A low-carbon building design decision-making method in hot summer and warm winter areas based on GAN and GA

Abstract

The invention relates to the technical field of building design, and discloses a low-carbon building design decision method for hot-summer and warm-winter areas based on GAN and GA. By setting a generator in a GAN model to hierarchically generate building external design parameters and internal design parameters, introducing a reinforcement learning algorithm, and setting an adaptive discriminator, the genetic algorithm GA is combined with a particle swarm optimization PSO. The method can achieve a balance between energy efficiency and building material costs while meeting low-carbon emission requirements, and fully realize the intelligence and dynamic response capabilities of building design.

Description

Technical Field

[0001] The present invention relates to the technical field of architectural design, and in particular to a low-carbon architectural design decision-making method for hot-summer and warm-winter areas based on GAN and GA. Background Art

[0002] The design of buildings not only needs to consider carbon emissions, but also needs to optimize energy consumption under different energy prices, which invisibly increases the complexity and uncertainty of building design decisions.

[0003] Current low-carbon building design methods mostly rely on the experience of designers and human consideration of factors such as building materials and energy systems. It is difficult to fully consider complex climate factors and energy usage patterns, and lack the ability to dynamically respond to environmental changes. Existing technologies cannot achieve true intelligence, dynamism and global optimization in building design. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a low-carbon building design decision method for hot-summer and warm-winter areas based on GAN and GA to solve the problems mentioned in the background technology, to strike a balance between energy efficiency and building material costs while meeting the requirements of low-carbon emissions, and to realize the intelligence and dynamic response capabilities of building design.

[0005] The present invention discloses a low-carbon building design decision method for hot-summer and warm-winter areas based on GAN and GA, the method comprising:

[0006] Determine initial building design parameters;

[0007] Generate multiple design schemes based on the initial building design parameters through a GAN model; wherein the GAN model is composed of a generator and an adaptive discriminator, including a multi-level structure of multiple generators and adaptive discriminators, the first-level generator is used to generate building appearance design parameters, the second-level generator is used to generate building internal structure design parameters, and the adaptive discriminator performs a preliminary carbon emission performance evaluation on the generated design schemes to obtain a preliminary carbon emission performance evaluation result;

[0008] Based on the design schemes generated by the GAN model and the preliminary evaluation results of carbon emission performance, the genetic algorithm GA is used to perform optimization operations on each design scheme to obtain the target design scheme; wherein the fitness function of the GA is a multi-objective optimization function, and the multi-objectives include carbon emissions, energy efficiency and building material costs;

[0009] The target design scheme is simulated and monitored and evaluated in real time according to external environmental data, and the target design scheme is dynamically adjusted based on the results of the real-time evaluation.

[0010] Furthermore, the initial building design parameters include size, material, energy system, historical meteorological data, and energy usage patterns.

[0011] Further, the building exterior design parameters include the shape of the building facade, material distribution, window layout and building orientation;

[0012] The building internal structure design parameters include room layout, partition wall materials, and internal arrangement of air conditioning and ventilation systems.

[0013] Furthermore, in the process of generating multiple design schemes based on the initial architectural design parameters through the GAN model, a reinforcement learning algorithm is used to learn the process of generating design schemes by the GAN model.

[0014] Furthermore, the reinforcement learning algorithm enables the GAN model to generate a better design solution through a reward mechanism.

[0015] Furthermore, the adaptive discriminator performs a preliminary carbon emission performance evaluation on the generated design solution, specifically including:

[0016] The adaptive discriminator determines the building design complexity level according to the overall building design parameters, and dynamically adjusts the evaluation criteria of the carbon emission performance based on the determined building design complexity level; wherein the overall building design parameters include initial building design parameters, building appearance design parameters and building internal structure design parameters;

[0017] Perform a preliminary carbon emission performance assessment on the generated design scheme based on the adjusted carbon emission performance assessment criteria.

[0018] Furthermore, the training process of the adaptive discriminator includes:

[0019] Collect historical data on completed building designs, including building dimensions, materials, energy systems, climate conditions, energy consumption data, and carbon emissions data;

[0020] Inputting the collected historical data into an initial discriminator for training, the initial discriminator forming an adaptive discriminator by learning the relationship between architectural design parameters of completed buildings and energy consumption and carbon emissions; wherein the architectural design parameters of the completed buildings include building size, materials, energy system and climate conditions;

[0021] Inputting the architectural design parameters of the built building into the GAN model, generating multiple virtual architectural design schemes through the GAN model generator, inputting the generated virtual architectural design schemes into the adaptive discriminator, and judging the rationality of the virtual architectural design schemes through the adaptive discriminator as the first rationality of the virtual architectural design schemes;

[0022] Taking the actual rationality of the completed building design scheme as the standard, marking the rationality of the generated virtual building design scheme, and obtaining the second rationality of the virtual building design scheme;

[0023] The adaptive discriminator is adjusted according to the difference between the first rationality and the second rationality of the virtual building design solution.

[0024] Furthermore, the training process of the adaptive discriminator also includes:

[0025] In the early stage of adaptive discriminator training, a set of initial building design complexity classification rules and evaluation criteria for evaluating carbon emission performance corresponding to different complexity levels are preset; wherein the building design complexity classification operation includes multiplying different overall building design parameters with their corresponding weighting coefficients based on weighting coefficients corresponding to different overall building design parameters, and then performing a summation operation; the initial building design complexity classification rules include an initial weighting coefficient combination;

[0026] The initial weighting coefficient combination is adjusted to obtain different weighting coefficient combinations, and based on the different weighting coefficient combinations, a building design complexity level classification operation is performed on multiple built buildings to determine the complexity levels of the building design parameters of the multiple built buildings, and the carbon emission performance of the multiple built buildings is evaluated according to the evaluation standard for evaluating the carbon emission performance corresponding to the determined complexity level, and the carbon emission performance results of the multiple built buildings obtained by the evaluation are compared with the actual carbon emission performance results of the multiple built buildings to obtain the evaluation accuracy corresponding to the different weighting coefficient combinations;

[0027] The optimal weighting coefficient combination is determined according to the evaluation accuracy corresponding to different weighting coefficient combinations, and the building design complexity level classification rule including the optimal weighting coefficient combination is used as the final building design complexity level classification rule.

[0028] Furthermore, in the process of performing optimization operations on each design scheme through the genetic algorithm GA, the method also includes combining the genetic algorithm GA with particle swarm optimization as a hybrid optimization algorithm, and accelerating the search process in the design scheme parameter space through the hybrid optimization algorithm.

[0029] Furthermore, the simulation monitoring and real-time evaluation of the target design scheme according to the external environment data and the dynamic adjustment of the target design scheme based on the result of the real-time evaluation specifically include:

[0030] Collect and input external environmental data related to building design, the external environmental data including climate data and energy price data;

[0031] Generate a BIM model of the target building corresponding to the target design solution through BIM technology;

[0032] Integrate the BIM model with energy consumption simulation software to simulate the energy consumption and carbon emissions of the target building under different external environmental data;

[0033] Monitor the energy consumption and carbon emission changes of target buildings in real time, and generate feedback data on building design performance during the monitoring process;

[0034] The architectural design parameters of the target design scheme are adjusted based on the feedback data.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] By combining generative adversarial networks (GAN) and genetic algorithms (GA) for multi-objective optimization, carbon emissions, energy efficiency, and material costs can be considered simultaneously, so that the design scheme can achieve the best balance among multiple objectives, and a scheme that better meets the requirements of low-carbon building design can be generated based on the multi-objective optimization mechanism;

[0037] By setting up an adaptive discriminator, the carbon emission assessment standard can be dynamically adjusted according to the complexity level of the building design, ensuring that design schemes of different complexities can be accurately and objectively evaluated, improving the assessment accuracy and making the optimization results more reliable;

[0038] Through the hierarchical generation structure of the generator, more complex design solutions can be generated while maintaining computational efficiency, avoiding the high-dimensional problems and computational complexity brought about by simultaneously dealing with exterior and interior designs in one generator; and by introducing reinforcement learning algorithms, the generation strategy can be continuously improved through the reward mechanism, and learning can be used to generate solutions that better meet the requirements of low-carbon buildings, helping the generator to generate more diverse design solutions that are more in line with actual application needs, solving the problem of the single design of traditional GAN ​​generators.

[0039] Combining GA with particle swarm optimization (PSO), the global search capability of PSO is used to accelerate the optimization process of the design scheme in the entire design space, effectively avoiding the trap of local optimal solutions and improving the search speed of the global optimal design scheme.

[0040] Combining BIM technology and energy consumption simulation, through real-time monitoring and dynamic adjustment of design parameters, we ensure that the performance of the architectural design scheme is always optimal in the actual environmental changes, thereby enhancing the adaptability and practical application effect of the design scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the economic application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0042] Figure 1 The present invention is a flowchart of a low-carbon building design decision method for hot-summer and warm-winter areas based on GAN and GA disclosed in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0044] Embodiment 1

[0045] The present invention discloses a low-carbon building design decision method for hot summer and warm winter areas based on generative adversarial network GAN and genetic algorithm GA. Figure 1 , Figure 1 : is a flow chart of a low-carbon building design decision method for hot summer and warm winter areas based on GAN and GA disclosed in an embodiment of the present invention, the method comprising:

[0046] Determine initial building design parameters.

[0047] Specifically, the initial building design parameters are the basic input parameters for generating the building design scheme, including but not limited to dimensions, materials, energy systems, historical meteorological data, and energy usage patterns. Among them, dimensions may include the height, area, and number of floors of the building; materials may include exterior wall materials, roof materials, floor materials, and the corresponding thermal insulation properties of different materials; energy systems may include heating, cooling, air conditioning, and renewable energy systems; historical meteorological data may include the climate conditions of the area, covering temperature, humidity, wind speed, and sunshine time; energy usage patterns may include energy demand patterns in different seasons or time periods. It can be understood that the above-mentioned initial building design parameters are only illustrative cases of this example, and the present invention is not limited to this.

[0048] Multiple design schemes are generated based on the initial architectural design parameters through the GAN model.

[0049] Among them, the GAN model consists of a generator and an adaptive discriminator, including a multi-level structure of multiple generators and adaptive discriminators. The first-layer generator is used to generate building appearance design parameters, and the second-layer generator is used to generate building internal structure design parameters.

[0050] Specifically, building exterior design parameters refer to design information related to the building exterior, including but not limited to the shape of the building facade, material distribution, window layout and building orientation. These parameters have an important impact on the building's energy efficiency, natural ventilation, lighting and other performance.

[0051] The design parameters of the building's internal structure include but are not limited to room layout, partition materials, and the internal arrangement of air conditioning and ventilation systems, which mainly affect the building's energy efficiency and the comfort of the indoor environment.

[0052] In the embodiment of the present invention, the purpose of generating the building appearance design parameters and the building internal structure design parameters respectively by different generators is to help the model optimize each design aspect more accurately when generating, so that the appearance and internal structure can be optimized for their specific goals respectively. That is, the appearance design pays more attention to the visual effect, external energy consumption performance and environmental adaptability (such as lighting, ventilation, and thermal insulation effects) of the building, and these factors are closely related to the overall shape, facade design, window arrangement and orientation of the building; the internal structure design focuses on the layout, functionality and energy system configuration of the internal space of the building (such as room distribution, partition materials, air conditioning system, etc.), so as to optimize energy efficiency and living comfort.

[0053] The exterior design is generated first, which determines the overall layout and orientation of the building and provides the external framework of the building. Based on this framework, the internal structure design can be adjusted in a more orderly manner. The internal structure design can be refined on the basis of the exterior design to ensure that the internal functional space of the building meets the constraints of the external design (such as the matching of window arrangement with room layout, and the rationality of ventilation and lighting). In other words, there are usually interrelated constraints between the exterior and internal structure of the building. Layered generation helps to avoid design conflicts and ensures the rationality of the design process.

[0054] Through this hierarchical generation structure, the model can generate more complex design solutions while maintaining computational efficiency, avoiding the high-dimensional problems and computational complexity brought by dealing with both exterior and interior design in one generator. Exterior design is mainly aimed at optimizing energy efficiency in interaction with the environment, such as the impact of exterior wall materials and window layout on lighting and natural ventilation, while interior structure design needs to optimize the functionality and energy efficiency of the building, such as the impact of the layout of air conditioning and ventilation systems on the energy consumption of the interior space. That is, the hierarchical generation structure enables the model to independently optimize these two types of design goals, so that the design of each part is more in line with its specific target requirements.

[0055] In an embodiment of the present invention, as a preferred implementation, in the process of generating multiple design solutions based on the initial architectural design parameters by using the GAN model, a reinforcement learning algorithm is used to learn the process of the GAN model generating design solutions. The reinforcement learning algorithm enables the GAN model to generate a better design solution through a reward mechanism.

[0056] The traditional GAN ​​generator generates data through adversarial training with the discriminator. The goal of the generator is to deceive the discriminator, but the generator in the GAN model often suffers from "mode collapse", that is, the generator may over-focus on generating certain specific types of design schemes and ignore other potential design possibilities, which will cause the generated design schemes to lack diversity and globality.

[0057] Specifically, in this embodiment, the introduction of a reinforcement learning reward mechanism can help the generator develop in a direction that better meets the requirements of multi-objective optimization during the training process. For example, by setting a reward function related to carbon emissions, energy efficiency, and material costs, reinforcement learning can guide the generator to generate a more low-carbon, energy-saving, and economical design solution.

[0058] In low-carbon building design, the generated scheme is not only required to be visually reasonable, but also to have excellent performance in practical indicators such as carbon emissions and energy efficiency. The traditional GAN ​​training process focuses on the appearance matching of the design scheme, but after the introduction of reinforcement learning, the generator can learn to generate design schemes that are more in line with these practical application indicators through additional reward mechanisms.

[0059] The reinforcement learning algorithm can provide feedback based on key indicators such as carbon emissions and energy efficiency of the generated solution, helping the generator to iterate faster towards optimization. In this way, the generated design solution is not only reasonable in appearance, but also meets the actual low-carbon requirements.

[0060] In the process of GAN model generating design solutions, the reinforcement learning algorithm evaluates the generated solutions by setting up an environment, including:

[0061] Set the state of the environment: The state of the environment includes the building design parameters generated by the generator (such as appearance design parameters and internal structure design parameters), which directly affect the energy consumption and carbon emission performance of the building.

[0062] Set the action space: During the generation process, the generator is equivalent to the "agent" in reinforcement learning. Its action space is the different combinations of design parameters. Each time a new plan is generated, it is equivalent to taking an action.

[0063] Set up a reward mechanism: Through the reward mechanism of reinforcement learning, positive or negative rewards are given according to whether the generated design scheme meets the requirements of low-carbon buildings (such as low carbon emissions, optimized energy efficiency, economic rationality, etc.). The reward value is determined by the target indicators of the building design, such as energy consumption, carbon emissions, and material costs.

[0064] Among them, the reward function is the core of reinforcement learning and is used to guide the learning process of the generator. Specifically, the reward function will give feedback based on the actual performance of the generated design solution. It includes:

[0065] Low carbon emissions. If the generated building design performs well in terms of carbon emissions (such as below the set threshold), the generator will receive a high positive reward.

[0066] Energy efficiency: rewards will be given for designs that are able to maintain high performance while reducing energy consumption.

[0067] Material cost optimization: if a material combination that meets low-carbon requirements and reduces the cost of building materials is selected in the generated plan, additional rewards will be obtained.

[0068] Through these goals, the generator continuously optimizes the design towards a direction that better meets low-carbon building design standards.

[0069] The training process of reinforcement learning is as follows:

[0070] Initial training: When the GAN model starts to generate design solutions, the generator and discriminator perform standard adversarial training, where the generator tries to generate design solutions and the discriminator tries to distinguish between generated solutions and real data.

[0071] Introducing reinforcement learning: During the initial training process, the reinforcement learning algorithm interacts with the generator and gradually adjusts the generator's generation strategy using a reward mechanism. The reinforcement learning algorithm gives rewards or penalties based on each generated design, and passes feedback to the generator to guide it to optimize its action strategy the next time it generates.

[0072] Generator optimization: The generator gradually optimizes the generated design in the process of continuously obtaining reward feedback, so that it not only meets the "realism" requirements of the discriminator, but also optimizes goals such as carbon emissions, energy efficiency, and material costs.

[0073] Iterative optimization: After multiple iterations of training, the reward mechanism of reinforcement learning enables the generator to continuously improve in the direction of generating better solutions. After each design solution is generated, the system will provide rewards in real time based on indicators such as energy consumption and carbon emissions. The generator gradually adjusts the generation strategy based on these feedbacks, and the generated solutions gradually approach the optimal low-carbon design.

[0074] Specifically, the generation process guided by reinforcement learning can be expressed as:

[0075]

[0076] in, is the parameter of the generator. The purpose of the above formula is to optimize the generator to maximize the reward function R(s,a) and the generated architectural design scheme G(z) containing construction design parameters. a is the action taken by the generator, that is, the decision of the generated design scheme, s is the state of the generated architectural design scheme G(z), and z is the initial architectural design parameter. is the noise distribution of the low-dimensional vector mapping of the initial architectural design parameters, E is the expectation of the generated design scheme under different input noises, is the score given by the discriminator D to the generated design solution. Reward function For evaluating generated architectural design options As a preferred embodiment of the present invention, the reward function includes evaluating the carbon emission content, energy efficiency and material cost of the comprehensive design solution.

[0077] It may be difficult for the generator to generate a multi-objective optimization design only through the feedback of the discriminator. In an embodiment of the present invention, by introducing reinforcement learning in the generator, the generator can learn how to generate a design that is more in line with the low-carbon goal through a reward mechanism, so that the low-carbon building design can simultaneously optimize carbon emissions, energy efficiency and cost.

[0078] The adaptive discriminator performs a preliminary carbon emission performance evaluation on the generated design scheme to obtain a preliminary carbon emission performance evaluation result.

[0079] It is understandable that the design solutions generated by the generative adversarial network include a variety of unreasonable or poorly performing designs in the early stage, that is, the GAN generator may generate a variety of solutions, some of which are visually reasonable but not feasible in terms of energy consumption and carbon emissions. In this embodiment, the adaptive discriminator performs a preliminary evaluation of carbon emission performance, which can quickly screen out those solutions that do not meet the requirements of low-carbon buildings. This step can effectively filter out designs with high carbon emissions or low energy efficiency, avoid unnecessary subsequent optimization, and thus improve the efficiency of the overall optimization process.

[0080] Carbon emission performance is one of the core goals of the low-carbon building design of the present invention. When optimizing the building design scheme, preliminary evaluation data on carbon emissions is required as a basis. The adaptive discriminator gives the carbon emission performance of each design scheme through preliminary evaluation, and the genetic algorithm in the present invention needs to combine multiple goals (such as carbon emissions, energy efficiency and material costs) for optimization. This preliminary evaluation result can become an important input in the multi-objective optimization of the genetic algorithm, helping the genetic algorithm GA to adjust the design parameters more accurately during the optimization process.

[0081] In an embodiment of the present invention, as a preferred implementation, the adaptive discriminator performs a preliminary carbon emission performance evaluation on the generated design solution, specifically including:

[0082] The adaptive discriminator determines the building design complexity level according to the overall building design parameters, and dynamically adjusts the evaluation criteria of the carbon emission performance based on the determined building design complexity level; wherein the overall building design parameters include initial building design parameters, building appearance design parameters and building internal structure design parameters;

[0083] Perform a preliminary carbon emission performance assessment on the generated design scheme based on the adjusted carbon emission performance assessment criteria.

[0084] Among them, the training process of the adaptive discriminator includes:

[0085] Collect historical data on completed building designs, including building dimensions, materials, energy systems, climate conditions, energy consumption data, and carbon emissions data;

[0086] Inputting the collected historical data into an initial discriminator for training, the initial discriminator forms an adaptive discriminator by learning the relationship between architectural design parameters of completed buildings and energy consumption and carbon emissions; wherein the architectural design parameters of the completed buildings include building size, materials, energy system and climate conditions;

[0087] Inputting the architectural design parameters of the built building into the GAN model, generating multiple virtual architectural design schemes through the GAN model generator, inputting the generated virtual architectural design schemes into the adaptive discriminator, and judging the rationality of the virtual architectural design schemes through the adaptive discriminator as the first rationality of the virtual architectural design schemes;

[0088] Taking the actual rationality of the completed building design scheme as the standard, marking the rationality of the generated virtual building design scheme, and obtaining the second rationality of the virtual building design scheme;

[0089] The adaptive discriminator is adjusted according to the difference between the first rationality and the second rationality of the virtual building design solution.

[0090] Furthermore, the training process of the adaptive discriminator also includes:

[0091] In the early stage of adaptive discriminator training, a set of initial building design complexity classification rules and evaluation criteria for evaluating carbon emission performance corresponding to different complexity levels are preset; wherein the building design complexity classification operation includes multiplying different overall building design parameters with their corresponding weighting coefficients based on weighting coefficients corresponding to different overall building design parameters, and then performing a summation operation; the initial building design complexity classification rules include an initial weighting coefficient combination;

[0092] The initial weighting coefficient combination is adjusted to obtain different weighting coefficient combinations, and based on the different weighting coefficient combinations, a building design complexity level classification operation is performed on multiple built buildings to determine the complexity levels of the building design parameters of the multiple built buildings, and the carbon emission performance of the multiple built buildings is evaluated according to the evaluation standard for evaluating the carbon emission performance corresponding to the determined complexity level, and the carbon emission performance results of the multiple built buildings obtained by the evaluation are compared with the actual carbon emission performance results of the multiple built buildings to obtain the evaluation accuracy corresponding to the different weighting coefficient combinations;

[0093] The optimal weighting coefficient combination is determined according to the evaluation accuracy corresponding to different weighting coefficient combinations, and the building design complexity level classification rule including the optimal weighting coefficient combination is used as the final building design complexity level classification rule.

[0094] It is understandable that the complexity of building design directly affects the carbon emission performance of the building. The size of the building, the diversity of materials, and the complexity of the energy system all determine the complexity level of the building design. This means that for a complex building design, a simple carbon emission assessment standard cannot accurately reflect the actual performance of the design, while an overly loose assessment standard may underestimate the carbon emissions of a simple building design. Therefore, in an embodiment of the present invention, an adaptive discriminator is set to dynamically adjust the assessment standard according to the complexity level of the building design, in order to ensure the accuracy of the assessment.

[0095] By setting rules for classifying the complexity of building designs, the appropriate carbon emission assessment standard can be automatically selected according to the complexity level of each design scheme, so that complex designs and simple designs can be evaluated using appropriate standards respectively.

[0096] Based on the design schemes generated by the GAN model and the preliminary evaluation results of carbon emission performance, the genetic algorithm GA is used to perform optimization operations on each design scheme to obtain the target design scheme; wherein the fitness function of the GA is a multi-objective optimization function, and the multi-objectives include carbon emissions, energy efficiency and building material costs;

[0097] Furthermore, in the process of performing optimization operations on each design solution through the genetic algorithm GA, the genetic algorithm is also combined with the particle swarm optimization to accelerate the search process in the design solution parameter space through the hybrid optimization algorithm.

[0098] The core of the genetic algorithm includes steps such as selection, crossover, and mutation. The goal is to optimize the design solution through iteration so that it performs better under the multi-objective optimization function. As a preferred implementation of the embodiment of the present invention, the multi-objective includes carbon emissions, energy efficiency, and building material costs, but it does not mean that it is limited to the above preferred objectives. The fitness function can be determined based on the multi-objective and its corresponding weight coefficients.

[0099] During each generation of optimization, candidate designs are scored according to the fitness function, and designs with high fitness are selected to enter the next round of crossover and mutation. Crossover operations are performed on the selected designs to generate new offspring solutions.

[0100] Particle swarm optimization (PSO) is an optimization algorithm based on group collaboration, which simulates the movement of particles to find the global optimal solution. The combination process of PSO and GA includes:

[0101] In PSO, each design solution is regarded as a particle, and the velocity and position of the particle are updated over time, where:

[0102] Location represents the position of the i-th design in the t-th generation (i.e., the parameters of the design), position It is a vector that includes all the parameters of the design solution. During the optimization process, these parameters are continuously adjusted to generate a design solution that meets the requirements of low-carbon design.

[0103] speed It represents the changing speed of the i-th building design scheme in the t-th generation, that is, the speed at which the parameters of the design scheme change between each generation. The adjustment of the speed determines the direction in which the design scheme approaches the optimal solution in the parameter space.

[0104] but:

[0105]

[0106]

[0107] Where w is the inertia weight, and They are local learning factor and global learning factor, and is a random number between [0,1], is the historical optimal position of example i, i.e., the best design parameter combination of the design scheme in the past iterations, The global optimal position is the parameter combination of the best performing solution among all the design solutions of the current generation in the entire particle swarm.

[0108] Multiple initial design solutions are generated through the GAN model, and the carbon emissions, energy efficiency and costs of these design solutions are calculated based on the preliminary evaluation results of the adaptive discriminator as the initial population of the genetic algorithm. The selection, crossover and mutation operations of the genetic algorithm GA are performed on the initial population, and the design solutions are optimized generation by generation to make them perform better in terms of carbon emissions, energy efficiency and cost. The design solutions optimized by the genetic algorithm GA are used as the initial particle swarm of PSO. PSO accelerates the search for the global optimal design solution through feedback of local and global optimal solutions. The combination of the genetic algorithm GA and PSO balances local search and global search, so that the genetic algorithm avoids falling into the local optimal solution. Through the feedback of the global optimal position, it ensures that the optimization process moves towards the global optimal solution.

[0109] In this embodiment, the above-mentioned whole process comprehensively evaluates carbon emissions, energy efficiency and material costs through the fitness function, and accelerates the search for the global optimal solution through PSO, so that the design scheme can approach the optimal low-carbon building design scheme faster and more accurately.

[0110] The target design scheme is simulated and monitored and evaluated in real time according to external environmental data, and the target design scheme is dynamically adjusted based on the results of the real-time evaluation.

[0111] Specifically, external environmental data refers to various external factors that affect building energy consumption, carbon emissions, and overall performance. These data not only affect the energy efficiency of the building in actual operation, but also determine whether the design scheme meets the climate, energy market and other conditions of a specific region. For the low-carbon building design in the present invention, external environmental data is used to help the system adjust the design parameters in real time to ensure that the building maintains optimal performance under different conditions. Including real-time climate data, historical meteorological data, energy prices and market fluctuations, socio-economic data, etc.

[0112] Furthermore, the target design scheme is simulated and monitored and evaluated in real time according to the external environment data, and the target design scheme is dynamically adjusted based on the result of the real-time evaluation, specifically including:

[0113] Collect and input external environmental data related to building design, the external environmental data including climate data and energy price data;

[0114] Generate a BIM model of the target building corresponding to the target design solution through BIM technology;

[0115] Integrate the BIM model with energy consumption simulation software to simulate the energy consumption and carbon emissions of the target building under different external environmental data;

[0116] Monitor the energy consumption and carbon emission changes of target buildings in real time, and generate feedback data on building design performance during the monitoring process;

[0117] The architectural design parameters of the target design scheme are adjusted based on the feedback data.

[0118] Finally, it should be noted that the low-carbon building design decision method for hot-summer and warm-winter areas based on generative adversarial network GAN and genetic algorithm GA disclosed in the embodiment of the present invention is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, ordinary technicians in this field should understand that the technical solutions recorded in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A low-carbon building design decision-making method for hot summer and warm winter areas based on GAN and GA, characterized by: The method comprises: Determine initial building design parameters; Generate multiple design schemes based on the initial building design parameters through a GAN model; wherein the GAN model is composed of a generator and an adaptive discriminator, including a multi-level structure of multiple generators and adaptive discriminators, the first-level generator is used to generate building appearance design parameters, the second-level generator is used to generate building internal structure design parameters, and the adaptive discriminator performs a preliminary carbon emission performance evaluation on the generated design schemes to obtain a preliminary carbon emission performance evaluation result; Based on the design schemes generated by the GAN model and the preliminary evaluation results of carbon emission performance, the genetic algorithm GA is used to perform optimization operations on each design scheme to obtain the target design scheme; wherein the fitness function of the GA is a multi-objective optimization function, and the multi-objectives include carbon emissions, energy efficiency and building material costs; Conduct simulation monitoring and real-time evaluation of the target design scheme according to external environment data, and dynamically adjust the target design scheme based on the results of real-time evaluation; The building exterior design parameters include the shape of the building facade, material distribution, window layout and building orientation; The building internal structure design parameters include room layout, partition wall materials, and internal arrangement of air conditioning and ventilation systems; The training process of the adaptive discriminator includes: Collect historical data on completed building designs, including building dimensions, materials, energy systems, climate conditions, energy consumption data, and carbon emissions data; Inputting the collected historical data into an initial discriminator for training, the initial discriminator forming an adaptive discriminator by learning the relationship between architectural design parameters of completed buildings and energy consumption and carbon emissions; wherein the architectural design parameters of the completed buildings include building size, materials, energy system and climate conditions; Inputting the architectural design parameters of the built building into the GAN model, generating multiple virtual architectural design schemes through the GAN model generator, inputting the generated virtual architectural design schemes into the adaptive discriminator, and judging the rationality of the virtual architectural design schemes through the adaptive discriminator as the first rationality of the virtual architectural design schemes; Taking the actual rationality of the completed building design scheme as the standard, marking the rationality of the generated virtual building design scheme, and obtaining the second rationality of the virtual building design scheme; adjusting the adaptive discriminator according to the difference between the first rationality and the second rationality of the virtual building design scheme; The simulation monitoring and real-time evaluation of the target design scheme according to the external environment data and the dynamic adjustment of the target design scheme based on the result of the real-time evaluation specifically include: Collect and input external environmental data related to building design, the external environmental data including climate data and energy price data; Generate a BIM model of the target building corresponding to the target design solution through BIM technology; Integrate the BIM model with energy consumption simulation software to simulate the energy consumption and carbon emissions of the target building under different external environmental data; Monitor the energy consumption and carbon emission changes of target buildings in real time, and generate feedback data on building design performance during the monitoring process; The architectural design parameters of the target design scheme are adjusted based on the feedback data.

2. The low-carbon building design decision-making method for hot-summer and warm-winter areas based on GAN and GA according to claim 1 is characterized in that: The initial building design parameters include size, materials, energy systems, historical weather data, and energy usage patterns.

3. The low-carbon building design decision-making method for hot-summer and warm-winter areas based on GAN and GA according to claim 1 is characterized in that: In the process of generating multiple design schemes based on the initial architectural design parameters through the GAN model, a reinforcement learning algorithm is used to learn the process of generating design schemes by the GAN model.

4. The low-carbon building design decision-making method for hot-summer and warm-winter areas based on GAN and GA according to claim 3 is characterized in that: The reinforcement learning algorithm enables the GAN model to generate a better design solution through a reward mechanism.

5. The low-carbon building design decision-making method for hot-summer and warm-winter areas based on GAN and GA according to claim 1 is characterized in that: The adaptive discriminator performs a preliminary carbon emission performance evaluation on the generated design solution, specifically including: The adaptive discriminator determines the building design complexity level according to the overall building design parameters, and dynamically adjusts the evaluation criteria of the carbon emission performance based on the determined building design complexity level; wherein the overall building design parameters include initial building design parameters, building appearance design parameters and building internal structure design parameters; Perform a preliminary carbon emission performance assessment on the generated design scheme based on the adjusted carbon emission performance assessment criteria.

6. The low-carbon building design decision-making method for hot-summer and warm-winter areas based on GAN and GA according to claim 5 is characterized in that: The training process of the adaptive discriminator also includes: In the early stage of adaptive discriminator training, a set of initial building design complexity classification rules and evaluation criteria for evaluating carbon emission performance corresponding to different complexity levels are preset; wherein the building design complexity classification operation includes multiplying different overall building design parameters with their corresponding weighting coefficients based on weighting coefficients corresponding to different overall building design parameters, and then performing a summation operation; the initial building design complexity classification rules include an initial weighting coefficient combination; The initial weighting coefficient combination is adjusted to obtain different weighting coefficient combinations, and based on the different weighting coefficient combinations, a building design complexity level classification operation is performed on multiple built buildings to determine the complexity levels of the building design parameters of the multiple built buildings, and the carbon emission performance of the multiple built buildings is evaluated according to the evaluation standard for evaluating the carbon emission performance corresponding to the determined complexity level, and the carbon emission performance results of the multiple built buildings obtained by the evaluation are compared with the actual carbon emission performance results of the multiple built buildings to obtain the evaluation accuracy corresponding to the different weighting coefficient combinations; The optimal weighting coefficient combination is determined according to the evaluation accuracy corresponding to different weighting coefficient combinations, and the building design complexity level classification rule including the optimal weighting coefficient combination is used as the final building design complexity level classification rule.

7. The low-carbon building design decision-making method for hot-summer and warm-winter areas based on GAN and GA according to claim 1 is characterized in that: In the process of performing optimization operations on each design scheme through the genetic algorithm GA, the method also includes combining the genetic algorithm GA with particle swarm optimization as a hybrid optimization algorithm, and accelerating the search process in the design scheme parameter space through the hybrid optimization algorithm.

Images