A method and system for urban renewal based on AI-based urban health assessment

Abstract

This application relates to the technical field of urban renewal, and discloses an urban renewal method and system based on AI-based urban health assessment. The method includes: constructing a dynamic health assessment indicator system based on reinforcement learning; analyzing multi-source data such as remote sensing time-series data and IoT real-time data; extracting core attributes to generate a unified semantic representation; building a dynamic mapping model through a graph neural network; inputting the data into a CIM database and recording it on a blockchain to form a traceable dataset; based on this dataset and multi-subject quantitative consensus rules, AI identifies indicator conflicts and generates reconciliation strategies, with edge nodes processed hierarchically; after reconciliation, a cache partition is divided, a digital twin engine is called to construct a coupled model, reinforcement learning optimizes parameters and performs incremental updates; the AI ​​engine transforms and verifies the rules to complete adaptability verification; and related models and strategies are iteratively optimized based on blockchain data. This application enables the refined and intelligent advancement of urban renewal.

Description

Technical Field

[0001] This application relates to the technical field of urban renewal, and in particular to an urban renewal method and system based on AI-based urban health assessment. Background Technology

[0002] Urban renewal is an important measure to optimize urban spatial structure and improve urban governance. Urban health checks, as a core preliminary step in urban renewal, directly impact the rationality and effectiveness of urban renewal plans through their scientific rigor and timeliness. Currently, urban health checks are gradually incorporating multi-source data to provide fundamental support for urban renewal through the integration and analysis of various data sources. However, existing data processing and indicator construction models still have significant limitations.

[0003] In the existing urban health assessment and urban renewal process, there is a lack of a unified integration and traceability mechanism for multi-source data. The health assessment indicator system is difficult to adapt to the dynamic development needs of the city. At the same time, the efficiency and accuracy of the indicator conflict handling, the simulation optimization of the renewal plan and the model iteration are insufficient, which cannot fully meet the requirements of the refined and intelligent development of urban renewal.

[0004] As can be seen from the above, the existing urban renewal methods based on urban health assessment have problems such as low accuracy and poor adaptability. How to achieve refined and intelligent advancement of urban renewal still needs to be solved. Summary of the Invention

[0005] To achieve refined and intelligent urban renewal, this application provides an urban renewal method and system based on AI-based urban health assessment.

[0006] Firstly, this application provides an urban renewal method based on AI-based urban health assessment, employing the following technical solution:

[0007] An urban renewal method based on AI-driven urban health assessment includes:

[0008] A dynamic health check indicator system is constructed based on reinforcement learning algorithms. Remote sensing time series data, IoT real-time data, multi-entity contribution data and historical health check cases are parsed synchronously at the data access end to extract core attributes and generate a unified semantic representation. The core attributes include indicator dynamic weights, area characteristic coefficients, entity contribution scores and hidden danger warning levels. A dynamic mapping model corresponding to indicators, areas and entities is built through graph neural networks. The model is transmitted to the CIM database through a distributed synchronization mechanism. Blockchain nodes synchronously record the entity's rights and responsibilities trajectory and indicator iteration logs to form a traceable dataset with dynamic tags.

[0009] Based on the traceable dataset and multi-agent quantitative consensus rules, the AI ​​model identifies indicator conflicts and disagreements, dynamically generates corresponding weight-oriented reconciliation strategies, and performs hierarchical processing by edge nodes. For high-priority security risk-related data, an expedited consensus calibration process is initiated, and for routine quality improvement indicator data, a multi-agent weighted fusion process is executed.

[0010] After reconciliation, in the edge-center collaborative architecture, the cache partitions are divided according to the physical examination level and update priority. The digital twin engine is called to build a coupled model corresponding to the physical examination index, update plan and simulation effect. The simulation parameters are optimized by combining reinforcement learning, and incremental synchronous updates are only performed on the index iteration and plan adjustment parts.

[0011] The AI ​​engine transforms physical examination standards and regional requirements into structured verification rules, completing the adaptation verification of updated plans and dynamic indicators; based on multi-subject feedback and implementation data recorded on the blockchain, the indicator weight model and reconciliation strategy are iteratively optimized.

[0012] Optionally, in the process of constructing multi-agent quantitative consensus rules and weight-oriented reconciliation strategies, the method also includes:

[0013] Based on the Shapley value algorithm, a multi-subject quantitative consensus rule is constructed, and three core parameters are extracted from the data contribution of each subject recorded by the blockchain node: the timeliness of opinion response and the accuracy of verification.

[0014] The Shapley value algorithm is used to calculate the weighted values ​​of three core parameters: data contribution, timeliness of opinion response, and verification accuracy. The weighted values ​​are dynamically updated according to the subsequent behavior of each subject.

[0015] After generating a weight-oriented reconciliation strategy, the rational opinions put forward by the subjects with higher weight coefficients are adopted first and used as the core basis of the reconciliation strategy.

[0016] By using an artificial intelligence model to perform full semantic analysis on the objections raised by subjects with lower weight coefficients, if the objections have compliance basis such as urban physical examination standards or relevant policy documents, a secondary consensus process is triggered until a unified and reconciled result is reached.

[0017] Optionally, in the process of scenario-based simulation and selection of the optimal update scheme for the digital twin coupled model, the method further includes:

[0018] A scenario-based simulation module is added to the digital twin coupling model. For three typical urban renewal scenarios, namely old residential area renovation, historical block protection, and industrial park upgrading, corresponding simulation parameter thresholds are preset. The simulation parameter thresholds for different scenarios are set according to the key requirements of urban health check for the corresponding scenarios.

[0019] Real-time acquisition of field data transmitted by IoT sensing devices, comparison of field data with preset simulation parameter thresholds, and dynamic correction of simulation parameters of digital twin coupling model;

[0020] Based on the corrected simulation parameters, simulations were performed on various candidate update schemes to generate a multi-scheme simulation comparison report, including the achievement of physical examination indicators, implementation cycle, and resource consumption of each scheme.

[0021] By analyzing the data in the simulation comparison report using reinforcement learning algorithms, the update scheme with the highest adaptability to the dynamic physical examination indicators is selected, and the final implementation scheme is determined.

[0022] Optionally, during the integration of smart contracts into blockchain nodes and the iterative optimization of the indicator weight model, the method also includes:

[0023] Integrate a smart contract module into the blockchain node and write the multi-subject contribution weight coefficients calculated by the Shapley value algorithm into the smart contract.

[0024] The smart contract automatically executes the profit distribution operation based on the contribution weight coefficient of each entity, and records the division of responsibilities of each entity in the implementation of the updated plan.

[0025] When the implementation of the updated plan fails to meet the threshold of the dynamic health check indicators, the smart contract automatically triggers a rectification reminder to the responsible party and links the rectification status with the subsequent assessment results of the party.

[0026] A multimodal feedback semantic parsing model is introduced to extract multi-subject feedback data and implementation data in three forms: text, voice, and visual annotation from blockchain records. The core demands of each subject and the problems in the implementation process are analyzed. Based on the analysis results, the calculation logic of the indicator weight model and the adaptation rules of the weight-oriented harmonization strategy are precisely optimized.

[0027] Optionally, during the preprocessing of multimodal feedback data at the edge and the linkage with smart contracts, the method further includes:

[0028] A lightweight multimodal fusion algorithm is introduced at the edge to preprocess feedback data in real time in three forms: text, voice, and visual annotation, retaining core feature data related to the optimization of physical examination indicators and the rectification of the plan;

[0029] A preset threshold for the quality of edge data preprocessing is set. When the integrity and accuracy of core feature data do not meet the threshold, the edge node is triggered to send a feedback reminder to the corresponding subject. The supplementary data is then uploaded to the central node after secondary preprocessing.

[0030] Add a feedback data validity verification clause to the smart contract, compare the preprocessed core feature data with the urban physical examination standard database, automatically mark invalid feedback data and record the corresponding subject information, and invalid feedback data will not be included in the optimization basis of the indicator weight model.

[0031] The core feature data is cross-validated through an edge-center collaborative verification mechanism, and the verified data is synchronously written to the blockchain and CIM database.

[0032] Optionally, in the process of scenario-based simulation and candidate update scheme generation, the method further includes:

[0033] Construct a cross-scenario collaborative constraint model, extract common indicators and differentiated requirements for three types of scenarios: renovation of old residential areas, protection of historical blocks, and upgrading of industrial parks, and set resource allocation, functional adaptation constraints and indicator priority association rules between scenarios;

[0034] In the candidate update scheme generation stage, a multi-objective optimization algorithm is introduced, with the achievement rate of physical examination indicators, resource consumption cost, and scenario collaboration adaptability as the core optimization objectives, to generate a set of candidate schemes that take into account both single scenario achievement and cross-scenario collaboration.

[0035] A scenario conflict early warning module has been added to identify cross-scenario resource and functional conflicts in the candidate solution set in advance, and output the conflict type, scope of impact and resolution suggestions as a reference for simulation.

[0036] Cross-scenario collaborative constraints are embedded into the simulation parameter system of the digital twin coupled model, and the threshold values ​​of simulation parameters for each scenario are corrected simultaneously.

[0037] Optionally, during the final execution plan determination and full-process iterative optimization process, the method also includes:

[0038] A spatiotemporal attention mechanism is introduced to dynamically weight the modified digital twin simulation parameters, strengthen the influence weight of core indicators in different time dimensions and spatial regions on the simulation results, and improve the spatiotemporal adaptation accuracy of the simulation results.

[0039] Once the final implementation plan is determined, a dynamic correlation model between the implementation progress of the plan and the achievement of the physical examination indicators is constructed. Implementation data is collected according to preset time nodes, and the deviation value of the indicator achievement rate is calculated in real time. When the deviation value exceeds the preset range, the plan fine-tuning process is automatically triggered.

[0040] Based on feedback data of cross-scenario collaboration effects, optimize the cross-scenario collaboration constraint model conditions and the weight allocation logic of the multi-objective optimization algorithm, and simultaneously update the benefit distribution and liability determination clauses related to cross-scenario collaboration in smart contracts.

[0041] A linkage and iteration mechanism is established between the indicator weight model and the cross-scenario collaborative constraint model. The result of indicator weight adjustment corrects the cross-scenario collaborative constraint conditions in reverse, and the synchronization of cross-scenario collaborative effect data serves as the core basis for the optimization of the indicator weight model.

[0042] Secondly, this application provides an urban renewal system based on AI-based urban health assessment, employing the following technical solution:

[0043] An urban renewal system based on AI-powered urban health assessment includes:

[0044] The dynamic indicator system and traceability dataset generation module constructs a dynamic health check indicator system based on reinforcement learning algorithms. At the data access end, it synchronously analyzes remote sensing time-series data, IoT real-time data, multi-entity contribution data, and historical health check cases to extract core attributes and generate a unified semantic representation. The core attributes include indicator dynamic weights, area characteristic coefficients, entity contribution scores, and hidden danger warning levels. A dynamic mapping model corresponding to indicators, areas, and entities is built through graph neural networks. The model is then transmitted to the CIM database through a distributed synchronization mechanism. Simultaneously, blockchain nodes record the entity's rights and responsibilities trajectory and indicator iteration logs, forming a traceable dataset with dynamic tags.

[0045] The indicator conflict identification and weight reconciliation module, based on the traceable dataset and multi-subject quantitative consensus rules, identifies indicator conflicts and disagreements through an AI model, dynamically generates corresponding weight-oriented reconciliation strategies, and performs hierarchical processing by edge nodes. It initiates an expedited consensus calibration process for high-priority security risk-related data and performs a multi-subject weighted fusion process for routine quality improvement indicator data.

[0046] After the digital twin modeling and incremental update module is reconciled, in the edge-center collaborative architecture, the cache partition is divided according to the physical examination level and update priority. The digital twin engine is called to build a coupled model corresponding to the physical examination index, update plan and simulation effect. The simulation parameters are optimized by combining reinforcement learning, and incremental synchronous updates are only performed on the index iteration and plan adjustment parts.

[0047] The scheme verification and model iteration optimization module uses an AI engine to transform physical examination standards and regional requirements into structured verification rules, completing the adaptation verification of updated schemes and dynamic indicators; based on multi-subject feedback and implementation data recorded on the blockchain, it iteratively optimizes the indicator weight model and reconciliation strategy.

[0048] Thirdly, this application provides an electronic device that adopts the following technical solution:

[0049] An electronic device includes a processor running a program for an urban renewal method based on AI-based urban health assessment as described in any one of the preceding claims.

[0050] Fourthly, this application provides a storage medium, which adopts the following technical solution:

[0051] A storage medium storing a program for an urban renewal method based on AI-based urban health assessment, as described in any one of the above-mentioned methods.

[0052] In summary, this application includes at least one of the following beneficial technical effects:

[0053] A dynamic health checkup indicator system is constructed based on reinforcement learning and graph neural networks. It accurately analyzes multi-source data and extracts core attributes. Combines regional characteristics and subject needs to generate a unified semantic representation. Data traceability is achieved through blockchain to ensure that health checkup indicators meet the actual needs of different regions and stages. The weights of multiple subjects are quantified by Shapley value algorithm. AI models identify indicator conflicts and generate targeted reconciliation strategies. Differentiated management is achieved through hierarchical processing of edge nodes. At the same time, cross-scenario collaborative constraints, conflict early warning and multi-objective optimization ensure that candidate solutions take into account both single-scenario adaptability and overall synergy. From indicator construction and data processing to solution generation, precise management is achieved at every stage.

[0054] By integrating technologies such as digital twins, reinforcement learning, and spatiotemporal attention mechanisms, a coupled model is constructed and simulation parameters are dynamically optimized. The model is then corrected using real-time on-site data to accurately simulate the implementation effect of the solution and select the optimal solution, replacing the traditional experience-based judgment mode. Through smart contracts, the rights, responsibilities, and benefits of multiple stakeholders are automatically managed. The edge terminals and central nodes are linked to complete data preprocessing and collaborative verification. A multimodal feedback semantic parsing model is used to accurately capture optimization needs. A linkage and iteration mechanism of indicator weights and cross-scenario collaborative constraint models is established to achieve full-process data-driven, automatic model optimization, and intelligent process management, which greatly improves the efficiency and accuracy of urban renewal and enables a deep integration of refinement and intelligence. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating an AI-based urban health assessment method for urban renewal, according to an exemplary embodiment.

[0056] Figure 2 This is a structural block diagram of an urban renewal system based on AI-based urban health assessment, illustrated according to an exemplary embodiment. Detailed Implementation

[0057] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0058] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0059] This application discloses an urban renewal method based on AI-based urban health assessment, referring to... Figure 1 ,include:

[0060] S100 constructs a dynamic health check indicator system based on reinforcement learning algorithms. At the data access end, it synchronously analyzes remote sensing time-series data, IoT real-time data, multi-entity contribution data, and historical health check cases to extract core attributes and generate a unified semantic representation. The core attributes include dynamic weight of indicators, regional characteristic coefficient, entity contribution score, and hidden danger warning level. A dynamic mapping model corresponding to indicators, regions, and entities is built through graph neural networks. The model is then transmitted to the CIM database through a distributed synchronization mechanism. Simultaneously, blockchain nodes record the entity's rights and responsibilities trajectory and indicator iteration logs to form a traceable dataset with dynamic tags.

[0061] Specifically, the steps include the following:

[0062] In practice, S101 first combines existing urban health check standards with the development positioning of different areas (such as the differentiated needs of old residential areas, historical blocks, and industrial parks) to initialize a set of basic health check indicators. Then, a reinforcement learning algorithm is introduced to take "adapting indicators to the dynamic development of the area", "accurately reflecting urban health check risks" and "meeting the needs of multiple entities" as the core optimization objectives, so that the algorithm can dynamically adjust the types, coverage and initial weights of basic indicators.

[0063] For example, for densely populated old residential areas, the algorithm will automatically strengthen the weight of relevant indicators such as infrastructure integrity rate and safety hazard investigation; for industrial parks, it will focus on optimizing indicators such as industrial adaptability and green and low-carbon, and finally form a dynamic health check indicator system that can be dynamically iterated with urban development and regional needs, rather than a fixed one.

[0064] S102. After the dynamic physical examination indicator system is initially constructed, the multi-source data synchronous parsing function of the data access terminal is started. At the same time, four types of core data are accessed and parsed separately to ensure the integrity and availability of the data.

[0065] Among them, remote sensing time-series data is mainly used to capture macro-level changes in urban spatial morphology and land use, such as dynamic changes in building density and green space coverage in a given area; IoT real-time data mainly comes from various urban sensing devices, such as real-time monitoring data on pipeline operation status, road traffic conditions, and environmental quality; multi-entity contribution data consists of feedback, suggestions, and related data submitted by different entities such as government departments, community residents, and enterprises, such as residents' complaints about aging community facilities and enterprises' requests for industrial upgrading; historical health check cases are reports, data, and relevant experience summaries generated from past urban health checkups, used to provide reference for this health checkup.

[0066] It should be noted that during the analysis process, invalid and redundant information in various data will be removed, and the core data content related to the physical examination indicators will be retained.

[0067] S103, after completing the synchronous analysis of the four types of data, further extracts four core attributes from the effective data after analysis: dynamic weight of indicators, regional characteristic coefficient, main contribution score, and hidden danger warning level.

[0068] Among them, the dynamic weight of the indicators is the dynamic weight value corresponding to each physical examination indicator after optimization by the reinforcement learning algorithm, which is used to distinguish the importance of different indicators in the physical examination assessment; the regional characteristic coefficient is a unique coefficient assigned to each region based on its development characteristics and positioning, which is used to adapt to the physical examination needs of different regions; the subject contribution score is a contribution score assigned to each subject based on the data submitted by each subject and the effectiveness and timeliness of the feedback, which is used to reflect the participation value of different subjects in the urban physical examination; the hidden danger warning level is based on the analyzed data to preliminarily judge the level of urban development hidden dangers in each region, providing a reference for subsequent hidden danger handling.

[0069] After extraction, these core attributes will undergo unified semantic standardization processing, transforming attribute data of different formats and types into a dataset with unified semantic representation, thus avoiding deviations in subsequent processing due to inconsistent data formats.

[0070] S104, based on the generated unified semantic representation of core attributes, introduces a graph neural network algorithm to build a dynamic mapping model corresponding to the three elements of indicator, region, and subject.

[0071] During the construction process, various physical examination indicators, different city districts, and various entities participating in the physical examination will be used as the core nodes of the model. Then, through the correlation between nodes (such as a certain type of indicator corresponding to a certain number of districts, or a certain entity corresponding to the physical examination feedback of a certain district), the connection relationship between nodes will be constructed. Finally, a mapping model that can accurately reflect the relationship between the three and can be dynamically adjusted with data updates will be formed to ensure that the indicators can accurately match the needs of the districts and meet the demands of the entities.

[0072] S105. After the dynamic mapping model is built, the model data and related core attribute data are synchronously transmitted to the CIM database (City Information Model Database) through a distributed synchronization mechanism to achieve centralized storage and management of data, facilitating quick access to relevant data in subsequent steps.

[0073] At the same time, the recording function of the blockchain nodes is activated to record the rights and responsibilities of each entity (such as an entity being responsible for submitting physical examination data for a certain area, or an entity being responsible for the preliminary review of indicators) and the iteration logs of indicators (such as the weight adjustment time, adjustment reason, and adjusted value of a certain type of indicator) in the blockchain nodes in real time. The immutable nature of the blockchain can ensure the authenticity and traceability of this data.

[0074] S106 After the data is transferred to the CIM database and recorded on the blockchain, all relevant data (core attribute data, mapping model data, subject responsibility data, indicator iteration data) are integrated, and corresponding dynamic tags (such as data source tags, data generation time tags, indicator associated area tags) are added to each data point, ultimately forming a complete dataset with dynamic tags and full traceability.

[0075] By constructing a dynamic physical examination indicator system, the problem of fixed traditional urban physical examination indicators that cannot adapt to the dynamic development needs of cities is solved, ensuring the scientific nature and relevance of the physical examination indicators. In addition, through multi-source data analysis, unified semantic processing, graph neural network mapping and blockchain recording, the unified integration and full traceability of multi-source data are achieved, solving the problems of inconsistent formats, unreliable data and lack of traceability in traditional multi-source data.

[0076] S200, based on the traceable dataset and multi-subject quantitative consensus rules, identifies indicator conflicts and disagreements through an AI model, dynamically generates corresponding weight-oriented reconciliation strategies, and performs hierarchical processing by edge nodes. For high-priority security risk-related data, an expedited consensus calibration process is initiated, and for routine quality improvement indicator data, a multi-subject weighted fusion process is executed.

[0077] Specifically, the steps include the following:

[0078] S201. First, directly retrieve the traceable dataset with dynamic tags generated by S100, and filter out the core data, multi-subject information and indicator correlations related to urban physical examination indicators to ensure that the data can be directly used for subsequent conflict identification and reconciliation.

[0079] At the same time, a pre-defined multi-stakeholder quantitative consensus rule is established. This rule is formulated in advance based on the responsibilities, data contribution capabilities, and participation rights of each participating entity (government departments, community residents, enterprises, testing institutions, etc.). The core of the rule includes the criteria for determining the weight of each entity's opinion, the scope of indicator conflicts, and the minimum threshold for reaching a consensus. This provides a unified basis for subsequent conflict identification and reconciliation, avoiding reconciliation deviations caused by the lack of rules.

[0080] S202, the retrieved traceable dataset and the clear multi-agent quantitative consensus rules are synchronously input into the preset AI model. The AI ​​model first performs a quick verification of the data, eliminating any missing or erroneous data that may occur during the data transmission process, to ensure the integrity and accuracy of the input data.

[0081] Meanwhile, the model completes the initial configuration of its own parameters according to the multi-subject quantitative consensus rules, such as setting the identification threshold for indicator conflicts, the judgment logic for opinion differences, and the core parameters for weight calculation, to ensure that the model can accurately meet the multi-subject collaborative needs of this urban health check and prepare for subsequent identification and reconciliation.

[0082] S203. After the AI ​​model is initialized, the automatic recognition process is initiated, carrying out recognition work in two scenarios. The first is indicator conflict recognition, which mainly targets various health check indicators in the traceable dataset to identify whether there are contradictions in the indicator values ​​and weights corresponding to different data sources and different areas. For example, the infrastructure integrity rate indicator in the same area may be inconsistent between the remote sensing time series data measurement results and the IoT real-time monitoring results, or the weight settings of the same type of indicator in different areas may not be in line with the development positioning of the area, resulting in conflicts. The second is opinion disagreement recognition, which mainly identifies the different opinions raised by various participating entities regarding health check indicators and data validity. For example, community residents may believe that the weight of a certain safety hazard indicator is too low, while government departments believe that the weight setting is reasonable in combination with the overall plan. Such opinion disagreements will be accurately captured and classified by the model, and the indicators involved in the disagreement, the relevant entities, and the core content of the disagreement will be marked.

[0083] S204 addresses all the conflicts and disagreements identified by the model by using an AI model that combines multi-stakeholder quantitative consensus rules to dynamically generate a unique weight-oriented reconciliation strategy. The core logic is "weight priority, compliance consideration, and meeting needs".

[0084] First, the model calculates the opinion weight of each participating entity based on the multi-entity quantitative consensus rules. The weight is determined by combining the entity's data contribution, the extent of its authority and responsibility, and the rationality of its opinion. Then, for different types of conflicts and disagreements, differentiated reconciliation plans are formulated. For example, in the case of conflicting indicator values, the data or calculation methods provided by the entity with the higher weight are used for calibration first, while fine-tuning is done by combining the opinions of other entities. In the case of disagreements, the rational opinions of the entity with the higher weight are adopted first as the core of reconciliation, while the rationality of the opinions of the entity with the lower weight is verified. This ensures that the reconciliation strategy not only complies with the consensus rules but also takes into account the demands of all entities, avoiding a "one-size-fits-all" approach to reconciliation.

[0085] After the S205 reconciliation strategy is generated, it is synchronously distributed to each edge node through the edge-center collaborative architecture. The edge node is responsible for the specific hierarchical processing and execution work. The core is to distinguish between "high-priority security risk related data" and "routine quality improvement indicator data" to avoid inefficiency or omission of risks due to unified processing.

[0086] For high-priority safety hazard related data, such as conflicting indicators and data deviations involving fire safety, pipeline leaks, and building structural safety, edge nodes immediately initiate an expedited consensus calibration process. This process quickly connects with relevant stakeholders, retrieves supplementary data, and completes the reconciliation of indicator conflicts and data calibration in the shortest possible time. This ensures the accuracy of safety hazard-related indicators and buys time for subsequent hazard handling. For routine quality improvement indicators, such as conflicts and disagreements related to green space coverage, environmental beautification, and facility improvement, edge nodes execute a multi-stakeholder weighted fusion process. Based on the weight of each stakeholder's opinion, different opinions and data are weighted and calculated to generate unified data and indicator setting schemes after fusion. This ensures that the reconciliation results meet the area's quality improvement needs.

[0087] S206 After the edge nodes complete the hierarchical processing, they will synchronously upload the reconciled indicator data, reconciliation process, relevant subject opinions and processing results to the central node, update the traceable dataset, and ensure that the indicator data and subject opinions in the dataset are the unified results after reconciliation. At the same time, the blockchain nodes will synchronously record the reconciliation process, reconciliation strategy and the participation of each subject to ensure that the reconciliation process is traceable and verifiable.

[0088] By accurately identifying conflicts and disagreements among indicators through AI models, the pain points of inconsistent data from multiple sources and conflicting demands from multiple stakeholders are resolved, avoiding deviations in the formulation of subsequent update plans due to data contradictions and disagreements. In addition, through a weight-oriented reconciliation strategy and hierarchical processing of edge nodes, the reasonable demands of all participating stakeholders are taken into account, while also emphasizing the principle of prioritizing the handling of security risks, ensuring that the reconciled indicator data is accurate, reliable, and compliant.

[0089] After S300 is harmonized, in the edge-center collaborative architecture, the cache partitions are divided according to the health check level and update priority. The digital twin engine is called to build a coupled model corresponding to the health check index, update plan and simulation effect. The simulation parameters are optimized by combining reinforcement learning, and incremental synchronous updates are only performed on the index iteration and plan adjustment parts.

[0090] Specifically, the steps include the following:

[0091] After S200 completes the hierarchical processing and synchronizes the data, S301 first receives the harmonized valid data uploaded by the edge nodes to the central node, including unified physical examination indicator data, physical examination levels of each area, update requirement list, and basic information confirmed by relevant entities.

[0092] Subsequently, the received data undergoes rapid pre-verification, focusing on checking the data's integrity and consistency to confirm that there are no unresolved conflicts or missing information. At the same time, the data format is verified to ensure it meets the access requirements of the digital twin engine, preventing subsequent modeling failures or simulation deviations due to data issues. After verification, the data is synchronously pushed to the corresponding processing nodes in the edge-center collaborative architecture to prepare for subsequent operations.

[0093] After data verification is passed, S302 leverages the distributed storage advantages of the edge-center collaborative architecture to divide the cache into two major dimensions, ensuring orderly data storage and efficient retrieval.

[0094] The first dimension is the health check level. Based on the harmonized data, each city area is divided into different levels according to the health check results (such as excellent, qualified, general hidden danger, major hidden danger). Data of different levels of areas are stored in corresponding cache partitions. The second dimension is the update priority. According to the health check level of the area, the severity of the hidden danger, the urgency of people's livelihood needs, etc., the update needs of each area are divided into three priorities: high, medium and low. Data of high priority (such as areas with major safety hazards) is allocated to a separate cache partition to ensure the efficiency of access.

[0095] Meanwhile, the caching roles of edge nodes and central nodes are clearly defined. Edge nodes mainly cache high-priority, frequently accessed data within their respective regions (such as physical examination indicators and update scheme templates for their regions), while central nodes cache all region data and global model data, achieving "rapid response at the edge and unified management at the center".

[0096] S303 After the cache partition is divided, the system automatically calls the preset digital twin engine. First, the divided cache partition data is synchronously loaded into the engine. At the same time, the core basic data such as the basic geographic information data, existing buildings, and infrastructure distribution of the corresponding city area are loaded to provide scene support for modeling.

[0097] Subsequently, based on the core needs of this urban health check, the initial configuration of the digital twin engine was completed. This included setting the simulation time dimension (such as short-term, medium-term, and long-term update simulations), spatial accuracy (such as district level, building level, and facility level), and evaluation dimensions of simulation effects (such as indicator compliance rate, scheme feasibility, and resource consumption). This ensured that the engine could accurately match the simulation needs of this urban renewal and avoid simulation result distortion caused by improper initialization parameters.

[0098] S304, after the engine initialization is completed, starts the coupled model construction process. The core is to establish a two-way correlation between "physical examination indicators - update scheme - simulation effect" to form a complete coupled model.

[0099] First, the harmonized health check indicators are used as input conditions for the model, clarifying the target threshold, current value, and gap for each indicator. Then, corresponding update plan templates are matched, and multiple targeted update plans are generated based on the actual situation of the area (such as old residential areas and historical blocks). Each plan clearly defines the specific implementation content, implementation steps, and resource investment. Finally, through a digital twin engine, each update plan is linked with the simulation effect to simulate the changes in each health check indicator, the improvement effect on urban spatial form, and the degree of improvement in people's experience after the implementation of the plan. This achieves a closed-loop association of "input indicators → matching plan → output simulation effect," ensuring that the model can accurately reflect the dynamic influence relationship between the three.

[0100] S305. After the initial construction of the coupled model is completed, a reinforcement learning algorithm is introduced to dynamically optimize the simulation parameters of the model. The core objective is to improve the accuracy of the simulation results, make them more in line with actual needs, and optimize the simulation efficiency.

[0101] First, the algorithm uses "the degree of fit between the simulation effect and the actual scenario," "the pass rate of the physical examination indicators," and "the rationality of resource consumption" as the core optimization objectives of reinforcement learning. The algorithm automatically iteratively adjusts the initial simulation parameters (such as simulation step size, indicator weight ratio, and scheme implementation cycle setting). Second, through multiple simulation iterations, the algorithm continuously learns the simulation effects corresponding to different parameter combinations, eliminates parameter combinations with large deviations, and retains the optimal parameter combinations. For example, for areas with major hidden dangers, the algorithm optimizes the simulation parameters related to safety hazard rectification to ensure that the simulation results can accurately reflect the changes in hazard indicators after rectification. Finally, the optimized simulation parameters are synchronously updated to the coupled model to complete the model optimization and calibration, ensuring that the simulation results can provide a reliable reference for subsequent scheme selection.

[0102] After the simulation parameters are optimized, the system starts the incremental synchronous update process. The core principle is to "only update the changes, not the whole", so as to avoid the problems of excessive bandwidth consumption and low efficiency caused by full updates.

[0103] Specifically, synchronous updates are performed on only two types of content: one is the content related to indicator iteration, namely, the parts of the harmonized data that have changed compared to the initial indicators, and the parts of the indicator weights adjusted after reinforcement learning optimization; the other is the content related to scheme adjustment, namely, the details of the updated scheme and implementation steps in the coupled model that need to be adjusted after simulation verification. During synchronous updates, the edge nodes first complete the updates of the relevant content in their respective regions, and then synchronize them to the central node. After the central node verifies and confirms the updated content, it updates the global coupled model and cached partition data to ensure that the data and models of the edge nodes and the central node are consistent. At this point, the entire execution process of S300 is completed.

[0104] By partitioning the cache into edges and centers using a collaborative architecture, the pain points of chaotic data storage and inefficient access across multiple regions and priorities are resolved, ensuring that data can quickly and accurately support modeling work. In addition, through the construction of digital twin coupled models and the optimization of reinforcement learning parameters, abstract health indicators and update plans are transformed into visualized and verifiable simulation effects, solving the problem that traditional urban renewal plans are "based on experience and cannot predict the effects in advance." At the same time, the incremental synchronous update mechanism significantly improves the efficiency of data and model updates and reduces system operating costs.

[0105] S400 uses an AI engine to transform physical examination standards and regional requirements into structured verification rules, completing the adaptation verification of updated plans and dynamic indicators; based on multi-subject feedback and implementation data recorded on the blockchain, it iteratively optimizes the indicator weight model and reconciliation strategy.

[0106] Specifically, the steps include the following:

[0107] After S401 and S300 completed simulation parameter optimization and incremental synchronization, S400 first initiated preparatory work, comprehensively sorting out two types of core criteria and organizing them into standardized data. One type is various urban health check standards, including national and local standards, safety regulations, and quality improvement requirements, covering all relevant areas such as infrastructure, safety hazards, ecological environment, and people's livelihood, ensuring that verification has a unified compliance basis. The other type is the specific requirements of each urban area, combining the area's development positioning, people's needs, and historical characteristics, such as requirements for improving convenience facilities in old residential areas, requirements for preserving the historical character of historical blocks, and requirements for industrial adaptation in industrial parks, ensuring that verification is relevant to the actual situation of the area and does not deviate from actual needs. After sorting, both types of criteria are organized into standardized data, removing redundant and ambiguous expressions to ensure that the data can be accurately recognized and analyzed by the AI ​​engine.

[0108] S402 synchronously inputs the standardized data of the physical examination specifications and regional requirements into the preset AI engine. The engine first performs semantic analysis on the data to clarify the core meaning, scope of application, and judgment criteria of each specification and requirement, so as to avoid verification deviations due to semantic ambiguity.

[0109] Subsequently, the AI ​​engine automatically transforms these unstructured specifications and requirements into structured verification rules. The structured rules adopt a clear format of "judgment conditions + verification standards + non-compliance handling prompts". For example, for the upgrading of fire protection facilities in old residential areas, the rules clearly state that "the number of fire protection facilities configured is greater than or equal to the population adaptation standard of the area" and "the facility integrity rate is greater than or equal to 95%". If these conditions are not met, the prompt will be "fire protection facilities need to be added and existing damaged facilities need to be rectified". All structured verification rules form a complete verification rule library, providing clear and implementable judgment basis for subsequent adaptability verification.

[0110] After the AI ​​engine generates the structured verification rules in S403, it simultaneously loads two types of core data: one is the final update plan after S300 optimization (including the specific update content, implementation steps, resource investment, etc. for each area); the other is the dynamic health check indicators after the entire iteration from S100 to S300 (including the optimized indicator weights, compliance thresholds, area characteristic coefficients, etc.).

[0111] Once loading is complete, the compatibility verification process is initiated. The AI ​​engine compares the updated plan with the dynamic health check indicators one by one according to each rule in the verification rule base, and conducts verification in two categories: one is compliance verification, which checks whether the updated plan complies with the city's health check standards, such as whether the safety hazard rectification measures in the updated plan meet the safety standard requirements; the other is compatibility verification, which checks whether the updated plan fits the dynamic health check indicators and the requirements of the area, such as whether the implementation of the updated plan can enable the area's health check indicators to reach the standard threshold, and whether it fits the development characteristics and people's livelihood demands of the area.

[0112] After the compatibility verification is completed (S404), the AI ​​engine automatically generates a detailed verification report, clearly indicating the verification results (pass, fail, require fine-tuning). For items that fail or require fine-tuning, the report details the problem, the corresponding verification rules, and specific rectification suggestions. It also marks the areas involved in the problem, the chapters of the updated solution, and relevant dynamic indicators, making it easier for staff to quickly locate the problem and promote rectification.

[0113] If the verification is successful, it means that the update plan fully complies with the specifications and indicator requirements and can directly enter the implementation stage. If there are any non-compliance items, the staff will adjust the update plan according to the rectification suggestions in the verification report. After the adjustment is completed, the plan will be re-entered into the AI ​​engine for a second verification until it passes the verification, ensuring that the final updated plan is compliant, adaptable and feasible.

[0114] In S405, while completing the compatibility verification of the updated plan, the system automatically retrieves relevant data recorded throughout the blockchain node process, focusing on organizing two core types of data: one type is multi-stakeholder feedback data, including feedback, suggestions, and satisfaction evaluations submitted by various participating entities (government, residents, enterprises, etc.) throughout S100-S400, such as residents' suggestions for improvement of the updated plan and enterprises' feedback on the feasibility of the plan implementation; the other type is implementation-related data, including data processing records, indicator iteration logs, reconciliation process records, simulation optimization data, etc. in S100-S300, as well as rectification records during the verification process. All of these data are fully traceable through the blockchain, ensuring the authenticity and integrity of the data and providing reliable support for subsequent model iterations.

[0115] S406: The compiled multi-subject feedback data and implementation data are simultaneously transmitted to the AI ​​engine, and the iterative optimization process is initiated based on the results of this compatibility verification.

[0116] The indicator weighting model was optimized by adjusting the calculation logic based on feedback from multiple stakeholders and implementation data. For example, the weight of indicators that received concentrated feedback from residents and affected the accuracy of physical examinations was appropriately increased. The area characteristic coefficients with poor adaptability were adjusted in combination with the implementation needs of the area to ensure that the optimized indicator weighting model is more in line with the actual physical examination needs and can accurately reflect the development of the area and the demands of multiple stakeholders.

[0117] The weight-oriented reconciliation strategy was optimized by adjusting the judgment logic and weight calculation standards based on reconciliation issues reflected in feedback from multiple stakeholders and reconciliation process records in the implementation data. For example, the priority determination method for reconciliation of disagreements was optimized, making the reconciliation strategy more considerate of the demands of all stakeholders and more efficient in resolving indicator conflicts and differences of opinion. After optimization, the updated indicator weight model and reconciliation strategy were synchronously uploaded to the central node and edge nodes, and updated to the CIM database and traceable dataset. This provides optimized model and strategy support for subsequent urban health checks and urban renewal work. At this point, the entire execution process of S400 was completed, and the entire urban renewal methodology formed a complete closed loop.

[0118] By using structured verification through an AI engine, the pain points of insufficient compliance and poor adaptability caused by the traditional urban renewal plan's "emphasis on formulation and neglect of verification" are resolved. This ensures that the final implemented plan is compliant, meets dynamic indicators and the needs of the area, avoids rework and rectification after the plan is implemented, and improves the efficiency and quality of urban renewal implementation.

[0119] Furthermore, leveraging the traceability advantages of blockchain data, the system integrates feedback and implementation data from multiple stakeholders throughout the entire process, enabling iterative optimization of indicator weighting models and reconciliation strategies. This solves the problem of traditional technical systems being "fixed and unable to adapt to dynamic needs," allowing the entire urban health check and urban renewal technical system to continuously adapt to urban development, regional needs, and the demands of multiple stakeholders, forming a complete closed loop of "verification-feedback-optimization-improvement." This not only ensures the accurate implementation of this urban renewal work but also provides more scientific and efficient technical support for subsequent urban health checks and urban renewal work, further enhancing the level of refined and intelligent governance in urban renewal.

[0120] In this embodiment of the application, the method specifically includes the following steps in constructing multi-agent quantitative consensus rules and implementing a weight-oriented reconciliation strategy:

[0121] Before the S200 was launched, the Shapley value algorithm was determined as the core (which can accurately quantify the contribution of each entity). Then, the data recorded by the blockchain nodes was retrieved, and three core parameters of each entity (government, residents, enterprises, testing institutions, etc.) were extracted: data contribution (the quantity, completeness, and validity of submitted data), response timeliness (the speed of response to system inquiries and data verification), and verification accuracy (the degree of consistency between submitted data and opinions and actual conditions and standards, such as the accuracy rate of verification of hidden danger data by testing institutions). These three factors directly determine the accuracy of the weight calculation.

[0122] Three core parameters are input into the algorithm, which calculates a unique weight coefficient for each subject based on preset parameter weights (data contribution is the highest, followed by verification accuracy, and finally response time). (The higher the weight, the stronger the credibility of the contribution and opinion.) A dynamic update mechanism is also set up to automatically adjust the weights based on the subject's participation performance (data integrity, response speed, and verification accuracy) to ensure that the weights reflect the subject's real-time participation.

[0123] After the S200AI model generates a preliminary reconciliation strategy, it combines the weight coefficients obtained in the second step and uses the rational opinions of high-weight subjects as the core basis for reconciliation. For example, the "priority rectification of major safety hazards" proposed by government departments (the coordinating body with high weight) and the "regional hazard indicator calibration suggestions" proposed by testing institutions (with high verification accuracy) are directly incorporated into the reconciliation strategy to ensure that the strategy is scientific and reasonable.

[0124] For objections from low-weighted entities, a full-scale semantic analysis using an AI model is conducted to extract core viewpoints and supporting evidence, which are then compared and verified against urban health check standards and relevant policies. If the objection has clear compliance support (such as residents' objections to the weighting of livelihood indicators conforming to local policies), a secondary consensus is triggered (organizing relevant entities to discuss and optimize strategies) until a consensus is reached; if there is no compliance basis, the reasons are provided, and a secondary consensus is not initiated to avoid inefficiency and internal friction.

[0125] By using dynamic weights and secondary consensus, we can ensure the enthusiasm of multiple stakeholders for collaboration, improve the efficiency and accuracy of reconciliation, and provide reliable support for the formulation of subsequent update plans.

[0126] In this embodiment of the application, the method specifically includes the following steps in the process of scenario-based simulation and selection of the optimal update scheme for the digital twin coupled model:

[0127] After initially constructing a coupled model of health check indicators, update schemes, and simulation effects in the S300, a scenario-based simulation module was added to adapt to different scenario requirements. This module focuses on three typical scenarios: renovation of old residential areas, preservation of historical districts, and upgrading of industrial parks, with corresponding preset simulation parameter thresholds for each. These thresholds are set in conjunction with the key aspects of urban health check for each scenario. For example, renovation of old residential areas emphasizes infrastructure and public welfare parameters; preservation of historical districts emphasizes the preservation of historical features and cultural relics; and upgrading of industrial parks emphasizes industrial adaptation and green, low-carbon parameters, ensuring that the simulation closely matches the actual needs of each scenario.

[0128] After the scenario-based module is set up, real-time data from various IoT sensing devices in the city is collected, including real-time data on the operational status of the area's infrastructure, environmental quality, and building conditions. These actual on-site data are compared one by one with the simulation parameter thresholds preset in the first step. If there is a data deviation (i.e., the on-site data does not match the preset threshold), the simulation parameters of the digital twin coupled model are automatically and dynamically corrected to ensure that the model simulation results are highly consistent with the actual on-site situation and improve the simulation accuracy.

[0129] After the simulation parameters are corrected, the multiple candidate update schemes generated by S300 are combined to conduct scenario-based simulations for each scheme. The simulations will show the achievement of each health check index, the required implementation period, and the consumption of various resources after each scheme is implemented. After the simulation is completed, a multi-scheme simulation comparison report will be automatically generated, which will clearly present the advantages, disadvantages and core index data of each scheme, providing an intuitive basis for subsequent selection.

[0130] The simulation comparison report of multiple schemes generated in the third step is retrieved. The core data in the report (achievement rate of physical examination indicators, implementation cycle, resource consumption, etc.) are input into the reinforcement learning algorithm. The algorithm takes "highest adaptability to dynamic physical examination indicators, most reasonable resource consumption, and most optimized implementation cycle" as its core objective, performs comprehensive analysis and ranking of all candidate schemes, and finally selects the scheme with the highest adaptability to dynamic physical examination indicators and the best fit to the actual scenario, which is then determined as the final execution scheme.

[0131] By using scenario-based simulation and reinforcement learning algorithms, the optimal renewal scheme is scientifically selected, providing accurate and reliable support for the adaptability verification of the S400 scheme and the implementation of subsequent schemes, further ensuring the rationality and feasibility of the urban renewal scheme.

[0132] In this embodiment of the application, the method specifically includes the following steps in the process of integrating smart contracts into blockchain nodes and iteratively optimizing the indicator weight model:

[0133] First, the smart contract and blockchain node are adapted and connected to ensure that the contract can read and record blockchain data normally. Then, the contribution weight coefficients of each subject calculated by the Shapley value algorithm are written into the smart contract one by one to achieve the immutable storage of the weight coefficients, providing the core basis for subsequent allocation of rights and responsibilities and calculation of benefits.

[0134] Based on the weight coefficients of each entity written in, the smart contract automatically completes the profit distribution operation for each entity according to the preset distribution rules (the higher the weight, the higher the corresponding profit share). At the same time, it records the specific division of responsibilities of each entity in the implementation of the updated plan in real time, clarifying who is responsible for which work and what the boundaries of responsibility are, ensuring that rights and responsibilities are traceable.

[0135] During the implementation of the updated plan, the effectiveness of the plan implementation is compared with the threshold of dynamic health check indicators in real time. If the implementation effectiveness does not reach the preset threshold, the smart contract will automatically send a rectification reminder to the corresponding responsible party, specifying the rectification requirements and time limit. At the same time, the rectification completion status and rectification effect will be directly linked to the subsequent assessment results of the party. If the rectification is in place, it will not affect the assessment. If the rectification is ineffective, the subsequent weight and assessment rating of the party will be reduced.

[0136] First, the model is started, and multi-subject feedback data and implementation data recorded by blockchain nodes throughout the process are retrieved. The focus is on extracting three forms of data: text feedback, voice suggestions, and visual annotations (such as annotations of potential hazards in photos and annotations of solution modifications). Through full semantic analysis of the model, the core demands of each subject, their opinions on the existing process, and the specific problems that arise during the implementation of the solution are accurately extracted. Based on these analysis results, the calculation logic of the indicator weight model is optimized in a targeted manner (such as adjusting the weight ratio of parameters). At the same time, the adaptation rules of the weight-oriented reconciliation strategy are optimized to ensure that the model and strategy can continuously meet the actual needs.

[0137] By using smart contracts, the rights, responsibilities, and benefits of multiple stakeholders can be automatically managed, ensuring fair distribution and clear responsibilities. At the same time, multimodal analysis can be used to accurately capture feedback issues, enabling precise iteration of indicator weight models and reconciliation strategies. This further improves the closed loop of the entire urban renewal technology system and enhances the efficiency of multi-stakeholder collaboration and the quality of solution implementation.

[0138] In this embodiment of the application, the method further includes the following steps during the preprocessing of multimodal feedback data at the edge and the linkage with smart contracts:

[0139] A lightweight multimodal fusion algorithm is introduced at the edge to process feedback data. Considering the limited computing power at the edge, a lightweight algorithm is selected to avoid excessive resource consumption. The algorithm receives multi-subject feedback data in three forms: text, voice, and visual annotation in real time, and performs preprocessing work simultaneously. The algorithm focuses on removing redundant information (such as invalid feedback and repetitive statements) that is not related to the optimization of physical examination indicators and the rectification of the plan, and retains only the core feature data to ensure that the data accurately serves the subsequent model optimization.

[0140] The system presets a data preprocessing quality threshold at the edge and triggers a supplementary reminder. It pre-sets the completeness and accuracy thresholds for core feature data and automatically verifies the data after preprocessing. If the data does not meet the threshold (e.g., missing key information or excessive data deviation), the edge node will automatically send a supplementary reminder to the subject that submitted the feedback, clearly informing them of the content to be supplemented and the time limit. After the subject supplements the data, the edge performs preprocessing again until the data meets the standards.

[0141] Add a validity verification clause for feedback data to the smart contract. The core feature data that has been preprocessed and meets the standards at the edge end is synchronously transmitted to the smart contract. The contract calls the city's physical examination standard database to compare the feedback data with the standard requirements one by one. It automatically marks invalid feedback data that does not meet the standards or has no practical reference value, and records the subject information of the data submission. It clarifies that invalid feedback data will not be included in the subsequent indicator weight model optimization, so as to avoid affecting the model optimization accuracy.

[0142] Cross-validation and data synchronization are completed through an edge-center collaborative verification mechanism. The edge pushes the core feature data that has passed the preprocessing and contract verification to the central node. The central node combines the historical data and existing physical examination index data in the CIM database to conduct cross-validation. After confirming that the data is without deviation or conflict, the data is synchronously written into the blockchain and the CIM database to achieve tamper-proof storage and centralized management of the data.

[0143] By employing edge preprocessing, contract verification, and collaborative validation to form multiple checks, the accuracy and reliability of the data used for model optimization are ensured. At the same time, the linkage between the edge and smart contracts, blockchain, and CIM database is realized, further improving the data management closed loop and providing solid support for the accurate iteration of indicator weight models and reconciliation strategies.

[0144] In this embodiment of the application, the method specifically includes the following steps in the process of scenario-based simulation and candidate update scheme generation:

[0145] First, we identify the core needs of three typical scenarios: renovation of old residential communities, protection of historical blocks, and upgrading of industrial parks. We then extract common indicators (such as general indicators for safety hazard rectification and infrastructure improvement) and differentiated needs (such as old residential communities focusing on adapting to people's livelihoods, historical blocks focusing on preserving their original appearance, and industrial parks focusing on industrial empowerment). Next, in conjunction with the overall urban renewal plan, we set resource allocation constraints between scenarios (such as the upper limit and priority of the allocation of human, material, and financial resources in different scenarios), functional adaptation constraints (such as avoiding conflicts in the functions of facilities in different scenarios and conforming to the overall urban functional layout), and indicator priority association rules (such as the weight association between common and differentiated indicators and the collaborative achievement requirements of cross-scenario indicators). This sets the collaborative boundaries for the generation of subsequent candidate solutions.

[0146] During the candidate update scheme generation stage, the multi-objective optimization algorithm is combined with the scenario-based simulation requirements. The three core optimization objectives are clearly defined as the achievement rate of health check indicators (ensuring that the scheme meets the health check standards), resource consumption cost (controlling update costs), and scenario collaboration adaptability (ensuring that the scheme takes into account both single scenario compliance and multi-scenario collaboration). The algorithm combines the previously preset scenario parameter thresholds and cross-scenario collaboration constraint models to automatically generate multiple sets of candidate update schemes, forming a set of candidate schemes. Each scheme takes into account both the personalized needs of a single scenario and the collaboration of multiple scenarios, avoiding the situation where the scheme only adapts to a single scenario and ignores the overall collaboration.

[0147] After the candidate solution set is generated, the newly added scenario conflict early warning module is activated. The module automatically scans all candidate solutions and identifies cross-scenario resource conflicts (such as different scenarios competing for the same batch of update resources) and functional conflicts (such as the facility functions of a certain scenario conflicting with the functional layout of adjacent scenarios). After identification, the module clearly outputs the conflict type, the scope of conflict impact (the scenarios and update links involved), and specific resolution suggestions (such as resource reallocation solutions and functional optimization adjustment directions). This information is used as a preliminary reference for subsequent scenario-based simulations to avoid invalid simulations and reduce the cost of subsequent solution adjustments.

[0148] The cross-scenario resource allocation, function adaptation constraints, and indicator priority association rules set in the first step are embedded one by one into the simulation parameter system of the digital twin coupling model. At the same time, combined with the suggestions output by the conflict warning module, the simulation parameter thresholds of three typical scenarios are simultaneously corrected to ensure that the simulation parameters not only adapt to the key points of a single scenario but also reflect the needs of cross-scenario collaboration, making the subsequent scheme simulation more in line with the actual situation of multi-scenario collaborative updates.

[0149] By employing collaborative constraints, multi-objective optimization, conflict warning, and parameter correction, the cross-scenario synergy and feasibility of candidate solutions are enhanced, providing precise support for subsequent simulation and optimal solution selection. This ensures that the final generated candidate solutions not only meet the update needs of a single scenario but also align with the overall urban renewal plan, thereby improving the efficiency and quality of urban renewal across multiple scenarios.

[0150] In this embodiment of the application, the method specifically includes the following steps during the final execution plan determination and full-process iterative optimization:

[0151] After the parameters of the digital twin coupled model are corrected, a spatiotemporal attention mechanism is introduced. This mechanism can accurately capture the differences in core health indicators in different time dimensions (such as short-term and medium-term update stages) and different spatial regions (such as different areas and different update scenarios). The corrected simulation parameters are dynamically weighted, focusing on strengthening the influence weight of core indicators in key time nodes and core spatial regions (such as safety indicators in areas with major hidden dangers and progress indicators in key update stages) on the simulation results, while weakening the interference of non-core indicators. This significantly improves the spatiotemporal adaptation accuracy of the simulation results, making the simulation results more in line with the actual update needs of different times and regions.

[0152] Once the final implementation plan is determined, a dynamic correlation model between the plan's implementation progress and the achievement of health indicators is immediately established. Multiple key time nodes (such as monthly and quarterly nodes) are set in advance, and actual data on the implementation of the plan (such as construction progress and indicator completion status) are collected in real time at these nodes. The model automatically compares the implementation data with the preset targets and calculates the deviation value of the health indicator achievement rate in real time. If the deviation value exceeds the preset range (such as the achievement rate of an indicator at a certain node being more than 5% lower than the preset value), the system will automatically trigger the plan fine-tuning process, clarify the direction of fine-tuning and specific measures (such as adjusting the construction schedule and optimizing resource allocation), to avoid the deviation from continuing to expand and ensure that the implementation of the plan always revolves around the health indicator target.

[0153] During the implementation of the solution, feedback data on the effectiveness of cross-scenario collaboration is continuously collected (such as the efficiency of collaborative updates in different scenarios, resource utilization, and the synergy of indicator achievement). Based on this data, the constraints of the cross-scenario collaboration constraint model are optimized in a targeted manner (such as adjusting the relevant rules for resource allocation and functional adaptation). At the same time, the weight allocation logic of the multi-objective optimization algorithm is optimized (such as increasing the weight of scenario collaboration adaptability in the algorithm). Meanwhile, the clauses related to cross-scenario collaboration in the smart contract are updated synchronously, including the profit distribution ratio of each subject in cross-scenario collaboration and the standards for liability determination, to ensure that the smart contract adapts to the optimized collaboration rules and to guarantee the fairness and standardization of multi-subject collaboration.

[0154] Establish a linkage and iteration mechanism between the two models, build a linkage and iteration channel between the indicator weight model and the cross-scenario collaborative constraint model, and form a two-way optimization closed loop:

[0155] The adjustment results of the weight model (such as increasing the weight of core indicators and optimizing the weight of differentiated indicators) will, in turn, correct the constraints of the cross-scenario collaborative constraint model, making the collaborative constraints more in line with the needs of indicator optimization.

[0156] Feedback data on cross-scenario collaboration effects will be used as the core basis for optimizing the indicator weight model. For example, if the adaptability of a certain type of collaboration indicator is consistently low, the proportion of that indicator in the weight model will be adjusted to ensure that the two models are optimized collaboratively and adapted to the dynamic needs of urban renewal.

[0157] By improving the closed-loop optimization of the entire process through a linkage and iteration mechanism, we can ensure that the final implementation plan is controllable and that the entire process technology system continuously adapts to the dynamic needs of urban renewal, thereby further improving the refinement, intelligence and overall implementation quality of urban renewal.

[0158] This application discloses an urban renewal system based on AI-based urban health assessment, referring to... Figure 2 ,include:

[0159] The Dynamic Indicator System and Traceability Dataset Generation Module 001 constructs a dynamic health check indicator system based on reinforcement learning algorithms. At the data access end, it synchronously analyzes remote sensing time-series data, IoT real-time data, multi-entity contribution data, and historical health check cases, extracting core attributes to generate a unified semantic representation. These core attributes include indicator dynamic weights, area-specific coefficients, entity contribution scores, and hazard warning levels. A dynamic mapping model corresponding to indicators, areas, and entities is built using a graph neural network. This model is then transmitted to the CIM database via a distributed synchronization mechanism, where blockchain nodes synchronously record the entity's rights and responsibilities trajectory and indicator iteration logs, forming a dynamically labeled traceable dataset.

[0160] The indicator conflict identification and weight reconciliation module 002, based on traceable datasets and multi-subject quantitative consensus rules, identifies indicator conflicts and disagreements through AI models, dynamically generates corresponding weight-oriented reconciliation strategies, and performs hierarchical processing by edge nodes. It initiates an expedited consensus calibration process for high-priority security risk-related data and performs a multi-subject weighted fusion process for routine quality improvement indicator data.

[0161] After the digital twin modeling and incremental update module 003 is harmonized, in the edge-center collaborative architecture, the cache partition is divided according to the physical examination level and update priority. The digital twin engine is called to build a coupled model corresponding to the physical examination index, update plan and simulation effect. The simulation parameters are optimized by combining reinforcement learning. Incremental synchronous updates are only performed on the index iteration and plan adjustment parts.

[0162] The scheme verification and model iteration optimization module 004 uses an AI engine to transform the physical examination standards and regional requirements into structured verification rules, and completes the adaptation verification of the updated scheme and dynamic indicators; based on the multi-subject feedback and implementation data recorded by blockchain, iteratively optimizes the indicator weight model and reconciliation strategy.

[0163] This application also discloses an electronic device, including a processor, wherein the processor runs a program of the urban renewal method based on AI urban health assessment as described in any one of the above embodiments.

[0164] This application also discloses a storage medium storing a program for the urban renewal method based on AI urban health assessment as described in any one of the above embodiments.

[0165] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. An urban renewal method based on AI-based urban health assessment, characterized in that, include: A dynamic health check indicator system is constructed based on reinforcement learning algorithms. Remote sensing time series data, IoT real-time data, multi-entity contribution data and historical health check cases are parsed synchronously at the data access end to extract core attributes and generate a unified semantic representation. The core attributes include indicator dynamic weights, area characteristic coefficients, entity contribution scores and hidden danger warning levels. A dynamic mapping model corresponding to indicators, areas and entities is built through graph neural networks. The model is transmitted to the CIM database through a distributed synchronization mechanism. Blockchain nodes synchronously record the entity's rights and responsibilities trajectory and indicator iteration logs to form a traceable dataset with dynamic tags. Based on the traceable dataset and multi-agent quantitative consensus rules, the AI ​​model identifies indicator conflicts and disagreements, dynamically generates corresponding weight-oriented reconciliation strategies, and performs hierarchical processing by edge nodes. For high-priority security risk-related data, an expedited consensus calibration process is initiated, and for routine quality improvement indicator data, a multi-agent weighted fusion process is executed. After reconciliation, in the edge-center collaborative architecture, the cache partitions are divided according to the health check level and update priority, and the digital twin engine is called to build a coupled model corresponding to the health check index, update scheme and simulation effect. A cross-scenario collaborative constraint model is constructed to extract common indicators and differentiated requirements for three types of scenarios: renovation of old residential areas, protection of historical blocks, and upgrading of industrial parks. Constraints on resource allocation and functional adaptation between scenarios and rules on indicator priority are set. In the candidate update scheme generation stage, a multi-objective optimization algorithm is introduced, with the achievement rate of physical examination indicators, resource consumption cost, and scenario collaborative adaptation as the core optimization objectives, to generate a set of candidate schemes that take into account both single scenario achievement and cross-scenario collaboration. A scenario conflict early warning module has been added to identify cross-scenario resource and functional conflicts in the candidate solution set in advance, and output the conflict type, scope of impact and resolution suggestions as a reference for simulation. Cross-scenario collaborative constraints are embedded into the coupled model simulation parameter system corresponding to physical examination indicators, update schemes, and simulation effects, and the threshold values ​​of simulation parameters for each scenario are corrected simultaneously. The coupled model of physical examination indicators, update plan and simulation effect includes a scenario-based simulation module. For three typical urban renewal scenarios, namely old community renovation, historical block protection and industrial park upgrading, corresponding simulation parameter thresholds are preset. The simulation parameter thresholds for different scenarios are set according to the key requirements of urban physical examination for the corresponding scenarios. Real-time acquisition of field data transmitted by IoT sensing devices; comparison of field data with preset simulation parameter thresholds; dynamic correction of simulation parameters of the coupled model corresponding to physical examination indicators, update schemes, and simulation effects; based on the corrected simulation parameters, simulation of multiple candidate update schemes is performed, generating a multi-scheme simulation comparison report including the achievement of physical examination indicators, implementation cycle, and resource consumption of each scheme. The data in the simulation comparison report are analyzed by reinforcement learning algorithm to select the update plan with the highest adaptability to the dynamic physical examination indicators and determine it as the final execution plan. By combining reinforcement learning to optimize simulation parameters, incremental synchronous updates are performed only on the parts related to index iteration and scheme adjustment. The AI ​​engine transforms physical examination standards and regional requirements into structured verification rules, completing the adaptation verification of updated plans and dynamic indicators; based on multi-subject feedback and implementation data recorded on the blockchain, the indicator weight model and reconciliation strategy are iteratively optimized.

2. The urban renewal method based on AI-based urban health assessment according to claim 1, characterized in that, In the step of dynamically generating the corresponding weight-oriented harmonic strategy, the method further includes: Based on the Shapley value algorithm, a multi-subject quantitative consensus rule is constructed, and three core parameters are extracted from the data contribution of each subject recorded by the blockchain node: the timeliness of opinion response and the accuracy of verification. The Shapley value algorithm is used to calculate the weighted values ​​of three core parameters: data contribution, timeliness of opinion response, and verification accuracy. The weighted values ​​are dynamically updated according to the subsequent behavior of each subject. After generating a weight-oriented reconciliation strategy, the rational opinions put forward by the subjects with higher weight coefficients are adopted first and used as the core basis of the reconciliation strategy. By using an artificial intelligence model to perform full semantic analysis on the objections raised by subjects with lower weight coefficients, if the objections have compliance basis such as urban physical examination standards or relevant policy documents, a secondary consensus process is triggered until a unified and reconciled result is reached.

3. The urban renewal method based on AI-based urban health assessment according to claim 1, characterized in that, In the steps of the multi-subject feedback and implementation data iterative optimization index weight model and reconciliation strategy based on blockchain records, the method further includes: Integrate a smart contract module into the blockchain node and write the multi-subject contribution weight coefficients calculated by the Shapley value algorithm into the smart contract. The smart contract automatically executes the profit distribution operation based on the contribution weight coefficient of each entity, and records the division of responsibilities of each entity in the implementation of the updated plan. When the implementation of the updated plan fails to meet the threshold of the dynamic health check indicators, the smart contract automatically triggers a rectification reminder to the responsible party and links the rectification status with the subsequent assessment results of the party. A multimodal feedback semantic parsing model is introduced to extract multi-subject feedback data and implementation data in three forms: text, voice, and visual annotation from blockchain records. The core demands of each subject and the problems in the implementation process are analyzed. Based on the analysis results, the calculation logic of the indicator weight model and the adaptation rules of the weight-oriented harmonization strategy are precisely optimized.

4. The urban renewal method based on AI-based urban health assessment according to claim 3, characterized in that, In the step of extracting multi-subject feedback data and implementation data in three forms—text, voice, and visual annotation—from blockchain records, the method further includes: A lightweight multimodal fusion algorithm is introduced at the edge to preprocess feedback data in real time in three forms: text, voice, and visual annotation, retaining core feature data related to the optimization of physical examination indicators and the rectification of the plan; A preset threshold for the quality of edge data preprocessing is set. When the integrity and accuracy of core feature data do not meet the threshold, the edge node is triggered to send a feedback reminder to the corresponding subject. The supplementary data is then uploaded to the central node after secondary preprocessing. Add a feedback data validity verification clause to the smart contract, compare the preprocessed core feature data with the urban physical examination standard database, automatically mark invalid feedback data and record the corresponding subject information, and invalid feedback data will not be included in the optimization basis of the indicator weight model. The core feature data is cross-validated through an edge-center collaborative verification mechanism, and the verified data is synchronously written to the blockchain and CIM database.

5. The urban renewal method based on AI-based urban health assessment according to claim 1, characterized in that, After generating a set of candidate solutions that balances single-scenario compliance with cross-scenario collaboration, the method also includes: A spatiotemporal attention mechanism is introduced to dynamically weight the modified digital twin simulation parameters, strengthen the influence weight of core indicators in different time dimensions and spatial regions on the simulation results, and improve the spatiotemporal adaptation accuracy of the simulation results. Once the final implementation plan is determined, a dynamic correlation model between the implementation progress of the plan and the achievement of the physical examination indicators is constructed. Implementation data is collected according to preset time nodes, and the deviation value of the indicator achievement rate is calculated in real time. When the deviation value exceeds the preset range, the plan fine-tuning process is automatically triggered. Based on feedback data of cross-scenario collaboration effects, optimize the cross-scenario collaboration constraint model conditions and the weight allocation logic of the multi-objective optimization algorithm, and simultaneously update the benefit distribution and liability determination clauses related to cross-scenario collaboration in smart contracts. A linkage and iteration mechanism is established between the indicator weight model and the cross-scenario collaborative constraint model. The result of indicator weight adjustment corrects the cross-scenario collaborative constraint conditions in reverse, and the synchronization of cross-scenario collaborative effect data serves as the core basis for the optimization of the indicator weight model.

6. An urban renewal system based on AI-driven urban health assessment, characterized in that, include: The dynamic indicator system and traceability dataset generation module constructs a dynamic health check indicator system based on reinforcement learning algorithms. At the data access end, it synchronously analyzes remote sensing time-series data, IoT real-time data, multi-entity contribution data, and historical health check cases to extract core attributes and generate a unified semantic representation. The core attributes include indicator dynamic weights, area characteristic coefficients, entity contribution scores, and hidden danger warning levels. A dynamic mapping model corresponding to indicators, areas, and entities is built through graph neural networks. The model is then transmitted to the CIM database through a distributed synchronization mechanism. Simultaneously, blockchain nodes record the entity's rights and responsibilities trajectory and indicator iteration logs, forming a traceable dataset with dynamic tags. The indicator conflict identification and weight reconciliation module, based on the traceable dataset and multi-subject quantitative consensus rules, identifies indicator conflicts and disagreements through an AI model, dynamically generates corresponding weight-oriented reconciliation strategies, and performs hierarchical processing by edge nodes. It initiates an expedited consensus calibration process for high-priority security risk-related data and performs a multi-subject weighted fusion process for routine quality improvement indicator data. After the digital twin modeling and incremental update module is reconciled, in the edge-center collaborative architecture, the cache partition is divided according to the health check level and update priority, and the digital twin engine is called to build a coupled model corresponding to the health check index, update plan and simulation effect. A cross-scenario collaborative constraint model is constructed to extract common indicators and differentiated requirements for three types of scenarios: renovation of old residential areas, protection of historical blocks, and upgrading of industrial parks. Constraints on resource allocation and functional adaptation between scenarios and rules on indicator priority are set. In the candidate update scheme generation stage, a multi-objective optimization algorithm is introduced, with the achievement rate of physical examination indicators, resource consumption cost, and scenario collaborative adaptation as the core optimization objectives, to generate a set of candidate schemes that take into account both single scenario achievement and cross-scenario collaboration. A scenario conflict early warning module has been added to identify cross-scenario resource and functional conflicts in the candidate solution set in advance, and output the conflict type, scope of impact and resolution suggestions as a reference for simulation. Cross-scenario collaborative constraints are embedded into the coupled model simulation parameter system corresponding to physical examination indicators, update schemes, and simulation effects, and the threshold values ​​of simulation parameters for each scenario are corrected simultaneously. The coupled model of physical examination indicators, update plan and simulation effect includes a scenario-based simulation module. For three typical urban renewal scenarios, namely old community renovation, historical block protection and industrial park upgrading, corresponding simulation parameter thresholds are preset. The simulation parameter thresholds for different scenarios are set according to the key requirements of urban physical examination for the corresponding scenarios. Real-time acquisition of field data transmitted by IoT sensing devices; comparison of field data with preset simulation parameter thresholds; dynamic correction of simulation parameters of the coupled model corresponding to physical examination indicators, update schemes, and simulation effects; based on the corrected simulation parameters, simulation of multiple candidate update schemes is performed, generating a multi-scheme simulation comparison report including the achievement of physical examination indicators, implementation cycle, and resource consumption of each scheme. The data in the simulation comparison report are analyzed by reinforcement learning algorithm to select the update plan with the highest adaptability to the dynamic physical examination indicators and determine it as the final execution plan. By combining reinforcement learning to optimize simulation parameters, incremental synchronous updates are performed only on the parts related to index iteration and scheme adjustment. The scheme verification and model iteration optimization module uses an AI engine to transform physical examination standards and regional requirements into structured verification rules, completing the adaptation verification of updated schemes and dynamic indicators; based on multi-subject feedback and implementation data recorded on the blockchain, it iteratively optimizes the indicator weight model and reconciliation strategy.

7. An electronic device, characterized in that, Includes a processor, wherein the processor runs a program for an urban renewal method based on AI-based urban health assessment as described in any one of claims 1-5.

8. A storage medium, characterized in that, The program stores the urban renewal method based on AI-based urban health assessment as described in any one of claims 1-5.

Images