible construction period as constraints. An improved multi-objective optimization algorithm is used to solve the Pareto optimal front for candidate schemes, and the optimal renovation scheme set is output. The final implementation plan is determined using a multi-attribute decision-making method based on preference aggregation. Phased construction management and dynamic carbon footprint monitoring: Develop phased construction plans according to the final implementation plan, deploy IoT construction monitoring equipment to collect hourly construction carbon emission data in real time during the construction phase, and use digital twin models to analyze and warn of construction carbon emission deviations.

2. The inclusive low-carbon renovation method for old residential communities according to claim 1, characterized in that, It also includes the following: after the transformation is completed, the inclusive indicators are re-evaluated at multiple time periods, and the carbon emissions after the transformation are calculated according to the accounting boundary, activity data and corresponding emission factors consistent with the carbon emission baseline. The transformation effect is evaluated by comparing with the baseline data, and the evaluation results are fed back to the multi-agent reinforcement learning model for policy iteration and update. The activity data includes the construction electricity consumption, centralized heating consumption, fuel consumption of category k, freight consumption, and daily maintenance material consumption of category q during the evaluation period.

3. The inclusive low-carbon renovation method for old residential communities according to claim 1, characterized in that, The construction of a multi-dimensional digital twin model based on the collected data includes the following: Building structure information collection: acquire point cloud data and orthophotos of building facades, process geometric parameters through point cloud data, and combine with urban construction archives to obtain building structure type, construction year, and existing insulation measures; Energy consumption data collection: collecting electricity consumption data, obtaining centralized heating heat data, and collecting energy consumption data of community public facilities; Traffic travel pattern data collection: Record the frequency and time distribution of entry and exit of motor vehicles, non-motor vehicles and pedestrians, and collect data on the distribution and utilization rate of parking facilities in the community; Data collection on resident demographic structure and preferences; Constructing a multi-dimensional digital twin model: The digital twin model adopts a three-layer architecture: the bottom layer is the BIM building information model layer, the middle layer is the GIS spatial geographic layer, and the top layer is the IoT real-time data layer.

4. The inclusive low-carbon renovation method for old residential communities according to claim 1, characterized in that, The method employs a multi-index comprehensive evaluation approach based on AHP-entropy weighting to score the target community's age-friendliness, accessibility, public space accessibility, and social integration, forming a four-dimensional score and an inclusive baseline index, including the following: Constructing an inclusive scoring system: Establishing a multi-level evaluation indicator system consisting of a target layer, a primary indicator layer, a secondary indicator layer, and a tertiary observation indicator layer; Composite weighting calculation: The AHP-entropy weighting method is used for combined weighting, and the subjective weight vector is obtained based on expert judgment. An objective weight vector is obtained based on the reference cell sample data. And a combined weight vector is obtained through game theory-based weighting. ; Using a game theory-based weighting model: The objective is to minimize the total deviation between the combined weights and the subjective and objective weights, and then solve for the combined weight vector. ,in, This is the subjective weight vector of AHP. Let α be the objective weight vector of the entropy weight method, and let α be the combination coefficient of the principal and objective weights. , Denotes the L2 norm; the combined weights satisfy... ,according to We weighted the homogenized and dimensionless index values ​​to obtain scores for four dimensions: age-friendliness, accessibility, public space accessibility, and social integration, as well as an inclusive baseline index. , As a benchmark for comparison of the inclusive improvement in subsequent MAPPO reward functions and multi-objective optimization; The establishment of a carbon emission baseline includes the following: carbon emission baseline Carbon emissions corresponding to purchased electricity Carbon emissions corresponding to centralized heating Carbon emissions corresponding to natural gas Carbon emissions related to motor vehicle travel in the community And the carbon emissions associated with routine maintenance and repair materials constitute: in, E represents the electricity consumption during the base period of construction. Electricity emission factors that are matched to the accounting year and region; H represents centralized heating. For heating emission factors; , For the k-th type of fuel consumption, Fuel emission factor; Calculated based on motor vehicle fuel consumption or vehicle driving activity data and freight emission factors; , This refers to the amount of materials used for routine maintenance and repair in category q. Let be the emission factor for material production, k and q be the category indices, and Σ represent the summation over all corresponding categories.

5. The inclusive low-carbon renovation method for old residential communities according to claim 1, characterized in that, The intelligent generation of the low-carbon transformation plan includes the following: The MAPPO multi-agent reinforcement learning framework is adopted, with 6 agents, each corresponding to one of the six transformation action dimensions. The six transformation action dimensions are used to describe the types of transformation measures that can be selected, and together with the four inclusive evaluation dimensions in step S2, they play different roles as "decision variables" and "evaluation results". To directly input the evaluation and calculation results of step S2 into the MAPPO model, the state space S is defined to include building thermal parameters, current energy consumption level, and carbon emission baseline. Inclusiveness four-dimensional score and inclusiveness baseline index The system considers resident preference distribution, remaining budget, and construction constraints. Each agent selects a corresponding modification action based on this state space, and the global reward function R is: Where R is the global reward value of the joint retrofit scheme; the digital twin model predicts the carbon emissions C(a) and the comprehensive inclusiveness index I(a) after the retrofit based on the joint actions, and calculates... , ; and The relative importance of carbon reduction targets and inclusiveness enhancement targets is controlled separately in the learning process to meet the following requirements. , through and Normalization is performed to avoid reward bias caused by differences in the dimensions and numerical scales of the two objectives; the larger R is, the better the overall performance of the joint transformation scheme under the two types of objectives. To take modification action a i Post-predicted carbon emissions To take modification action a i The post-forecasted comprehensive index of inclusiveness; ΔC and ΔI represent the carbon emission reduction and inclusiveness improvement relative to the baseline, respectively; Each agent is based on the current global state s in the state space S ( ) Observe and independently output the transformation actions in their respective dimensions. i ( ), through MAPPO's centralized critic network The value of joint actions is evaluated to guide agents in learning cooperative strategies that benefit the global objective. Here, C represents the joint action value function output by the centralized critic network, and s represents the current global state. to Together they constitute a joint action.

6. The inclusive low-carbon renovation method for old residential communities according to claim 5, characterized in that, Of the six intelligent agents, the first is responsible for energy-saving renovation of the building envelope, the second for deploying renewable energy systems, the third for designing green transportation micro-circulation systems, the fourth for rainwater harvesting and sponge city transformation, the fifth for configuring age-friendly facilities, and the sixth for building an accessible environment.

7. The inclusive low-carbon renovation method for old residential communities according to claim 1, characterized in that, The establishment of a multi-objective optimization model with the objectives of maximizing the carbon emission reduction rate and the inclusive comprehensive index improvement rate, and with the total renovation budget and the feasible construction period as constraints, includes the following: An improved multi-objective optimization algorithm is used for multi-objective optimization, and the objective function is defined as: in, The objective function is carbon emission reduction. This represents the carbon emission reduction of the retrofit plan relative to the baseline. This represents the baseline value for carbon emissions; by taking the negative value for the reduction rate, the problem of "maximizing the carbon emission reduction rate" is transformed into a minimization problem. in, To improve the inclusiveness of the objective function, The transformation plan relative to the inclusive baseline index The increase in the comprehensive inclusiveness index; through the analysis of / Taking negative values ​​transforms the problem of "maximizing the rate of increase in the inclusive comprehensive index" into a minimization problem. Where st represents the constraint condition. Let Σ be the cost of the i-th modification measure, where i is the modification measure number. The total cost of all renovation measures, To meet the budget ceiling for renovation; in, The estimated construction period for the renovation plan, The maximum allowable construction period; The improved multi-objective optimization algorithm is used to solve the Pareto optimal frontier of candidate schemes and output the optimal modification scheme set. It includes the following: Based on the standard NSGA-III, an adaptive reference point dynamic adjustment mechanism is introduced: After each generation of evolution, the distribution density of the current Pareto frontier in the two objective dimensions is analyzed. The reference point density is reduced in areas with excessively dense distribution and increased in areas with sparse distribution to ensure uniform coverage of the solution set on the frontier. At the same time, an elite retention strategy is introduced, and the best 10% of individuals in each generation are directly introduced into the next generation. All solutions on the Pareto optimal frontier constitute the optimal modification scheme set.

8. The inclusive low-carbon renovation method for old residential communities according to claim 1, characterized in that, The method of determining the final implementation scheme using a multi-attribute decision-making approach based on preference aggregation includes the following: The resident-participatory scheme confirmation and interaction platform pushes the best renovation scheme set to residents of the target community through a visual simulation effect. The functional architecture of the resident-participatory scheme confirmation and interaction platform includes: a 3D visualization engine layer to realize real-time 3D rendering and virtual roaming of renovation schemes; an age-friendly interactive interface layer; a scheme display and comparison layer; an opinion collection and statistical analysis layer; and a preference aggregation decision layer that combines residents' preference weights to rank the schemes. The platform is launched through both WeChat mini-program and APP, allowing residents to view renovation plans and provide feedback anytime, anywhere. After the collection period, the resident support rate for each plan is calculated, and the final implementation plan is determined by combining the preferences and aggregated decision-making results.

9. A method for inclusive low-carbon renovation of old residential communities according to claim 1, characterized in that, Carbon emissions per hour of construction The calculation is as follows: in, Let t represent the electricity consumption for construction within the t-th hour; This represents the fuel consumption for the k-th type of construction within the t-th hour. This refers to the amount of routine maintenance materials for category m within hour t. and These refer to the corresponding material transportation weight and transportation distance; , , and These are emission factors from electricity, fuel, materials production, and freight transportation, respectively.

10. An inclusive low-carbon renovation system for old residential communities, characterized in that, Includes the following modules: Data acquisition and modeling module: integrates UAV data acquisition subsystem, IoT sensor network management subsystem and BIM / GIS data fusion subsystem, and constructs multi-dimensional digital twin model based on the acquired data; Inclusivity Assessment and Carbon Accounting Module: Employing a multi-indicator comprehensive evaluation method based on AHP-entropy weighting, the module scores the age-friendliness, accessibility, public space accessibility, and social integration of the target community, generating four-dimensional scores and an inclusive baseline index; and establishing a carbon emission baseline. Intelligent solution generation module: It uses the scores of each dimension of inclusiveness, the inclusiveness baseline index and the carbon emission baseline as the state input and reward calculation benchmark of the multi-agent reinforcement learning model to generate multiple sets of candidate low-carbon transformation solutions; Multi-objective optimization module: It establishes a multi-objective optimization model with the objectives of maximizing carbon emission reduction rate and maximizing the comprehensive inclusive index improvement rate, and with the total renovation budget and construction feasible period as constraints. It uses an improved multi-objective optimization algorithm to solve the Pareto optimal front for candidate schemes and outputs the optimal renovation scheme set. Resident participation module: Provides mobile application and web-based management backend, offering feedback data statistical analysis and preference aggregation decision-making functions; Construction monitoring module: This module integrates IoT construction monitoring equipment management, real-time carbon footprint calculation, and early warning information push functions. It supports linkage with digital twin models, mapping construction progress and hourly construction carbon emission data into the three-dimensional model in real time, thereby realizing visualized control of the construction process. The effect evaluation module provides evaluation task management, data comparison and analysis, and evaluation report generation functions. It supports automatic feedback of evaluation results to the multi-agent reinforcement learning model of the intelligent solution generation module to realize iterative updates of strategies. This module provides trend analysis charts to intuitively show the changing trend of the transformation effect over time.

Images