Multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II

Abstract

The present invention provides a multi-objective optimization method for university campus planning and design based on cGAN and NSGA‑II, and the method specifically includes: obtaining a large amount of existing different campus planning and design related data, and associating the corresponding carbon emission and incremental cost data to form a model data set; inputting the model data set into a pre-built cGAN model for training, and optimizing the initial campus planning and design scheme output by the cGAN model through the NSGA‑II algorithm; optimizing and adjusting the cGAN model based on the target campus planning and design, and visualizing the scheme. The collaborative optimization model of carbon emission level and incremental cost of university campuses constructed by the present invention provides scientific decision-making support for architects to fully consider the balance between carbon emission reduction performance and economic benefits in the scheme stage, which is helpful to further realize the "campus carbon neutrality" construction plan.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II. Background Art

[0002] As an urban planning unit that integrates teaching, scientific research, production and life and has a highly dense population, the carbon neutrality experience of university campuses provides an important reference for the development of low-carbon cities. In the process of exploring the "carbon neutrality" path of universities, how to achieve a balance between carbon emission performance and economic benefits has become a key issue.

[0003] However, pursuing a balance between emission reduction measures and controlling incremental costs is a complex and arduous task facing campus planning and design. Traditional low-carbon campus planning and design methods mainly rely on the personal experience of architects and simple mathematical models. These methods are somewhat lacking in scientificity when dealing with complex campus planning issues. They are difficult to take into account both emission reduction performance and economic efficiency, two somewhat contradictory goals, at the same time, resulting in unsatisfactory planning and design results. In addition, it is difficult to efficiently generate campus planning and design plans using traditional methods, and architects cannot intuitively and quickly pre-evaluate the effects of the plans.

[0004] In order to overcome the limitations of traditional methods, some multi-objective optimization methods, such as genetic algorithms and particle swarm optimization, have been tried in the field of campus planning. These methods have improved the scientific nature of planning and design to a certain extent, but there are still many shortcomings. For example, they often find it difficult to fully explore the complex solution space, resulting in the optimization results may not be the global optimal solution. At the same time, these methods lack the effective use of problem-specific domain knowledge, resulting in the results may not meet the actual needs of campus planning. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-objective optimization method and system for university campus planning and design based on cGAN and NSGA-II, which not only synergistically optimizes the carbon emission level and incremental cost of university campuses, but also can quickly generate intuitive campus planning and design plans, provide scientific decision-making support for architects in the planning stage, and help further realize the "campus carbon neutrality" construction plan to solve at least one of the above-mentioned prior art problems.

[0006] In a first aspect, the present invention provides a multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II, and the method specifically includes:

[0007] Acquire existing campus planning and design data, as well as carbon emission level index data and incremental cost data under different campus plans, and associate the existing campus planning and design data, the carbon emission level index data and the incremental cost data to form a model data set;

[0008] The model data set is input into a pre-built cGAN model for training, and the initial campus planning and design scheme output by the cGAN model is optimized by the NSGA-II algorithm to obtain a target campus planning and design scheme;

[0009] Based on the target campus planning and design scheme, the cGAN model is optimized and adjusted, and the target campus planning and design scheme is visualized.

[0010] Furthermore, the cGAN model includes a multi-scale generator, a multi-task discriminator and a loss function;

[0011] The multi-scale generator is used to generate a number of initial campus planning and design schemes according to the model data set;

[0012] The multi-task discriminator is used to evaluate the authenticity and consistency of several initial campus planning and design schemes;

[0013] The loss function is used to evaluate the loss of the cGAN model during the model training process.

[0014] Further, the multi-scale generator includes a first input layer, an encoder layer, a self-attention mechanism, a decoder layer and a first output layer;

[0015] The first input layer is used to input the model data set;

[0016] The encoder layer is used to extract features of the campus planning and design image of the model data set using multiple convolutional layers;

[0017] The self-attention mechanism is used to enhance the ability to capture features of different scales;

[0018] The decoder layer is used to restore the spatial dimension of the planning design feature map output by the encoder layer using transposed convolution;

[0019] The first output layer is used to output a planning design feature map.

[0020] Further, the multi-task discriminator includes a main discriminator, a first auxiliary discriminator and a second auxiliary discriminator;

[0021] The main discriminator is used to evaluate the overall authenticity of the planning design feature map;

[0022] The first auxiliary discriminator is used to evaluate whether the carbon emission level of the planning design characteristic diagram meets the expected target;

[0023] The second auxiliary discriminator is used to evaluate whether the cost-effectiveness of the planning design feature map meets the expected goal.

[0024] In a second aspect, the present invention provides a multi-objective optimization system for university campus planning and design based on cGAN and NSGA-II, and the system specifically includes:

[0025] The first planning and design optimization module is used to obtain existing campus planning and design data, as well as carbon emission level index data and incremental cost data under different campus plans, and associate the existing campus planning and design data, the carbon emission level index data and the incremental cost data to form a model data set;

[0026] The second planning and design optimization module is used to input the model data set into the pre-built cGAN model for training, and optimize the initial campus planning and design scheme output by the cGAN model through the NSGA-II algorithm to obtain the target campus planning and design scheme;

[0027] The third planning and design optimization module is used to optimize and adjust the cGAN model based on the target campus planning and design scheme, and visualize the target campus planning and design scheme.

[0028] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor and a computer program stored in the memory. When the computer program is executed on the processor, a multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II as described in any one of the above methods is implemented.

[0029] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II as described in any one of the above methods is implemented.

[0030] Compared with the prior art, the present invention has at least one of the following technical effects:

[0031] 1. This paper constructs a new multi-objective optimization framework by cleverly combining conditional generative adversarial networks (cGAN) and non-dominated sorting genetic algorithm II (NSGA-II). It not only synergistically optimizes the carbon emission level and incremental cost of university campuses, but also quickly generates intuitive campus planning and design solutions, providing scientific decision-making support for architects in the planning stage.

[0032] 2. The present invention designs a cGAN model that can generate high-quality and diversified campus planning and design solutions, and effectively integrates it with the NSGA-II algorithm. Through an innovative two-way feedback mechanism, the co-evolution of the cGAN generation solution and the NSGA-II optimization process is achieved, thereby significantly improving the optimization effect.

[0033] 3. The present invention is committed to developing a complete multi-objective optimization method for campus planning and design, covering the entire process from data processing, algorithm design, model training to result visualization, providing a full range of intelligent support tools for university campus planning and design decision-making.

[0034] 4. The cGAN structure of multi-scale generator and multi-task discriminator is adopted, so that the generated planning and design scheme can more realistically reflect the complex campus environment, and effectively capture the planning and design features of different scales during the generation process. Compared with the traditional generation model, it can generate more diverse and high-quality planning and design schemes.

[0035] 5. The present invention uses the adaptive crossover and mutation mechanism of NSGA-II, combined with dynamic reference point adjustment, to respond to the diversity changes of the population in real time, thereby improving the adaptability and search efficiency of the algorithm.

[0036] 6. By establishing a two-way feedback mechanism between cGAN and NSGA-II, the optimization results can be used to guide the improvement of the generation model, forming a closed-loop optimization, which helps to continuously improve the quality of the solution. However, the existing technology often ignores this information flow.

[0037] 7. The present invention adopts a multi-stage strategy from coarse optimization to fine optimization, which can focus on solving specific problems at different stages. Compared with the single-stage optimization method of the prior art, it can more effectively explore the solution space and improve the optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of a multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II provided by an embodiment of the present invention;

[0039] Figure 2 is a schematic diagram of the structure of a cGAN model provided by an embodiment of the present invention;

[0040] Figure 3 It is a structural schematic diagram of a multi-objective optimization system for university campus planning and design based on cGAN and NSGA-II provided by an embodiment of the present invention;

[0041] Figure 4 It is a structural schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0043] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0044] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0045] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0046] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0047] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0048] In the embodiment of the present application, the execution subject of the process includes a terminal device, which includes but is not limited to: a server, a computer, a smart phone, a tablet computer, and other devices capable of executing the method disclosed in the present application. Figure 1 A flow chart of a multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II disclosed in an embodiment of the present invention is shown, and the details are as follows:

[0049] S101, obtaining existing campus planning and design data, as well as carbon emission level indicator data and incremental cost data under different campus planning, and associating the existing campus planning and design data, the carbon emission level indicator data and the incremental cost data to form a model data set.

[0050] In this embodiment, existing campus planning and design data are collected to obtain data related to carbon emission levels and costs (such as carbon emission level indicators and incremental costs under different plans), including: roads, such as road width and road grid (carbon emissions per unit length of transportation) (cost per unit area of ​​transportation); landscape, such as green area, shape and location (carbon absorption per unit area of ​​green space) (cost per unit area of ​​landscape); water area, shape and location (cost per unit area of ​​water body); buildings, such as floor area ratio, height, location and shape (construction cost per unit area) (carbon emissions per unit area of ​​building / year); transportation, such as station location and service facility location (transportation station per unit area, cost of service facilities).

[0051] The collected data is cleaned to remove noise and outliers, ensure the accuracy and consistency of the data, and normalize the data to facilitate model training and improve the convergence speed of the algorithm.

[0052] In this embodiment, the model data set formed by associating the existing campus planning and design data, carbon emission level index data and incremental cost data contains rich information, which helps the cGAN model to generate more diverse initial campus planning and design schemes. At the same time, these data also provide a more comprehensive optimization target for the NSGA-II algorithm, so that the final target campus layout scheme is more in line with actual needs.

[0053] S102, inputting the model data set into a pre-built cGAN model for training, and optimizing the initial campus planning and design scheme output by the cGAN model through the NSGA-II algorithm to obtain a target campus planning and design scheme.

[0054] In this embodiment, the pre-constructed cGAN model is trained using the pre-processed model data set. During the training process, the cGAN model learns how to generate a campus planning and design scheme that meets these conditions according to given conditions (such as carbon emission limits, construction cost budgets, etc.). During the training process, by continuously adjusting the model parameters, the generated campus planning and design scheme gradually approaches the real and optimized layout. After the training is completed, the cGAN model can generate multiple initial campus planning and design schemes according to the given conditions. These schemes meet the preset constraints in terms of carbon emission levels, incremental costs, etc.

[0055] The initial campus planning and design scheme generated by the cGAN model is used as the input of the NSGA-II algorithm. The NSGA-II algorithm performs multi-objective optimization on these schemes, considering multiple indicators such as carbon emissions, incremental costs, and rationality of campus functional layout. Through the selection, crossover, mutation and other operations in the genetic algorithm, continuous iterative optimization is performed to finally obtain a set of target campus planning and design schemes. The target campus planning and design scheme output by the NSGA-II algorithm is evaluated, including carbon emissions, incremental costs, campus environment, and rationality of functional layout. According to the evaluation results, the optimal campus planning and design scheme is selected as the final implementation plan.

[0056] In this embodiment, the cGAN model is used to quickly generate multiple initial campus planning and design schemes, which greatly shortens the planning cycle. The initial scheme generated by the cGAN model has met the preset conditions in terms of carbon emission level, incremental cost, etc., providing a good basis for optimization. The NSGA-II algorithm performs multi-objective optimization on the initial scheme, further improving the planning efficiency. The NSGA-II algorithm uses multi-objective optimization to make the final campus planning and design scheme achieve optimal or near-optimal results in many aspects.

[0057] S103, based on the target campus planning and design scheme, optimizing and adjusting the cGAN model, and visually displaying the target campus planning and design scheme.

[0058] In this embodiment, a feedback mechanism is established between cGAN and NSGA-II, and the samples generated by cGAN are used for the optimization of NSGA-II, and the generation strategy of cGAN is adjusted according to the optimization results of NSGA-II (the target value of the optimal solution). After each iteration, the optimization results of NSGA-II are used as conditional information learning feedback to cGAN. This mechanism realizes the organic combination of optimization and generation by feeding back the optimization results to the generation model to guide its continuous improvement. At the same time, a multi-stage optimization strategy is adopted, and the gradual process from rough optimization to fine optimization ensures that the quality and adaptability of the solution can be continuously improved at each stage. Through this systematic feedback and optimization cycle, the model can quickly adapt to changing needs, improve the practicality and advantages of the generated layout, and ultimately achieve more efficient design and optimization results. In each generation, cGAN is executed to generate layouts, and the intermediate results are periodically evaluated and the parameters of cGAN and NSGA-II are continuously adjusted to optimize the generation effect.

[0059] By generating a Pareto front diagram, the trade-off between carbon emission levels and incremental costs of different options is displayed, which helps to analyze the pros and cons of each option. Detailed planning maps are generated for typical options on the Pareto front, including the location, shape and height of buildings, road networks, green space and water distribution, and the location of transportation stations and service facilities. The generated campus layout is visualized to facilitate further analysis by designers.

[0060] In some embodiments, reference Figure 2 , the cGAN model includes a multi-scale generator, a multi-task discriminator and a loss function;

[0061] The multi-scale generator is used to generate a number of initial campus planning and design schemes according to the model data set;

[0062] The multi-task discriminator is used to evaluate the authenticity and consistency of several initial campus planning and design schemes;

[0063] The loss function is used to evaluate the loss of the cGAN model during the model training process.

[0064] In this embodiment, in the improvement of cGAN, a multi-scale generator is used, combined with the U-Net structure and the attention mechanism, so that the network can more effectively capture the different scale features of the layout, thereby generating a more detailed and layered campus plan. In addition, the conditional self-attention embedding technology is used to optimize the utilization efficiency of conditional information in the generation process. This allows the generated layout to not only meet basic design requirements, but also better adapt to specific conditions and constraints.

[0065] A multi-task discriminator is also introduced, which can simultaneously evaluate the authenticity, emission reduction and cost-effectiveness of the generated planning design. This design not only improves the comprehensive performance of planning scheme generation, but also ensures the feasibility and optimization effect of the generated planning scheme under multiple objectives.

[0066] Further, the multi-scale generator includes a first input layer, an encoder layer, a self-attention mechanism, a decoder layer and a first output layer;

[0067] The first input layer is used to input the model data set;

[0068] The encoder layer is used to extract features of the campus planning and design image of the model data set using multiple convolutional layers;

[0069] The self-attention mechanism is used to enhance the ability to capture features of different scales;

[0070] The decoder layer is used to restore the spatial dimension of the planning design feature map output by the encoder layer using transposed convolution;

[0071] The first output layer is used to output a planning design feature map.

[0072] In this embodiment, the multi-scale generator adopts a U-Net structure combined with a self-attention mechanism to enhance the ability to capture features of different scales. The input of the multi-scale generator includes random noise and conditional information (including layout type, carbon emission targets, cost constraints, etc.).

[0073] In the first input layer, the input is random noise and condition information The input conditional information is converted into a vector of the same dimension as the random noise through the fully connected layer, and concatenated with the random noise to form a new input .

[0074] In the encoder layer, multiple convolutional layers are used for feature extraction. Assume that the size of the input feature map is , the number of channels is C, then the first layer of convolution operation is as follows:

[0075]

[0076] represents the output feature map after the first convolution operation, is the random noise after conditional information embedding, and its size is , this layer will generate 64 different feature maps (i.e. 64 convolution kernels), each of which has a size of 4×4. The convolution kernel slides on the input feature map with a stride of 2 to reduce the spatial dimension of the feature map; zeros are padded at the edges of the input feature map to ensure that the size of the output feature map is the same as the input feature map.

[0077] Increase the number of convolution kernels layer by layer, reduce the size of the feature map, Leaky ReLU ( ) activation function further improves the nonlinear expression ability of the model.

[0078] In the self-attention mechanism, the self-attention mechanism is used after some convolutional layers in the form of:

[0079]

[0080] in, They are query, key, and value representations, respectively. express The transpose of is the dimension of the key vector.

[0081] In the decoder layer part, upsampling layers (transposed convolution) are used to gradually restore the spatial dimensions of the feature maps.

[0082] When decoding each layer, skip connections are used to concatenate the corresponding encoder layer features with the decoder layer features to preserve spatial information.

[0083] In the first output layer, the planning design feature map generated by a convolutional layer output is assumed to be G(z,c):

[0084]

[0085] The output feature map size is , G(z,c) represents the generated planning design feature graph.

[0086] Further, the multi-task discriminator includes a main discriminator, a first auxiliary discriminator and a second auxiliary discriminator;

[0087] The main discriminator is used to evaluate the overall authenticity of the planning design feature map;

[0088] The first auxiliary discriminator is used to evaluate whether the carbon emission level of the planning design characteristic diagram meets the expected target;

[0089] The second auxiliary discriminator is used to evaluate whether the cost-effectiveness of the planning design feature map meets the expected goal.

[0090] In this embodiment, a main discriminator is designed to evaluate the authenticity of the overall layout, and two auxiliary discriminators are set up to evaluate the carbon emission level and cost-effectiveness respectively. By introducing the design of the main discriminator and two auxiliary discriminators, the multi-task discriminator can comprehensively evaluate the generated campus planning and design scheme, and on the basis of ensuring the overall authenticity, it is refined to the evaluation of carbon emission level and cost-effectiveness. This design not only improves the reliability of the generated model, but also enhances the model's adaptability to complex needs.

[0091] Conditional embedding mechanism: implements the self-attention mechanism, optimizes the utilization efficiency of conditional information, and injects conditional information into each layer of the multi-scale generator.

[0092] The multi-task discriminator combines the input planning design feature map G(z,c) and conditional information c (conditional information includes target carbon emissions, budget constraints, etc.) to perform multi-task learning.

[0093] In the main discriminator, the input feature map and conditional information first pass through a convolutional layer to extract features using multiple convolution kernels.

[0094]

[0095] After a certain layer of the main discriminator, the conditional information is embedded into the model and fused with the convolutional features through addition or concatenation operations.

[0096] Continue to use multiple convolutional layers to extract deeper features and apply the Leaky ReLU activation function.

[0097]

[0098] Finally, a probability value D(G(z,c),c) is output through a fully connected layer to judge the authenticity of the generated planning and design scheme.

[0099]

[0100] The inputs of the two auxiliary discriminators are also the generated planning design feature map G(z,c) and conditional information c, and the structure is similar to that of the main discriminator.

[0101] Construct convolutional layers and feature fusion:

[0102]

[0103]

[0104] Fusion condition information, extraction of features:

[0105]

[0106]

[0107] Output layer:

[0108]

[0109]

[0110] Carbon emissions for output assessment, Cost effectiveness for output evaluation.

[0111] Furthermore, the loss function includes an adversarial loss function, a reconstruction loss function, a performance loss function and a constraint loss function;

[0112] The adversarial loss function is used to ensure that the multi-scale generator can generate a planning design feature map that is considered to be true by the multi-task discriminator;

[0113] The reconstruction loss function is used to ensure that the planning design feature map generated by the multi-scale generator is consistent with the real sample;

[0114] The performance loss function is used to measure the performance loss between the carbon emission level and cost-effectiveness of the planning design feature map generated by the multi-scale generator;

[0115] The constraint loss function is used to ensure that the planning design feature map generated by the multi-scale generator meets the planning requirements.

[0116] In this embodiment, in order to effectively train cGAN, a comprehensive custom loss function is required, which combines adversarial loss, reconstruction loss, performance loss and constraint loss to ensure that the generated campus planning design plan meets authenticity, input consistency, performance standards and planning requirements at the same time.

[0117] Specifically, the adversarial loss function includes a generator loss function and a discriminator loss function; the generator loss function is , the discriminator loss function is ,in, Indicates conditional information. represents random noise, represents the real sample, Represents the random noise distribution Next The expected calculation of Represents the real data distribution The following input The expected calculation of represents the planning design feature map generated by the multi-scale generator, Represents the true probability value output by the multi-task discriminator for judging the planning design feature map;

[0118] The reconstruction loss function is , where represents the generated sample;

[0119] The performance loss function satisfies

[0120]

[0121]

[0122]

[0123] Among them, it means represents the performance loss function, represents the carbon emission loss function, represents the incremental cost loss function, is the target carbon emissions, is the actual carbon emissions generated by the planning and design solution. is the actual cost of the generated planning and design solution, is the budget ceiling. and represents the weight coefficient;

[0124] The constraint loss function satisfies

[0125]

[0126]

[0127] in, represents the constraint loss function, represents the penalty loss function, Indicates the need to satisfy Constraints, It is the first constraints.

[0128] In this embodiment, the adversarial loss ensures that the generator is able to generate samples that the discriminator considers to be real. The adversarial loss in the standard GAN is as follows:

[0129] Generator loss:

[0130]

[0131] Discriminator loss:

[0132]

[0133] Represents the random noise distribution Next Expected calculation of ; Represents the real data distribution The following input Expected calculation.

[0134] The reconstruction loss ensures that the generated layout is consistent with the input condition. The L1 loss (mean absolute error) is used to measure the generated G(z,c) and the real sample The consistency between them.

[0135]

[0136] The performance loss is based on the weighted loss of carbon emissions and incremental costs. The loss of carbon emissions and cost benefits are first defined.

[0137] Carbon emission losses:

[0138]

[0139] in, is the target carbon emissions, It is the actual carbon emissions of generating planning and design solutions.

[0140] Cost loss:

[0141]

[0142] in, is the actual cost of generating the planning and design solution, It's the budget cap.

[0143] Weighted performance loss:

[0144]

[0145] In order to ensure that the generated planning and design scheme meets various planning requirements, constraint losses can be designed. Due to the existence of constraints (minimum green area, building spacing, etc.), constraint losses can be defined as follows:

[0146]

[0147] Function: When a constraint is not met, the penalty loss can be defined as

[0148]

[0149] in, Indicates the need to satisfy Constraints, It is the first constraints.

[0150] Finally, combining all the above losses, we define the comprehensive loss function for:

[0151]

[0152] In some embodiments, in the above step S102, the initial campus planning and design scheme output by the cGAN model is optimized by the NSGA-II algorithm to obtain the target campus planning and design scheme, specifically including:

[0153] Setting the number of individuals in the initial population, each individual in the initial population represents a parameter vector of an initial campus planning and design scheme;

[0154] By calculating the dominance relationship between every two individuals in the initial population, dividing the initial population into multiple levels;

[0155] The initial population divided into multiple levels is subjected to selection, crossover and mutation operations, local search is performed through simulated annealing algorithm, and the reference point is dynamically adjusted according to the distribution of the current Pareto front to generate the target campus planning and design plan.

[0156] In this embodiment, the basic structure of NSGA-II is constructed, including population initialization, non-dominated sorting and congestion calculation.

[0157] The size of the initial population is set as N, and each individual can be represented as a parameter vector of campus planning, including building location, area, green area, etc.

[0158] Non-dominant hierarchy: The population is divided into multiple hierarchies by calculating the dominance relationship between each pair of individuals.

[0159] Dominance relationship: Individual Dominant individual B

[0160] Calculate the domination number (the number of individuals dominated) and crowding degree (the distance to neighboring individuals) of each individual.

[0161] Selection mechanism: Using tournament selection, individuals are selected from each non-dominated layer for crossover and mutation.

[0162] Crowding degree comparison: When selecting, individuals with higher crowding degrees are given priority to improve population diversity.

[0163] Dynamically adjust crossover and mutation probabilities according to population diversity to improve solution quality and search efficiency.

[0164] Adaptive crossover: Dynamically adjust the crossover rate according to the fitness of the current population. It can be expressed as:

[0165]

[0166] in, is the basic crossover probability, is the fitness adjustment factor, is the fitness function.

[0167] Adaptive mutation: mutation rate It is also dynamically adjusted according to the fitness, using uniform mutation as the mutation strategy. The mutation rate calculation formula is expressed as:

[0168]

[0169] After each generation of evolution, local search is applied to the best individuals to improve the quality of the solution. A simulated annealing algorithm is added to perform a detailed search near the current solution.

[0170] As the optimization process progresses, the reference point is dynamically adjusted according to the distribution of the current Pareto front to improve the coverage of the target space. The reference point update formula is as follows:

[0171]

[0172] in, , are the old and new reference points, respectively. is the update factor, is the target value set of the current Pareto front.

[0173] The design architecture of NSGA-II combines key technologies such as non-dominated sorting, crossover and mutation operations, local search, and dynamic reference point adjustment to provide an effective multi-objective optimization solution. After integration with cGAN, it can generate campus planning and design solutions while optimizing multiple objectives such as carbon emission levels and costs, providing an innovative solution for sustainable development.

[0174] The improved NSGA-II optimization pseudo code is shown in Table 1.

[0175] Table 1

[0176]

[0177] In the improvement of NSGA-II, firstly, adaptive evolution parameters are introduced to dynamically adjust the crossover and mutation probabilities according to the diversity of the population. This innovation significantly improves the adaptability and efficiency of the algorithm, enabling it to operate the population more effectively at different stages of evolution. Secondly, the quality of the solution is further improved by integrating the local search strategy. This strategy ensures the superiority of the final solution and enhances the exploration ability of the algorithm. In addition, the dynamic reference point adjustment mechanism is incorporated into the optimization process of the target space to improve the distribution and coverage of the solution, so that the generated solution can more comprehensively cover different areas in the target space, thereby improving the overall optimization effect.

[0178] Reference Figure 3 An embodiment of the present invention provides a multi-objective optimization system 3 for university campus planning and design based on cGAN and NSGA-II, and the system 3 specifically includes:

[0179] The first planning and design optimization module 301 is used to obtain existing campus planning and design data, as well as carbon emission level index data and incremental cost data under different campus plans, and associate the existing campus planning and design data, the carbon emission level index data and the incremental cost data to form a model data set;

[0180] The second planning and design optimization module 302 is used to input the model data set into the pre-built cGAN model for training, and optimize the initial campus planning and design scheme output by the cGAN model through the NSGA-II algorithm to obtain the target campus planning and design scheme;

[0181] The third planning and design optimization module 303 is used to optimize and adjust the cGAN model based on the target campus planning and design scheme, and visualize the target campus planning and design scheme.

[0182] It is understandable that if Figure 1 The contents of the embodiment of the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II are applicable to the embodiment of the multi-objective optimization system for university campus planning and design based on cGAN and NSGA-II. The functions specifically implemented by the embodiment of the multi-objective optimization system for university campus planning and design based on cGAN and NSGA-II are the same as those in the embodiment of the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II. Figure 1 The multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II is the same as the embodiment shown in FIG. Figure 1 The beneficial effects achieved by the embodiment of the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II are also the same.

[0183] It should be noted that the information interaction, execution process and other contents between the above-mentioned systems are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0184] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0185] Reference Figure 4 The embodiment of the present invention further provides a computer device 4, comprising: a memory 402 and a processor 401 and a computer program 403 stored in the memory 402. When the computer program 403 is executed on the processor 401, the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II as described in any one of the above methods is implemented.

[0186] The computer device 4 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 4 may include, but is not limited to, a processor 401 and a memory 402. Those skilled in the art will appreciate that Figure 4 It is only an example of the computer device 4 and does not constitute a limitation on the computer device 4. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0187] The processor 401 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 gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0188] In some embodiments, the memory 402 may be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 402 may also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 4. Further, the memory 402 may also include both an internal storage unit and an external storage device of the computer device 4. The memory 402 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 402 may also be used to temporarily store data that has been output or is to be output.

[0189] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II as described in any one of the above methods is implemented.

[0190] In this embodiment, if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the camera / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, USB flash drive, mobile hard disk, disk or optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0191] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0192] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0193] In the embodiments disclosed in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0194] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Claims

1. A multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II, characterized by: The method specifically comprises: Acquire existing campus planning and design data, as well as carbon emission level index data and incremental cost data under different campus plans, and associate the existing campus planning and design data, the carbon emission level index data and the incremental cost data to form a model data set; The model data set is input into a pre-built cGAN model for training, and the initial campus planning and design scheme output by the cGAN model is optimized by the NSGA-II algorithm to obtain a target campus planning and design scheme; Based on the target campus planning and design scheme, the cGAN model is optimized and adjusted, and the target campus planning and design scheme is visualized; The cGAN model includes a multi-scale generator, a multi-task discriminator and a loss function; the loss function includes an adversarial loss function, a reconstruction loss function, a performance loss function and a constraint loss function; the adversarial loss function includes a generator loss function and a discriminator loss function; the generator loss function is , the discriminator loss function is ,in, Indicates conditional information. represents random noise, represents the real sample, Represents the random noise distribution Next The expected calculation of Represents the real data distribution The following input The expected calculation of represents the planning design feature map generated by the multi-scale generator, Represents the true probability value output by the multi-task discriminator for judging the planning design feature map; The reconstruction loss function is , The performance loss function satisfies: Among them, it means represents the performance loss function, represents the carbon emission loss function, represents the incremental cost loss function, is the target carbon emissions, is the actual carbon emissions generated by the planning and design solution. is the actual cost of the generated planning and design solution, is the budget ceiling. and represents the weight coefficient; The constraint loss function satisfies: in, represents the constraint loss function, represents the penalty loss function, Indicates the need to satisfy Constraints, It is the first constraints.

2. The multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II according to claim 1 is characterized in that: The multi-scale generator is used to generate a number of initial campus planning and design schemes according to the model data set; The multi-task discriminator is used to evaluate the authenticity and consistency of several initial campus planning and design schemes; The loss function is used to evaluate the loss of the cGAN model during the model training process.

3. The multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II according to claim 2 is characterized in that: The multi-scale generator includes a first input layer, an encoder layer, a self-attention mechanism, a decoder layer and a first output layer; The first input layer is used to input the model data set; The encoder layer is used to extract features of the campus planning and design image of the model data set using multiple convolutional layers; The self-attention mechanism is used to enhance the ability to capture features of different scales; The decoder layer is used to restore the spatial dimension of the planning design feature map output by the encoder layer using transposed convolution; The first output layer is used to output a planning design feature map.

4. The multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II according to claim 3 is characterized in that: The multi-task discriminator includes a main discriminator, a first auxiliary discriminator and a second auxiliary discriminator; The main discriminator is used to evaluate the overall authenticity of the planning design feature map; The first auxiliary discriminator is used to evaluate whether the carbon emission level of the planning design characteristic diagram meets the expected target; The second auxiliary discriminator is used to evaluate whether the cost-effectiveness of the planning design feature map meets the expected goal.

5. The multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II according to claim 4 is characterized in that: The adversarial loss function is used to ensure that the multi-scale generator can generate a planning design feature map that is considered to be true by the multi-task discriminator; The reconstruction loss function is used to ensure that the planning design feature map generated by the multi-scale generator is consistent with the real sample; The performance loss function is used to measure the performance loss between the carbon emission level and cost-effectiveness of the planning design feature map generated by the multi-scale generator; The constraint loss function is used to ensure that the planning design feature map generated by the multi-scale generator meets the planning requirements.

6. The multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II according to any one of claims 1 to 5, characterized in that: The initial campus planning and design scheme output by the cGAN model is optimized by the NSGA-II algorithm to obtain the target campus planning and design scheme, specifically including: Setting the number of individuals in the initial population, each individual in the initial population represents a parameter vector of an initial campus planning and design scheme; By calculating the dominance relationship between every two individuals in the initial population, dividing the initial population into multiple levels; The initial population divided into multiple levels is subjected to selection, crossover and mutation operations, local search is performed through simulated annealing algorithm, and the reference point is dynamically adjusted according to the distribution of the current Pareto front to generate the target campus planning and design plan.

7. A system for implementing the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II as described in any one of claims 1 to 6, characterized in that: The system specifically comprises: The first planning and design optimization module is used to obtain existing campus planning and design data, as well as carbon emission index data and incremental cost data under different campus plans, and associate the existing campus planning and design data, the carbon emission level index data and the incremental cost data to form a model data set; The second planning and design optimization module is used to input the model data set into the pre-built cGAN model for training, and optimize the initial campus planning and design scheme output by the cGAN model through the NSGA-II algorithm to obtain the target campus planning and design scheme; A third planning and design optimization module is used to optimize and adjust the cGAN model based on the target campus planning and design scheme, and visualize the target campus planning and design scheme; The cGAN model includes a multi-scale generator, a multi-task discriminator and a loss function; the loss function includes an adversarial loss function, a reconstruction loss function, a performance loss function and a constraint loss function; the adversarial loss function includes a generator loss function and a discriminator loss function; the generator loss function is , the discriminator loss function is ,in, Indicates conditional information. represents random noise, represents the real sample, Represents the random noise distribution Next The expected calculation of Represents the real data distribution The following input The expected calculation of represents the planning design feature map generated by the multi-scale generator, Represents the true probability value output by the multi-task discriminator for judging the planning design feature map; The reconstruction loss function is , The performance loss function satisfies: Among them, it means represents the performance loss function, represents the carbon emission loss function, represents the incremental cost loss function, is the target carbon emissions, is the actual carbon emissions generated by the planning and design solution. is the actual cost of the generated planning and design solution, is the budget ceiling. and represents the weight coefficient; The constraint loss function satisfies: in, represents the constraint loss function, represents the penalty loss function, Indicates the need to satisfy Constraints, It is the first constraints.

8. A computer device, characterized in that: include: A memory and a processor and a computer program stored in the memory, when the computer program is executed on the processor, implements the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by the processor, the multi-objective optimization method for university campus planning and design based on cGAN and NSGA-II as described in any one of claims 1 to 6 is implemented.

Images