A block ventilation design management and control and multi-objective decision system for urban renewal projects

Abstract

The application discloses a kind of urban renewal project's street ventilation design management and control and multi-objective decision system, the present application relates to wisdom city planning and building environment digitization design technical field, including: data acquisition and twin construction module, for obtaining and fusing multiple-source city data, constructs and includes the digital twin of street three-dimensional form parameter and wind environment probability model;Ventilation efficiency measure and threshold management module.The urban renewal project's street ventilation design management and control and multi-objective decision system, support man-machine cooperation's parameterized scheme generation and multiple rounds, multi-objective optimization, it is changed into a kind of closed-loop intelligent decision-making mode based on quantitative analysis, automatic feedback iteration, from the planning mode of traditional experience judgment, each link disconnection, to effectively co-ordinate ventilation, sunshine, development intensity and so on multiple targets in scheme design initial stage, improve the scientificity and comprehensiveness of scheme, reduce environmental risk in later implementation.

Description

Technical Field

[0001] This invention relates to the field of smart city planning and digital design technology of built environment, specifically a street ventilation design control and multi-objective decision-making system for urban renewal projects. Background Technology

[0002] Currently, regarding the design and management of street ventilation in urban renewal projects, among the existing technologies, the invention patent "Intelligent Planning and Evaluation System for Urban Renewal Space Based on Digital Twin" (authorized publication number CN120688928A) provides a closed-loop system integrating data acquisition, intelligent planning, simulation, evaluation, and monitoring feedback. This system utilizes digital twin technology to dynamically reflect urban conditions and employs artificial intelligence to generate candidate solutions for multi-dimensional simulation and evaluation, effectively improving the scientific rigor and dynamic adaptability of urban renewal. However, this system primarily focuses on macro- or meso-scale urban renewal spatial planning, emphasizing a comprehensive balance of multiple objectives such as spatial efficiency, environmental quality, and social equity, without specifically addressing in-depth mechanistic integration and specialized design control for three-dimensional ventilation performance at the street scale. Specifically, its limitations in street ventilation are as follows: It lacks an efficient predictive model for the three-dimensional wind field of the street, making it difficult to quickly assess ventilation efficiency under different morphological schemes; it lacks a mechanism for deriving safety thresholds and discretionary ranges for street ventilation, failing to translate minimum ventilation performance requirements into specific morphological control index flexibility ranges; in the scheme generation and optimization stages, it does not specifically consider the linkage optimization of ventilation performance with other planning control indicators, such as floor area ratio and density, and the weight and achievement path of ventilation objectives in multi-objective decision-making are unclear; although its simulation module covers environmental dimensions, it typically relies on general models, and the simulation accuracy and efficiency for key ventilation details such as the internal wind field of the street and the wind pressure on the building surface may be insufficient, making it difficult to support the ventilation gain design of individual buildings. Therefore, there is an urgent need for a system specifically designed for street ventilation design control and multi-objective decision-making in urban renewal projects to overcome the shortcomings of the aforementioned existing technologies. Summary of the Invention

[0003] The purpose of this invention is to provide a street ventilation design control and multi-objective decision-making system for urban renewal projects, in order to solve the problems mentioned in the background art.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a street ventilation design control and multi-objective decision-making system for urban renewal projects, comprising:

[0005] The data acquisition and digital twin construction module is used to acquire and fuse multi-source urban data to construct a digital twin that includes three-dimensional morphological parameters of the blocks and a wind environment probability model.

[0006] The ventilation efficiency measurement and threshold management module is connected to the data acquisition and twin construction module. It is used to manage the ventilation efficiency measurement index set and dynamically set or adjust the ventilation performance safety threshold of the block as a whole and individual buildings based on the digital twin and standard data.

[0007] The ventilation performance prediction and control indicator discretion range derivation module is connected to the ventilation efficiency measurement and threshold management module. It is used to run an integrated prediction model based on the digital twin and the safety threshold to quickly evaluate ventilation performance and reverse derive the flexible discretion range of urban planning control indicators that meet ventilation requirements.

[0008] The human-machine collaborative solution generation and multi-objective optimization decision module is connected to the ventilation performance prediction and control planning index discretion interval derivation module. It is used to receive the flexible discretion interval, support designers to input space prototypes, and automatically derive detailed solutions that meet the constraints. Then, the solution pool is screened and iteratively optimized through a multi-objective optimization algorithm to output the Pareto optimal solution set.

[0009] The control rule feedback and system self-learning module is connected to the human-machine collaborative solution generation and multi-objective optimization decision-making module. It is used to analyze the final selected solution, generate design control details suggestions, and send the project's entire process data back to the system knowledge base to update the integrated prediction model and threshold setting logic.

[0010] Furthermore, the data acquisition and twin construction module is specifically used for:

[0011] Automatically extract multiple morphological parameters such as building outline, height, density, floor area ratio, street opening ratio, and building surface complexity from integrated air-ground mapping data;

[0012] Based on long-term meteorological monitoring data, a wind environment probability model was constructed that includes the probability of different wind direction and wind speed combinations.

[0013] Socially perceived data was used to cross-validate the initially constructed static 3D model, and inconsistencies in the data were identified and marked.

[0014] Furthermore, the ventilation efficiency measurement and threshold management module is specifically used for:

[0015] Construct a two-layer index library containing basic ventilation indicators and system-derived combined indicators, wherein the combined indicators include indicators reflecting the uniformity of ventilation distribution;

[0016] A machine learning model was used to establish a mapping relationship between street morphology parameter combinations, basic ventilation index values, and environmental performance targets.

[0017] If most of the proposed solutions fail to reach the initial safety threshold during the simulation process, a threshold sensitivity analysis will be initiated, and an adjustable safety threshold range will be recommended in conjunction with local climate adaptability objectives.

[0018] Furthermore, the ventilation performance prediction and control indicator discretion interval derivation module includes a prediction unit, a verification unit, and an interval generation unit.

[0019] The prediction unit is used to run at least two ventilation performance prediction models built on different principles, including a fast prediction model based on machine learning training and an auxiliary prediction model based on a simplified ventilation mechanism, to make parallel predictions under the same planning conditions.

[0020] The verification unit is connected to the prediction unit and is used to compare the prediction results of the fast prediction model and the auxiliary prediction model. If the difference exceeds the preset tolerance, the verification process based on computational fluid dynamics simulation is triggered, and the reliability of the model is determined or the prediction results are weighted and fused according to the verification results.

[0021] The interval generation unit is connected to the verification unit. It is used to solve all possible parameter combinations of urban planning control indicators that meet ventilation requirements by using a multi-objective inverse optimization algorithm, based on the consistency of model prediction results or reliable fusion, with ventilation safety threshold as the constraint condition, thereby automatically generating a multi-dimensional indicator flexible discretion interval.

[0022] The interval generation unit is also connected to a rule reasoning engine, which automatically compares and filters the generated discretionary interval with local planning regulations, sunlight and fire protection distance constraints, and outputs the final discretionary interval with joint constraints of ventilation and compliance.

[0023] Furthermore, the fast prediction model is a lightweight neural network model trained on a broad-spectrum street wind field database, with the input being rasterized or graph-structured street morphology data; the auxiliary prediction model is a simplified mechanism model based on fuzzy interval estimation or Bayesian inference, with the input being a subset of key morphological parameters.

[0024] Furthermore, the human-machine collaborative solution generation and multi-objective optimization decision-making module is specifically used for:

[0025] Receive the spatial prototype drawn by the designer based on the final discretionary range;

[0026] Using constraint satisfaction algorithms and Monte Carlo tree search methods, the spatial prototype is automatically derived within the final discretion interval to generate multiple refined morphological schemes that satisfy all given constraints, forming an initial scheme pool.

[0027] A multi-objective genetic algorithm is used to perform the first round of optimization on the initial scheme pool. The optimization objectives include at least maximizing the overall ventilation efficiency of the block, maximizing the total sunshine duration, and optimizing the economic volume within the allowable range, to obtain the first generation Pareto front scheme set.

[0028] For the schemes in the first generation Pareto frontier scheme set, the ventilation performance is recalculated by calling the ventilation performance prediction and control index discretion interval derivation module, and verified by using a rapid solar radiation simulation tool.

[0029] Analyze the recalculation and verification results to identify schemes with uneven ventilation distribution or localized sunlight defects, and automatically generate specific morphological adjustment suggestions;

[0030] Based on the aforementioned morphological adjustment recommendations, a second round of optimization was initiated. From the final Pareto scheme set that emerged victorious after optimization, several representative schemes were selected for high-fidelity computational fluid dynamics and solar radiation coupling simulation, serving as the authoritative verification basis for the final performance.

[0031] Furthermore, the proposed morphological adjustments include adjusting the height of specific buildings, adding or widening ventilation corridors between specific buildings, and modifying the location or size of openings in the building layout.

[0032] Furthermore, the control rule feedback and system self-learning module is specifically used for:

[0033] Analyze the correspondence between the actual morphological parameters of the final selected scheme and the final discretionary range, and summarize the parameter combination rules for successful implementation;

[0034] Based on the aforementioned parameter combination rules, generate or optimize design guidelines for this block.

[0035] The digital twin, final solution, high-fidelity simulation results, and decision logic data of this project will be stored as new samples in the system knowledge base for incremental training and parameter updates of the machine learning model in the ventilation efficiency measurement and threshold management module and the prediction model in the ventilation performance prediction and regulatory index discretion interval derivation module.

[0036] Furthermore, the multi-source urban data also includes historical urban planning documents, real-time data from IoT sensors, and traffic flow data.

[0037] Furthermore, the various modules of the system exchange data and schedule processes through application programming interfaces, forming a complete closed-loop workflow from data input, analysis and prediction, scheme generation and optimization to rule feedback and self-learning.

[0038] This invention provides a street block ventilation design control and multi-objective decision-making system for urban renewal projects. It has the following beneficial effects:

[0039] The urban renewal project's street ventilation design control and multi-objective decision-making system deeply integrates the scientific assessment of street ventilation performance with urban planning control indicators by constructing an integrated data-driven workflow. It utilizes multi-source data to build a dynamically updated digital twin and introduces parallel prediction models and procedural verification mechanisms to ensure the reliability of ventilation assessment results. Based on this, the system can automatically generate flexible control zones that integrate ventilation safety baselines with multiple regulatory constraints, and supports human-machine collaborative parameterized scheme generation and multi-round, multi-objective optimization. This process transforms the traditional planning model, which relies on experience-based judgment and is disconnected between different stages, into a closed-loop intelligent decision-making model based on quantitative analysis and automatic feedback iteration. This effectively coordinates multiple objectives such as ventilation, sunlight, and development intensity from the initial design stage, improving the scientific rigor and comprehensiveness of the plan and reducing environmental risks during later implementation.

[0040] The urban renewal project's street ventilation design control and multi-objective decision-making system establishes a positive feedback loop of continuous evolution from project practice to system capabilities. It not only outputs optimized solutions and specific morphological control guidelines for individual renewal projects, but also adaptively learns and updates its internal predictive models and evaluation thresholds by accumulating data throughout the entire project process. This self-improving mechanism allows the system's decision support capabilities to continuously accumulate and evolve with the development of more projects, providing increasingly precise and localized technical support for urban renewal projects of different regions and types. Ultimately, the system provides planning management departments and design teams with a set of operable and sustainable intelligent tools, helping to implement refined low-carbon ecological design requirements at the street scale and improve the overall physical environmental quality and sustainable development resilience of the urban built environment. Attached Figure Description

[0041] Figure 1 This is a module data flow diagram of a street ventilation design control and multi-objective decision-making system for urban renewal projects according to the present invention;

[0042] Figure 2 This invention presents a decision-making process state diagram for a multi-objective decision-making system for street ventilation design control in urban renewal projects. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Please see Figure 1 and Figure 2 This invention provides a technical solution: a street ventilation design control and multi-objective decision-making system for urban renewal projects, comprising:

[0045] The data acquisition and digital twin construction module is used to acquire and fuse multi-source urban data to construct a digital twin that includes three-dimensional morphological parameters of the blocks and a wind environment probability model.

[0046] The ventilation efficiency measurement and threshold management module connects to the data acquisition and digital twin construction module. It is used to manage the set of ventilation efficiency measurement indicators and dynamically set or adjust the ventilation performance safety thresholds of the entire block and individual buildings based on the digital twin and standard data.

[0047] The ventilation performance prediction and planning indicator discretion range derivation module is connected to the ventilation efficiency measurement and threshold management module. It is used to run an integrated prediction model based on digital twins and safety thresholds to quickly evaluate ventilation performance and reverse derive the flexible discretion range of urban planning control indicators that meet ventilation requirements.

[0048] The human-machine collaborative solution generation and multi-objective optimization decision-making module is connected to the ventilation performance prediction and control planning index discretion interval derivation module. It is used to receive flexible discretion intervals, support designers to input space prototypes, and automatically generate detailed solutions that meet the constraints. Then, the solution pool is screened and iteratively optimized through multi-objective optimization algorithms to output the Pareto optimal solution set.

[0049] The control rule feedback and system self-learning module connects to the human-machine collaborative solution generation and multi-objective optimization decision-making module. It is used to analyze the final selected solution, generate design control details suggestions, and send the project's entire process data back to the system knowledge base for updating the integrated prediction model and threshold setting logic.

[0050] It should be further explained that the implementation of this system relies on the collaborative work of five interrelated functional modules, forming a complete closed loop from data perception to decision feedback. First, the data acquisition and twin construction module is responsible for integrating multi-source heterogeneous data from surveying and mapping, meteorological monitoring, the Internet of Things, and social sensing platforms. It automatically extracts multiple morphological parameters from surveying and mapping data, including building outlines, height, density, floor area ratio, street opening ratio, and building surface complexity. Unlike the conventional approach of using a single dominant wind direction, this module constructs a wind environment probability model based on long-term meteorological data. This model describes the probability of different wind direction and speed combinations, providing input conditions that better reflect the randomness of reality for subsequent analysis.

[0051] Meanwhile, this module utilizes socially sensed data, such as pedestrian traffic heat maps in specific areas, to cross-validate the initially generated static 3D model, identifying and marking potential data conflict areas, thus providing a layer of assurance for data quality. Secondly, the ventilation efficiency measurement and threshold management module establishes and manages a two-layer ventilation index library, which not only includes basic indicators such as building surface wind pressure difference and street air exchange rate, but also systematically derives combined indicators such as ventilation uniformity index. This module not only references fixed thresholds in the standards but also establishes complex mapping relationships between morphological parameter combinations, basic ventilation index values, and environmental goals such as human comfort through training machine learning models. This supports dynamic assessment and rationalization recommendations of safety thresholds based on data analysis in specific projects. The core functionality is implemented by the ventilation performance prediction and planning indicator discretion range derivation module.

[0052] This module embeds at least two parallel prediction models: a lightweight neural network model trained on a large amount of historical wind field simulation data for rapid evaluation; and a simplified mechanism model based on fuzzy interval estimation or Bayesian inference. The two models simultaneously predict the same planning conditions, and their outputs are compared by a built-in verification unit. If the difference in results exceeds a preset tolerance, a rapid verification process based on computational fluid dynamics is automatically triggered to determine the model's reliability or guide the fusion of results. Based on reliable predictions, the module employs a multi-objective inverse optimization algorithm, using ventilation safety thresholds as constraints, to solve for all possible combinations of morphological parameters that meet the requirements, thereby automatically generating multi-dimensional flexible discretionary intervals for urban planning control indicators such as building density and height distribution. The generated intervals are further filtered by a rule-based inference engine to ensure compatibility with hard constraints in local regulations such as sunlight and fire protection, ultimately outputting a "ventilation-compliance" joint discretionary interval.

[0053] The rule-based reasoning engine interfaces with local planning regulations through structured storage. The mandatory requirements in local planning regulations, such as sunlight standards, fire safety distances, and building setbacks, are broken down into structured data items and stored in the regulations database in the format of "indicator type - constraint condition - value range". The comparison logic between the discretionary interval and the regulations is an indicator-by-indicator comparison. First, the value range of each regulatory indicator in the flexible discretionary interval is extracted, and then matched with the constraint conditions of the corresponding indicator in the regulations database. If some parameter combinations within the interval violate the regulations, a conflict filtering mechanism is triggered, removing parameter segments that violate the regulations and retaining the parameter range that meets both ventilation requirements and legal constraints. This ultimately forms the final discretionary interval with joint constraints of ventilation and compliance, ensuring that the output discretionary interval can be directly used in design without additional manual verification.

[0054] The core logic of the multi-objective inverse optimization algorithm is based on the ventilation safety threshold as the core constraint. It focuses on solving the feasible parameter combinations of urban planning control indicators. The objective function is designed around "meeting ventilation performance requirements" while maximizing the flexibility of the indicators. The constraints clearly include the legal value boundaries and technical feasibility ranges of each planning indicator. In the solution process, key planning indicators such as plot ratio, building density, and height distribution are first identified as decision variables. Then, based on the basic street data provided by the digital twin, initial parameter combinations that meet the ventilation safety threshold are selected. Through iterative elimination of conflicting combinations and supplementation of marginal feasible combinations, a multi-dimensional flexible discretionary range covering all effective parameter combinations is finally formed, ensuring that the parameter combinations in each range can meet the ventilation baseline requirements.

[0055] Then, the human-machine collaborative scheme generation and multi-objective optimization decision-making module begins operation. Designers can input initial spatial form concepts based on the aforementioned joint discretionary range within the parametric design environment provided by this module. The system then automatically generates a large number of specific form schemes that satisfy all constraints using constraint satisfaction algorithms and Monte Carlo tree search, forming an initial scheme pool. Next, the system uses a multi-objective genetic algorithm to perform the first round of optimization on this scheme pool. The objective function typically covers the overall ventilation performance of the block, total sunshine duration, and economic volume. The first-generation Pareto scheme set obtained from the optimization is immediately subjected to ventilation performance recalculation and rapid sunshine verification.

[0056] The system analyzes these results, automatically identifies schemes with defects such as severely uneven ventilation distribution or insufficient local sunlight, and generates specific morphological adjustment suggestions such as "adjusting the height of a building" or "adding ventilation corridors of a specific width." Based on these suggestions, the system initiates a second round of optimization to generate an improved final Pareto scheme set, and extracts representative schemes from it for high-fidelity coupled simulation as authoritative confirmation of the final performance.

[0057] Finally, the control rule feedback and system self-learning module analyzes the final selected scheme, summarizes the correspondence between its actual morphological parameters and the initial discretionary range, and generates specific clauses that can be incorporated into the design guidelines. Simultaneously, the entire process data, including the project's digital twin, the final scheme, simulation results, and decision-making logic, is encapsulated into new knowledge samples and fed back to the system's central knowledge base. These samples will be used for regular incremental training and parameter updates of the aforementioned ventilation prediction model and threshold management model, enabling the system to continuously improve its prediction accuracy and decision-making adaptability when dealing with new urban renewal projects. The entire system achieves data flow and process scheduling between modules through application programming interfaces, ensuring fully automated and intelligent operation from data input to intelligent feedback.

[0058] The data acquisition and twin building module is specifically used for:

[0059] Automatically extract multiple morphological parameters such as building outline, height, density, floor area ratio, street opening ratio, and building surface complexity from integrated air-ground mapping data;

[0060] Based on long-term meteorological monitoring data, a wind environment probability model was constructed that includes the probability of different wind direction and wind speed combinations.

[0061] Socially perceived data was used to cross-validate the initially constructed static 3D model, and inconsistencies in the data were identified and marked.

[0062] It should be further explained that the data acquisition and digital twin construction module constructs a high-quality urban digital twin by executing a series of specific processing procedures. This module first connects to an integrated air-ground mapping data source, using image recognition and geometric analysis algorithms to automatically extract several key morphological parameters from the mapping data. These parameters include, but are not limited to, the building's outer contour polygon, roof elevation, base area, building density, the ratio of total building area to land area of ​​the block, the location and width of openings in the block's outer walls, and a complexity index calculated based on the tortuosity of the building surface. Regarding wind environment characterization, this module calls upon a historical meteorological monitoring database to statistically analyze years of wind direction and speed observation records for the target area. Using a probability distribution function fitting method, it constructs a multi-dimensional joint probability model. This model can output the probability of specific wind conditions, such as "east wind, wind speed of three meters per second," occurring in a specific season, thus replacing the traditional assumption of a single dominant wind direction.

[0063] To improve the accuracy of the initial twin model, this module also integrates social sensing data streams, such as mobile phone signaling location density or shared bicycle parking heatmaps within a specific time period. Through spatial overlay analysis, the distribution of human activity intensity reflected by the sensing data is compared with the initially generated static 3D model. If areas marked as open spaces in the model but with consistently low actual human activity are found, or points marked as buildings but with abnormally active signals are found, the system will automatically mark these spatial units as areas to be verified and generate a data inconsistency report, prompting manual review or triggering a task to collect higher-precision data. This cross-validation step provides consistent data for subsequent analysis.

[0064] The ventilation efficiency measurement and threshold management module is specifically used for:

[0065] Construct a two-layer indicator library that includes basic ventilation indicators and system-derived combined indicators. The combined indicators include those that reflect the uniformity of ventilation distribution.

[0066] A machine learning model was used to establish a mapping relationship between street morphology parameter combinations, basic ventilation index values, and environmental performance targets.

[0067] If most of the proposed solutions fail to reach the initial safety threshold during the simulation process, a threshold sensitivity analysis will be initiated, and an adjustable safety threshold range will be recommended in conjunction with local climate adaptability objectives.

[0068] It should be further explained that the ventilation efficiency measurement and threshold management module achieves refined performance target control by constructing a two-layer indicator library and a dynamic threshold management mechanism. The indicator library built into this module first integrates basic ventilation efficiency indicators such as wind pressure difference at building surfaces, street air exchange rate, and average wind speed at pedestrian height. Based on this, the system automatically generates a series of derived combined indicators through algorithms, such as the "ventilation uniformity index." This index is calculated by first obtaining simulated wind pressure difference values ​​at the surfaces of all representative buildings in the street, and then calculating the coefficient of variation of these values ​​to quantify the degree of uniformity of ventilation performance in spatial distribution.

[0069] The quantitative assessment of the ventilation uniformity index involves selecting representative buildings from different areas within the block, including core functional areas, peripheral areas, and areas surrounding open spaces. For each area, 3-5 buildings with typical forms are selected as samples, and wind pressure difference data are collected at key locations on the surface of each sample building. The uniformity of ventilation distribution is judged based on the dispersion of this data; the lower the dispersion, the better the uniformity. Weighting is determined according to the functional attributes of the block, with representative buildings in core functional areas, such as residential and office areas, having a higher weight than those in peripheral areas. The weight of areas surrounding open spaces is appropriately adjusted based on pedestrian activity to ensure that the assessment results reflect differences in ventilation experience under actual usage scenarios.

[0070] In terms of safety threshold management, this module employs machine learning methods, specifically the gradient boosting tree algorithm, using a large number of historical computational fluid dynamics simulation cases as the training set. The input features for each case include combinations of morphological parameters such as floor area ratio, building density, average height, and windward density. The output labels are the corresponding basic ventilation index values ​​and whether specific environmental performance targets are met, such as a binary label for human thermal comfort based on standards. Through training, this module establishes a predictive mapping relationship from morphological parameters to ventilation performance and the probability of meeting targets.

[0071] In practical project applications, when a batch of generated solutions is pre-evaluated based on the initial specification thresholds and it is found that the proportion of compliant solutions is insufficient, this module will automatically initiate threshold sensitivity analysis: that is, under the current design constraints, the system simulates gradually relaxing or tightening the ventilation safety thresholds, observes the change curve of the compliance rate of the solutions, and combines it with the climate characteristic data of the project location, such as the number of high-temperature days in summer and the frequency of calm winds, to comprehensively calculate and provide a set of threshold adjustment range suggestions that can both ensure basic environmental quality and have realistic accessibility. This suggestion is used as a key decision reference input into the subsequent optimization process.

[0072] In threshold sensitivity analysis, the step size for threshold adjustment is set based on the initial safety threshold, typically 5%-10% of the initial threshold. The threshold is gradually loosened or tightened, and the corresponding proportion of compliant schemes is calculated. The compliance rate is calculated as "number of schemes meeting the ventilation safety threshold / total number of schemes." By plotting the compliance rate as a function of the threshold, the threshold critical point where the compliance rate changes significantly is identified. The quantitative indicators for climate adaptability targets refer to meteorological statistics of the project location over the past 10 years, including the number of days with a maximum daily temperature ≥35℃ in summer (June-August) and the frequency of calm winds (wind speed <1m / s). If the number of high-temperature days in summer exceeds 90 days or the frequency of calm winds is higher than 30%, the adjustable range of the safety threshold is appropriately expanded during threshold adjustment to balance environmental quality and scheme feasibility.

[0073] The ventilation performance prediction and control plan indicator discretion interval derivation module includes a prediction unit, a verification unit, and an interval generation unit.

[0074] The prediction unit is used to run at least two ventilation performance prediction models built on different principles, including a fast prediction model trained based on machine learning and an auxiliary prediction model built based on a simplified ventilation mechanism, to make predictions in parallel under the same planning conditions.

[0075] The verification unit is connected to the prediction unit and is used to compare the prediction results of the fast prediction model and the auxiliary prediction model. If the difference exceeds the preset tolerance, the verification process based on computational fluid dynamics simulation is triggered. The reliability of the model is determined or the prediction results are weighted and fused based on the verification results.

[0076] The interval generation unit connects to the verification unit. Based on the consistency of model prediction results or reliable fusion, it uses the ventilation safety threshold as a constraint and employs a multi-objective inverse optimization algorithm to solve all possible parameter combinations of urban planning control indicators that meet ventilation requirements, thereby automatically generating multi-dimensional indicator flexible discretion intervals.

[0077] The interval generation unit is also connected to a rule reasoning engine, which automatically compares and filters the generated discretionary intervals with local planning regulations, sunlight and fire safety distance constraints, and outputs the final discretionary interval with joint constraints of ventilation and compliance.

[0078] It should be further explained that the ventilation performance prediction and control indicator discretion interval derivation module, through its closely collaborative prediction unit, verification unit, and interval generation unit, achieves reliable assessment of ventilation performance and intelligent generation of control indicators. The prediction unit runs two models with different construction principles in parallel: one is a convolutional neural network model trained on a broad-spectrum street three-dimensional wind field database. This model takes the street's top-view two-dimensional morphological raster data and building height information as input, and directly outputs the predicted values ​​of key ventilation indicators through multi-layer nonlinear transformation; the other is a simplified mechanism model based on fuzzy interval estimation theory. This model selects only a limited number of key parameters such as plot ratio, building density, average height, and windward area ratio under the prevailing wind direction as input, and calculates and outputs the estimated interval of ventilation performance through predefined fuzzy rules and linear relationships.

[0079] The verification unit receives the parallel outputs of the two models in real time and calculates the absolute or relative difference between their predicted values. If the difference exceeds the preset tolerance value, the verification unit automatically calls a verified computational fluid dynamics standard verification process with an appropriate grid size to quickly simulate the typical morphology of the current block and uses the simulation results as a benchmark to determine which model's prediction is more reliable. Alternatively, the verification unit can dynamically assign weights and fuse the outputs of the two models based on the verification results to obtain a more reliable final prediction value that has been verified through a program.

[0080] Based on this reliable prediction, the interval generation unit sets the ventilation safety threshold as an insurmountable hard constraint. It adopts a multi-objective inverse optimization algorithm, such as the inverse solution mode of the non-dominated sorting genetic algorithm, to search for the set of all parameter combinations that can satisfy the ventilation constraint in the parameter space composed of various urban planning control indicators, such as building density, height limit, green space ratio, etc.

[0081] The boundaries of this set define the flexible discretionary ranges for each indicator. To ensure the generated ranges are directly implementable, the range generation unit is connected to a rule-based reasoning engine that embeds local technical regulations. This engine automatically compares and performs logical AND operations with mandatory specifications such as daylighting standards, building setbacks, and fire lane widths, eliminating all range segments with compliance conflicts. Finally, it outputs a "ventilation-compliance" joint discretionary range that integrates ventilation performance requirements and various statutory planning constraints and can be directly used for design guidance.

[0082] The fast prediction model is a lightweight neural network model trained on a broad-spectrum street wind field database, with input being rasterized or graph-structured street morphology data; the auxiliary prediction model is a simplified mechanism model based on fuzzy interval estimation or Bayesian inference, with input being a subset of key morphological parameters.

[0083] It should be further explained that in the prediction unit, the fast prediction model trained based on machine learning is specifically implemented as a lightweight convolutional neural network. The input layer of this network receives preprocessed street morphology raster data, which divides the two-dimensional plane of the street into regular grid cells. Each cell contains two channels of information: building coverage and the average height of buildings within that grid. The network structure includes several convolutional layers for extracting spatial features, and subsequent fully connected layers map these features into predicted values ​​for one or more target ventilation efficiency indicators. The model uses a training set consisting of a large number of historical computational fluid dynamics simulation cases for supervised learning. Each case includes its corresponding morphological raster data and validated simulation result labels. The network weights are optimized through a backpropagation algorithm, ultimately achieving a fast mapping capability from morphological features to ventilation performance.

[0084] Meanwhile, the auxiliary prediction model based on a simplified ventilation mechanism employs a fuzzy interval estimation method. The model's input consists of a finite number of key morphological parameters extracted from the current street's digital twin, such as floor area ratio, building footprint area ratio, average height, and windward width. The model internally uses a pre-defined rule base, containing rules such as "if the floor area ratio is within a certain range and the building density is high, then the membership function of ventilation efficiency is a specific triangular distribution." Through a fuzzy inference engine, the precise input parameters are converted into fuzzy quantities, relevant rules are activated, and after defuzzification, a predicted interval or expected value for ventilation efficiency is output. The model's parameters, such as the condition intervals and conclusion distributions of the rules, can be calibrated using historical data. Its calculation process does not rely on large-scale matrix operations, thus achieving rapid estimation based on physical experience, parallel to the data-driven model.

[0085] The lightweight neural network model's network structure includes 3-5 convolutional layers and 2-3 fully connected layers. The convolutional layers are used to extract spatial features from the street morphology raster data, while the fully connected layers are responsible for mapping the features to predicted ventilation index values. The activation function is the ReLU function. The input is two-dimensional planar rasterized data of the street, with a grid size of 1 meter × 1 meter and the average building height data within the corresponding grid. The output is key ventilation indicators such as the wind pressure difference on the building surface and the air exchange rate of the street. During training, the mean squared error is used as the loss function, and the Adam optimizer is used. The number of iterations is set to 500-1000. Training stops when the loss function value tends to stabilize or reaches the preset number of iterations.

[0086] The broad-spectrum street wind field database draws data from computational fluid dynamics simulations of street wind fields in different climate zones and morphological types (high-density streets, low-density streets, mixed-use streets, etc.), with a sample size of no less than 1,000 sets. It covers the parameter range of common streets with building heights of 10-100 meters, floor area ratios of 0.5-3.0, and building densities of 20%-60%. During data preprocessing, outliers are first removed, and then morphological parameters are normalized to ensure that the data format is consistent and compatible with the model input.

[0087] In the simplified mechanism model of fuzzy interval estimation, the fuzzy rule base is constructed based on ventilation mechanisms known to those skilled in the art and a large amount of experimental data. For example, mapping rules such as "when the floor area ratio is between 1.0 and 2.0 and the building density is between 30% and 40%, the ventilation efficiency is likely to be in the medium range" and "if the windward area ratio under the prevailing wind is greater than 0.6, the ventilation efficiency is likely to be at a low level" are used. The membership function adopts a triangular distribution, and the distribution range is set according to the correlation strength between morphological parameters and ventilation performance. For example, the peak point of the membership function of floor area ratio corresponding to ventilation efficiency is set at around 1.5.

[0088] In the simplified mechanism model of Bayesian inference, the prior distribution is set based on the correlation data of morphological parameters and ventilation performance of similar historical blocks. The plot ratio, building density, average height, and windward area ratio are selected as key input parameters. The prior probability distribution of each parameter is determined by statistical analysis of historical data. During the inference process, the posterior probability is dynamically adjusted in combination with the specific basic data of the current block, and finally the estimated range of ventilation performance is output.

[0089] The human-machine collaborative solution generation and multi-objective optimization decision-making module is specifically used for:

[0090] Receive the spatial prototype drawn by the designer based on the final discretionary range;

[0091] Using constraint satisfaction algorithms and Monte Carlo tree search methods, the spatial prototype is automatically derived within the final discretion interval to generate multiple refined morphological schemes that satisfy all given constraints, forming an initial scheme pool.

[0092] A multi-objective genetic algorithm was used to optimize the initial pool of schemes in the first round. The optimization objectives included at least maximizing the overall ventilation efficiency of the block, maximizing the total sunshine duration, and optimizing the economic volume within the allowable range, thus obtaining the first generation Pareto front scheme set.

[0093] For the schemes in the first generation Pareto frontier scheme set, the ventilation performance prediction and control index discretion interval derivation module is called to recalculate the ventilation performance, and the rapid solar radiation simulation tool is used for verification.

[0094] Analyze the recalculation and verification results to identify schemes with uneven ventilation distribution or localized sunlight defects, and automatically generate specific morphological adjustment suggestions;

[0095] Based on the morphological adjustment recommendations, a second round of optimization was initiated. From the final Pareto scheme set that emerged victorious after optimization, several representative schemes were selected for high-fidelity computational fluid dynamics and solar radiation coupling simulations, serving as authoritative verification of the final performance.

[0096] It should be further explained that the specific implementation process of the human-machine collaborative solution generation and multi-objective optimization decision-making module is as follows: First, the module provides a parametric design interface, allowing urban planners or designers to load the target block base and input initial spatial morphology concepts, such as outlining the general layout of the building complex or specifying the location of the main open spaces to form several spatial prototypes. After receiving these prototypes, the system parses them into a series of adjustable design parameters (such as the width, depth, height, orientation angle, and setback distance of each building), and then calls the built-in constraint satisfaction algorithm.

[0097] Based on the "ventilation-compliance" joint discretionary range provided by the aforementioned modules and basic specifications such as the "Technical Regulations for Construction Project Planning Management," the algorithm sets a reasonable value range for each design parameter. On this basis, it uses the Monte Carlo tree search method to perform heuristic exploration in a vast parameter combination space, automatically generating hundreds or thousands of detailed three-dimensional morphological schemes that differ in parameter values ​​but all strictly satisfy all preset constraints, thus constructing an initial scheme pool.

[0098] The Monte Carlo tree search method sets the search depth to match the number of regulatory indicators, with each parameter dimension corresponding to a search depth level. The expansion strategy prioritizes unexplored parameter combinations, sorting explored combinations by "constraint satisfaction + number of unexplored branches," and prioritizing the expansion of paths with higher potential. Pruning conditions include parameter combinations that clearly violate the legal boundaries of regulatory indicators, have ventilation performance far below the lower limit of the safety threshold, or conflict with sunlight or fire protection constraints. Pruning reduces invalid searches, improves the generation efficiency of the initial scheme pool, and ensures that each refined morphological scheme in the scheme pool satisfies all given constraints.

[0099] Next, the module uses the non-dominated sorting genetic algorithm NSGA-II to perform the first round of multi-objective optimization on the scheme pool. The optimization objective function is typically set to maximize the predicted value of the overall ventilation efficiency of the block, maximize the cumulative effective sunshine duration of all residential or office buildings, and approach the optimal value of development benefits within the floor area ratio discretion range. This round of optimization generates a first-generation Pareto front scheme set.

[0100] The population size of the multi-objective genetic algorithm is set to 100-200 schemes, with a crossover probability ranging from 0.6 to 0.8 and a mutation probability ranging from 0.05 to 0.15. The iteration termination condition is set as follows: after 20-30 consecutive generations of optimization, the Pareto front scheme set shows no significant change, or the number of iterations reaches 300 generations. The weight setting logic of the objective function is dynamically adjusted according to project needs and local regulations. If the project is located in a hot climate region, the weight of the overall ventilation efficiency of the block can be appropriately increased; if the project involves development benefit balance requirements, the weight of the economic volume can be fine-tuned within the allowable range to ensure that each optimization objective is both mutually accommodating and meets the priority requirements of the actual application scenario.

[0101] Subsequently, the system automatically recalculates ventilation indicators for each scheme in the set using a ventilation performance prediction model, and simultaneously verifies solar compliance using a rapid solar radiation analysis tool based on a radiative transfer model. By analyzing these calculation results, the system identifies localized areas where overall ventilation performance is significantly lower than the average level of the neighborhood, or individual buildings where cumulative solar radiation duration does not meet regulatory requirements. Based on this, the system automatically generates specific and actionable morphological adjustment suggestions, such as "reduce the height of building B5 by a certain number of meters," "create an east-west ventilation corridor with a width not less than a specific value between buildings A2 and A3," and "adjust the planar torsion angle of building C1." These suggestions are then transformed into new constraints or optimization target weight adjustments, driving the system to initiate a second round of multi-objective optimization.

[0102] The second round of optimization searches the solution space incorporating the new constraints, producing an improved final Pareto solution set. To ensure the reliability of the decision-making basis, the system selects several high-performing solutions from the final set and performs high-fidelity computational fluid dynamics and building solar radiation coupled numerical simulations with fine meshes and complete physical processes. The simulation results are then used as the final authoritative basis for evaluating the environmental performance of the solutions, and are output along with all optimization process data for decision-makers to comprehensively compare and select the best solution.

[0103] In high-fidelity computational fluid dynamics coupled simulation with solar radiation, the mesh generation standard is determined based on the building size and flow field gradient. A denser mesh (no larger than 0.5 meters) is used on the building surface and surrounding areas with drastic flow field changes, while a sparser mesh (5-10 meters) is used in areas far from the building to ensure a balance between flow field simulation accuracy and computational efficiency. The standard k-ε model is selected for the turbulence model, which is suitable for street-scale wind field simulation scenarios. The time step for solar radiation simulation is set in hours, covering the local effective solar radiation period, typically 8:00-18:00. Combining the latitude, longitude, and seasonal characteristics of the project location, the solar radiation duration and irradiance of the building surface and site are calculated within each time step, achieving collaborative simulation and verification of wind field and solar radiation.

[0104] Form adjustment suggestions include adjusting the height of specific buildings, adding or widening ventilation corridors between specific buildings, and modifying the location or size of openings in the building layout. It should be further noted that the form adjustment suggestions generated by this module are based on in-depth analysis of the scheme's performance data, specifically manifested in several operation types that can directly guide design modifications. For adjustments to individual buildings, the system will identify buildings whose wind pressure values ​​are consistently lower than the block average based on the specific values ​​of the wind pressure difference on the building surface in the ventilation recalculation results, and accordingly suggest a specific reduction in the roof height or standard floor height of the building, such as "adjusting the target building height from fifty meters to forty-five meters."

[0105] For optimizing the airflow within the block, the system analyzes the simulated wind speed flow field cloud map, identifies local areas with severe airflow obstruction, and precisely suggests creating or widening a linear open space with a clear direction and specific width between two or more specific building units as a ventilation corridor. For example, "Add an east-west oriented corridor with a width of not less than ten meters between the north facade of building A and the south facade of building B."

[0106] For adjustments involving building layout and openings, the system analyzes the permeability of the windward side of the building complex based on the prevailing wind direction data in the wind environment probability model. If it finds that the windward side is excessively blocked by continuous solid interfaces, it will suggest adjusting the plan position or rotation angle of specific buildings to form effective air inlets, or modifying the boundary shape of the ground floor open area to guide airflow penetration. These suggestions are all output in the form of text descriptions with specific spatial positioning and quantitative parameters, and can be directly mapped to the corresponding parameters in the parametric design model, serving as precise input conditions for the next round of automated solution generation and optimization.

[0107] The control rule feedback and system self-learning module is specifically used for:

[0108] Analyze the correspondence between the actual morphological parameters of the final selected scheme and the final discretionary range, and summarize the parameter combination rules for successful implementation;

[0109] Based on the rules of parameter combination, generate or optimize design guidelines for this block.

[0110] The digital twin, final solution, high-fidelity simulation results, and decision logic data of this project will be stored as new samples in the system knowledge base for incremental training and parameter updates of the machine learning model in the ventilation efficiency measurement and threshold management module and the prediction model in the ventilation performance prediction and regulatory index discretion interval derivation module.

[0111] It should be further explained that the control rule feedback and system self-learning module performs specific data extraction and knowledge update processes during the project closure phase. This module first analyzes the implementation plan selected in the final decision, maps and compares its detailed morphological parameter set with the previously generated "ventilation-compliance" joint discretionary range, and uses statistical induction methods to identify high-frequency parameter combination patterns that are stably associated with excellent ventilation performance. For example, "when the building density is in the middle of the range and combined with a specific range of opening ratio, the ventilation uniformity index performs better."

[0112] Based on these summarized patterns, the module automatically transforms them into clause-based design control recommendations that can guide similar projects in the future. For example, the guidelines specify that "in the prevailing wind direction, it is recommended that the ratio of the total width of the openings of continuous building interfaces to the total length of the interfaces should not be less than a certain specific value."

[0113] Simultaneously, the module structures and encapsulates the complete data package of the current project, including but not limited to the initial state of the input digital twin, the prediction and simulation data throughout the process, the multi-round optimization scheme set, the finally selected scheme model, and all performance evaluation reports, into a standard-format knowledge sample. These knowledge samples are transmitted and stored in the system's central knowledge base.

[0114] This knowledge base connects the machine learning models in the ventilation efficiency measurement and threshold management module with the rapid prediction models in the ventilation performance prediction module. The system is equipped with timed or triggered model update tasks. When a certain number of new samples are accumulated or a preset time node is reached, these new samples are used to incrementally train or fine-tune the parameters of the existing model. This allows the model to continuously evolve and improve its prediction accuracy and adaptability to local planning conditions when facing new neighborhood renewal projects, completing a full cycle from single project application to system capability iteration.

[0115] Multi-source urban data also includes historical urban planning documents, real-time data from IoT sensors, and traffic flow data.

[0116] It should be further explained that, in the system's data acquisition and digital twin construction module, the acquired and integrated multi-source urban data, in addition to basic surveying and mapping and meteorological information, also includes other key data categories to enhance the integrity of the digital twin and the accuracy of its real-world mapping. Historical urban planning documents, especially approved control detailed planning drawings and texts, are extracted using document parsing and spatial registration technologies to reveal original land use, floor area ratio limits, building density, green space ratio, and other control indicators. This information serves as the background layer for understanding the planning history and initial design conditions of the area.

[0117] The real-time data from IoT sensors comes from various environmental monitoring devices deployed in the neighborhood, such as miniature sensors for temperature, humidity, wind speed and direction, and air quality installed on lampposts or building facades. This data is continuously accessed through IoT protocols and dynamically associated with the corresponding three-dimensional spatial location in the digital twin based on its geographical coordinates, providing the model with high spatiotemporal resolution feedback on the real environmental status.

[0118] Traffic flow data, obtained by accessing checkpoint vehicle records, floating car trajectories, or bus card swipe data from the city's traffic management platform, is anonymized and aggregated to analyze traffic flow, speed, and direction patterns on roads surrounding the block at different times. This dynamic traffic pattern information helps infer the intensity and spatiotemporal distribution of human activities around the block and complements and cross-validates social perception data such as mobile phone signaling. Together, they are used to assess and calibrate the functional vitality of the block and its potential impact on the local wind and heat environment, thus providing multi-dimensional dynamic data support for building a digital twin of the city that is closer to the actual operating state.

[0119] The various modules of the system exchange data and schedule processes through application programming interfaces, forming a complete closed-loop workflow from data input, analysis and prediction, scheme generation and optimization to rule feedback and self-learning.

[0120] It should be further explained that the various functional modules of the system do not operate in isolation. Instead, they exchange data and schedule processes in an orderly manner through a set of predefined application programming interface specifications, thus forming an automated and intelligent closed-loop workflow. The system is designed with a central workflow engine, which coordinates the startup, execution, and data transfer of each module according to the preset business logic sequence.

[0121] For example, once the data acquisition and twin construction module completes the construction and verification of the digital twin of the target block, it will automatically call the input interface of the ventilation efficiency measurement and threshold management module through its output interface according to the agreed data format (such as a data package containing a set of morphological parameters and a wind environment probability model) to trigger the threshold analysis task.

[0122] Data exchange between modules is based on a unified structured data contract, such as using JSON or Protocol Buffers format to encapsulate input parameters and output results, ensuring accurate transmission and parsing of data semantics.

[0123] The workflow engine monitors the execution status of each module. After the ventilation performance prediction module completes parallel prediction and verification, it automatically packages the verified prediction results together with the safety threshold and passes them to the interval generation unit of the control planning indicator discretion interval derivation module through interface calls.

[0124] Similarly, the generated joint discretionary range serves as a key input parameter, activating the solution derivation and optimization process of the human-machine collaborative solution generation and multi-objective optimization decision-making module via API calls. Throughout the multi-round optimization and decision-making process, intermediate solutions, performance evaluation results, and adjustment suggestions are transferred in real time between modules via API.

[0125] Once the final solution is selected, the workflow engine will drive the control rule feedback and system self-learning modules to perform knowledge extraction tasks, and inject the generated new knowledge samples into the system's central knowledge storage through the data feedback interface, thereby providing input for possible subsequent incremental updates of the model.

[0126] This architecture, based on standardized interfaces and central scheduling, ensures a seamless, automatically driven closed loop from raw data input to final control rule output and system self-update, effectively reducing manual intervention and improving the overall system's processing efficiency and reliability.

[0127] This system deeply integrates the scientific assessment of street ventilation performance with urban planning and control indicators by constructing an integrated data-driven workflow. It utilizes multi-source data to build a dynamically updated digital twin and introduces parallel prediction models and procedural verification mechanisms to ensure the reliability of ventilation assessment results. Based on this, the system can automatically generate flexible control zones that integrate ventilation safety baselines with multiple regulatory constraints, and supports human-machine collaborative parameterized scheme generation and multi-round, multi-objective optimization. This process transforms the traditional planning model, which relies on experience-based judgment and involves disconnected processes, into a closed-loop intelligent decision-making model based on quantitative analysis and automatic feedback iteration. This effectively coordinates multiple objectives such as ventilation, sunlight, and development intensity from the initial design stage, improving the scientific rigor and comprehensiveness of the plan and reducing environmental risks during later implementation.

[0128] This system establishes a positive feedback loop from project practice to continuous evolution of system capabilities. It not only outputs optimized solutions and specific morphological control guidelines for individual urban renewal projects, but also adaptively learns and updates its internal predictive models and evaluation thresholds by accumulating data throughout the entire project process. This self-improving mechanism allows the system's decision support capabilities to continuously accumulate and evolve with the development of more projects, providing increasingly precise and localized technical support for urban renewal projects of different regions and types. Ultimately, the system provides planning management departments and design teams with a set of operable and sustainable intelligent tools, helping to implement refined low-carbon ecological design requirements at the block scale and improve the overall physical environmental quality and sustainable development resilience of the urban built environment.

[0129] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0130] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A street block ventilation design control and multi-objective decision-making system for urban renewal projects, characterized in that, include: The data acquisition and digital twin construction module is used to acquire and fuse multi-source urban data to construct a digital twin that includes three-dimensional morphological parameters of the blocks and a wind environment probability model. The ventilation efficiency measurement and threshold management module is connected to the data acquisition and twin construction module. It is used to manage the ventilation efficiency measurement index set and dynamically set or adjust the ventilation performance safety threshold of the block as a whole and individual buildings based on the digital twin and standard data. The ventilation performance prediction and control indicator discretion range derivation module is connected to the ventilation efficiency measurement and threshold management module. It is used to run an integrated prediction model based on the digital twin and the safety threshold to quickly evaluate ventilation performance and reverse derive the flexible discretion range of urban planning control indicators that meet ventilation requirements. The human-machine collaborative solution generation and multi-objective optimization decision module is connected to the ventilation performance prediction and control planning index discretion interval derivation module. It is used to receive the flexible discretion interval, support designers to input space prototypes, and automatically derive detailed solutions that meet the constraints. Then, the solution pool is screened and iteratively optimized through a multi-objective optimization algorithm to output the Pareto optimal solution set. The control rule feedback and system self-learning module is connected to the human-machine collaborative solution generation and multi-objective optimization decision-making module. It is used to analyze the final selected solution, generate design control details suggestions, and send the project's entire process data back to the system knowledge base to update the integrated prediction model and threshold setting logic.

2. The street ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 1, characterized in that: The data acquisition and twin construction module is specifically used for: Automatically extract multiple morphological parameters such as building outline, height, density, floor area ratio, street opening ratio, and building surface complexity from integrated air-ground mapping data; Based on long-term meteorological monitoring data, a wind environment probability model was constructed that includes the probability of different wind direction and wind speed combinations. Socially perceived data was used to cross-validate the initially constructed static 3D model, and inconsistencies in the data were identified and marked.

3. The street ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 2, characterized in that: The ventilation efficiency measurement and threshold management module is specifically used for: Construct a two-layer index library containing basic ventilation indicators and system-derived combined indicators, wherein the combined indicators include indicators reflecting the uniformity of ventilation distribution; A machine learning model was used to establish a mapping relationship between street morphology parameter combinations, basic ventilation index values, and environmental performance targets. If most of the proposed solutions fail to reach the initial safety threshold during the simulation process, a threshold sensitivity analysis will be initiated, and an adjustable safety threshold range will be recommended in conjunction with local climate adaptability objectives.

4. The street ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 3, characterized in that: The ventilation performance prediction and control indicator discretion interval derivation module includes a prediction unit, a verification unit, and an interval generation unit. The prediction unit is used to run at least two ventilation performance prediction models built on different principles, including a fast prediction model based on machine learning training and an auxiliary prediction model based on a simplified ventilation mechanism, to make parallel predictions under the same planning conditions. The verification unit is connected to the prediction unit and is used to compare the prediction results of the fast prediction model and the auxiliary prediction model. If the difference exceeds the preset tolerance, the verification process based on computational fluid dynamics simulation is triggered, and the reliability of the model is determined or the prediction results are weighted and fused according to the verification results. The interval generation unit is connected to the verification unit. It is used to solve all possible parameter combinations of urban planning control indicators that meet ventilation requirements by using a multi-objective inverse optimization algorithm, based on the consistency of model prediction results or reliable fusion, with ventilation safety threshold as the constraint condition, thereby automatically generating a multi-dimensional indicator flexible discretion interval. The interval generation unit is also connected to a rule reasoning engine, which automatically compares and filters the generated discretionary interval with local planning regulations, sunlight and fire protection distance constraints, and outputs the final discretionary interval with joint constraints of ventilation and compliance.

5. The street ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 4, characterized in that: The rapid prediction model is a lightweight neural network model trained on a broad-spectrum street wind field database, with the input being rasterized or graph-structured street morphology data; the auxiliary prediction model is a simplified mechanism model based on fuzzy interval estimation or Bayesian inference, with the input being a subset of key morphological parameters.

6. The street ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 5, characterized in that: The human-machine collaboration scheme generation and multi-objective optimization decision-making module is specifically used for: Receive the spatial prototype drawn by the designer based on the final discretionary range; Using constraint satisfaction algorithms and Monte Carlo tree search methods, the spatial prototype is automatically derived within the final discretion interval to generate multiple refined morphological schemes that satisfy all given constraints, forming an initial scheme pool. A multi-objective genetic algorithm is used to perform the first round of optimization on the initial scheme pool. The optimization objectives include at least maximizing the overall ventilation efficiency of the block, maximizing the total sunshine duration, and optimizing the economic volume within the allowable range, to obtain the first generation Pareto front scheme set. For the schemes in the first generation Pareto frontier scheme set, the ventilation performance prediction and control index discretion interval derivation module is called to recalculate the ventilation performance, and the rapid solar radiation simulation tool is used for verification. Analyze the recalculation and verification results to identify schemes with uneven ventilation distribution or localized sunlight defects, and automatically generate specific morphological adjustment suggestions; Based on the aforementioned morphological adjustment recommendations, a second round of optimization was initiated. From the final Pareto scheme set that emerged victorious after optimization, several representative schemes were selected for high-fidelity computational fluid dynamics and solar radiation coupling simulation, serving as the authoritative verification basis for the final performance.

7. A street block ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 6, characterized in that: The proposed morphological adjustments include adjusting the height of specific buildings, adding or widening ventilation corridors between specific buildings, and modifying the location or size of openings in the building layout.

8. A street block ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 7, characterized in that: The control rule feedback and system self-learning module is specifically used for: Analyze the correspondence between the actual morphological parameters of the final selected scheme and the final discretionary range, and summarize the parameter combination rules for successful implementation; Based on the aforementioned parameter combination rules, generate or optimize design guidelines for this block. The digital twin, final solution, high-fidelity simulation results, and decision logic data of this project will be stored as new samples in the system knowledge base for incremental training and parameter updates of the machine learning model in the ventilation efficiency measurement and threshold management module and the prediction model in the ventilation performance prediction and regulatory index discretion interval derivation module.

9. A street block ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 8, characterized in that: The multi-source urban data also includes historical urban planning documents, real-time data from IoT sensors, and traffic flow data.

10. A street block ventilation design control and multi-objective decision-making system for urban renewal projects according to claim 9, characterized in that: The various modules of the system exchange data and schedule processes through application programming interfaces, forming a complete closed-loop workflow from data input, analysis and prediction, scheme generation and optimization to rule feedback and self-learning.

Images