Multi-objective optimized urban development boundary land intelligent configuration simulation method

Abstract

The invention discloses a multi-objective optimized urban development boundary land intelligent configuration simulation method, which comprises the following steps: S1, acquiring multi-source original data of a to-be-simulated city, and preprocessing the multi-source original data; s2, based on the preprocessed data, dividing urban land basic units, extracting feature information of each unit, and constructing an urban development boundary configuration basic data set; and S3, building a city development scene generation model of a generative model architecture, and taking the basic data set as input. The invention relates to the technical field of city planning and simulation. According to the multi-objective optimized intelligent configuration simulation method for the urban development boundary land, a generative model and a deep learning technology are adopted, a virtual development boundary scheme is generated according to multi-source data, the virtual development boundary scheme and actual urban land data are compared and analyzed, a deviation value is calculated, model parameters are automatically adjusted, iterative optimization is carried out, and the intelligent configuration simulation method for the urban development boundary land is achieved. And the development scheme gradually tends to be optimal.

Description

Technical Field

[0001] This invention relates to the field of urban planning and simulation, and more specifically, to a multi-objective optimization simulation method for intelligent allocation of urban development boundary land use. Background Technology

[0002] With the acceleration of urbanization, the rational planning and allocation of urban development boundaries has become a crucial aspect of sustainable urban development. Traditional urban planning methods often rely on experience and manual design, making it difficult to comprehensively consider multiple objectives such as ecological protection, resource utilization efficiency, and economic development. Furthermore, traditional methods are inefficient when processing large-scale, high-dimensional urban data and struggle to quickly generate and evaluate various planning schemes.

[0003] In recent years, the development of big data, artificial intelligence, and computer simulation technologies has provided new solutions for urban planning. In particular, the application of deep learning models such as generative adversarial networks has made it possible to generate high-quality, diverse virtual urban development boundary schemes. However, how to effectively utilize these technologies, combined with multi-objective optimization methods, to achieve intelligent allocation of urban development boundary land remains a key issue that needs to be addressed in the field of urban planning.

[0004] While some research on urban planning simulation exists in the current technology, most focuses on optimizing single objectives, such as land use efficiency or ecological protection, while neglecting other important factors. Furthermore, existing methods often lack effective feedback mechanisms when generating and evaluating virtual planning schemes, leading to significant deviations between the generated schemes and actual needs. Therefore, developing an urban development boundary planning method that can comprehensively consider multiple objectives and achieve intelligent configuration and simulation is of significant practical importance. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-objective optimization intelligent configuration simulation method for urban development boundary land use. This method solves the problems of low efficiency in traditional methods when processing large-scale, high-dimensional urban data, and difficulty in quickly generating and evaluating multiple planning schemes. In the existing technology, although there have been some studies on urban planning simulation, most of them focus on the optimization of a single objective, only considering land use efficiency or ecological protection, and lack an effective feedback mechanism, resulting in a large deviation between the generated scheme and actual needs.

[0006] This invention achieves the above objectives through the following technical solution: a multi-objective optimization intelligent configuration simulation method for urban development boundary land use, comprising the following steps: S1. Obtain multi-source raw data of the city to be simulated, and preprocess the multi-source raw data; S2. Based on the preprocessed data, divide the urban land use into basic units, extract the feature information of each unit, and construct a basic dataset for urban development boundary configuration. S3. Build a city development scenario generation model with a generative model architecture. Using the basic dataset as input, generate multiple sets of virtual city development boundary schemes through the generation module. At the same time, use the discrimination module to evaluate the rationality and authenticity of the generated schemes to form a model training loop. S4. Introduce a multi-objective optimization index system to conduct a comprehensive evaluation of the generated virtual city development boundary schemes and select candidate schemes that meet the preset requirements. S5. Compare and analyze the candidate schemes with the current status data of urban land use, calculate the deviation between the schemes and the actual situation, adjust the parameters of the generative model based on the deviation feedback, and iteratively optimize the virtual development boundary scheme. S6. Output the optimized urban development boundary land use configuration scheme and generate the corresponding visual simulation scene.

[0007] Furthermore, in step S1: The multi-source raw data includes, but is not limited to, remote sensing image data, land use status data, topographic elevation data, transportation network data, and population and economic statistics. The preprocessing operations include data cleaning, standardization, and spatial alignment. Data cleaning is used to remove outliers and redundant data, standardization is used to unify the value range of numerical data, and spatial alignment is used to ensure the consistency of various types of data in spatial location.

[0008] Furthermore, in step S2: The urban land use basic unit is divided using a grid division method, and the unit size is flexibly set according to the scale of the city to be simulated and the simulation accuracy requirements. The feature information includes land use attribute features and spatial association features. The land use attribute features include, but are not limited to, land use type, plot ratio, building density and ecological sensitivity index. The spatial association features include, but are not limited to, spatial adjacency relationship, transportation accessibility and resource sharing degree. The basic dataset is formed by integrating the comprehensive feature vectors of each land use basic unit. The comprehensive feature vectors are obtained by fusing the land use attribute feature vectors and the spatial correlation feature vectors.

[0009] Furthermore, in step S3: The generative model is a generative adversarial network, and the generation module is a generator that adopts an encoder-decoder architecture. By transforming and reconstructing the input features, it outputs a virtual city development boundary scheme vector. The discrimination module is a discriminator that adopts a deep learning network architecture. Its input is a real urban development boundary data sample or a virtual scheme sample output by the generator, and its output is the probability value that the sample belongs to the real data. The model training loop is implemented by alternately training the generator and the discriminator. During the training process, an adversarial loss function is set to minimize the discriminator's ability to recognize virtual schemes and maximize the discriminator's ability to distinguish between real and virtual samples, until the model converges.

[0010] Furthermore, the generator loss function is used to minimize the discriminator's ability to recognize virtual schemes, and its expression is:

[0011] in Configure a basic dataset for urban development boundaries. For the first The comprehensive feature vector of each land use basic unit, This is the output function of the generator. The output function of the discriminator. For expectation operators; The discriminator loss function is used to maximize the ability to distinguish between real and virtual samples, and its expression is:

[0012] in To develop boundary data samples for real cities, Develop a dataset of real-world city boundaries.

[0013] Furthermore, in step S4: The multi-objective optimization index system includes ecological protection indicators, resource utilization efficiency indicators, and economic development indicators. The ecological protection indicators are calculated based on the proportion of ecological land within the development boundary and the avoidance rate of sensitive ecological areas; the resource utilization efficiency indicators are calculated based on land development intensity and infrastructure matching rate; and the economic development indicators are calculated based on the GDP output potential and the predicted number of jobs within the development boundary. The comprehensive evaluation is achieved by calculating the comprehensive evaluation score of the virtual development boundary scheme. The comprehensive evaluation score is obtained by weighted summation of the standardized values ​​of each indicator and their corresponding weights. Schemes with a comprehensive evaluation score greater than or equal to a preset threshold are selected as candidate schemes.

[0014] Furthermore, the weights of each indicator are determined using the analytic hierarchy process (AHP), and the sum of the weights of all indicators is 1. The standardization process is used to eliminate the influence of different dimensions between indicators. The expression for the comprehensive evaluation score is:

[0015] in To optimize the number of indicators in a multi-objective indicator system, For the first The weight of each indicator, For the first The standardized value of each indicator.

[0016] Furthermore, in step S5: The real-world urban land use status data includes the spatial range of the real development boundary, land use structure and indicator parameters. A real-world comparison dataset is constructed based on the real-world urban land use status data. The deviation value is calculated using a distance calculation formula, which includes the Euclidean distance formula. The deviation value is used as a feedback signal to introduce into the loss function of the generation module, update the parameters of the generative model and retrain the model, iteratively generate new virtual development boundary schemes until the calculated deviation value is less than the preset deviation threshold.

[0017] Furthermore, the expression for the Euclidean distance formula is as follows:

[0018] in, The number of basic urban land use units. Dimensions for configuring parameters for development boundaries. For the virtual development boundary scheme, the first The land use basic unit in the first The values ​​that can be obtained under each configuration parameter For the real-world comparison dataset, the first The land use basic unit in the first The values ​​that can be obtained under each configuration parameter; The updated generator module loss function expression is as follows:

[0019] in This is the deviation penalty coefficient.

[0020] Furthermore, in step S1, the standardization process employs a max-min standardization method, expressed as:

[0021] in For the first The first type of multi-source raw data One sample, , The first Minimum and maximum values ​​of multi-source raw data; In step S2, the spatial correlation characteristics of each land use basic unit are modeled using a graph neural network to generate spatial correlation feature vectors.

[0022] The beneficial effects of this invention are as follows: 1. By establishing a multi-objective optimization indicator system that considers ecological protection, resource utilization efficiency, and economic development, and taking into account various factors, the advantages and disadvantages of virtual development boundary schemes can be comprehensively evaluated, ensuring the feasibility and efficiency of the development schemes in multiple dimensions.

[0023] 2. By employing generative models and deep learning techniques, a virtual development boundary scheme is generated based on multi-source data. This scheme is then compared and analyzed with actual urban land use data to calculate deviation values. The model parameters are automatically adjusted and iteratively optimized to ensure that the development scheme gradually approaches the optimal level.

[0024] 3. By preprocessing multi-source raw data and modeling spatial correlation features, we can provide more scientific and accurate urban development boundary configuration schemes, providing strong data support for urban planning and decision-making.

[0025] 4. It takes into account ecological protection and resource utilization efficiency, effectively avoids over-exploitation and ecological damage, promotes sustainable development, and has high flexibility, allowing for customized simulation and optimization according to the actual needs of different cities.

[0026] 5. The optimized development boundary scheme can generate a visual simulation scenario, intuitively display the development results, help decision-makers better understand the planning effect, and improve the feasibility of planning implementation and public acceptance. Attached Figure Description

[0027] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a diagram of the generative model architecture of the present invention; Figure 3 The feedback optimization closed-loop diagram for this invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Example 1: Please see Figure 1-3 This invention provides a technical solution: a multi-objective optimization intelligent configuration simulation method for urban development boundary land use, the method comprising the following steps: S1. Obtain multi-source raw data of the city to be simulated, and perform data cleaning, standardization and spatial alignment preprocessing. Among them, multi-source raw data refers to raw information from different data sources, such as geographic information systems, remote sensing imagery, and socioeconomic statistics. This data is diverse and complex. Data cleaning involves checking and correcting the acquired raw data to remove noise, erroneous data, and incomplete data, thereby improving data quality. Data standardization converts data of different formats and dimensions into a unified standard form, making the data comparable and consistent, facilitating subsequent analysis and processing. Spatial alignment preprocessing ensures that different spatial data have consistency in terms of geographic coordinates, projections, etc., enabling them to be accurately analyzed and simulated within the same spatial framework. S2. Based on the preprocessed data, divide the urban land use into basic units, extract the land use attribute characteristics and spatial correlation characteristics of each unit, and construct a basic dataset for urban development boundary configuration. Among them, the basic urban land use unit divides the urban area into several basic units. These units form the basis for urban land use analysis and configuration. They can be regular grid units or irregular units divided according to factors such as urban function and topography. Land use attribute characteristics describe the properties of the urban land use unit itself, such as land use (residential, commercial, industrial, etc.), land area, topography, etc. Spatial correlation characteristics reflect the spatial relationships between urban land use units, such as adjacency, distance, and spatial accessibility. The basic dataset for urban development boundary configuration consists of the divided basic urban land use units and their extracted land use attribute characteristics and spatial correlation characteristics, providing data support for subsequent urban development boundary configuration. S3. Build a city development scenario generation model with a generative adversarial network architecture. Using a basic dataset as input, the generator generates multiple sets of virtual city development boundary schemes. At the same time, the discriminator evaluates the rationality and authenticity of the generated schemes, forming an adversarial training loop. The Generative Adversarial Network (GAN) architecture is a deep learning model architecture consisting of a generator and a discriminator. The generator is responsible for generating virtual data, while the discriminator is responsible for determining whether the input data is real data or virtual data generated by the generator. The two are continuously optimized through adversarial training. The urban development scenario generation model is a model built on the GAN architecture to generate virtual urban development boundary schemes to simulate urban land use layouts under different conditions. The generator, as part of the GAN, generates multiple sets of virtual urban development boundary schemes based on the input dataset. The discriminator, as the other part of the GAN, evaluates the rationality and realism of the virtual urban development boundary schemes generated by the generator and judges their similarity to real urban development scenarios. The adversarial training loop is a process in which the generator and discriminator compete against each other and continuously train. The generator attempts to generate more realistic virtual schemes to deceive the discriminator, while the discriminator continuously improves its judgment ability to accurately distinguish between real and virtual schemes. This iterative training improves the model's performance. S4. Introduce a multi-objective optimization index system to comprehensively evaluate the generated virtual city development boundary schemes and select candidate schemes that meet the needs of ecological protection, resource utilization efficiency and economic development. The multi-objective optimization index system is a system composed of multiple indicators reflecting different objectives, used to comprehensively evaluate urban development boundary schemes. In this method, these objectives include ecological protection, resource utilization efficiency, and economic development needs, and the indicators may involve aspects such as the degree of protection of ecologically sensitive areas, land resource utilization efficiency, and economic development potential. The comprehensive evaluation is to use the multi-objective optimization index system to conduct a comprehensive and systematic assessment of the generated virtual urban development boundary schemes, considering the achievement of each objective, and giving a comprehensive evaluation result. The candidate schemes are virtual urban development boundary schemes that meet the conditions of ecological protection, resource utilization efficiency, and economic development needs after comprehensive evaluation. These schemes will be used as objects for further analysis and optimization. S5. Compare and analyze the candidate schemes with the current status data of urban land use, calculate the deviation value between the schemes and the actual situation, adjust the model parameters of the generative adversarial network based on the deviation feedback, and iteratively optimize the virtual development boundary scheme. Among them, the current urban land use data reflects the actual land use situation in the city, including information such as land use type, distribution, and development intensity; comparative analysis compares the candidate schemes with the current urban land use data to identify the differences and similarities between the two; the deviation value is the degree of difference between the candidate scheme and the actual situation calculated through comparative analysis, and is expressed numerically as the degree of deviation between the scheme and reality; model parameter adjustment modifies and optimizes the model parameters of the generative adversarial network based on deviation feedback to improve the accuracy and rationality of the generated virtual development boundary schemes; iterative optimization continuously improves the virtual development boundary schemes by repeatedly generating schemes, evaluating and screening, conducting comparative analysis, and adjusting parameters to make them closer to the actual situation and meet planning requirements; S6. Output the optimized urban development boundary land use configuration scheme and generate the corresponding visual simulation scene to provide a basis for urban planning decisions; Among them, the optimized urban development boundary land allocation scheme is the final urban development boundary land allocation scheme obtained after iterative optimization. This scheme takes into account multiple factors such as ecology, resources, and economy, and has high rationality and feasibility. The visualization simulation scenario displays the optimized urban development boundary land allocation scheme in intuitive graphic and image forms, enabling decision-makers to have a clearer understanding of the scheme's layout and effects, and providing intuitive reference for urban planning decisions.

[0030] It should be noted that during the process, multi-source data acquisition and preprocessing are conducted to ensure data quality and consistency, providing a reliable foundation for subsequent analysis. Land use units are divided and a basic dataset is constructed to accurately grasp the characteristics of urban land use, providing detailed basis for configuration schemes. A generative adversarial network model is built, and diverse and reasonable virtual schemes are generated through adversarial training between the generator and discriminator. A multi-objective optimization index system is introduced to comprehensively consider ecological, resource, and economic needs, and high-quality candidate schemes are selected. The candidate schemes are compared and analyzed with real data and iteratively optimized to improve the fit between the schemes and the actual situation. The optimized schemes are output and a visualized scene is generated, providing an intuitive, scientific, and reasonable basis for urban planning decisions. This helps to achieve intelligent and efficient allocation of urban development boundary land and promote sustainable urban development.

[0031] In one embodiment, multi-source raw data of the city to be simulated is acquired, and the data is preprocessed by cleaning, standardization, and spatial alignment, including: Obtain multi-source raw data of the city to be simulated, including remote sensing imagery, land use status data, topographic elevation data, transportation network data, and population and economic statistics, and construct a dataset. ,in Indicates the number of data types. Indicates the first class of raw data; The raw data is cleaned to remove outliers and redundant data. The min-max standardization method is then used to process the numerical data. The standardization expression is as follows:

[0032] in Indicates the first The first in the class data One sample, , They represent the first Minimum and maximum values ​​of class data; Geographic Information System (GIS) tools are used to spatially align multi-source data, unify coordinate systems and spatial resolution, and ensure the consistency of spatial location for various data types, thus forming a preprocessed dataset. .

[0033] This design acquires multi-source data, such as remote sensing imagery, and constructs a dataset. After cleaning and removing abnormal and redundant data, numerical data is processed using max-min standardization. Spatial alignment is then achieved using GIS tools. The multi-source data comprehensively reflects urban characteristics, while cleaning and standardization ensure data quality and comparability. Spatial alignment ensures accurate matching of various data in spatial location. This provides a high-quality, unified, and accurate data foundation for subsequent urban development boundary land use allocation, avoiding configuration errors caused by data issues. This makes the simulation method reliable and scientific from the source, improving the accuracy of the entire intelligent configuration process.

[0034] In one embodiment, based on the preprocessed data, urban land use basic units are divided, and the land use attribute characteristics and spatial correlation characteristics of each unit are extracted to construct a basic dataset for urban development boundary configuration, including: Based on the preprocessed dataset The city area to be simulated is divided into several basic urban land use units of equal area using a grid partitioning method. The unit size is set according to the city size and simulation accuracy requirements. Extract the attribute characteristics of each land use basic unit, including land use type, plot ratio, building density, and ecological sensitivity index, to form an attribute feature vector. ,in Represents the dimension of attribute features; The spatial relationship characteristics between each land use unit and its surrounding units are calculated, including spatial adjacency, accessibility, and resource sharing. Spatial relationships are modeled using graph neural networks to generate spatial relationship feature vectors. ; Fusion attribute feature vector Spatial associated feature vector Construct a comprehensive feature vector for each land use unit, integrate the comprehensive feature vectors of all units, and form a basic dataset for urban development boundary configuration. ,in Indicates the number of basic land use units. Indicates the first The comprehensive feature vector of each unit.

[0035] This design divides the preprocessed dataset into basic urban land units of equal area, extracts attribute features to form vectors, calculates spatial correlation features and models them to generate vectors, and integrates them to construct a comprehensive feature vector to form a basic dataset. The grid division method can flexibly adjust the precision according to the actual situation of the city. The fully extracted attributes and spatial correlation features can accurately characterize the characteristics of each unit. The integrated feature vector after fusion covers a variety of information. The constructed basic dataset is complete and accurate, providing a rich, detailed and reliable basis for generating virtual city development boundary schemes, making the scheme more in line with the actual situation of the city.

[0036] In one embodiment, a city development scenario generation model with a generative adversarial network architecture is constructed. Using a basic dataset as input, a generator produces multiple sets of virtual city development boundary schemes. Simultaneously, a discriminator evaluates the rationality and realism of the generated schemes, forming an adversarial training loop, including: A generator module for a generative adversarial network is constructed, employing an encoder-decoder architecture, based on a fundamental dataset. The unit-synthesized feature vector is taken as input, and the features are transformed and reconstructed through convolutional and deconvolutional layers to output a virtual urban development boundary scheme vector. ,in This indicates the dimension of the development boundary configuration parameters. Indicates the first The unit in the first The values ​​that can be obtained under each configuration parameter; A discriminator module of a generative adversarial network (GAN) is constructed, employing a convolutional neural network architecture. The input consists of real urban development boundary data samples or virtual scheme samples output by the generator. The output, passed through a fully connected layer, is the probability value that the sample represents real data. ,in Indicates the input sample; Define the adversarial loss function and the generator loss function. ,in Configure a basic dataset for urban development boundaries. For the first The comprehensive feature vector of each land use basic unit, This is the output function of the generator. The output function of the discriminator. This is the expectation operator, used to minimize the discriminator's ability to identify virtual schemes; Discriminator loss function:

[0037] in To develop boundary data samples for real cities, Develop boundary datasets for real cities to maximize the ability to distinguish between real and virtual samples; The generator and discriminator parameters are updated using an alternating training method, with the generator minimizing the parameters via gradient descent. The discriminator minimizes the threshold value through gradient descent. This forms an adversarial training loop until the model converges, generating multiple stable virtual city development boundary schemes.

[0038] This design constructs a generative adversarial network (GAN). The generator uses an encoder-decoder architecture to generate virtual scheme vectors, while the discriminator uses a convolutional neural network to output the true probability of samples. An adversarial loss function is set, and the parameters are updated alternately during training. The GAN can continuously optimize the quality of generated schemes through adversarial training between the generator and the discriminator. The generator strives to generate realistic schemes, while the discriminator improves its discrimination ability. The two promote each other, and the adversarial loss function guides the model training direction. Alternating training enables the model to converge quickly, ultimately generating multiple stable and reasonable virtual city development boundary schemes, providing a wealth of choices for subsequent selection and optimization.

[0039] In one embodiment, a multi-objective optimization index system is introduced to comprehensively evaluate the generated virtual city development boundary schemes and select candidate schemes that meet the needs of ecological protection, resource utilization efficiency, and economic development, including: A multi-objective optimization indicator system is constructed, which includes ecological protection indicators, resource utilization efficiency indicators, and economic development indicators. Among them, the ecological protection indicators are calculated by the proportion of ecological land within the development boundary and the avoidance rate of sensitive ecological areas; the resource utilization efficiency indicators are calculated by the land development intensity and infrastructure matching rate; and the economic development indicators are calculated by the GDP output potential within the development boundary and the predicted number of jobs. Each indicator is standardized to eliminate the influence of dimensions, and the weights of each indicator are determined using the analytic hierarchy process (AHP). ,in Indicates the number of indicators, and ; Calculate the overall evaluation score for each virtual development boundary scheme. ,in Indicates the first The standardized values ​​of each indicator; Set a threshold for the overall evaluation score Filter out those with scores greater than or equal to The virtual solution is presented as a candidate solution to meet multiple objectives.

[0040] This design constructs a multi-objective optimization indicator system, covering ecological, resource, and economic indicators. After standardization, the weights are determined using the analytic hierarchy process (AHP), and a comprehensive evaluation score is calculated to screen candidate solutions. The multi-objective indicator system comprehensively considers the needs of urban development in all aspects, avoiding the one-sidedness caused by a single objective. Standardization eliminates the influence of dimensions, making the indicators comparable. The AHP scientifically determines the weights, accurately reflecting the importance of each indicator. The comprehensive evaluation score can objectively measure the merits of the solutions. The selected candidate solutions meet the needs of multiple objectives, providing a more realistic and comprehensive selection of solutions for urban planning.

[0041] In one embodiment, candidate schemes are compared and analyzed with real-world urban land use data to calculate the deviation between the schemes and the actual situation. Based on the deviation feedback, the model parameters of the generative adversarial network are adjusted to iteratively optimize the virtual development boundary scheme, including: Obtain real-world land use data for the city to be simulated, extract the spatial extent of the current development boundary, land use structure, and indicator parameters, and construct a real-world comparison dataset. ; Compare each candidate solution with the real-world dataset. The deviation value is calculated using the Euclidean distance formula:

[0042] in, The number of basic urban land use units. Dimensions for configuring parameters for development boundaries. For the virtual development boundary scheme, the first The land use basic unit in the first The values ​​that can be obtained under each configuration parameter For the real-world comparison dataset, the first The land use basic unit in the first Values ​​under each configuration parameter Deviation value As a feedback signal, the generator's loss function is introduced, and the updated generator loss function is:

[0043] in This is the deviation penalty coefficient; Based on the updated loss function, the generative adversarial network is retrained, and new virtual development boundary schemes are iteratively generated. The deviation calculation and model adjustment steps are repeated until the deviation between the candidate scheme and the real data is less than a preset threshold. The solution was optimized.

[0044] This design involves acquiring real-world land use data to construct a comparison set, using the Euclidean distance formula to calculate the deviation between candidate schemes and reality, incorporating this deviation into the generator's loss function, and iteratively optimizing the scheme. The real-world comparison dataset provides a true reference, and the Euclidean distance formula accurately quantifies the gap between the scheme and reality. By incorporating the deviation value into the loss function, the generator can adjust according to real-world conditions during training. Through iteration, the deviation from reality is continuously reduced, and the final optimized scheme is closer to the actual situation. This improves the practicality and reliability of the virtual development boundary scheme and provides a more valuable reference for urban planning decisions.

[0045] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0046] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A multi-objective optimization intelligent allocation simulation method for urban development boundary land use, characterized in that, Includes the following steps: S1. Obtain multi-source raw data of the city to be simulated, and preprocess the multi-source raw data; S2. Based on the preprocessed data, divide the urban land use into basic units, extract the feature information of each unit, and construct a basic dataset for urban development boundary configuration. S3. Build a city development scenario generation model with a generative model architecture. Using the basic dataset as input, generate multiple sets of virtual city development boundary schemes through the generation module. At the same time, use the discrimination module to evaluate the rationality and authenticity of the generated schemes to form a model training loop. S4. Introduce a multi-objective optimization index system to conduct a comprehensive evaluation of the generated virtual city development boundary schemes and select candidate schemes that meet the preset requirements. S5. Compare and analyze the candidate schemes with the current status data of urban land use, calculate the deviation between the schemes and the actual situation, adjust the parameters of the generative model based on the deviation feedback, and iteratively optimize the virtual development boundary scheme. S6. Output the optimized urban development boundary land use configuration scheme and generate the corresponding visual simulation scene.

2. The method according to claim 1, characterized in that, In step S1: The multi-source raw data includes, but is not limited to, remote sensing image data, land use status data, topographic elevation data, transportation network data, and population and economic statistics. The preprocessing operations include data cleaning, standardization, and spatial alignment. Data cleaning is used to remove outliers and redundant data, standardization is used to unify the value range of numerical data, and spatial alignment is used to ensure the consistency of various types of data in spatial location.

3. The method according to claim 1, characterized in that, In step S2: The urban land use basic unit is divided using a grid division method, and the unit size is flexibly set according to the scale of the city to be simulated and the simulation accuracy requirements. The feature information includes land use attribute features and spatial association features. The land use attribute features include, but are not limited to, land use type, plot ratio, building density and ecological sensitivity index. The spatial association features include, but are not limited to, spatial adjacency relationship, transportation accessibility and resource sharing degree. The basic dataset is formed by integrating the comprehensive feature vectors of each land use basic unit. The comprehensive feature vectors are obtained by fusing the land use attribute feature vectors and the spatial correlation feature vectors.

4. The method according to claim 1, characterized in that, In step S3: The generative model is a generative adversarial network, and the generation module is a generator that adopts an encoder-decoder architecture. By transforming and reconstructing the input features, it outputs a virtual city development boundary scheme vector. The discrimination module is a discriminator that adopts a deep learning network architecture. Its input is a real urban development boundary data sample or a virtual scheme sample output by the generator, and its output is the probability value that the sample belongs to the real data. The model training loop is implemented by alternately training the generator and the discriminator. During the training process, an adversarial loss function is set to minimize the discriminator's ability to recognize virtual schemes and maximize the discriminator's ability to distinguish between real and virtual samples, until the model converges.

5. The method according to claim 4, characterized in that: The generator loss function is used to minimize the discriminator's ability to recognize virtual schemes, and its expression is: ； in Configure a basic dataset for urban development boundaries. For the first The comprehensive feature vector of each land use basic unit, This is the output function of the generator. The output function of the discriminator. For expectation operators; The discriminator loss function is used to maximize the ability to distinguish between real and virtual samples, and its expression is: ； in To develop boundary data samples for real cities, Develop a dataset of real-world city boundaries.

6. The method according to claim 1, characterized in that, In step S4: The multi-objective optimization index system includes ecological protection indicators, resource utilization efficiency indicators, and economic development indicators. The ecological protection indicators are calculated based on the proportion of ecological land within the development boundary and the avoidance rate of sensitive ecological areas; the resource utilization efficiency indicators are calculated based on land development intensity and infrastructure matching rate; and the economic development indicators are calculated based on the GDP output potential and the predicted number of jobs within the development boundary. The comprehensive evaluation is achieved by calculating the comprehensive evaluation score of the virtual development boundary scheme. The comprehensive evaluation score is obtained by weighted summation of the standardized values ​​of each indicator and their corresponding weights. Schemes with a comprehensive evaluation score greater than or equal to a preset threshold are selected as candidate schemes.

7. The method according to claim 6, characterized in that: The weights of each indicator are determined by the analytic hierarchy process, and the sum of the weights of all indicators is 1. The standardization process is used to eliminate the influence of different dimensions between indicators. The expression for the comprehensive evaluation score is: ； in To optimize the number of indicators in a multi-objective indicator system, For the first The weight of each indicator, For the first The standardized value of each indicator.

8. The method according to claim 1, characterized in that, In step S5: The real-world urban land use status data includes the spatial range of the real development boundary, land use structure and indicator parameters. A real-world comparison dataset is constructed based on the real-world urban land use status data. The deviation value is calculated using a distance calculation formula, which includes the Euclidean distance formula. The deviation value is used as a feedback signal to introduce into the loss function of the generation module, update the parameters of the generative model and retrain the model, iteratively generate new virtual development boundary schemes until the calculated deviation value is less than the preset deviation threshold.

9. The method according to claim 8, characterized in that, The expression for the Euclidean distance formula is: ； in, The number of basic urban land use units. Dimensions for configuring parameters for development boundaries. For the virtual development boundary scheme, the first The land use basic unit in the first The values ​​that can be obtained under each configuration parameter For the real-world comparison dataset, the first The land use basic unit in the first The values ​​that can be obtained under each configuration parameter; The updated generator module loss function expression is as follows: ； in This is the deviation penalty coefficient.

10. The method according to any one of claims 1-9, characterized in that, In step S1, the standardization process employs the max-min standardization method, expressed as: ； in For the first The first type of multi-source raw data One sample, , The first Minimum and maximum values ​​of multi-source raw data; In step S2, the spatial correlation characteristics of each land use basic unit are modeled using a graph neural network to generate spatial correlation feature vectors.

Images