Block green space structure intelligent optimization method and system

Abstract

The invention discloses an intelligent optimization method and system for a green space structure of a block, and relates to the technical field of urban planning, landscape gardens and artificial intelligence, and the method comprises the steps: obtaining a planning region, carrying out the unit division based on the planning region, obtaining a plurality of block units, obtaining the related index data of each block unit, and carrying out the calculation of the related index data; inputting the related index data of each block unit into a pre-established regression model, and outputting to obtain an optimized regression model; calculating variable importance for the optimized regression model, and performing weight correction and coupling degree calculation based on the variable importance to obtain a corrected weight and a block unit coupling degree; vacant land images are obtained, grid identification is carried out on the vacant land images, then superposition is carried out to obtain a vacant land expansion area, land grid vacancy probability prediction is carried out based on the optimized regression model, and a vacant land probability is obtained; generating a block initial database based on the vacant land use probability, generating an initial population for the block initial database based on a genetic algorithm, and determining a block green space structure optimization result through the fitness of an optimal individual in each generation of initial population based on a reinforcement learning algorithm; and diversified choices can be provided for urban green space planning.

Description

Technical Field

[0001] The present invention relates to the fields of urban planning, landscape gardening and artificial intelligence technology, and specifically to a method and system for intelligently optimizing a block green space structure. Background Art

[0002] In recent years, the phenomenon of urban shrinkage has spread across China and will become the norm in China's urban development in the future. In this process, the number of vacant urban land has increased, causing multiple problems in the ecological environment, social economy and other aspects. Green space has been proven to be one of the effective means to solve the problems caused by vacant urban land, and it has the functions of improving the level of ecosystem services, alleviating social problems and driving economic development.

[0003] At present, existing technologies have explored intelligent planning methods for green spaces or other similar planning elements (such as urban open spaces). However, this type of method lacks both the definition of the preconditions for green space planning and the evaluation of the coupling degree between the generated scheme and the built environment. Therefore, it is difficult to meet the planning requirements of limited spatial resources under stock planning. At the same time, the random method is used to layout green spaces in intelligent planning, which has high computing power cost and slow calculation speed. Summary of the invention

[0004] In order to solve the deficiencies mentioned in the above background technology, the purpose of the present invention is to provide a method and system for intelligent optimization of block green space structure.

[0005] In the first aspect, the purpose of the present invention can be achieved by the following technical solution: a method for intelligent optimization of block green space structure, the method comprising the following steps:

[0006] Obtaining a planning area, demarcating units based on the planning area to obtain a plurality of block units, obtaining relevant indicator data of each block unit, inputting the relevant indicator data of each block unit into a pre-established regression model, and outputting an optimized regression model;

[0007] The relevant indicator data of each block unit includes the corresponding target performance indicator and potential influencing factor indicator of each block unit, and the regression model is constructed based on the random forest regression algorithm;

[0008] Calculate the importance of variables for the optimized regression model, perform weight correction and coupling calculation based on the importance of variables, and obtain the corrected weight and block unit coupling; obtain vacant land images, perform grid recognition on the vacant land images, and then superimpose them to obtain the vacant land expansion area, and predict the vacancy probability of land grids based on the optimized regression model to obtain the vacant land probability;

[0009] The initial database of the block is generated based on the probability of vacant land, the initial population is generated from the initial database of the block based on the genetic algorithm, and the optimization result of the green space structure of the block is determined by the fitness of the best individual in each generation of the initial population based on the reinforcement learning algorithm.

[0010] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the process of obtaining relevant indicator data of each block unit:

[0011] Calculate the corresponding target performance indicators and potential influencing factor indicators for each block unit in each period, and subtract the indicators of the previous year from the indicators of the current year to obtain the target performance change value {ΔP} and the potential influencing factor indicator change value {ΔE}. For each block unit i, it includes {ΔP, ΔE}; perform correlation analysis on all the potential influencing factor indicator change values, eliminate the influencing factors with multicollinear relationships, and summarize the remaining indicators of all block units into the database I{ΔP, ΔE}.

[0012] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the formula of the random forest regression algorithm is as follows:

[0013]

[0014] Where E={E1,E2,…E n} is a feature vector containing n potential influencing index change values, M is the number of decision trees, T m (X) is the predicted value of the mth decision tree for the input feature vector X. Each decision tree T m (X) Process the input features according to its own tree structure and splitting rules to give the final prediction value;

[0015] The evaluation indicators of the regression model include mean square error (MSE), mean absolute error (RMSE) and R square (R 2 ; The formula is as follows:

[0016]

[0017] In the formula, y i is the actual value of the test set, is the value calculated based on the random forest regression model, is the average value of the test set, and n is the number of samples in the test set.

[0018] In combination with the first aspect, in some implementations of the first aspect, the method further includes: calculating the variable importance VIM of the optimized regression model e The calculation process:

[0019]

[0020] Where M is the number of decision trees of the output regression model M, T m is the set of all nodes of the mth decision tree, ΔMSE t,e is the reduction in impurity caused by feature e at node t;

[0021] The process of weight correction and coupling calculation based on variable importance:

[0022] For each target performance model, select the variables whose independent variable importance VIM is greater than the preset threshold, use the independent variable importance VIM as the initial weight W, and modify the initial weight W to obtain the modified weight W ′ As the coupling model weight, for each block unit i, the coupling degree D for target J ij The model formula is the sum of the products of variables greater than the threshold and the modified weights;

[0023]

[0024] Among them, D ij is the coupling degree of block unit i at target j, E im is the value of variable m in block unit i whose variable importance VIM is greater than the threshold, W′ m is the modified weight;

[0025] The weight correction method is:

[0026]

[0027] Among them, W m VIM is the importance of independent variables in the target performance model e The importance score VIM of variables m greater than the threshold e , n is the total number of variables m.

[0028] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the optimized regression model RF expression is as follows:

[0029]

[0030] Where X={X1,X2,…X n} is a feature vector containing n driving factors, M is the number of decision trees finally determined, and decision tree T m The depth of (X) and the number of node split samples are determined according to the training results.

[0031] In combination with the first aspect, in certain implementations of the first aspect, the method also includes: the vacant land is completely vacant land with no development purpose within the urban area, as well as land that is abandoned, vacant or underutilized for construction purposes, and the vacant land categories include bare land, natural grassland, natural shrubs, inefficient construction land, areas under construction and natural wetlands.

[0032] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a process of generating the initial block database is as follows:

[0033] Construct an M×M matrix D0 corresponding to the block grid division, where each element V ij represents the grid at the i-th row and j-th column in the block, where 1≤i≤M,1≤j≤M;

[0034]

[0035] Each element V ij All are expressed in the form of attribute vectors. For an element V containing n attributes ij = {V ij1 , V ij2 …V ijn};

[0036] V ij1 is the current land use type, V ij1 ∈{0,1,2,3…,9}, representing bare land, natural grassland, artificial grassland, natural shrub, forest land, residential land, commercial land, industrial land, hard square and water body respectively;

[0037] V ij2 is the probability of vacant land, ranging from 0≤V ij2 ≤1.

[0038] V ij3 Indicates whether green space supplementation is carried out, 0 means no supplementation, 1 means supplementation;

[0039] V ij4 represents the type of green space addition, 0 is natural green space and 1 is artificial green space;

[0040] V ij5 is the land conversion cost, including land acquisition cost C1, land remediation cost C2, according to the current land type V ij1 Determined according to the project area conditions, each land type corresponds to a land conversion cost. The land acquisition cost C1 of each type is C11, C12...C110, and the land remediation cost C2 is C21, C22,...C210;

[0041] Vij6 is the greenfield construction cost, according to V ij4 The type of green space addition is determined, and the green space construction costs of natural green space and artificial green space are C31 and C32 respectively;

[0042] V ij7 is the total cost of green space construction, and its calculation formula is

[0043]

[0044] V ij8 is the construction potential value of the green space, and the calculation formula is:

[0045] V ij8 =V ij2 *(1-V ij5 ′)

[0046] Among them, V ij5 ′ is the maximum and minimum values ​​after normalization of V ij5 The value of

[0047] Determine the demand for each planning target of different land use types, and calculate the weight of each planning target based on the grid proportion of land use types in the block unit:

[0048] Set the demand degree of each land use type l for each planning target m in, <1, for each land use type, the sum of the demand degrees of all planning objectives is 1, that is Calculate the area ratio P of each land use type in the block unit area_ln , for the entire block unit, the weight W of the planning target m m For all land use types L n The proportion of occupied land units multiplied by their corresponding demand The sum of is calculated as:

[0049]

[0050] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: the process of generating an initial population for the block initial database based on the genetic algorithm, and determining the block green space structure optimization result through the fitness of the best individual in each generation of the initial population based on the reinforcement learning algorithm is as follows:

[0051] Initialization of the green space supplement population in the block: determine the number of individuals N in the initial population, and based on the principle of genetic algorithm, generate an initial population containing N schemes based on the initial database of the block. The principle is: according to the green space construction potential value of each grid, determine its green space supplement probability, use a random number to compare with the supplement probability, and determine whether to supplement the green space. The grid that confirms the supplement is randomly assigned the type of green space supplement, and based on the result, update the green space construction cost and the total cost of green space construction;

[0052] Green space supplementary cost accounting: Calculate the total cost of all grid construction of each individual in the initial population, exclude individuals with over-budget costs, re-initialize the population and generate new plans until the number of individuals that meet the budget cost reaches N;

[0053] Generate scheme fitness evaluation: Use the green space coupling degree calculation formula to calculate the sub-goal coupling degree of each individual, multiply the sub-goal coupling degree by the corresponding planning goal weight to obtain the fitness of each individual;

[0054] Reward function calculation: Based on the principle of reinforcement learning, the reward function is calculated to guide the direction of the genetic algorithm. The calculation method is to compare the fitness of the individuals generated by the new generation with the fitness of the best individual in the previous generation. If the fitness of each individual in the current generation is improved compared with the highest fitness of all individuals in the previous generation, a positive reward is given; otherwise, a negative reward is given. For the first operation, it is compared with the fitness of the original state.

[0055] Iteration of selection, crossover, and mutation schemes: Random selection is performed in the initial population according to the tournament selection method, and the individual with the highest probability of being a candidate in the group is selected as the next generation population until N individuals are selected, where the probability of being a candidate is proportional to the fitness and reward function; the selected N individuals are randomly selected in pairs to generate 2 / N parent individual pairs, and the crossover probability is adjusted according to the sum of the rewards of each individual pair, and compared with the random number to determine whether to perform a crossover operation. If a crossover operation is performed, the attribute fragments of the two parent individuals are randomly exchanged to generate two child schemes to replace the parent scheme, where the crossover probability is negatively correlated with the sum of the parent individual reward functions; the child individuals updated by the crossover operation are mutated, the mutation probability is adjusted according to the individual reward value, and compared with the random value to determine whether to change the attribute: whether to perform green space supplementation and the type of green space supplementation, and then the attribute is mutated to generate a new scheme;

[0056] Supplementary cost calculation: Calculate the supplementary cost of the N offspring individuals after mutation, remove the over-budget schemes, and if the number of schemes is less than N, return to the previous step to re-select, crossover, and mutate until N individual schemes are generated; at this time, the generated N schemes are the individual group results of this round of iteration;

[0057] Optimization of block green space structure scheme: Calculate the fitness F of the best individual in each generation of the population. If the change in fitness of the best individual in multiple consecutive generations is less than a preset threshold, it is considered that the fitness has converged; output the last round of schemes, which is the optimal scheme.

[0058] In a second aspect, in order to achieve the above-mentioned purpose, the present invention discloses a block green space structure intelligent optimization system, comprising:

[0059] The model optimization module is used to obtain the planning area, delineate units based on the planning area, obtain multiple block units, obtain relevant indicator data of each block unit, input the relevant indicator data of each block unit into a pre-established regression model, and output an optimized regression model;

[0060] The relevant indicator data of each block unit includes the corresponding target performance indicator and potential influencing factor indicator of each block unit, and the regression model is constructed based on the random forest regression algorithm;

[0061] The probability prediction module is used to calculate the importance of variables in the optimized regression model, perform weight correction and coupling calculation based on the importance of variables, and obtain the corrected weight and the coupling degree of the block unit; obtain the vacant land image, perform grid recognition on the vacant land image, and then superimpose it to obtain the vacant land expansion area, and perform land grid vacancy probability prediction based on the optimized regression model to obtain the vacant land probability;

[0062] The structural optimization module is used to generate an initial block database based on the probability of vacant land, generate an initial population for the initial block database based on a genetic algorithm, and determine the optimization results of the block green space structure through the fitness of the best individuals in each generation of the initial population based on a reinforcement learning algorithm.

[0063] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a terminal device is disclosed, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the memory stores a computer program capable of running on the processor, and when the processor loads and executes the computer program, the above-mentioned method for intelligent optimization of block green space structure is adopted.

[0064] Beneficial effects of the present invention:

[0065] The present invention can effectively make up for the defects of strong subjectivity and randomness in the current urban green space construction. By constructing the structure of the block database, it can realize the formulation of intelligent green space supplement plans in high-restriction urban environments; based on the current land use type composition of the city, the urban green space planning target needs are determined according to local conditions, and the optimized structure is generated according to the needs; and based on reinforcement learning and genetic algorithms, the above key steps are integrated to generate a series of optimization plans in multiple rounds of iterations, providing diversified choices for urban green space planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0067] Figure 1 It is a schematic flow chart of the method of the present invention;

[0068] Figure 2 It is a technical flow chart of the intelligent generation of block green space structure optimization scheme based on reinforcement learning and genetic algorithm of the present invention;

[0069] Figure 3 It is a schematic diagram of the system structure of the present invention;

[0070] Figure 4 is a diagram of a block vacancy probability calculation process according to an embodiment of the present invention;

[0071] Figure 5 It is a partial initial individual graph generated by the embodiment of the present invention based on the genetic algorithm;

[0072] Figure 6 It is the final preferred individual graph obtained according to the fitness evaluation in the embodiment of the present invention. DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] Embodiment 1:

[0075] like Figure 1 As shown, a method for intelligent optimization of block green space structure includes the following steps:

[0076] Obtaining a planning area, demarcating units based on the planning area to obtain a plurality of block units, obtaining relevant indicator data of each block unit, inputting the relevant indicator data of each block unit into a pre-established regression model, and outputting an optimized regression model;

[0077] The relevant indicator data of each block unit includes the corresponding target performance indicator and potential influencing factor indicator of each block unit, and the regression model is constructed based on the random forest regression algorithm;

[0078] Specifically, in this embodiment, after the planning area is determined, the study area is divided into square block-scale planning units according to the actual situation of the planning area and planning needs, hereinafter referred to as block units. In this example, Hengyang City, Hunan Province is selected as the planning area, and the planning area is divided into 1km*1km block units according to the general road spacing distance and block size in the planning area.

[0079] The target performance P{P1,P2,...,Pm} is selected according to the planning orientation, and the potential influencing factors E{E1,E2,En...,En} are listed according to the relationship between the green space structure and the built environment.

[0080] In this example, combined with the coordination requirements for production, living and ecological space in national land space planning, P1 living performance, P2 production performance and P3 ecological performance are proposed.

[0081] Furthermore, according to relevant literature, block activity, average house price and block green volume are used to reflect P1 living performance, P2 production performance and P3 ecological performance respectively.

[0082] Furthermore, for the block activity, the calculation method is the average of the cumulative daily population in the study unit within a week. The higher the value, the higher the living performance; for the average house price, the calculation method is the average of the house prices of all communities in the study unit. The higher the value, the higher the production performance; for the three-dimensional green volume, the calculation method is the sum of the NDVI values ​​in the study unit. The higher the value, the higher the production performance.

[0083] Furthermore, according to previous studies, the relationship between green space structure and built environment includes scale, form, and spatial layout, and a series of representative indicators are selected as potential influencing factors. In this example, the scale is selected, such as per capita green area (E1), per capita natural green area (E2), per capita artificial green area (E3); the form is selected from artificial green space synergy entropy (E4) and natural green space synergy entropy (E5); the spatial layout is selected from green space 5-minute service coverage (E6); the nearest green space distance mean (E7); the nearest artificial green space distance mean (E8); the nearest natural green space distance mean (E9).

[0084] The specific calculation method is shown in Table 1 below:

[0085] Table 1 Corresponding indicators and calculation methods of potential influencing factors

[0086]

[0087]

[0088] According to the target performance and potential influencing factors, the current year and the previous year data of the study area planning are collected, and the corresponding target performance indicators and potential influencing factor indicators of each block unit in each period are calculated. The current year indicator is subtracted from the previous year indicator as the target performance change value {ΔP} and the potential influencing factor indicator change value {ΔE}. For each block unit i, it contains {ΔP, ΔE}. A correlation analysis is performed on the change values ​​of all potential influencing factor indicators to eliminate the influencing factors with multicollinear relationships. The remaining indicators of all block units are summarized into the database I{ΔP, ΔE}.

[0089] In this example, considering data availability and urban changes, 2024 and 2019 were selected as target years to collect data for these two years. The block activity comes from Baidu heat map real-time population data, the average house price comes from Lianjia, Anjuke and other housing price data platforms, the block green volume comes from NDVI data, the green space distribution data and building distribution data come from remote sensing image data, and the number of block units comes from census data.

[0090] Taking the target performance change value (ΔP) as the dependent variable and the potential influencing factor index change value (ΔE) as the independent variable, a regression model M is constructed for each target performance. Output the regression model with better explanatory power according to the evaluation index. Calculate the variable importance of each independent variable.

[0091] In this example, a random forest algorithm is used to build a regression model. In the random forest regression model, database I is randomly divided into a training set and a test set according to a ratio of 8:2, with the training set as the training set and the test set as the test set to evaluate the training results.

[0092] The importance of the independent variables can be extracted by principal component regression and least squares regression. The random forest model is selected because it has the following advantages:

[0093] 1. Stability: Random forest integrates multiple decision trees and has good anti-overfitting ability. Least squares method and principal component regression are prone to overfitting when the amount of data is small or there are too many features, resulting in a decrease in the generalization ability of the model on the test set;

[0094] 2. Scalability: When processing large-scale data, principal component regression and least squares regression have high computational complexity, long training time, and relatively complex implementation of parallel computing. Their efficiency and scalability are not as good as the random forest regression algorithm.

[0095] 3. Adapt to high-dimensional data: Random forest uses a random feature selection method in the process of building a decision tree to automatically screen out important features and reduce the problem of collinearity between features; principal component regression needs to perform principal component analysis to reduce the dimension before processing the data, which may cause loss of original data; the least squares regression model is unstable in the case of high-dimensional data and multicollinearity.

[0096] During the training process, the training set is used for training, and the performance of the random forest regression model is adjusted by adjusting the number of decision trees (n_estimators), the maximum depth of the decision tree (max_depth), the minimum number of samples for node splitting (min_samples_split), the minimum number of samples for leaf nodes (min_samples_leaf), and the number of features for node splitting (max_features).

[0097] The basic formula of the random forest regression algorithm is:

[0098]

[0099] Where E={E1,E2,…E n} is a feature vector containing n potential influencing index change values, M is the number of decision trees, T m (X) is the predicted value of the mth decision tree for the input feature vector X. Each decision tree T m (X) Processes the input features according to its own tree structure and splitting rules to give the final prediction value.

[0100] In the actual training process, the RandomForestRegressor function of the Python scikitlearn library is directly called for calculation.

[0101] The test set was used to evaluate the explanatory power of the constructed regression model through indicators such as mean square error (MSE), mean absolute error (RMSE), and R square (R2).

[0102] Among them, the calculation formulas for mean square error (MSE), mean absolute error (RMSE), and R square (R2) are:

[0103]

[0104] In the formula, y i is the actual value of the test set, is the value calculated based on the random forest regression model, is the average value of the test set, and n is the number of samples in the test set.

[0105] In this example, it is considered that the regression model has better explanatory power when the R square is greater than 0.7 and the MSE and RMSE indicators are smaller, and the regression model M at this time is output.

[0106] Calculate the importance of variables for the optimized regression model, perform weight correction and coupling calculation based on the importance of variables, and obtain the corrected weight and block unit coupling; obtain vacant land images, perform grid recognition on the vacant land images, and then superimpose them to obtain the vacant land expansion area, and predict the vacancy probability of land grids based on the optimized regression model to obtain the vacant land probability;

[0107] Based on the output regression model M, the importance of independent variables is calculated using variable importance measures (VIM).

[0108] For each independent variable e, the variable importance score (VIMe) is calculated as:

[0109]

[0110] Where M is the number of decision trees of the output regression model M, T m is the set of all nodes of the mth decision tree, ΔMSE t,e is the reduction in impurity caused by feature e at node t.

[0111] In this example, Python tools are used to call the forest.feature_importances function in the sklearn package to directly calculate the importance of the independent variable.

[0112] For each target performance model, select the variables whose independent variable importance (VIM) is greater than a certain threshold, use the independent variable importance (VIM) as the initial weight (W), and modify the initial weight. ′ As the coupling model weight. For each block unit i, the coupling degree D for target J ij The model formula is the sum of the variables greater than the threshold and their corrected weights.

[0113]

[0114] Among them, D ij is the coupling degree of block unit i at target j, E im is the value of variable m in block unit i whose variable importance (VIM) is greater than a certain threshold, W′ m is the corrected weight.

[0115] The weight correction method is:

[0116]

[0117] Among them, W m The importance of independent variables in the target performance model (VIM) e ) is the importance score (VIMe) of variable m that is greater than a certain threshold, and n is the total number of variables m.

[0118] In this example, the VIM threshold is selected as 10%.

[0119] According to the definition of urban vacant land and the actual situation of the research site, the category of vacant land is determined, and the land use situation of each grid in the current planning year and the previous year of the research area is identified.

[0120] In this example, based on the definition of vacant land, that is, completely vacant land without development purposes within the city, as well as abandoned, vacant or underutilized land from construction purposes, combined with the actual situation of the research site, it is determined that the vacant land categories include the following categories: bare land, natural grassland, natural shrubs, and inefficient construction land. In other research areas, vacant land may also include areas under construction, natural wetlands, etc.

[0121] Determine the grid size for land identification, vacancy probability prediction, land supplement cost assessment, and land supplement potential assessment according to research needs. The grid size is recommended to be between 10m and 20m in length. In this example, a grid size of 20m is used. In other different research sites, the size should be selected according to needs.

[0122] Furthermore, differentiated identification methods are adopted for different types of vacant land features. For bare land, natural grassland, and natural shrubs, machine learning algorithms are used to interpret remote sensing images to obtain urban land distribution maps, and then the types of bare land, natural grassland, and natural shrubs are extracted. For inefficient construction land, remote sensing images are first interpreted to obtain the distribution area of ​​construction land, and then the inefficient land evaluation index is used to extract the inefficient construction land area. After completing land identification, the land type of each grid is obtained using the majority resampling method.

[0123] In this example, a 17-level Google map remote sensing image with a spatial resolution of 2.15m was selected. Other high-resolution data such as Esri World Image, Tiandi Map, Gaofen Series, GeoEye, Quickbird, Ikonos, etc. can also be selected. Using ArcGIS Pro image classification tools, the number of samples is set to 50 for each type, and each sample is evenly distributed in the target area. Support vector machine is selected as the classifier. Random forest, maximum likelihood method or VGG-16, U-net and other machine learning or deep learning algorithms with image recognition classification can also be selected as classifiers; the recognition results are evaluated using the accuracy assessment tool. In this example, the output model accuracy reaches more than 80% of the land type recognition results, and 10 types of bare land, natural grassland, artificial grassland, natural shrubs, woodland, residential land, commercial land, industrial land, hard square, water body, etc. are identified. The land data is resampled to the grid size using the majority resampling method. The distribution areas of bare land, natural grassland and natural shrubs are extracted.

[0124] Extract construction land grids such as residential land, commercial land, and industrial land. For residential land, commercial land, and industrial land, use the land inefficiency evaluation index to evaluate the inefficiency of each type of land, select land with inefficiency below a certain threshold, and include it in vacant land. The calculation method of inefficiency is as follows: set n evaluation indicators (X1, X2, ...X n ), for this type of land use grid, the grid inefficiency is the grid evaluation index value (x1, x2, ...x n ), the specific expression is as follows:

[0125]

[0126] In the formula, C represents the inefficiency of the construction land type grid, and n represents the number of evaluation indicators;

[0127] In this example, the evaluation indicators of residential land include road intersection density, service facility density, and night light index. Among them, the road intersection data comes from the OSM road network, and the intersections of the road network are extracted through ArcGIS. The service facility density comes from Amap, and the POIs of "catering services, shopping services, life services, sports and leisure services, and medical care services" are screened out. The night light index comes from NPP / VIIRS data. The evaluation indicators of commercial land include popularity index and commercial facility density. Among them, the popularity index comes from Baidu heat map, and the commercial facility density data comes from Amap. The evaluation indicators of industrial land include building density and GDP. Among them, the building density comes from remote sensing satellite images, and the GDP data comes from the resource and environmental science data platform. After completing the inefficiency evaluation of each land, the inefficiency is normalized respectively, and then the units with the lowest 10% of inefficiency evaluation of each type are selected as inefficient construction land.

[0128] The obtained vacant land distribution positions in the two phases are superimposed to obtain the area of ​​vacant land expansion. The relationship between driving factors and vacant land expansion is constructed through regression algorithm. For the research grid i, its driving factors constitute an n-dimensional vector All driving factors are normalized. The research grid of the expanded vacant land is assigned a value of 1, and the rest of the research grids are assigned a value of 0. The regression relationship between the n-dimensional vector and the expansion of vacant land is constructed.

[0129] In this example, ArcGIS was used to apply the Erase tool to output the location of vacant land expansion. The normalized driving factors were converted into a raster layer.

[0130] Furthermore, in this example, referring to the relevant research on the evolution mechanism of construction land changes in shrinking cities, slope (x1), water sensitivity (x2), distance to the city center (x3), commercial facility density (x4), transportation facility density (x5), housing prices (x6), population density (x7), and hard surface proportion (x8) were selected as driving factors.

[0131] The regression algorithm used can be a random forest regression algorithm, a logistic regression algorithm, or an artificial neural network algorithm. m} and its corresponding weight {W x1 ,W x2, …W xm}.

[0132] In this example, the random forest regression algorithm is selected considering its advantage in accurately analyzing complex nonlinear relationships.

[0133] In addition to the random forest regression algorithm, the Markov transition matrix algorithm or the artificial neural network algorithm can also be selected. However, the artificial neural network algorithm takes a long time to train and is prone to overfitting. It is not suitable for conditions with limited data and it is difficult to explain the impact of different factors on the vacancy probability. The Markov transition matrix algorithm predicts the conversion of different land use types to vacant land in the next period based on the transfer probability of the previous period. This prediction method ignores the impact of other factors on land vacancy except land use factors. Therefore, the random forest regression algorithm is used in this example to estimate the future land vacancy status.

[0134] The regression model RF is as follows:

[0135]

[0136] Where X={X1,X2,…X n} is a feature vector containing n driving factors, M is the number of decision trees finally determined, and decision tree T m The depth of (X), the number of node splitting samples, etc. are determined based on the training results.

[0137] The initial database of the block is generated based on the probability of vacant land, the initial population is generated from the initial database of the block based on the genetic algorithm, and the optimization result of the green space structure of the block is determined by the fitness of the best individual in each generation of the initial population based on the reinforcement learning algorithm.

[0138] Specifically, the generation process of the initial block database is as follows:

[0139] First, construct an M×M matrix D0 corresponding to the block grid division, where each element V ij (1≤i≤M,1≤j≤M) represents the grid at the i-th row and j-th column in the block.

[0140]

[0141] Each element V ij All are expressed in the form of attribute vectors. For an element V containing n attributes ij = {V ij1 , V ij2 …V ijn}.

[0142] In this example, each element contains 8 attributes. In other cases, you can add attributes as needed (such as terrain, slope, etc., to reflect the basic conditions of land use), but at least the following attributes must be included, and the specific values ​​within each attribute can be adjusted as needed.

[0143] Specifically, V ij1 is the current land use type, the data source is the land use data interpreted by S31, Vij1 ∈{0,1,2,3…,9}, representing 10 types of land use, including bare land, natural grassland, artificial grassland, natural shrub, forest land, residential land, commercial land, industrial land, hard square, and water body. In other application cases, the type should be adjusted according to the actual situation.

[0144] V ij2 is the vacancy probability, the data source is S33 land vacancy probability, and the value range is 0≤V ij2 ≤1.

[0145] V ij3 Indicates whether green space is supplemented, 0 means no supplementation, 1 means supplementation.

[0146] V ij4 Represents the type of green space addition, 0 is natural green space and 1 is artificial green space.

[0147] V ij5 It is the land conversion cost, including land acquisition cost (C1) and land remediation cost (C2). ij1 According to the project area conditions, each land type corresponds to a land conversion cost, that is, the land acquisition costs (C1) of 10 types such as bare land, natural grassland, artificial grassland, natural shrubs, woodland, residential land, commercial and office land, industrial land, hard square and water body are C11, C12...C110 respectively, and the land improvement costs (C2) are C21, C22,...C210 respectively.

[0148] V ij6 is the greenfield construction cost, according to V ij4 The types of green space additions are determined, and the green space construction costs for natural green space and artificial green space are C31 and C32 respectively.

[0149] V ij7 is the total cost of green space construction, and its calculation formula is

[0150]

[0151] V ij8 is the construction potential value of the green space, and its calculation formula is:

[0152] V ij8 =V ij2 *(1-V ij5 ′).

[0153] Among them, V ij5 ′ is the maximum and minimum values ​​after normalization of V ij5 The value of .

[0154] Figure 4This is a diagram of the calculation process of the vacancy probability of an example block.

[0155] Determine the demand for each planning objective of different land use types, and calculate the weight of each planning objective based on the grid proportion of land use types within the block unit.

[0156] First, set the demand degree of each land use type l for each planning target m For example, for land use type l1, the demand for planning target m1 is expressed as For each land use type, the sum of the demand degrees of all planning objectives is 1, that is, = 1. Next, calculate the area ratio P of each land use type in the block unit area_ln For the entire block unit, the weight of the planning target m is the weight of all land use types L n The proportion of occupied land units multiplied by their corresponding demand The calculation formula is:

[0157]

[0158] Specifically, the scheme of the present invention is further described below through an embodiment: the process of generating an initial population from the initial database of a block based on a genetic algorithm, and determining the optimization result of the block green space structure through the fitness of the best individual in each generation of the initial population based on a reinforcement learning algorithm is as follows:

[0159] Based on the principle of genetic algorithm, the initial population (scheme set) D1 is generated based on the initial database (D0) of the block. The basic principle is to select the green space construction potential value (V ij8 ) to determine whether to add green space (i.e. determine V ij3 ) and then determine the form of green space addition (i.e. V ij4 ), and update other corresponding attribute values. The specific operations are as follows:

[0160] Determine the number of individuals (schemes) N for generating the initial population. Each individual is an M×M matrix identical to the initial block database D0, where each element is a vector containing 8 attributes. In this example, the number of individuals (schemes) for generating the initial population is set to 30. Other numbers can also be selected in other schemes, but are usually set between 20-50.

[0161] Figure 5 Some initial individuals (schemes) generated based on genetic algorithms.

[0162] According to the green space construction potential value (V ij8 ), determine the probability of green space supplementation (P ij), and then determine whether to add green space (V ij3 ). For element V ij , which performs green space supplement probability (P ij ) is calculated as:

[0163]

[0164] Among them, S is the sum of the green space construction potential values ​​of all current individual elements, and its calculation formula is:

[0165]

[0166] Use the random number generator to generate a random number r between 0 and 1. If r ≤ P ij , then for element V ij , whose property V ij3 If set to 1, it indicates that green space is added; otherwise, it remains at the current value (0).

[0167] When green space augmentation is performed (i.e. V ij3 =1), use the random number generator to randomly generate a value of 0 or 1 and assign it to V ij4 When green space is not supplemented (i.e. V ij3 =0)V ij4 is 0.

[0168] According to the green space construction type (V ij4 ) Update green space construction cost (V ij6 ), and according to Formula Update V ij7 .

[0169] Green space supplementary costing

[0170] Calculate each individual (scheme) D in the initial population (scheme set) D1 1_n Calculate its total optimization cost C D1n The calculation method is the total green space construction cost V of all elements in the current individual ij7 The sum of is calculated as:

[0171]

[0172] Compare the total cost of the supplement with the budgeted cost C budge If C D1n ≤C budge , then keep the generated individuals (schemes), otherwise exclude the generated individuals (schemes) and regenerate new individuals (schemes). Repeat the above steps until the number of generated individuals reaches N.

[0173] Generate solution fitness evaluation

[0174] The green space structure coupling degree calculation formula is used to calculate the sub-target coupling degree of each individual (scheme), and the overall coupling degree is calculated according to the planning bottom line weight of the planning target, and the overall coupling degree is used as the fitness F of the individual (scheme). The calculation formula is:

[0175]

[0176] Among them, D j is the coupling degree of sub-target j, and its calculation method refers to the structural coupling degree measurement model, W j is the planning weight for sub-goal j;

[0177] Reward function calculation

[0178] Calculate the reward function to guide the direction of the genetic algorithm change. The reward value is calculated based on the change in fitness of the previous and next generations of the genetic algorithm. For example, if the fitness of each individual in the current generation is improved compared to the highest fitness of all individuals in the previous generation, a positive reward is given; if the fitness decreases, a negative reward is given. The reward value is calculated as follows. For the first operation, compare with the original state.

[0179] R = α × (F current -F previosu_best )

[0180] Among them, α is an adjustment coefficient constant greater than zero, F current is the fitness of the individuals in the current generation, F previosu_best is the highest fitness among all individuals in the previous generation. When the reward function is calculated for the first time, F previosu_best is the fitness of the initial state of the block.

[0181] Iteration of solutions based on selection, crossover, and mutation

[0182] According to the "tournament selection" method, K individuals (K is less than N, usually 3-5) are randomly selected from the population as the tournament group each time, and the individual with the highest probability of being a candidate is selected as the next generation population until N individuals are selected. The probability of individual k being a candidate is P k The calculation formula is as follows:

[0183]

[0184] In a tournament group of size K, F k is the fitness of individual k, obtained by calculation, R k is the reward value of individual k, obtained according to the reward function. β is the adjustment coefficient constant, which is used to control the influence of the reward coefficient.

[0185] After selecting the N individuals, randomly select the parent individuals in pairs to form N / 2 groups of parents. According to the sum of the rewards R of the individual pairs (the two individuals involved in the crossover) sum To adjust the crossover probability P C , and its calculation formula is:

[0186]

[0187] Among them, R sum is the sum of the rewards of the two individuals participating in the crossover, is the sum of the highest reward values ​​of all individual pairs, and δ is the adjustment coefficient constant.

[0188] Generate a random number r, if r≤P C , perform crossover operations on individuals, otherwise keep the individuals directly.

[0189] The crossover operation is to randomly select n different crossover points, divide the individual into n+1 parts, and the newly generated offspring is composed of the parent generation. For example, the elements of the offspring individual C are composed of the first fragment of the parent generation A, the second fragment of B, the third fragment of A, and the fourth fragment of B; the offspring individual D is the opposite, consisting of the first fragment of the parent generation B, the second fragment of A, the third fragment of B, and the fourth fragment of A. This operation is performed on N / 2 groups of parent individuals, and each group of parents crosses to generate two offspring individuals, generating a total of N offspring individuals.

[0190] Perform mutation operation on N offspring individuals after crossover operation. Set the initial mutation probability P m0 , and adjust the mutation probability P according to the individual reward value m0 ′. The adjustment method is as follows:

[0191]

[0192] Among them, P m0 is the initial mutation probability, R j is the reward value of individual j, It is the maximum reward value among the offspring individuals. If the individual reward value is low, its mutation probability is increased; if the individual reward value is high, its mutation probability is reduced.

[0193] For each element of each individual, generate a random number r1 between 0 and 1. If r1≤P m0 ′, then V ij3 Perform 0 / 1 conversion. Generate a random number r2 between 0 and 1. If r2≤P m0 ′ and V ij3 =1, then V ij4 Perform 0 / 1 conversion. Calculate V again based on the value at this time ij7.

[0194] According to the additional cost of the mutated individuals, the schemes that exceed the budget are deleted. If the number of schemes is less than N, the individuals are reselected, crossed, and mutated until N individual schemes are generated. At this time, the generated N schemes are the individual group results of this round of iteration.

[0195] Optimal selection of green space structure scheme for blocks

[0196] Calculate the fitness F of the best individual in each generation of the population. If the change in the fitness of the best individual for multiple consecutive generations (for example, 20 generations) is less than a preset threshold, the fitness is considered to have converged. Output the last round of solutions, which is the best solution.

[0197] Figure 6 is the final preferred individual (scheme) obtained based on fitness evaluation.

[0198] Embodiment 2: In the second aspect, as Figure 3 As shown, in order to achieve the above-mentioned purpose, the present invention discloses a block green space structure intelligent optimization system, comprising:

[0199] The model optimization module 11 is used to obtain the planning area, perform unit demarcation based on the planning area, obtain multiple block units, obtain relevant indicator data of each block unit, input the relevant indicator data of each block unit into a pre-established regression model, and output an optimized regression model;

[0200] The relevant indicator data of each block unit includes the corresponding target performance indicator and potential influencing factor indicator of each block unit, and the regression model is constructed based on the random forest regression algorithm;

[0201] The probability prediction module 12 is used to calculate the importance of variables for the optimized regression model, perform weight correction and coupling calculation based on the importance of variables, and obtain the corrected weight and the coupling degree of the block unit; obtain the vacant land image, perform grid recognition on the vacant land image, and then superimpose it to obtain the vacant land expansion area, and perform land grid vacancy probability prediction based on the optimized regression model to obtain the vacant land probability;

[0202] The structural optimization module 13 is used to generate an initial block database based on the probability of vacant land, generate an initial population for the initial block database based on a genetic algorithm, and determine the optimization result of the block green space structure through the fitness of the best individual in each generation of the initial population based on a reinforcement learning algorithm.

[0203] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0204] It needs to be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to execute the above method. The storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0205] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0206] The above shows and describes the basic principles, main features and advantages of the present disclosure. Those skilled in the art should understand that the present disclosure is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, the present disclosure may have various changes and improvements, and these changes and improvements fall within the scope of the present disclosure to be protected.

Claims

1. A method for intelligent optimization of block green space structure, characterized in that: The method comprises the following steps: Obtaining a planning area, demarcating units based on the planning area to obtain a plurality of block units, obtaining relevant indicator data of each block unit, inputting the relevant indicator data of each block unit into a pre-established regression model, and outputting an optimized regression model; The relevant indicator data of each block unit includes the corresponding target performance indicator and potential influencing factor indicator of each block unit, and the regression model is constructed based on the random forest regression algorithm; Calculate the importance of variables for the optimized regression model, perform weight correction and coupling calculation based on the importance of variables, and obtain the corrected weight and block unit coupling; obtain vacant land images, perform grid recognition on the vacant land images, and then superimpose them to obtain the vacant land expansion area, and predict the vacancy probability of land grids based on the optimized regression model to obtain the vacant land probability; The initial database of the block is generated based on the probability of vacant land, the initial population is generated from the initial database of the block based on the genetic algorithm, and the optimization result of the green space structure of the block is determined by the fitness of the best individual in each generation of the initial population based on the reinforcement learning algorithm.

2. The intelligent optimization method for block green space structure according to claim 1 is characterized in that: The process of obtaining relevant indicator data for each block unit: Calculate the corresponding target performance indicators and potential influencing factor indicators for each block unit in each period, and subtract the indicators of the previous year from the indicators of the current year to obtain the target performance change value {ΔP} and the potential influencing factor indicator change value {ΔE}. For each block unit i, it includes {ΔP, ΔE}; perform correlation analysis on all the potential influencing factor indicator change values, eliminate the influencing factors with multicollinear relationships, and summarize the remaining indicators of all block units into the database I{ΔP, ΔE}.

3. The intelligent optimization method for block green space structure according to claim 1 is characterized in that: The formula of the random forest regression algorithm is as follows: Where E={E1,E2,…E n } is a feature vector containing n potential influencing index change values, M is the number of decision trees, T m (X) is the predicted value of the mth decision tree for the input feature vector X. Each decision tree T m (X) Process the input features according to its own tree structure and splitting rules to give the final prediction value; The evaluation indicators of the regression model include mean square error (MSE), mean absolute error (RMSE) and R square (R 2 ; The formula is as follows: In the formula, y i is the actual value of the test set, is the value calculated based on the random forest regression model, is the average value of the test set, and n is the number of samples in the test set.

4. The intelligent optimization method for block green space structure according to claim 1 is characterized in that: The optimized regression model is used to calculate the variable importance VIM e The calculation process: Where M is the number of decision trees of the output regression model M, T m is the set of all nodes of the mth decision tree, ΔMSE t,e is the reduction in impurity caused by feature e at node t; The process of weight correction and coupling calculation based on variable importance: For each target performance model, select the variables whose independent variable importance VIM is greater than the preset threshold, use the independent variable importance VIM as the initial weight W, and modify the initial weight W to obtain the modified weight W ′ As the coupling model weight, for each block unit i, the coupling degree D for target J ij The model formula is the sum of the products of variables greater than the threshold and the modified weights; Among them, D ij is the coupling degree of block unit i at target j, E im is the value of variable m in block unit i whose variable importance VIM is greater than the threshold, W ′ m is the modified weight; The weight correction method is: Among them, W m VIM is the importance of independent variables in the target performance model e The importance score VIM of variables m greater than the threshold e , n is the total number of variables m.

5. The intelligent optimization method for block green space structure according to claim 4 is characterized in that: The optimized regression model RF expression is as follows: Where X={X1,X2,…X n } is a feature vector containing n driving factors, M is the number of decision trees finally determined, and decision tree T m The depth of (X) and the number of node split samples are determined according to the training results.

6. The intelligent optimization method for block green space structure according to claim 1 is characterized in that: The vacant land refers to the completely vacant land with no development purpose within the city, as well as the land that is abandoned, vacant or underutilized for construction purposes. The categories of vacant land include bare land, natural grassland, natural shrubs, inefficient construction land, areas under construction and natural wetlands.

7. The intelligent optimization method for block green space structure according to claim 1 is characterized in that: The generation process of the block initial database is as follows: Construct an M×M matrix D0 corresponding to the block grid division, where each element V ij represents the grid at the i-th row and j-th column in the block, where 1≤i≤M,1≤j≤M; Each element V ij All are expressed in the form of attribute vectors. For an element V containing n attributes ij = {V ij1 , V ij2 …V ijn }; V ij1 is the current land use type, V ij1 ∈{0,1,2,3…,9}, representing bare land, natural grassland, artificial grassland, natural shrub, forest land, residential land, commercial land, industrial land, hard square and water body respectively; V ij2 is the probability of vacant land, ranging from 0≤V ij2 ≤1; V ij3 Indicates whether green space supplementation is carried out, 0 means no supplementation, 1 means supplementation; V ij4 represents the type of green space addition, 0 is natural green space and 1 is artificial green space; V ij5 is the land conversion cost, including land acquisition cost C1, land remediation cost C2, according to the current land type V ij1 Determined according to the project area conditions, each land type corresponds to a land conversion cost. The land acquisition cost C1 of each type is C11, C12...C110, and the land remediation cost C2 is C21, C22,...C210; V ij6 is the greenfield construction cost, according to V ij4 The type of green space addition is determined, and the green space construction costs of natural green space and artificial green space are C31 and C32 respectively; V ij7 is the total cost of green space construction, and its calculation formula is V ij8 is the construction potential value of the green space, and the calculation formula is: V ij8 =V ij2 *(1-V ij5 ′) Among them, V ij5 ′ is the maximum and minimum values ​​after normalization of V ij5 The value of Determine the demand for each planning target of different land use types, and calculate the weight of each planning target based on the grid proportion of land use types in the block unit: Set the demand degree of each land use type l for each planning target m in, <1, for each land use type, the sum of the demand degrees of all planning objectives is 1, that is Calculate the area ratio P of each land use type in the block unit area_ln , for the entire block unit, the weight W of the planning target m m For all land use types L n The proportion of occupied land units multiplied by their corresponding demand The sum of is calculated as:

8. The method for intelligent optimization of block green space structure according to claim 1, characterized in that: The process of generating an initial population from the initial database of a block based on a genetic algorithm and determining the optimization result of the block green space structure through the fitness of the best individual in each generation of the initial population based on a reinforcement learning algorithm is as follows: Initialization of the green space supplement population in the block: determine the number of individuals N in the initial population, and based on the principle of genetic algorithm, generate an initial population containing N schemes based on the initial database of the block. The principle is: according to the green space construction potential value of each grid, determine its green space supplement probability, use a random number to compare with the supplement probability, and determine whether to supplement the green space. The grid that confirms the supplement is randomly assigned the type of green space supplement, and based on the result, update the green space construction cost and the total cost of green space construction; Green space supplementary cost accounting: Calculate the total cost of all grid construction of each individual in the initial population, exclude individuals with over-budget costs, re-initialize the population and generate new plans until the number of individuals that meet the budget cost reaches N; Generate scheme fitness evaluation: Use the green space coupling degree calculation formula to calculate the sub-goal coupling degree of each individual, multiply the sub-goal coupling degree by the corresponding planning goal weight to obtain the fitness of each individual; Reward function calculation: Based on the principle of reinforcement learning, the reward function is calculated to guide the direction of the genetic algorithm. The calculation method is to compare the fitness of the individuals generated by the new generation with the fitness of the best individual in the previous generation. If the fitness of each individual in the current generation is improved compared with the highest fitness of all individuals in the previous generation, a positive reward is given; otherwise, a negative reward is given. For the first operation, it is compared with the fitness of the original state. Iteration of selection, crossover, and mutation schemes: Random selection is performed in the initial population according to the tournament selection method, and the individual with the highest probability of being a candidate in the group is selected as the next generation population until N individuals are selected, where the probability of being a candidate is proportional to the fitness and reward function; the selected N individuals are randomly selected in pairs to generate 2 / N parent individual pairs, and the crossover probability is adjusted according to the sum of the rewards of each individual pair, and compared with the random number to determine whether to perform a crossover operation. If a crossover operation is performed, the attribute fragments of the two parent individuals are randomly exchanged to generate two child schemes to replace the parent scheme, where the crossover probability is negatively correlated with the sum of the parent individual reward functions; the child individuals updated by the crossover operation are mutated, the mutation probability is adjusted according to the individual reward value, and compared with the random value to determine whether to change the attribute: whether to perform green space supplementation and the type of green space supplementation, and then the attribute is mutated to generate a new scheme; Supplementary cost calculation: Calculate the supplementary cost of the N offspring individuals after mutation, remove the over-budget schemes, and if the number of schemes is less than N, return to the previous step to re-select, crossover, and mutate until N individual schemes are generated; at this time, the generated N schemes are the individual group results of this round of iteration; Optimization of block green space structure scheme: Calculate the fitness F of the best individual in each generation of the population. If the change in fitness of the best individual in multiple consecutive generations is less than a preset threshold, it is considered that the fitness has converged; output the last round of schemes, which is the optimal scheme.

9. An intelligent optimization system for green space structure in a block, characterized in that: include: The model optimization module is used to obtain the planning area, delineate units based on the planning area, obtain multiple block units, obtain relevant indicator data of each block unit, input the relevant indicator data of each block unit into a pre-established regression model, and output an optimized regression model; The relevant indicator data of each block unit includes the corresponding target performance indicator and potential influencing factor indicator of each block unit, and the regression model is constructed based on the random forest regression algorithm; The probability prediction module is used to calculate the importance of variables in the optimized regression model, perform weight correction and coupling calculation based on the importance of variables, and obtain the corrected weight and the coupling degree of the block unit; obtain the vacant land image, perform grid recognition on the vacant land image, and then superimpose it to obtain the vacant land expansion area, and perform land grid vacancy probability prediction based on the optimized regression model to obtain the vacant land probability; The structural optimization module is used to generate an initial block database based on the probability of vacant land, generate an initial population for the initial block database based on a genetic algorithm, and determine the optimization results of the block green space structure through the fitness of the best individuals in each generation of the initial population based on a reinforcement learning algorithm.

10. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, a method for intelligent optimization of block green space structure according to any one of claims 1 to 8 is adopted.

Images