Zero-carbon building auxiliary design method and device

Abstract

The embodiment of the present invention provides a zero-carbon building auxiliary design method and device, which relates to the field of building design optimization technology. The method includes: determining multiple optimization objectives and design parameters of the zero-carbon building; simulating multiple optimization target values ​​corresponding to different design parameter values ​​according to the geometric model, multiple optimization objectives and design parameters of the zero-carbon building to obtain a target sample set; training the proxy model according to the target sample set to obtain a trained proxy model; using the trained proxy model as the objective function to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters are used to assist the design of the zero-carbon building. The method of the embodiment of the present invention realizes the multi-objective optimization design of the zero-carbon building, making the multi-objective design process of the zero-carbon building more scientific and standardized.

Description

Technical Field

[0001] The present invention relates to the technical field of building design optimization, and in particular to a zero-carbon building auxiliary design method and device. Background Art

[0002] As one of the three major energy-consuming sectors, the construction industry is associated with a large amount of energy consumption and carbon emissions.

[0003] Traditional architectural design is primarily led by architects, often focusing on planning requirements, building facades, and functional design as primary factors for initial design, with coordination with mechanical and electrical engineering professionals. However, traditional architectural design typically prioritizes building energy consumption, while ignoring multiple objectives such as carbon emissions. Therefore, multi-objective optimization design for zero-carbon buildings is a pressing technical challenge facing those skilled in the art. Summary of the Invention

[0004] In response to the problems in the prior art, embodiments of the present invention provide a zero-carbon building auxiliary design method and device.

[0005] Specifically, the embodiments of the present invention provide the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a zero-carbon building auxiliary design method, comprising:

[0007] Determine multiple optimization objectives and design parameters for zero-carbon buildings;

[0008] Based on the geometric model of the zero-carbon building, multiple optimization objectives and design parameters, multiple optimization target values ​​corresponding to different design parameter values ​​are simulated to obtain a target sample set;

[0009] The proxy model is trained based on the target sample set to obtain a trained proxy model. The proxy model is established based on the bidirectional long short-term memory network BiLSTM optimized by a meta-heuristic algorithm.

[0010] The trained proxy model is used as the objective function to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters are used to assist the design of the zero-carbon building.

[0011] In a second aspect, an embodiment of the present invention further provides a zero-carbon building auxiliary design device, comprising:

[0012] A selection module for selecting multiple optimization objectives and design parameters for zero-carbon buildings;

[0013] A sample module is used to simulate multiple optimization target values ​​corresponding to different design parameter values ​​based on the geometric model of the zero-carbon building, multiple optimization objectives and design parameters, and obtain a target sample set;

[0014] The training module is used to train the proxy model based on the target sample set to obtain the trained proxy model; the proxy model is established based on the meta-heuristic algorithm and the bidirectional long short-term memory network BiLSTM;

[0015] The optimization module is used to use the trained proxy model as the objective function to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters are used to assist the design of the zero-carbon building.

[0016] In a third aspect, an embodiment of the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the program, the steps of the zero-carbon building auxiliary design method as described in the first aspect are implemented.

[0017] In a fourth aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the zero-carbon building auxiliary design method as described in the first aspect.

[0018] In a fifth aspect, an embodiment of the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the zero-carbon building auxiliary design method as described in the first aspect.

[0019] The zero-carbon building auxiliary design method and device provided by the embodiment of the present invention innovatively establish an agent model based on the combination of a metaheuristic algorithm and a bidirectional long short-term memory network BiLSTM. Compared with agent models of other types and structures, it can more quickly, accurately and effectively determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters, thereby realizing the multi-objective optimization design of the zero-carbon building, making the multi-objective design process of the zero-carbon building more scientific and standardized. At the same time, it reduces the call to the EnergyPlus software during the optimization process, shortens the optimization time, and effectively improves the collaborative optimization and design of the annual operating carbon emissions per unit area of ​​the building, the annual operating heating energy consumption per unit area of ​​the building, the annual operating cooling energy consumption per unit area of ​​the building, the incremental cost per unit area, the annual number of uncomfortable hours of the building, the natural satisfaction rate of indoor lighting in the building, the self-sufficiency rate of photovoltaic power generation, and the rooftop photovoltaic power generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 1 is a flow chart of a zero-carbon building auxiliary design method provided by an embodiment of the present invention;

[0022] Figure 2 Schematic diagram of a zero-carbon building auxiliary design decision-making device provided by an embodiment of the present invention;

[0023] Figure 3 Schematic diagram of the working principle of the selection module provided by an embodiment of the present invention;

[0024] Figure 4 Schematic diagram of the working principle of the input module provided by an embodiment of the present invention;

[0025] Figure 5 Schematic diagram of the working principle of the training module provided by an embodiment of the present invention;

[0026] Figure 6 Schematic diagram of the working principle of the test module provided by an embodiment of the present invention;

[0027] Figure 7 Schematic diagram of the working principle of the optimization module provided by an embodiment of the present invention;

[0028] Figure 8 is a schematic diagram of a Pareto front solution provided by an embodiment of the present invention;

[0029] Figure 9 Schematic diagram of the structure of a zero-carbon building auxiliary design device provided by an embodiment of the present invention;

[0030] Figure 10 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0031] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0032] The method of the embodiment of the present invention can be applied to architectural design scenarios, realizing the multi-objective optimization design of zero-carbon buildings, making the multi-objective design process of zero-carbon buildings more scientific and standardized.

[0033] Traditional architectural design typically prioritizes energy consumption, while neglecting multiple objectives, such as carbon emissions. Therefore, optimizing the design of zero-carbon buildings under multiple objectives is a pressing technical challenge.

[0034] The zero-carbon building auxiliary design method of the embodiment of the present invention innovatively establishes an agent model based on a metaheuristic algorithm and a bidirectional long short-term memory network BiLSTM. Compared with agent models of other types and structures, it can more quickly, accurately and effectively determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters, thereby realizing the multi-objective optimization design of the zero-carbon building, making the multi-objective design process of the zero-carbon building more scientific and standardized. At the same time, it reduces the call to the EnergyPlus software during the optimization process, shortens the optimization time, and effectively improves the collaborative optimization and design of the annual operating carbon emissions per unit area of ​​the building, the annual operating heating energy consumption per unit area of ​​the building, the annual operating cooling energy consumption per unit area of ​​the building, the incremental cost per unit area, the annual number of uncomfortable hours of the building, the natural satisfaction rate of indoor lighting in the building, the self-sufficiency rate of photovoltaic power generation, and the rooftop photovoltaic power generation.

[0035] The following combination Figures 1-10 The technical solution of the present invention is described in detail with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0036] Figure 1 This is a flow chart of an embodiment of the zero-carbon building auxiliary design method provided by the present invention. Figure 1 As shown, the method provided in this embodiment includes:

[0037] Step 101: Determine multiple optimization objectives and design parameters for a zero-carbon building.

[0038] Specifically, in order to achieve the multi-objective optimization design of zero-carbon buildings, the embodiment of the present application first selects and determines the optimization objectives required for zero-carbon building design and the design parameters and constraints corresponding to the optimization objectives. Among them, the optimization objectives include but are not limited to the annual operating carbon emissions per unit area of ​​the building, the annual operating heating energy consumption per unit area of ​​the building, the annual operating cooling energy consumption per unit area of ​​the building, the incremental cost per unit area, the annual number of uncomfortable hours of the building, the natural satisfaction rate of indoor lighting in the building, the self-sufficiency rate of photovoltaic power generation, the rooftop photovoltaic power generation, etc.; the design parameters corresponding to the objectives include the external window heat transfer coefficient, the ground heat transfer coefficient, the roof heat transfer coefficient, the east wall window-to-wall ratio, the west wall window-to-wall ratio, the south wall window-to-wall ratio, the north wall window-to-wall ratio, the external window shading coefficient, the wall heat transfer coefficient, the solar heat gain coefficient (SHGC), the number of ventilation times, the building orientation, the permeability, the roof photovoltaic module installation angle, the roof photovoltaic module installation area, the roof photovoltaic module installation orientation, the roof photovoltaic cell type, and the energy storage battery type.

[0039] Optionally, the upper and lower limits of some design parameters of zero-carbon buildings are as follows: external window heat transfer coefficient (1.0≤x1≤1.5W / m 2 ·K), ground heat transfer coefficient (0.25≤x2≤0.4W / m 2 ·K), roof heat transfer coefficient (0.1≤x3≤0.3W / m 2 ·K), east wall window-to-wall ratio (0.2≤x4≤0.4), west wall window-to-wall ratio (0.2≤x5≤0.4), south wall window-to-wall ratio (0.2≤x6≤0.5), north wall window-to-wall ratio (0.2≤x7≤0.5), exterior window shading coefficient (200≤x8≤1000), wall heat transfer coefficient (0.1≤x9≤0.3W / m 2 ·K), Solar Heat Gain Coefficient (SHGC) value (0.45≤x 10 ≤0.52), ventilation frequency (0.4≤x 11 ≤1.0 times / hour) and building orientation (0≤x 12 ≤360 degrees), roof photovoltaic tilt angle (0≤x 13 ≤60 degrees), roof photovoltaic orientation (0≤x 14 ≤360 degrees), energy storage battery power (1≤x 14 ≤500kw) etc.

[0040] For example, the optimization objectives include but are not limited to the annual carbon emissions per unit area of ​​the building, the annual heating energy consumption per unit area of ​​the building, the annual cooling energy consumption per unit area of ​​the building, the incremental cost per unit area, the annual discomfort hours of the building, the natural satisfaction rate of indoor lighting in the building, the self-sufficiency rate of photovoltaic power generation, the rooftop photovoltaic power generation, etc., which can be expressed by the following formula:

[0041] T=(T1,T2,T3,T4,T5,T6,T7,T8)

[0042] Wherein, T1 to T8 represent the number of targets to be optimized.

[0043] The constructed design parameter sample matrix is ​​shown in the following formula:

[0044]

[0045] Among them, m and n are the design parameter number and sample size respectively.

[0046] Step 102: Based on the geometric model of the zero-carbon building, multiple optimization objectives, and design parameters, multiple optimization objective values ​​corresponding to different design parameter values ​​are simulated to obtain a target sample set.

[0047] Specifically, after determining multiple optimization objectives and design parameters of zero-carbon buildings, the embodiment of the present application further establishes a numerical simulation model of the building based on the target building to be optimized, its geometric dimensions and enclosure structure parameters, and the energy consumption simulation tool and the photovoltaic power generation simulation tool. The numerical simulation model is imported into the building energy consumption simulation tool and the photovoltaic power generation simulation tool to calculate the optimization target results under different design parameter conditions. The different parameters are sampled and combined for simulation to finally obtain a target sample set. Optionally, the target sample set is preprocessed and further divided into a training set and a test set for subsequent input. For example, in the process of generating the target sample set, the zero-carbon building design parameter sample matrix, the building geometric model and the building optimization target matrix are first constructed; wherein, the construction of the zero-carbon building optimization parameter sample matrix includes determining the distribution form of the optimization parameters, determining the sampling method of the sample, determining the size of the sample capacity, executing the sampling process and forming the matrix. The sampling method is to use Simlab software and Latin hypercube sampling to select 1,300 sets of design schemes. The design parameters and optimization objectives correspond one to one. The annual operating carbon emissions per unit area of ​​the building, the annual operating heating energy consumption per unit area of ​​the building, the annual operating cooling energy consumption per unit area of ​​the building, the incremental cost per unit area, the annual number of uncomfortable hours of the building, the natural satisfaction rate of indoor lighting in the building, the self-sufficiency rate of photovoltaic power generation, and the amount of rooftop photovoltaic power generation are calculated under different schemes.

[0048] The constructed design parameter sample matrix is ​​shown in the following formula:

[0049]

[0050] Among them, m and n are the design parameter number and sample size respectively.

[0051] Building envelope construction parameters include the building's wall structure, roof structure, floor structure, lighting power density, lighting schedule, equipment density, number of occupants, activity level, occupancy rate, cooling control, heating control, and heating, ventilation, and air conditioning (HVAC) systems. Table 1 lists some building envelope construction parameters.

[0052] Table 1

[0053]

[0054]

[0055] Import the design parameter sample matrix into JePlus+EnergyPlus to calculate the corresponding target values ​​and form the optimization target matrix, as shown below:

[0056]

[0057] Among them, y mn It represents the selected mth optimization objective and nth calculation condition.

[0058] Step 103: Train the proxy model according to the target sample set to obtain a trained proxy model; the proxy model is established based on a meta-heuristic algorithm and a bidirectional long short-term memory network BiLSTM.

[0059] Specifically, after simulating multiple optimization target values ​​corresponding to different design parameter values ​​based on the geometric model of the zero-carbon building, multiple optimization objectives, and design parameters, and obtaining a target sample set, the proxy model can be further trained using the target sample set to obtain a trained proxy model. Optionally, the proxy model in the embodiment of the present application is established based on a metaheuristic algorithm and a bidirectional long short-term memory network (BiLSTM). Optionally, the metaheuristic algorithm includes, but is not limited to, at least one of the following: a genetic algorithm, simulated annealing, tabu search, particle swarm optimization, and an ant colony algorithm.

[0060] For example, the training set in the target sample set is imported into the metaheuristic algorithm and deep learning algorithm corresponding to the proxy model for training, a mapping relationship between the design parameters and the optimization objectives is constructed, and the mapping relationship is used as a zero-carbon building multi-objective optimization function proxy model. The metaheuristic algorithm optimizes the structural parameters of the deep learning algorithm, and further checks the proxy model during the optimization process to evaluate the model training effect and whether it meets the requirements. If it meets the requirements, the proxy model after training is output. Optionally, if the requirements are not met, the deep learning algorithm model parameters are cyclically updated and enter cyclic training until the trained model meets the accuracy requirements and the cycle ends. Optionally, after the training is completed, the trained proxy model is tested using the test set in the target sample set to obtain the performance evaluation index of the test set; wherein, the performance evaluation index includes the root mean square error RMSE, the mean absolute error MAE and R 2 Indicators; Based on the performance evaluation indicators, the trained proxy model is evaluated and finally the trained proxy model is obtained.

[0061] For example,

[0062]

[0063]

[0064] Among them, y i represents the simulated value of the i-th optimization target, f i represents the predicted value of the i-th optimization target, n represents the total number of samples, and mean(f) represents the average value of the optimization target simulation value.

[0065] For example, the target sample set is split into a 70% training set and a 30% test set, resulting in 910 data sets used for model training and 390 data sets used for model validation. A 5-fold cross-validation approach is used for the sample data, where the data set is evenly divided into five groups, with one group selected as the test sample at a time and the remaining four groups used as the training samples. Optionally, the training set is used to train a proxy model, resulting in a trained proxy model. Specifically, the trained metaheuristic algorithm - Bidirectional Long Short-Term Memory (BiLSTM) network model is trained, resulting in a trained metaheuristic algorithm - Bidirectional Long Short-Term Memory (BiLSTM) network model.

[0066] Step 104: Using the trained proxy model as an objective function, determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters are used to assist in the design of the zero-carbon building.

[0067] Specifically, after obtaining the trained proxy model, in the embodiment of the present application, the proxy model is used as the objective function of the multi-objective optimization algorithm, and the multi-objective optimization algorithm is adopted to optimize the zero-carbon building design parameters and optimization targets to obtain the Pareto optimization result; the Pareto optimization result includes the optimized design parameters and the optimization targets corresponding to the optimized design parameters; the design parameters and optimization targets that deviate from the actual project in the Pareto optimization result are eliminated, and the passive design parameters and optimization targets of the zero-carbon building are determined to assist the building design in zero-carbon building design.

[0068] It should be noted that proxy model optimization is an optimization method that combines numerical simulation and proxy model, which can accelerate the optimization process and find the optimal solution. It is widely used in practical engineering and scientific research, and can help improve design efficiency and the accuracy of optimization results. For the problem of multi-objective optimization design of zero-carbon buildings under multiple objectives, there is no prior knowledge of the specific objective function, so it is impossible to know which proxy model is the most accurate. In the embodiment of the present application, an innovative proxy model is established based on a metaheuristic algorithm and a bidirectional long short-term memory network BiLSTM. Compared with proxy models of other types and structures, it can more quickly, accurately and effectively determine the optimized design parameters of zero-carbon buildings and the optimization targets corresponding to the optimized design parameters. For example, compared with the prediction results of the ordinary LSTM model, the RMSE of the annual operating carbon emissions per unit area of ​​the building of this application is 0.056, R 2 =0.877, MAE=0.052; RMSE of incremental cost per unit area=0.066, R 2 =0.850, MAE=0.073; RMSE of annual building discomfort hours=0.089, R 2 =0.841, MAE=0.086, RMSE of rooftop photovoltaic power generation=0.045, R 2 =0.873, MAE=0.054, and the accuracy is better than the existing prediction results.

[0069] The method of the above embodiment innovatively establishes an agent model based on a metaheuristic algorithm and a bidirectional long short-term memory network BiLSTM. Compared with agent models of other types and structures, it can more quickly, accurately and effectively determine the optimized design parameters of zero-carbon buildings and the optimization targets corresponding to the optimized design parameters, thereby realizing the multi-objective optimization design of zero-carbon buildings, making the multi-objective design process of zero-carbon buildings more scientific and standardized. At the same time, it reduces the call to the EnergyPlus software during the optimization process, shortens the optimization time, and effectively improves the collaborative optimization and design of the annual operating carbon emissions per unit area of ​​the building, the annual operating heating energy consumption per unit area of ​​the building, the annual operating cooling energy consumption per unit area of ​​the building, the incremental cost per unit area, the annual number of uncomfortable hours of the building, the natural satisfaction rate of indoor lighting in the building, the self-sufficiency rate of photovoltaic power generation, and the rooftop photovoltaic power generation.

[0070] In one embodiment, the proxy model is established based on an improved grey goose optimization algorithm and a bidirectional long short-term memory network (BiLSTM). The process of updating individual positions in the improved grey goose optimization algorithm includes:

[0071] Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide the global optimal solution; the positions of the exploration group individuals are updated based on the following method:

[0072] X i+1 =X i -M×|N×X best -X i |

[0073] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model, that is, the optimal solution of a set of zero-carbon building design parameters and optimization objectives; X best Indicates the location of the current best solution; X i+1 Indicates the position of the individual after update; M indicates the first step length; N indicates the second step length; the first step length and the second step length are used to dynamically adjust the distance and direction of individual movement.

[0074] Specifically, in the embodiment of the present application, the agent model is established based on the improved Grey Goose Optimization Algorithm and the Bidirectional Long Short-Term Memory (BiLSTM) network. Optionally, in the embodiment of the present application, the individual position update process in the Grey Goose Optimization Algorithm is improved, that is, the process of optimizing the design parameters and optimization objectives of the zero-carbon building is improved, including:

[0075] Update the location of the exploration group individuals based on the following method:

[0076] X i+1 =X i-M×|N×X best -X i |

[0077] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 Represents the position of the individual after update; M represents the first step length; N represents the second step length; the first step length and the second step length are used to dynamically adjust the distance and direction of individual movement; thus, an improvement in the optimization process of zero-carbon building design parameters and optimization objectives is achieved, thereby maximizing the development of the search space, improving the efficiency and accuracy of the optimization solution, and thus achieving a more accurate and reasonable design of zero-carbon building multi-objective optimization, making the zero-carbon building multi-objective design process more scientific and accurate.

[0078] For example, the specific steps in the improved grey goose optimization algorithm are as follows:

[0079] S1: Initialize the goose population. Randomly generate a group of individuals, each of which represents the optimal solution for a set of BiLSTM algorithm parameters, that is, the optimal solution for a set of zero-carbon building design parameters and optimization objectives. The initial population can be expressed as Calculate , where n is the population size.

[0080] S2: Calculate fitness. Define an objective function f(x) to evaluate the quality of each individual, where x is the position of the individual.

[0081] S3: Dynamic grouping. Individuals are divided into an exploration group and an exploitation group. Initially, the ratio of the two groups is set to 50% exploration and 50% exploitation.

[0082] S4: Update position. The position of the goose population is updated, and the update step length is calculated as follows:

[0083] M=2aα1-m

[0084] N=2α2

[0085] Where M and N are the first and second update steps, respectively. a is the iterative nonlinear change vector, which varies linearly from 2 to 0 during the iterations. α1 and α2 vary randomly within the range [0, 1]. M and N dynamically adjust the distance and direction of individual movement to maintain exploratory behavior during global search.

[0086] Update the location of the exploration group individuals to:

[0087] X i+1 =X i -M·|N·X best-X i |

[0088] Among them, X i is the current position of the individual goose. best is the position of the best solution found so far. X i+1 is the position of the individual after update.

[0089] S5: Select the best solution. For each individual in the exploration group and the exploitation group in the goose population, calculate its fitness using the objective function. Compare the fitness of all individuals and find the individual with the highest fitness value. Set the position of this best individual X best Record it as the best solution for the current iteration.

[0090] S6: Iterative Update. Based on the results of the exploration and exploitation operations, the positions of all individuals in the goose population are updated. The fitness of each individual in the updated population is evaluated, and the new optimal individual is selected. This process is repeated until the predetermined number of iterations is reached or the stopping condition is met. The optimal solution is ultimately returned and used as the parameters of the BiLSTM model. Based on the model parameters, a proxy model is obtained.

[0091] The specific steps in the BiLSTM algorithm are as follows:

[0092] S1: Define the network structure. This includes the input layer, BiLSTM layer, output layer, and activation function. The number of neurons in the output layer typically matches the dimension of the target variable being optimized.

[0093] S2: Training the model. The model is trained using the training set. The model calculates the predicted value through forward propagation and adjusts the weights through optimization algorithms such as backpropagation and gradient descent to minimize the difference between the predicted value and the true value.

[0094] S3: Parameter Tuning. Based on the model's performance on the test set, adjust the model's parameters and use the improved Grey Goose Optimization (GGO) algorithm to optimize the model parameters. The relevant parameters to be optimized include the number of hidden layer nodes, initial learning rate, learning rate reduction factor, regularization coefficient, etc.

[0095] S4: Model evaluation: Evaluate the model performance on the test set to check its generalization ability and prevent overfitting. Different evaluation metrics are used, such as mean square error (MSE), root mean square error (RMSE), or mean absolute error (MAE).

[0096]

[0097]

[0098]

[0099] Among them, y i represents the simulated value of the i-th optimization target, f i represents the predicted value of the i-th optimization target, n represents the total number of samples, and mean(f) represents the average value of the optimization target simulation value.

[0100] For example, a proxy model was established using the improved Grey Goose Optimization (GGO) algorithm coupled with a BiLSTM (bidirectional long short-term memory) network. This algorithm was used to optimize hyperparameters. The selected optimization objectives were to minimize annual operating carbon emissions per unit area, minimize incremental costs per unit area, minimize annual discomfort hours, and maximize rooftop photovoltaic power generation. The GGO-BiLSTM model had 20 input layers, 4 output layers, and 40 hidden layers. The initial learning rate was 0.01, and the cross-validation value was 5.

[0101] The RMSE of annual carbon emissions per unit area of ​​buildings is 0.034, R 2 =0.954, MAE=0.032; RMSE of incremental cost per unit area=0.036, R 2 =0.920, MAE=0.065; RMSE of annual building discomfort hours=0.077, R 2 =0.901, MAE=0.071, RMSE of rooftop photovoltaic power generation=0.025, R 2 =0.973, MAE=0.021, and the prediction accuracy of the GGO-BiLSTM model established meets the requirements.

[0102] Compared with the prediction results of the common LSTM model, the RMSE of the annual carbon emissions per unit area of ​​the building is 0.056, R 2 =0.877, MAE=0.052; RMSE of incremental cost per unit area=0.066, R 2 =0.850, MAE=0.073; RMSE of annual building discomfort hours=0.089, R 2 =0.841, MAE=0.086, RMSE of rooftop photovoltaic power generation=0.045, R 2 =0.873, MAE=0.054, the prediction accuracy of the GGO-BiLSTM model established in the present invention is better than the existing prediction results.

[0103] The method of the above embodiment improves the process of optimizing the design parameters and optimization objectives of zero-carbon buildings by improving the individual position update process in the Grey Goose Optimization Algorithm. Compared to existing technologies, this method can maximize the exploitation of the search space, improve the efficiency and accuracy of the optimization solution, and thus achieve more accurate and reasonable multi-objective optimization design of zero-carbon buildings. It also improves the flexibility, diversity, and randomness of the optimization process for zero-carbon building design parameters and optimization objectives, making the multi-objective design process of zero-carbon buildings more scientific and accurate.

[0104] In one embodiment, the agent model is established based on a meta-heuristic algorithm and a bidirectional long short-term memory network (BiLSTM), including:

[0105] The proxy model is based on the improved Grey Goose Optimization Algorithm and the BiLSTM network. The process of updating individual positions in the improved Grey Goose Optimization Algorithm includes:

[0106] Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide the global optimal solution; the positions of the exploration group individuals are updated based on the following method:

[0107] X (i+1) =γ1×X best +γ2×(X rand1 -X rand )+(1-γ3)×X i

[0108] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after the update; γ1, γ2, and γ3 represent random factors used to control the degree of randomness in the individual update process; X rand1 and X rand2 represents the positions of two different individuals chosen randomly.

[0109] Specifically, in the embodiment of the present application, the agent model is established based on the improved Grey Goose Optimization Algorithm and the Bidirectional Long Short-Term Memory (BiLSTM) network. Optionally, in the embodiment of the present application, the individual position update process in the Grey Goose Optimization Algorithm is improved, that is, the process of optimizing the design parameters and optimization objectives of the zero-carbon building is improved, including:

[0110] Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide the global optimal solution; the positions of the exploration group individuals are updated based on the following method:

[0111] X (i+1) =γ1×X best +γ2×(X rand1 -X rand2 )+(1-γ3)×X i

[0112] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after the update; γ1, γ2, and γ3 represent random factors, which vary in the range of [0, 2] and are used to control the degree of randomness in the individual update process. rand1 and X rand2 It is the position of two different individuals randomly selected from the population, which further increases randomness and diversity, and thus realizes the improvement of the optimization process of zero-carbon building design parameters and optimization objectives, so as to maximize the development of the search space and improve the efficiency and accuracy of the optimization solution, thereby achieving a more accurate and reasonable design of zero-carbon building multi-objective optimization, making the zero-carbon building multi-objective design process more scientific and accurate.

[0113] The method of the above embodiment improves the process of optimizing the design parameters and optimization objectives of zero-carbon buildings by improving the individual position update process in the Grey Goose Optimization Algorithm. Compared to existing technologies, this method can maximize the exploitation of the search space, improve the efficiency and accuracy of the optimization solution, and thus achieve more accurate and reasonable multi-objective optimization design of zero-carbon buildings. It also improves the flexibility, diversity, and randomness of the optimization process for zero-carbon building design parameters and optimization objectives, making the multi-objective design process of zero-carbon buildings more scientific and accurate.

[0114] In one embodiment, the agent model is established based on a meta-heuristic algorithm and a bidirectional long short-term memory network (BiLSTM), including:

[0115] The proxy model is based on the improved Grey Goose Optimization Algorithm and the BiLSTM network. The process of updating individual positions in the improved Grey Goose Optimization Algorithm includes:

[0116] Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide the global optimal solution; the position of the development group individuals is updated based on the following method:

[0117] X (i+1) =X i +D×(1+ζ)×γ×(Xbest -X i )

[0118] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after update; D represents the update step size, which is used to dynamically adjust the search step size; ζ and γ represent random factors.

[0119] Specifically, in the embodiment of the present application, the agent model is established based on the improved Grey Goose Optimization Algorithm and the Bidirectional Long Short-Term Memory (BiLSTM) network. Optionally, in the embodiment of the present application, the individual position update process in the Grey Goose Optimization Algorithm is improved, that is, the process of optimizing the design parameters and optimization objectives of the zero-carbon building is improved, including:

[0120] Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide the global optimal solution; the position of the development group individuals is updated based on the following method:

[0121] X (i+1) =X i +D×(1+ζ)×γ×(X best -X i )

[0122] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 Represents the position of the individual after update; D represents the update step size, which is used to dynamically adjust the search step size; ζ and γ represent random factors; thus, the optimization process of zero-carbon building design parameters and optimization objectives is improved, so that the global optimal solution can be obtained and the accuracy of the optimization solution can be improved, thereby achieving a more accurate design of zero-carbon building multi-objective optimization, making the zero-carbon building multi-objective design process more scientific and accurate.

[0123] For example, the specific steps in the improved grey goose optimization algorithm are as follows:

[0124] S1: Initialize the goose population. Randomly generate a group of individuals, each of which represents an optimized solution for a set of BiLSTM algorithm parameters. The initial population can be expressed as Calculate , where n is the population size.

[0125] S2: Calculate fitness. Define an objective function f(x) to evaluate the quality of each individual, where x is the position of the individual.

[0126] S3: Dynamic grouping. Individuals are divided into an exploration group and an exploitation group. Initially, the ratio of the two groups is set to 50% exploration and 50% exploitation.

[0127] S4: Update position. The position of the goose population is updated, and the update step length is calculated as follows:

[0128] M=2αα1-m

[0129] N=2α2

[0130] Where M and N are the first and second update steps, respectively. a is the iterative nonlinear change vector, which varies linearly from 2 to 0 during the iterations. α1 and α2 vary randomly within the range [0, 1]. M and N dynamically adjust the distance and direction of individual movement to maintain exploratory behavior during global search.

[0131] Update the location of the exploration group individuals to:

[0132] X i+1 =X i -M·|N·X best -X i |

[0133] Among them, X i is the current position of the individual goose. best is the position of the best solution found so far. X i+1 is the position of the individual after update.

[0134] Update the location of the development team individuals to:

[0135] X (i+1) =X i +D·(1+ζ)·γ·(X best -X i )

[0136] D is the update step size, a decreasing function related to the number of iterations, used to dynamically adjust the search step size. ζ is a random factor, ranging in value from [0, 1], used to increase randomness during the search process. γ is a random factor, ranging in value from [0, 1], used to adjust the intensity of the movement toward the optimal solution.

[0137] S5: Select the best solution. For each individual in the exploration group and the exploitation group in the goose population, calculate its fitness using the objective function. Compare the fitness of all individuals and find the individual with the highest fitness value. Set the position of this best individual X best Record it as the best solution for the current iteration.

[0138] S6: Iterative Update. Based on the results of the exploration and exploitation operations, the positions of all individuals in the goose population are updated. The fitness of each individual in the updated population is evaluated, and the new optimal individual is selected. This process is repeated until the predetermined number of iterations is reached or the stopping condition is met. The optimal solution is ultimately returned and used as the parameters of the BiLSTM model. Based on the model parameters, a proxy model is obtained.

[0139] The method of the above embodiment improves the process of optimizing the design parameters and optimization objectives of zero-carbon buildings by improving the individual position update process in the Grey Goose Optimization Algorithm. This allows for a quick and accurate global optimal solution, improving the accuracy of the optimization solution, and thus enabling more accurate multi-objective optimization of zero-carbon buildings. It also enhances the flexibility, diversity, and randomness of the optimization process for zero-carbon building design parameters and optimization objectives, making the multi-objective design process of zero-carbon buildings more scientific and accurate.

[0140] In one embodiment, the proxy model is established based on an improved grey goose optimization algorithm and a bidirectional long short-term memory network (BiLSTM). The process of updating individual positions in the improved grey goose optimization algorithm includes:

[0141] Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide the global optimal solution; the position of the development group individuals is updated based on the following method:

[0142]

[0143] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after the update; D represents the update step size, which is used to dynamically adjust the search step size; γ represents the random factor; X sentry and X sentr Represents the two individuals in the goose flock that are ranked first in addition to the current best solution.

[0144] Specifically, in the embodiment of the present application, the agent model is established based on the improved Grey Goose Optimization Algorithm and the Bidirectional Long Short-Term Memory (BiLSTM) network. Optionally, in the embodiment of the present application, the individual position update process in the Grey Goose Optimization Algorithm is improved, that is, the process of optimizing the design parameters and optimization objectives of the zero-carbon building is improved, including:

[0145] Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide the global optimal solution; the position of the development group individuals is updated based on the following method:

[0146]

[0147] Among them, X i Indicates the position of an individual in the current flock; the position of an individual represents the optimal solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after the update; D represents the update step size, which is used to dynamically adjust the search step size; γ represents the random factor; X sentry1 and X sentry2 It represents the top two individuals in the goose flock except for the current best solution; it also realizes the improvement of the optimization process of zero-carbon building design parameters and optimization objectives, so that the global optimal solution can be obtained quickly and accurately, the accuracy of the optimization solution can be improved, and then a more accurate zero-carbon building multi-objective optimization design can be achieved.

[0148] The method of the above embodiment improves the process of optimizing the design parameters and optimization objectives of zero-carbon buildings by improving the individual position update process in the Grey Goose Optimization Algorithm. This allows for a quick and accurate global optimal solution, improving the accuracy of the optimization solution, and thus enabling more accurate multi-objective optimization of zero-carbon buildings. It also enhances the flexibility, diversity, and randomness of the optimization process for zero-carbon building design parameters and optimization objectives, making the multi-objective design process of zero-carbon buildings more scientific and accurate.

[0149] In one embodiment, the trained agent model is used as an objective function to determine the optimized design parameters of the zero-carbon building and the optimization objectives corresponding to the optimized design parameters, including:

[0150] The trained proxy model is used as the objective function, and the improved rime optimization algorithm is used to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the improved rime optimization algorithm is obtained by combining the rime optimization algorithm with the non-dominated sorting algorithm; the improved rime optimization algorithm is used to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters.

[0151] Specifically, the rime optimization algorithm is not effective for multi-objective optimization. In the embodiment of the present application, the rime optimization algorithm is combined with the non-dominated sorting algorithm to obtain an improved rime optimization algorithm. The improved rime optimization algorithm can be used to more efficiently and accurately determine the optimal solution in the multi-objective optimization process, and thus the optimized design parameters of the zero-carbon building and the optimization objectives corresponding to the optimized design parameters can be determined more efficiently and accurately.

[0152] For example, the proxy model is used as the objective function of the improved rime algorithm NSROA, and the objective function is used to calculate the fitness function value. The optimization process includes the following steps:

[0153] S1: Optimize model hyperparameters, including competition scale, population size, crossover and mutation probabilities, and maximum number of evolutionary iterations;

[0154] S2: Set the design parameters and ranges, and set the proxy model to the Non-Dominated Rime Optimization Algorithm (NSROA) based on the non-dominated sorting strategy, which is a combination of the rime optimization algorithm and the non-dominated sorting strategy.

[0155] S3: Randomly generate the first generation of population, which is a zero-carbon building design scheme. Each individual in the population represents a potential solution to the problem.

[0156] S4: Calculate the fitness function value of each individual according to the fitness function;

[0157] S5: According to the fitness function value of each individual, the individuals in the population are sorted by non-domination, and the individuals are divided into multiple levels; the first level contains individuals that are not dominated by other individuals;

[0158] S6: Calculate the dominating set and the number of dominated times of each individual; calculate the crowding degree between individuals in each level;

[0159] S7: Based on the non-dominated sorting and crowding degree, simulate the growth process of rime to update the individual position, generate new individuals, merge the newly generated individuals into the current population to form the next generation population, and repeat the above steps until the predetermined maximum number of evolutionary iterations or convergence conditions are met; among which, individuals with high non-dominated levels are preferred, and individuals with low crowding degrees are selected within the same level;

[0160] S8: Evaluate the target value of the optimization target generated for each solution, and stop by iterative calculation until the maximum number of evolutionary iterations or the convergence condition (average change is less than 0.00001) is met.

[0161] S9: The individuals in the final population are used as a set of Pareto frontier solutions found by the NSROA algorithm to obtain the Pareto optimization results. The Pareto non-dominated solution set after the iteration is completed includes a total of 100 sets of design parameters and optimization objectives.

[0162] According to the actual project situation, the design parameters and optimization objectives that deviate from the actual project in the Pareto optimization results are eliminated, and the passive design parameters and optimization objectives of the zero-carbon building are determined, including:

[0163] S1: The optimal value of the optimization objective on the Pareto frontier constitutes an ideal point;

[0164] S2: Use the following formula to calculate the distance from the ideal point to each solution point in the Pareto optimization result;

[0165]

[0166] Among them, t b1 , t b2 , t bn is the coordinate value of each solution point in the Pareto optimization result, t b , t b2 , t bn is the coordinate value of the ideal point;

[0167] S3: The solution point with the smallest distance is taken as the optimal solution, and the design parameters and optimization objectives of the zero-carbon building are determined based on the optimal solution.

[0168] The method of the above embodiment combines the rime optimization algorithm with the non-dominated sorting algorithm to obtain an improved rime optimization algorithm, so that the improved rime optimization algorithm can be used to more efficiently and accurately determine the optimal solution in the multi-objective optimization process, and thus can more efficiently and accurately determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters, thereby improving the efficiency and accuracy of the auxiliary design of zero-carbon buildings.

[0169] For example, the present application also provides a zero-carbon building auxiliary design decision device such as Figure 2 As shown. It includes selection module, input module, training module, update module, test module and optimization module. Figure 3The selection module shown is used to determine the optimization objectives and design parameters of zero-carbon buildings; the multi-objective optimization model includes optimization objectives, design parameters and constraints, and the constraints represent the value range of the design parameters; the optimization objectives include annual operating carbon emissions per unit area of ​​the building, annual operating heating energy consumption per unit area of ​​the building, annual operating cooling energy consumption per unit area of ​​the building, incremental cost per unit area, annual discomfort hours of the building, natural satisfaction rate of indoor lighting in the building, photovoltaic power generation self-sufficiency rate, and rooftop photovoltaic power generation; the design parameters include external window heat transfer coefficient, ground heat transfer coefficient, roof heat transfer coefficient, east wall window-to-wall ratio, west wall window-to-wall ratio, south wall window-to-wall ratio, north wall window-to-wall ratio, external window shading coefficient, wall heat transfer coefficient, solar heat gain coefficient SHGC value, ventilation times, building orientation, permeability, roof photovoltaic module installation inclination angle, roof photovoltaic module installation area, roof photovoltaic module installation orientation, roof photovoltaic cell type, and energy storage battery type.

[0170] like Figure 4 The input module shown is used to establish a geometric model of the building according to the geometric dimensions and envelope structure parameters of the building, and import the geometric model into the building energy consumption simulation software to simulate the optimization target under different design parameter conditions to obtain a data sample set.

[0171] like Figure 5 The training module shown is used to import the training set into the improved Grey Goose Optimization (GGO) and BiLSTM algorithms for training. It establishes a mapping relationship between design parameters and optimization objectives, and uses this mapping relationship as a proxy model for the zero-carbon building multi-objective optimization function. The improved Grey Goose Optimization (GGO) algorithm optimizes the structural parameters of the BiLSTM algorithm. During the optimization process, the proxy model is further checked to evaluate the training results and determine whether they meet the requirements. If they do not meet the requirements, the update module is used. If they do meet the requirements, the trained model is output.

[0172] When the zero-carbon building multi-objective optimization function proxy model established by the training module does not meet the requirements, the update module is entered to cyclically update the deep learning algorithm model parameters and enter cyclic training until the trained model meets the usage accuracy requirements and the cycle ends.

[0173] like Figure 6 The test module shown is used to test the trained proxy model using the test set to obtain the performance evaluation indicators of the test set; among them, the performance evaluation indicators include root mean square error RMSE, mean absolute error MAE and R 2 Indicators; Based on the performance evaluation indicators, the trained proxy model is evaluated and finally the proxy model is obtained.

[0174] like Figure 7The optimization module shown is used to use the surrogate model as the objective function of the multi-objective optimization algorithm, and use the non-dominated sorting rime optimization algorithm to optimize the zero-carbon building design parameters and optimization objectives to obtain Pareto optimization results; the Pareto optimization results include the optimized design parameters and the optimization objectives corresponding to the optimized design parameters; the design parameters and optimization objectives that deviate from the actual project in the Pareto optimization results are eliminated to determine the design parameters and optimization objectives of the zero-carbon building, which are used to assist architectural designers in zero-carbon building design. Optionally, the surrogate model is used as the objective function of the non-dominated sorting rime optimization algorithm (NSROA), and the improved rime NSROA algorithm is used to optimize the zero-carbon building design parameters and optimization objectives to obtain Pareto optimization results; the Pareto optimization results include the optimized design parameters and the optimization objectives corresponding to the optimized design parameters. That is, the improved rime NSROA algorithm is used to optimize the zero-carbon building design parameters and optimization objectives, and the final Pareto optimization results are a set of Pareto frontier solutions.

[0175] like Figure 8 The schematic diagram of the Pareto frontier solution provided by the present invention is shown. This set of solutions is non-dominated under multiple objective functions, that is, no solution to any objective function is better than it, and no solution to any other objective function is inferior to it.

[0176] The device of the above embodiment innovatively establishes an agent model based on a metaheuristic algorithm and a bidirectional long short-term memory network BiLSTM. Compared with agent models of other types and structures, it can more quickly, accurately and effectively determine the optimized design parameters of the zero-carbon building and the optimization objectives corresponding to the optimized design parameters, thereby realizing the multi-objective optimization design of the zero-carbon building and making the multi-objective design process of the zero-carbon building more scientific and accurate.

[0177] The zero-carbon building auxiliary design device provided by the present invention is described below. The zero-carbon building auxiliary design device described below and the zero-carbon building auxiliary design method described above can be referenced to each other.

[0178] Figure 9 Schematic diagram of the structure of the zero-carbon building auxiliary design device provided by the present invention. The zero-carbon building auxiliary design device provided in this embodiment includes:

[0179] A selection module 910 is used to determine multiple optimization objectives and design parameters of a zero-carbon building;

[0180] The sample module 920 is used to simulate multiple optimization target values ​​corresponding to different design parameter values ​​based on the geometric model of the zero-carbon building, multiple optimization targets and design parameters, and obtain a target sample set;

[0181] A training module 930 is configured to train the proxy model based on the target sample set to obtain a trained proxy model; the proxy model is established based on a meta-heuristic algorithm and a bidirectional long short-term memory network (BiLSTM);

[0182] The optimization module 940 is used to use the trained proxy model as the objective function to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters are used to assist in the design of the zero-carbon building.

[0183] The device of the embodiment of the present invention is used to execute the method in any of the aforementioned method embodiments. Its implementation principle and technical effects are similar and will not be repeated here.

[0184] Figure 10 The following is a schematic diagram of the physical structure of an electronic device, which may include: a processor (processor) 1010, a communication interface (communications interface) 1020, a memory (memory) 1030, and a communication bus 1040. The processor 1010, the communication interface 1020, and the memory 1030 communicate with each other via the communication bus 1040. The processor 1010 can call the logic instructions in the memory 1030 to execute the zero-carbon building auxiliary design method.

[0185] In addition, the logic instructions in the memory 1030 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0186] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the zero-carbon building auxiliary design method provided by the above methods.

[0187] In yet another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is configured to execute the above-mentioned zero-carbon building auxiliary design methods when executed by a processor.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. 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 zero-carbon building auxiliary design method, characterized in that: include: Determining multiple optimization objectives and design parameters for a zero-carbon building; the optimization objectives include at least one of the following: a carbon emission target, an energy consumption index, a cost index, and a photovoltaic power generation index; and the design parameters include at least one of the following: a heat transfer coefficient, a window-to-wall ratio, and photovoltaic module installation parameters; According to the geometric model of the zero-carbon building, the multiple optimization objectives and the design parameters, simulating multiple optimization target values ​​corresponding to different design parameter values ​​to obtain a target sample set; Training the proxy model according to the target sample set to obtain a trained proxy model; The agent model is established based on a meta-heuristic algorithm and a bidirectional long short-term memory network BiLSTM; Using the trained agent model as an objective function, determining the optimized design parameters of the zero-carbon building and the optimization objectives corresponding to the optimized design parameters; The optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters are used to assist in the design of the zero-carbon building; The method of using the trained proxy model as an objective function to determine the optimized design parameters of the zero-carbon building and the optimization objectives corresponding to the optimized design parameters includes: The trained proxy model is used as the objective function, and the improved rime optimization algorithm is used to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; wherein the improved rime optimization algorithm is obtained by combining the rime optimization algorithm with the non-dominated sorting algorithm; the improved rime optimization algorithm is used to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the improved rime optimization algorithm is obtained by combining the rime optimization algorithm with the non-dominated sorting algorithm, including: performing non-dominated sorting on individuals in the population according to the fitness function value of each individual, and dividing the individuals into multiple levels; calculating the dominating set and the number of dominated times of each individual; calculating the crowding degree between individuals in each level; simulating the growth process of rime according to the non-dominated sorting and crowding degree to update the individual positions, generate new individuals, and merge the newly generated individuals into the current population to form the next generation population; the population is a zero-carbon building design scheme, and each individual in the population represents a potential solution.

2. The zero-carbon building auxiliary design method according to claim 1, characterized in that: The proxy model is established based on the improved Grey Goose optimization algorithm and the bidirectional long short-term memory network BiLSTM; The process of updating individual positions in the improved grey goose optimization algorithm includes: Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide a global optimal solution; the positions of the exploration group individuals are updated based on the following method: X i+1 =X i -M×|N×X best -X i | Among them, X i represents the position of an individual in the current flock; the position of the individual represents the optimized solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 Represents the position of the individual after update; M represents the first step length; N represents the second step length; the first step length and the second step length are used to dynamically adjust the distance and direction of individual movement.

3. The zero-carbon building auxiliary design method according to claim 1, characterized in that: The proxy model is established based on the improved Grey Goose optimization algorithm and the bidirectional long short-term memory network BiLSTM; The process of updating individual positions in the improved grey goose optimization algorithm includes: Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide a global optimal solution; the positions of the exploration group individuals are updated based on the following method: X (i+1) =γ1×X best +γ2×(X rand1 -X rand2 )+(1-γ3)×X i Among them, X i represents the position of an individual in the current flock; the position of the individual represents the optimized solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after the update; γ1, γ2 and γ3 represent random factors, which are used to control the degree of randomness in the individual update process; X rand1 and X rand2 represents the positions of two different individuals chosen randomly.

4. The zero-carbon building auxiliary design method according to claim 1, characterized in that: The proxy model is established based on the improved Grey Goose optimization algorithm and the bidirectional long short-term memory network BiLSTM; The process of updating individual positions in the improved grey goose optimization algorithm includes: Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide a global optimal solution; the positions of the development group individuals are updated based on the following method: X (i+1) =X i +D×(1+ζ)×γ×(X best -X i ) Among them, X i represents the position of an individual in the current flock; the position of the individual represents the optimized solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after update; D represents the update step size, which is used to dynamically adjust the search step size; ζ and γ represent random factors.

5. The zero-carbon building auxiliary design method according to claim 1, characterized in that: The proxy model is established based on the improved Grey Goose optimization algorithm and the bidirectional long short-term memory network BiLSTM; The process of updating individual positions in the improved grey goose optimization algorithm includes: Individuals in the grey goose optimization algorithm are divided into exploration group individuals and development group individuals; the exploration group is used to maximize the development of the search space; the development group is used to provide a global optimal solution; the positions of the development group individuals are updated based on the following method: Among them, X i represents the position of an individual in the current flock; the position of the individual represents the optimized solution of a set of BiLSTM algorithm parameters in the agent model; X best Indicates the location of the current best solution; X i+1 represents the position of the individual after the update; D represents the update step size, which is used to dynamically adjust the search step size; γ represents the random factor; X sentry1 and X sentry2 Represents the two individuals in the goose flock that are ranked first in addition to the current best solution.

6. A zero-carbon building auxiliary design device, characterized in that: include: A selection module is configured to determine multiple optimization objectives and design parameters for a zero-carbon building; the optimization objectives include at least one of the following: a carbon emission target, an energy consumption index, a cost index, and a photovoltaic power generation index; and the design parameters include at least one of the following: a heat transfer coefficient, a window-to-wall ratio, and photovoltaic module installation parameters; A sample module is used to simulate multiple optimization target values ​​corresponding to different design parameter values ​​according to the geometric model of the zero-carbon building, the multiple optimization targets and the design parameters, to obtain a target sample set; A training module, configured to train the proxy model according to the target sample set to obtain a trained proxy model; the proxy model is established based on a meta-heuristic algorithm and a bidirectional long short-term memory network (BiLSTM); an optimization module, configured to use the trained proxy model as an objective function to determine optimized design parameters of the zero-carbon building and optimization objectives corresponding to the optimized design parameters; The optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters are used to assist in the design of the zero-carbon building; The method of using the trained proxy model as an objective function to determine the optimized design parameters of the zero-carbon building and the optimization objectives corresponding to the optimized design parameters includes: The trained proxy model is used as the objective function, and the improved rime optimization algorithm is used to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; wherein the improved rime optimization algorithm is obtained by combining the rime optimization algorithm with the non-dominated sorting algorithm; the improved rime optimization algorithm is used to determine the optimized design parameters of the zero-carbon building and the optimization targets corresponding to the optimized design parameters; the improved rime optimization algorithm is obtained by combining the rime optimization algorithm with the non-dominated sorting algorithm, including: performing non-dominated sorting on individuals in the population according to the fitness function value of each individual, and dividing the individuals into multiple levels; calculating the dominating set and the number of dominated times of each individual; calculating the crowding degree between individuals in each level; simulating the growth process of rime according to the non-dominated sorting and crowding degree to update the individual positions, generate new individuals, and merge the newly generated individuals into the current population to form the next generation population; the population is a zero-carbon building design scheme, and each individual in the population represents a potential solution.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the zero-carbon building auxiliary design method according to any one of claims 1 to 5 are implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the zero-carbon building auxiliary design method according to any one of claims 1 to 5 are implemented.

Images