Urban form intelligent iterative optimization method for relieving heat island effect

Abstract

The invention discloses an urban form intelligent iterative optimization method for heat island effect alleviation. Comprising the following steps of urban multi-source big database construction, urban heat island effect and urban form characteristic measurement, urban form automatic iterative optimization rule setting, urban form multi-target iterative optimization based on reinforcement learning and automatic output and interactive display. According to the method, firstly, a target city multi-source big database is constructed through collected building vector data, road network data and remote sensing images, the average heat island intensity is measured under the optimal grid scale, and then action rules, reward rules and limiting conditions for form adjustment are set with each single building as a unit; and finally, multi-target iterative optimization of the urban form is carried out based on reinforcement learning. According to the method, urban form multi-target automatic optimization under the guidance of heat island effect relief can be considered, and digital and intelligent application conversion of urban planning work is promoted.

Description

Technical Field

[0001] This invention belongs to the field of urban design application and transformation technology, and in particular relates to an intelligent iterative optimization method for urban form aimed at mitigating the heat island effect. Background Technology

[0002] The urban heat island effect is a prevalent environmental problem in the process of urbanization. It is mainly manifested in the fact that urban areas are significantly hotter than surrounding rural areas, impacting residents' comfort and the sustainable development of cities. Mitigating the urban heat island effect can not only improve residents' living environment but also optimize urban spatial structure and enhance the city's ecological benefits. Research and analysis of the urban heat island effect can provide a clear understanding of the mechanisms by which urban morphology influences it, thus offering a scientific basis for urban planning and architectural design.

[0003] Mitigating the urban heat island effect can promote the sustainable development of the urban ecological environment. The heat island effect leads to increased urban temperatures, increases residents' energy consumption, and exacerbates air pollution and ecological degradation. Optimizing urban form design can effectively reduce the intensity of the heat island effect, improve the urban microclimate, reduce energy consumption and pollution emissions, and improve the quality of the urban environment. Furthermore, mitigating the heat island effect can improve residents' living comfort. A well-designed urban form can improve ventilation and the thermal environment within the city by optimizing building layout, adjusting building height and orientation, thereby enhancing residents' living experience.

[0004] Urban form optimization can also promote social equity and the rational allocation of resources. The heat island effect is typically more pronounced in densely populated urban areas with high building density, areas that are often also relatively resource-scarce. Scientific urban form optimization can reduce the intensity of the heat island effect in these areas, improve residents' living environment, and enhance social equity. Simultaneously, urban form optimization can achieve a balance between economic and environmental benefits. Through reasonable adjustments to building forms, the heat island effect can be mitigated while ensuring reasonable changes in urban floor area ratio, thus promoting sustainable urban economic development.

[0005] Currently, mitigating the urban heat island effect mainly relies on methods such as urban form optimization and ecological design. Traditional urban form optimization methods are typically single-objective oriented, neglecting the balance between economy, rationality, and comfort, making it difficult to achieve multi-objective optimization. Furthermore, traditional methods are highly dependent on data; data redundancy or inaccuracy can affect the reliability of optimization results. In addition, traditional methods lack automation and intelligent means, resulting in low optimization efficiency and failing to meet the needs of rapid modern urban development. Summary of the Invention

[0006] Purpose of the invention: The purpose of this invention is to provide an intelligent iterative optimization method for urban morphology aimed at mitigating the urban heat island effect, effectively alleviating the urban heat island effect, improving the quality of the urban ecological environment, optimizing spatial structure, and promoting sustainable urban development.

[0007] Technical solution: To achieve the above objectives, the intelligent iterative optimization method for urban morphology described in this invention includes the following steps: S1. Collect current building vector data, road network data, and remote sensing image data of the target city, unify data dimensions, eliminate redundant data, and construct a multi-source big data database for the target city. S2. Divide the target city area into uniformly distributed grids of equal size; calculate the surface temperature data of each grid using remote sensing image data and a single-window inversion algorithm; calculate the urban morphology index of each grid using building vector data; calculate the average temperature difference between each grid and the surrounding grids as the heat island effect intensity of each grid, and statistically analyze the average heat island intensity of all grids within the target city. S3. Define the action rules, reward rules, and constraints for urban form adjustment. The action rules are based on individual buildings, which can be stretched vertically in three dimensions, rotated in orientation, and have their major and minor axes adjusted in planar plane. In each iteration, the building's form parameters are randomly adjusted. The reward rule uses the reduction of the average heat island intensity across all grids in the target city as the reward. After each iteration, the surface temperature is simulated, and the average heat island intensity across all grids in the target city is recalculated. If the heat island intensity decreases, a positive reward is given; otherwise, a negative reward is given. The reward value is proportional to the change in heat island intensity. The reward value is fed back into the iterative optimization to guide the generation of the next round of building adjustment actions. The constraints for urban form adjustment are that the change in urban floor area ratio must not exceed a threshold, ensuring that building spacing meets sunlight requirements. If these requirements are not met, the adjustment action is rejected. S4. Determine the multi-objectives for iterative optimization of urban form, including economic, rational and comfort objectives. Use reinforcement learning algorithms to iteratively optimize urban form, simulate new surface temperature distribution based on the adjusted urban form, recalculate the average heat island intensity, adjust the parameters of iterative optimization based on the reward value, generate the next round of building adjustment actions, and stop iterative when the heat island effect mitigation does not exceed the set threshold for multiple consecutive iterations. S5. Establish a holographic projection sand table to interactively display the optimal urban form scheme with multiple objectives.

[0008] Optionally, step S2, which divides the target city area into uniformly distributed grids of equal size, specifically includes the following steps: First, import the boundary vector data of the target city according to a set format and ensure that it has been projected onto the UTM projection coordinate system; then, divide the target city area using different grid sizes, with grid sizes of 300m×300m, 400m×400m, and 500m×500m, and visualize the surface temperature data after dividing the area into different grids to observe whether the spatial distribution of the heat island effect is smooth or has obvious details, so as to determine the appropriate grid size.

[0009] Optionally, in step S2, the land surface temperature data of each grid is calculated using remote sensing image data and a single-window inversion algorithm, and the urban morphology index of each grid is calculated using building vector data. Specifically, this includes the following steps: First, the remote sensing image data is projected to ensure that the remote sensing image data and the target urban area are in the same coordinate system and spatially superimposed on the grid. The pixels within each grid are averaged to obtain the land surface temperature of the corresponding grid. Then, the urban morphology index corresponding to each grid is calculated, including building density. Average building height and green space coverage ; , , , in, It is represented as the base area of ​​each building within the grid. Represented as the area of ​​the grid; Represented as the height of each building within the grid. Represented as the number of buildings within the grid; It is represented as the area of ​​each green space within the grid.

[0010] Optionally, in step S2, the average temperature difference between each grid and its surrounding grids is calculated as the heat island effect intensity of each grid, and the average heat island intensity of all grids within the target city is statistically analyzed. Specifically, this includes the following steps: First, determining the adjacency relationship of each grid and extracting the surface temperature of each grid from its surrounding adjacent grids; then, calculating the average temperature difference between each grid and its surrounding grids. And calculate the average heat island intensity of all grids within the target city. ; , , in, This represents the average temperature difference of the current grid, which in turn represents the intensity of the urban heat island effect of the current grid. This is represented as the current surface temperature of the grid. It is represented as the surface temperature of the surrounding grid cells; The average heat island intensity of the target city is expressed as... This represents the total number of grid cells in the target city.

[0011] Optionally, the morphological parameters in step S3 include building height. Building orientation Major axis width With minor axis width In each iteration, a certain number of buildings are randomly selected, and their morphological parameters are randomly adjusted. The adjusted morphological parameters must meet the following constraints: , , , , in, This represents the adjusted building height. This represents the original building height. Represented as a height adjustment amount, randomly generated, with a value range that can be set to... ; This indicates the adjusted building orientation. This indicates the original building orientation. This indicates that the orientation is adjusted. Randomly generated, The range of values ​​can be set to ; , This is represented by the adjusted major and minor axis widths. , Represented as the original major and minor axis widths, , This is expressed as the adjustment amount for the major and minor axes. , Randomly generated, , The range of values ​​can be set to , .

[0012] Optionally, the reward rules for urban morphology adjustment in step S3 include the following steps: First, calculate the average heat island intensity of all grids within the target city. If the average heat island intensity decreases after the urban form is adjusted, that is... A positive reward is given if the heat island intensity increases, and a negative reward is given if the heat island intensity decreases. The reward value is directly proportional to the change in heat island intensity; if the heat island intensity decreases, a positive reward is given. Then the reward value If the intensity of the heat island increases, Then the reward value Finally, the reward value Feedback is fed into the iterative optimization process to guide the generation of the next round of building form adjustment actions; the optimization objective of the iterative optimization is to maximize the cumulative reward value. ,in , , in, Represented as reward value, Represented as a reward coefficient, A positive value is used to control the size of the reward; This is expressed as the change in average heat island intensity. Represented as the first The reward value of each iteration, Represented as the total number of iterations; The average heat island intensity of all grid cells within the target city after urban morphology adjustment. The average heat island intensity of all grids within the target city before urban morphology adjustment.

[0013] Optionally, the restriction conditions for urban form adjustment in step S3 include the following steps: calling the building database and setting the restriction conditions for urban form adjustment, then randomly adjusting the building form parameters according to the action rules for urban form adjustment, and checking whether the adjusted building form meets the restriction conditions for urban form adjustment, that is, the change limit of urban plot ratio does not exceed the threshold, and the building spacing meets the sunlight requirements. If not, the adjustment action is rejected. , , , in, Expressed as floor area ratio, Represented as the first The floor area of ​​each building, Represented as the first The height of the building. This is expressed as the total area of ​​the land parcel; Indicated as building spacing, Indicated as building height, It is expressed as the solar altitude angle, which is determined based on the local latitude and season; The plot ratio after the adjustment of the urban form; The plot ratio before the adjustment of the urban form; This represents the total number of buildings within the grid.

[0014] Optionally, the multi-objective determination of urban morphology iterative optimization in step S4 specifically includes the following steps: The economic objective is quantified by calculating the sum of the absolute values ​​of the changes in floor area ratio before and after optimization, expressed as: , in, The total number of buildings within the grid. and They represent the first The floor area ratio of a building before and after optimization; Quantified values ​​for economic objectives; The rationality objective is measured by the cosine of the angle between the normal vector of the main facade and the due south direction, expressed as: , in, For the first The azimuth angle of the main facade normal of a building, i.e., its orientation. This is the azimuth angle for due south, i.e., the optimal orientation; The quantified value of the reasonableness target; The comfort target is characterized by the magnitude of surface temperature decrease, calculated jointly using remote sensing inversion of surface temperature and shadow cover, as expressed in the following expression: , in, To optimize the surface temperature difference before and after, This represents the initial highest surface temperature. The area obscured by the building's projected image. The total area of ​​the site , For the weighting coefficients, take... =0.7, =0.3; Quantified values ​​for comfort targets; The reward function for reinforcement learning is designed as a multi-objective weighted sum: , in, This is a multi-objective weighted sum; the weight coefficients satisfy... The system sets constraints and outputs a Pareto front solution set in each iteration. Finally, the optimal solution is selected from the non-dominated solutions using the TOPSIS decision method.

[0015] Optionally, the constraints in step S4 are: the change in the plot ratio of a single building shall not exceed 50% of the original value; the building spacing and greening rate shall meet the standard requirements, and the greening rate shall not be less than 30%.

[0016] Optionally, in step S4, a reinforcement learning algorithm is used to iteratively optimize the urban morphology. The iteration stops when the degree of mitigation of the heat island effect does not exceed a set threshold for multiple consecutive iterations. Specifically, this includes the following steps: (1) Rule embedding in state space: The constraints are transformed into digital constraints and input into the state representation, including the following data: spatial constraint matrix: Boolean matrix of building spacing, setback distance and solar radiation coefficient; regulatory index vector: upper limit of plot ratio, lower limit of green space ratio and building density threshold; historical action trajectory: records the satisfaction of the constraints by the previous k steps; (2) Multi-objective reward hierarchical architecture: A three-level reward structure is designed to achieve objective synergy, expressed as: , in, For the total reward, For hard-constraint rewards, when all rules are satisfied Take 1, otherwise Set to 0; As an economic reward, ; This represents the change in floor area ratio. This is the initial floor area ratio; , For the orientation angle of each building, For optimal orientation; + , For comfort rewards, This represents the change in the General Thermal Climate Index. This represents the change in the sky view factor; , and All are weighting coefficients; (3) Multi-objective Pareto solution set selection: The constraint-type non-dominated sorting algorithm C-NSGA-II is adopted. The first sorting criterion is the degree of constraint violation ∑violation; the second sorting criterion is the non-dominated level of the objective function; the solution that satisfies violation=0 and is located on the Pareto front is retained; (4) Dynamic weight adjustment strategy: The target weights are automatically adjusted according to the optimization stage. , in For learning rate, Improve the rate of the objective function to achieve asymptotic optimization that balances multiple objectives. This represents the dynamic weight at time t; Initial weights; (5) Iterative optimization stopping strategy: When the overall heat island effect mitigation does not exceed the set threshold after multiple consecutive iterations, the iterative optimization process is stopped and the results are output.

[0017] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: (1) This invention can construct a target city database based on multi-source data, accurately measure the intensity of urban heat island effect and urban morphological characteristics, use reinforcement learning algorithm to realize multi-objective intelligent iterative optimization of urban morphology, and output the optimal urban morphology scheme through automated simulation and interactive display, thereby effectively alleviating urban heat island effect, improving urban ecological environment quality, optimizing spatial structure, and promoting sustainable urban development. (2) This invention overcomes the bottlenecks of data silos and single indicators in traditional heat island mitigation methods by constructing a multi-source big data database and a precise measurement system for cities; by unifying the dimensional processing of building vector data, road network data and remote sensing image data, it eliminates data redundancy and heterogeneity barriers, and achieves precise quantification of the intensity of urban heat island effect and urban morphological characteristics; at the same time, it improves the accuracy and stability of heat island effect measurement by using grid division and average temperature difference calculation, and provides a scientific basis for optimization strategies. (3) The present invention provides an intelligent iterative optimization framework for urban morphology based on reinforcement learning, which solves the problems of single-objective orientation and insufficient intelligence in traditional optimization methods. By setting building adjustment action rules, reward rules and constraints, the reinforcement learning algorithm drives multi-objective iterative optimization of urban morphology to achieve a balance between economy, rationality and comfort. The optimization process can automatically adjust building morphology parameters to continuously alleviate the heat island effect, while ensuring the rationality of changes in urban plot ratio and sunshine requirements. (4) This invention realizes the digital closed loop and interactive display of the optimization process. By simulating the change of the heat island effect through ENVI-met software, combined with holographic projection sand table and gesture recognition, the optimized urban form scheme is dynamically displayed. The system supports real-time segmentation and attribute query of the scheme, forming a complete intelligent closed loop from "measurement-optimization-display-iteration", which significantly improves the scientific nature and visualization capability of planning decisions and empowers intelligent decision support for urban ecological environment improvement and spatial structure optimization. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a diagram showing the division of urban morphology and heat island intensity units in this invention; Figure 3 This is a schematic diagram of the urban form individual unit adjustment rules in this invention; Figure 4 This is an automatically generated and interactive display diagram of the urban morphology iteration optimization results in this invention. Detailed Implementation

[0019] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0020] like Figure 1 As shown, the present invention provides an intelligent iterative optimization method for urban morphology aimed at mitigating the urban heat island effect, comprising the following steps: S1. Construction of a Multi-Source Urban Big Data Platform Collect current building vector data, road network data, and Landsat8 OLI_TIRS remote sensing image data of the target city, unify data units, eliminate redundant data, and construct a multi-source large database for the target city.

[0021] Building vector data includes total building area, floor area ratio, building density and standard deviation of building height; road network data includes road centerline and road width; Landsat8 OLI_TIRS remote sensing image data includes multispectral bands and spatial resolution, among which multi-source big data is shown in Table 1. Table 1

[0022] S2. Urban Heat Island Effect and Measurement of Urban Morphological Characteristics The target city area was divided into uniformly distributed 500m×500m grids of equal size using the fishing net tool in ArcGIS. The surface temperature data of each grid was calculated using Landsat8 OLI_TIRS remote sensing image data and a single-window inversion algorithm. The urban morphology index of each grid was calculated using building vector data. The average temperature difference between each grid and the eight surrounding grids was calculated as the heat island effect intensity of each grid, and the average heat island intensity of all grids in the target city was statistically analyzed.

[0023] Step S2 involves dividing the target city area into uniformly distributed 500m×500m grids of equal size. Specifically, this includes the following steps: First, import the boundary vector data of the target city in shapefile format and ensure that it has been projected onto the UTM projection coordinate system for distance measurement. Then, divide the target city area using different grid sizes: 300m×300m, 400m×400m, and 500m×500m. Visualize the surface temperature data after dividing the area into different grids to observe whether the spatial distribution of the heat island effect is smooth or has obvious details, in order to determine the appropriate grid size as 500m×500m.

[0024] Step S2 utilizes Landsat8 OLI_TIRS remote sensing image data and a single-window inversion algorithm to calculate the land surface temperature data for each grid cell, and uses building vector data to calculate the urban morphology indicators for each grid cell. Specifically, this includes the following steps: First, the Landsat8 OLI_TIRS remote sensing image data is projected to ensure that the remote sensing image data and the target urban area are in the same coordinate system and spatially overlaid with the grid cells. The pixels within each grid cell are averaged to obtain the land surface temperature for that grid cell. Then, ArcGIS is used to calculate the urban morphology indicators corresponding to each grid cell, including building density. Average building height and green space coverage ; , , , in, It is represented as the base area of ​​each building within the grid. Represented as the area of ​​the grid, for example ; Represented as the height of each building within the grid. Represented as the number of buildings within the grid; It is represented as the area of ​​each green space within the grid.

[0025] like Figure 2 As shown, step S2 calculates the average temperature difference between each grid and its eight surrounding grids as the intensity of the urban heat island effect for each grid, and statistically analyzes the average urban heat island intensity of all grids within the target city. Specifically, this includes the following steps: First, using ArcGIS spatial analysis tools, the adjacency relationships of each grid are determined, and the surface temperature of each grid's eight surrounding adjacent grids is extracted; then, the average temperature difference between each grid and its eight surrounding grids is calculated. And calculate the average heat island intensity of all grids within the target city. ; , , in, This represents the average temperature difference of the current grid, which in turn represents the intensity of the urban heat island effect of the current grid. This is represented as the current surface temperature of the grid. It is represented as the surface temperature of the eight adjacent grid cells around the Earth. The average heat island intensity of the target city is expressed as... This represents the total number of grid cells in the target city.

[0026] S3. Rules for Automatic Iterative Optimization of Urban Form The action rules, reward rules, and restrictions for adjusting the city form are set. The action rules for adjusting the city form are based on each individual building as a unit. Each individual building can be stretched vertically in three dimensions, rotated in orientation, and its major and minor axes can be adjusted in planar plane. In each iteration, the form parameters of the building are randomly adjusted. The reward rule for urban morphology adjustment is to use the reduction of the average heat island intensity of all grids within the target city as the reward. After each round of iterative optimization, the entire urban morphology model is input into ENVI-met software to simulate the surface temperature and remeasure the average heat island intensity of all grids within the target city. If the heat island intensity decreases, a positive reward is given; otherwise, a negative reward is given. The reward value is proportional to the change in heat island intensity. The reward value is fed back into the iterative optimization to guide the generation of the next round of building adjustment actions. The restrictions on urban form adjustment are that the change in urban plot ratio is limited to no more than 10%; and that the building spacing must meet the sunshine requirements based on building height and local solar altitude angle. If the restrictions on urban form adjustment are not met, the adjustment will be rejected.

[0027] like Figure 3 As shown, the morphological parameters in step S3 include building height. Building orientation Major axis width With minor axis width In each iteration, a certain number of buildings are randomly selected, for example, 10% or 20%, and their morphological parameters are randomly adjusted. The adjusted morphological parameters must meet the following constraints: , , , , in, This represents the adjusted building height. This represents the original building height. Represented as a height adjustment amount, randomly generated, with a value range that can be set to... ; This indicates the adjusted building orientation. This indicates the original building orientation. This indicates that the orientation is adjusted. Randomly generated, The range of values ​​can be set to ; , This is represented by the adjusted major and minor axis widths. , Represented as the original major and minor axis widths, , This is expressed as the adjustment amount for the major and minor axes. , Randomly generated, , The range of values ​​can be set to , .

[0028] Step S3 sets the reward rules for city morphology adjustment, specifically including the following steps: First, calculate the average heat island intensity of all grids within the target city. If the average heat island intensity decreases after the urban form is adjusted, that is... A positive reward is given if the heat island intensity increases, and a negative reward is given if the heat island intensity decreases. The reward value is directly proportional to the change in heat island intensity; if the heat island intensity decreases, a positive reward is given. Then the reward value If the intensity of the heat island increases, Then the reward value Finally, the reward value Feedback is fed into the iterative optimization process to guide the generation of the next round of building form adjustment actions; the optimization objective of the iterative optimization is to maximize the cumulative reward value. ,in , , in, Represented as reward value, Represented as a reward coefficient, A positive value is used to control the size of the reward; This is expressed as the change in average heat island intensity. Represented as the first The reward value of each iteration, Represented as the total number of iterations; The average heat island intensity of all grid cells within the target city after urban morphology adjustment. The average heat island intensity of all grids within the target city before urban morphology adjustment.

[0029] Step S3 sets the restrictions on urban form adjustment, specifically including the following steps: calling the building database and setting the restrictions on urban form adjustment, then randomly adjusting the building form parameters according to the action rules of urban form adjustment, and checking whether the adjusted building form meets the restrictions on urban form adjustment, that is, the change limit of urban plot ratio is no more than 10%, and the building spacing meets the sunlight requirements. If not, the adjustment action is rejected. , , , in, Expressed as floor area ratio, Represented as the first The floor area of ​​each building, Represented as the first The height of the building. This is expressed as the total area of ​​the land parcel; Indicated as building spacing, Indicated as building height, It is expressed as the solar altitude angle, which is determined based on the local latitude and season; The plot ratio after the adjustment of the urban form; The plot ratio before the adjustment of the urban form; This represents the total number of buildings within the grid.

[0030] Step S4: Multi-objective iterative optimization of urban morphology based on reinforcement learning The iterative optimization of urban morphology is based on multiple objectives, including economic, rational, and comfort objectives. The economic objective is the minimum change in floor area ratio; the rational objective is the optimal overall building orientation; and the comfort objective is the maximum mitigation of the urban heat island effect. A reinforcement learning algorithm is used to iteratively optimize the urban morphology. The adjusted urban morphology is input into ENVI-met software to simulate the new surface temperature distribution and recalculate the average heat island intensity. Based on the reward value, the parameters for iterative optimization are adjusted to generate the next round of building adjustments. This process is repeated until the reduction in heat island effect does not exceed 5% for 10 consecutive iterations, at which point the iteration stops.

[0031] Step S4 determines the multi-objectives of urban morphology iterative optimization, specifically including the following steps: The economic objective is quantified by calculating the sum of the absolute values ​​of the changes in floor area ratio before and after optimization, expressed as: , in, The total number of buildings within the grid. and They represent the first The floor area ratio of a building before and after optimization; Quantified values ​​for economic objectives; The rationality objective is measured by the cosine of the angle between the normal vector of the main facade and the due south direction, expressed as: , in, For the first The azimuth angle of the main facade normal of a building, i.e., its orientation. This is the azimuth angle for due south, i.e., the optimal orientation; The quantified value of the reasonableness target; The comfort target is characterized by the magnitude of surface temperature decrease, calculated jointly using remote sensing inversion of surface temperature and shadow cover, as expressed in the following expression: , in, To optimize the surface temperature difference before and after, This represents the initial highest surface temperature. The area obscured by the building's projected image. The total area of ​​the site , For the weighting coefficients, take... =0.7, =0.3; Quantified values ​​for comfort targets; The reward function for reinforcement learning is designed as a multi-objective weighted sum: , in, This is a multi-objective weighted sum; the weight coefficients satisfy... The following constraints are set: the plot ratio of a single building shall not change by more than 50% of the original value; the building spacing and greening rate shall meet the requirements of the "Urban Residential Area Planning and Design Standard" GB50180-2018, and the greening rate shall not be less than 30%; the optimization process is implemented through the Python + TensorFlow framework, and each iteration outputs the Pareto front solution set, and finally the optimal solution is selected from the non-dominated solutions through the TOPSIS decision method.

[0032] Step S4 uses a reinforcement learning algorithm to iteratively optimize the urban morphology. The iteration stops when the reduction in the heat island effect does not exceed 5% for 10 consecutive iterations. The specific steps include the following: (1) Rule embedding in state space: Convert the constraints into digital constraints and input them into the state representation, including the following data: spatial constraint matrix: Boolean matrix of building spacing, setback distance and sunshine coefficient; regulatory index vector: upper limit of plot ratio, lower limit of green space ratio and building density threshold; historical action trajectory: record the satisfaction of the constraints by the previous k steps.

[0033] (2) Multi-objective reward hierarchical architecture: A three-level reward structure is designed to achieve objective synergy, expressed as: , in, For the total reward, For hard-constraint rewards, when all rules are satisfied Take 1, otherwise Set to 0; As an economic reward, ; This represents the change in floor area ratio. This is the initial floor area ratio; , For the orientation angle of each building, For optimal orientation; + , For comfort rewards, This represents the change in the General Thermal Climate Index. This represents the change in the sky view factor; , and All are weighting coefficients.

[0034] (3) Multi-objective Pareto solution set selection: The constraint-type non-dominated sorting algorithm C-NSGA-II is adopted. The first sorting criterion is the degree of constraint violation ∑violation; the second sorting criterion is the non-dominated level of the objective function; and the solution that satisfies violation=0 and is located on the Pareto front is retained.

[0035] (4) Dynamic weight adjustment strategy: The target weights are automatically adjusted according to the optimization stage. , in For learning rate, Improve the rate of the objective function to achieve asymptotic optimization that balances multiple objectives. This represents the dynamic weight at time t; Initial weights.

[0036] (5) Iterative optimization stopping strategy: When the overall heat island effect mitigation does not exceed 5% after 10 consecutive iterations, the iterative optimization process is stopped and the results are output.

[0037] S5. Automatic Output and Interactive Display like Figure 4 As shown, a holographic projection sand table is built to interactively display the optimal urban form scheme with multiple objectives. The required equipment includes a three-dimensional physical sand table, a digital holographic projector, and a gesture recognition device.

Claims

1. A city form intelligent iterative optimization method for heat island effect mitigation, characterized in that, Comprise the following steps: S1. Collect the target city present situation building vector data, road network data, remote sensing image data, unify data dimension, clear redundant data, build target city multi-source database; S2. The target city area is divided into equal size and uniform distribution grid; using remote sensing image data and single window inversion algorithm to calculate the land surface temperature data of each grid; using building vector data to calculate the city form index of each grid; calculate the average temperature difference between each grid and the surrounding grid, as the heat island effect intensity of each grid, and statistics the average heat island intensity of all grids in the target city; S3. Set the action rule of city form adjustment, the reward rule of city form adjustment and the limit condition of city form adjustment, the action rule of city form adjustment is to take each single building as a unit, the single building can be stretched up and down in three-dimensional height, can be rotated in direction, can adjust the width of long axis and short axis in plane; in each iteration, the form parameters of the building are adjusted randomly; the reward rule of city form adjustment is to take the average heat island intensity relief of all grids in the target city as the reward, after each iteration optimization, the land surface temperature is simulated, and the average heat island intensity of all grids in the target city is recalculated; If the heat island intensity is reduced, positive reward is given, otherwise negative reward is given, the size of reward value is proportional to the change of heat island intensity; the reward value is fed back to the iterative optimization, guiding the generation of the next round of building adjustment action; the limit condition of city form adjustment is that the variation limit of city volume rate does not exceed the threshold value, and the building spacing meets the sunshine requirement, if not, the adjustment action is refused; S4. Determine the multi-objective of city form iterative optimization, the multi-objective includes economic target, rationality target and comfort target, use reinforcement learning algorithm to optimize the city form, simulate new land surface temperature distribution according to the adjusted city form, and recalculate the average heat island intensity, adjust the parameters of iterative optimization according to the reward value, generate the next round of building adjustment action, when the heat island effect relief amplitude does not exceed the set threshold value for continuous multiple iterations, the iteration stops; S5. Establish holographic projection sand table, interactive display the optimal city form scheme of multi-objective.

2. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The step S2 in the target city area is divided into equal size and uniform distribution grid, specifically comprising the following steps: first, import the boundary vector data of the target city according to the set format and ensure that it has been projected to UTM projection coordinate system; then use different grid size to divide the target city area, and visualize the land surface temperature data after different grid division, observe whether the spatial distribution of heat island effect is smooth or has obvious details, to determine the appropriate grid size.

3. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The step S2 utilizes remote sensing image data and a single-window inversion algorithm to calculate the ground temperature data of each grid, and utilizes building vector data to calculate the urban form index of each grid, specifically including the following steps: firstly, the remote sensing image data is projected to ensure that the remote sensing image data and the target urban area are located in the same coordinate system and are spatially overlapped with the grid, the pixels in each grid are averaged to obtain the ground temperature corresponding to the grid; then the urban form index corresponding to each grid is calculated, the urban form index includes building density , building height mean value , and green coverage ; ， ， ， wherein, represents the bottom area of each building within the grid, represents the area of the grid; represents the height of each building within the grid, represents the number of buildings within the grid; represents the area of each green space within the grid.

4. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The average temperature difference between each grid and the surrounding grids in the step S2 is calculated as the heat island intensity of each grid, and the average heat island intensity of all grids in the target city is counted, specifically including the following steps: first, the adjacency relationship of each grid is determined, and the ground surface temperature of the surrounding adjacent grids of each grid is extracted; then, the average temperature difference between each grid and its surrounding grids is calculated , and the average heat island intensity of all grids in the target city is calculated . ， ， wherein, represents the average temperature difference of the current grid, i.e., the heat island intensity of the current grid, represents the surface temperature of the current grid, represents the surface temperature of the four surrounding adjacent grids; represents the average heat island intensity of the target city, represents the total number of grids of the target city.

5. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The morphological parameters in the step S3 include building height , building orientation , long axis width , and short axis width In each round of iteration, a certain number of buildings are randomly selected, and the morphological parameters of the buildings are randomly adjusted. The adjusted morphological parameters of the buildings need to meet the restriction conditions: ， ， ， ， wherein, is represented as the adjusted building height, is represented as the original building height, is represented as the height adjustment amount, which is randomly generated, and the value range can be set as ; is represented as the adjusted building orientation, is represented as the original building orientation, is represented as the orientation adjustment angle, is randomly generated, the value range can be set as ; , is represented as the adjusted long axis and short axis width, , is represented as the original long axis and short axis width, , is represented as the adjustment amount of the long axis and short axis, , is randomly generated, , the value range can be set as , .

6. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The step S3 of setting the reward rule of the urban form adjustment specifically comprises the following steps: firstly, calculating the average heat island intensity of all grids in the target city If the average heat island intensity decreases after the urban form adjustment, i.e. , a positive reward is given; otherwise, a negative reward is given; the reward value is proportional to the change of the heat island intensity, if the heat island intensity decreases, , the reward value ; If the heat island intensity increases, then the reward value ; finally, the reward value is fed back to the iterative optimization for guiding the generation of the next round of building form adjustment actions; the optimization goal of the iterative optimization is to maximize the cumulative reward value , wherein ， ， wherein, is represented as a reward value, is represented as a reward coefficient, is a positive value for controlling the size of the reward; is represented as a change in average heat island intensity, is represented as the reward value of the th iteration, is represented as the total number of iterations; is the average heat island intensity of all grids in the target city after adjustment of the urban form, is the average heat island intensity of all grids in the target city before adjustment of the urban form.

7. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The step S3 in the limit condition of city form adjustment, specifically comprising the following steps: call the building database and set the limit condition of city form adjustment, then adjust the building form parameters randomly according to the action rule of city form adjustment, and check whether the adjusted building form meets the limit condition of city form adjustment, that is, the variation limit of city volume rate does not exceed the threshold value, the building spacing meets the sunshine requirement, if not, the adjustment action is refused; ， ， ， in, This is expressed as floor area ratio. Represented as the first The floor area of ​​each building, Represented as the first The height of the building. This is expressed as the total area of ​​the land parcel; Indicated as building spacing, Indicated as building height, It is expressed as the solar altitude angle, which is determined based on the local latitude and season; The plot ratio after adjustments to the urban form; The floor area ratio before the adjustment of the urban form; This represents the total number of buildings within the grid.

8. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The step S4 determines the multi-objective of the urban form iterative optimization, and specifically includes the following steps. The economy target is quantified by calculating the sum of absolute values of the volume rate change before and after optimization, and the expression is: ， wherein, is the total number of buildings in the grid, and respectively represent the volume rate of the first building before and after optimization; is the quantitative value of the economic target; The rationality target is measured by the cosine value of the angle between the normal vector of the main facade and the south direction, and the expression is: ， wherein, is the first is the normal azimuth of the main facade of the building, i.e. the orientation, is the positive south direction azimuth, i.e. the optimal orientation; is the quantification of the rationality target; The comfort target is represented by the surface temperature cooling amplitude, and is calculated by the remote sensing inversion surface temperature and shadow coverage rate, and the expression is: ， wherein, is the optimized difference between the front and back surface temperature, is the initial maximum surface temperature, is the building projected shading area, is the total area of the site, , is the weight coefficient, taken as = 0.7, = 0.3; is the quantified value of the comfort target; The reward function of the reinforcement learning is designed as a multi-objective weighted sum: ， wherein, is a multi-objective weighted sum value; the weight coefficients satisfy and a constraint condition is set; a Pareto front solution set is output in each round of iteration, and the optimal scheme is finally selected from the non-dominated solutions by a TOPSIS decision method.

9. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 8, characterized in that, The constraint condition in the step S4 is that the volume rate change of the single building should not exceed 50% of the original value, and the building spacing and green rate meet the standard requirements, and the green rate is not less than 30%.

10. The urban form intelligent iterative optimization method for heat island effect alleviation according to claim 1, characterized in that, The step S4 uses the reinforcement learning algorithm to iteratively optimize the urban form, and stops iteration when the heat island effect relief amplitude does not exceed the set threshold in continuous multiple iterations, and specifically includes the following steps: (1) Rule embedding state space: convert the limit conditions into digital constraint conditions and input the state representation, including the following data: spatial constraint matrix: Boolean matrix of building spacing, setback distance and sunshine coefficient; regulation index vector: upper limit of volume rate, lower limit of green rate and building density threshold; historical action trajectory: record the satisfaction of the limit conditions by the previous k steps of action; (2) Multi-objective reward layered architecture: design a three-level reward structure to realize the coordination of targets, and the expression is: ， wherein, is the total reward, is the hard constraint reward, when all rules are satisfied is 1, otherwise is 0; is the economy reward, ; is the change in the volume rate, is the initial volume rate; , is the orientation angle of each building, is the optimal orientation; , is the comfort reward, is the change in the general thermal climate index, is the change in the sky view factor; , and are weight coefficients; (3) Multi-objective Pareto solution set screening: use the constraint non-dominated sorting algorithm C-NSGA-II, the first sorting criterion: constraint violation ∑violation; the second sorting criterion: target function non-dominated level; reserve the solutions that meet violation=0 and are located on the Pareto frontier; (4) Dynamic weight adjustment strategy: automatically adjust the target weight according to the optimization stage: ， wherein is a learning rate, is a target function improvement rate, and the gradual optimization of multi-objective balance is realized. represents a dynamic weight at time t; initial weight (5) Iterative optimization stopping strategy: when the overall heat island effect relief amplitude does not exceed the set threshold after continuous multiple iterations, stop the iterative optimization process and output the result.

Images