A method for building performance analysis and optimization by fusing local climate zones
By generating high-precision meteorological files through local climate zone classification and measured data, and combining CNN-MLP models and multi-objective optimization algorithms, the problems of climate input bias and coupling in building performance analysis are solved, enabling rapid and accurate optimization and parameter recommendation of building design.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV OF TECH
- Filing Date
- 2026-07-07
- Publication Date
- 2026-08-04
AI Technical Summary
Existing building performance analysis tools suffer from several drawbacks when dealing with complex urban environments, including homogenization bias in climate input data, weak coupling between urban form and building performance, high optimization costs and lack of interpretability, and a lack of climate adaptability in design tools. These issues lead to inaccurate analysis of building energy consumption and thermal comfort.
By combining local climate zone classification with measured data, high-precision meteorological files are generated through an urban weather generator. Urban morphological features are extracted using a CNN-MLP model, and a building performance proxy model is constructed. Combined with multi-objective optimization algorithms and SHAP analysis, rapid multi-objective optimization and parameter recommendation are achieved.
It achieves high accuracy and low cost in building performance simulation, provides design parameter recommendations for different local climate zones, meets the need for rapid decision-making in the early stages of design, and improves the climate adaptability and interpretability of the design.
Smart Images
Figure CN122509033A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of building performance optimization and urban climate simulation technology, specifically involving a method for building performance analysis and optimization that integrates local climate zones. Background Technology
[0002] With the acceleration of urbanization, changes in the properties of urban underlying surfaces have led to significant urban heat island (UHI) effects and ventilation obstruction, resulting in high spatiotemporal heterogeneity in building energy consumption and thermal comfort. As major energy consumers, cities face immense pressure to reduce carbon emissions in their building sectors. However, existing building performance analysis and design methods still encounter the following technical challenges when dealing with complex urban environments: 1. Homogenization bias of climate input data
[0003] Current building performance simulation (BPS) tools generally rely on typical meteorological year (TMY) data. TMY data usually comes from suburban or airport weather stations and cannot capture the microclimate differences within cities caused by variations in building density, height, and surface materials.
[0004] 2. Weak coupling analysis of urban morphology and building performance
[0005] Existing studies mostly use statistical methods to establish the relationship between urban morphology (such as floor area ratio and sky visibility factor SVF) and building energy consumption, but lack coupled models driven by physical mechanisms. Traditional methods treat urban morphology as a static parameter and fail to quantify its moderating effect on near-surface temperature, humidity, and wind speed through microclimate models, resulting in a lack of cross-scale coordination between buildings, neighborhoods, and the city.
[0006] 3. Building performance optimization typically involves trade-offs among multiple objectives such as energy consumption, thermal comfort, and carbon emissions, requiring extensive simulation and evaluation. However, high-fidelity simulations (such as EnergyPlus) are time-consuming per run, and massive iterations cannot meet the rapid decision-making needs in the early stages of design. Although surrogate models (such as ANNs) are used to accelerate optimization, existing research mostly adopts a "black box" approach, lacking interpretable analysis. Designers cannot understand the impact mechanism of key parameters (such as window-to-wall ratio and green space configuration) on performance, making it difficult to implement optimization results.
[0007] 4. Lack of climate adaptability in design tools
[0008] Existing architectural design tools do not integrate the Local Climate Zone (LCZ) classification system and cannot provide differentiated parameter recommendations for different LCZ types.
[0009] In summary, there is an urgent need for an integrated approach that combines LCZ classification, microclimate simulation, and intelligent optimization to overcome the bottlenecks of traditional technologies, such as homogeneous climate input, weakly coupled analysis, high-cost optimization, and lack of climate adaptability, in order to achieve accurate prediction of building performance and low-carbon design at the urban scale. Summary of the Invention
[0010] To address the aforementioned problems in the existing technology, the purpose of this invention is to provide a method for analyzing and optimizing building performance that integrates local climate zones.
[0011] This invention provides the following technical solution: a method for analyzing and optimizing building performance based on local climate zones, comprising the following steps: Step 1: Based on the Local Climate Zone (LCZ) classification system, obtain and classify the urban morphology data of the target area to determine the local climate zone type of the target area; Step 2: Collect measured meteorological data of the target area and obtain macro-meteorological data corresponding to the target area; Step 3: Based on the Urban Weather Generator (UWG) model, local climate and meteorological files are generated by combining macro-meteorological data and urban morphology data. These files are then validated using measured meteorological data. The local climate and meteorological files with the highest degree of agreement with the measured meteorological data are selected as the target climate input data. Step 4: Construct a parametric idealized building model, set the range of values for building design parameters, conduct building performance simulation based on target climate input data, and obtain building performance indicators under different combinations of design parameters; Step 5: Construct a training dataset based on the combination of architectural design parameters and their corresponding architectural performance indicators, and train the architectural performance proxy model. Step 6: Based on the building performance proxy model, a multi-objective optimization algorithm is used to find the Pareto optimal solution set; Step 7: Construct an LCZ-building performance database based on the results of multi-scenario simulations, and provide parameter recommendations for building designs under the corresponding LCZ type based on the database.
[0012] Furthermore, the specific process of obtaining and classifying the urban morphology data of the target area in step 1 is as follows: 1) Acquire remote sensing image data and building vector data of the target area; 2) Construct a CNN-MLP hybrid model, which includes a Convolutional Neural Network (CNN) module and a Multilayer Perceptron (MLP) module; 3) Use the CNN module to extract the spatial structural features of buildings from remote sensing image data, and use the MLP module to extract the geometric features of buildings from building vector data; 4) Integrate spatial structural features and geometric features, output building form classification results through the classification decision layer, and determine the local climate zone type of the target area based on the building form classification results.
[0013] Furthermore, the macro-meteorological data includes typical meteorological year TMY data and measured meteorological year AMY data; the specific process of processing the macro-meteorological data and urban morphology data based on the urban weather generator UWG model in step 3 to generate local climate and meteorological files is as follows: TMY and AMY data were used as rural reference meteorological inputs, and combined with parameters such as building density, building height, facade-to-site ratio, and vegetation coverage from urban morphology data, simulations were performed using the UWG model to generate TMY+UWG meteorological files and AMY+UWG meteorological files.
[0014] Furthermore, the specific process of using the measured meteorological data to verify the local climate meteorological files in step 3, and selecting local climate meteorological files whose accuracy meets the preset requirements as the target climate input data, is as follows: 1) Calculate the error index between measured meteorological data and multiple candidate meteorological files; 2) Based on error indices, the meteorological file with the highest degree of agreement with the measured meteorological data is selected as the target climate input data. The error indices include root mean square error (RMSE) and mean absolute percentage error (MAPE).
[0015] Furthermore, the building performance indicators in step 4 include greenhouse gas emissions (GHG), outdoor discomfort duration (PODH), and indoor discomfort duration (PIDH); the outdoor discomfort duration (PODH) is calculated based on the Universal Thermal Climate Index (UTCI), the indoor discomfort duration (PIDH) is calculated based on the Advanced Thermal Comfort Model (ACM), and the greenhouse gas emissions (GHG) include carbon emissions during the building's operation phase.
[0016] Furthermore, the specific process of constructing a training dataset based on the combination of building design parameters and the corresponding building performance indicators in step 5, and training the building performance proxy model based on the training dataset, is as follows: 1) The Latin hypercube sampling method is used to sample architectural design parameters and generate multiple sets of architectural design parameter combinations; 2) Use a building energy consumption simulation engine to simulate the building performance of each combination of building design parameters, obtain the corresponding building performance indicators, and form a training dataset; 3) Based on the training dataset, train an artificial neural network (ANN) model or an extreme gradient boosting (XGBoost) model to obtain a building performance proxy model.
[0017] Furthermore, after obtaining the Pareto optimal solution set in step 6, the process also includes: The SHAP interpretability analysis framework is used to analyze the building performance proxy model and quantify the global sensitivity and interactive effects of each building design parameter on the building performance index. Based on the results of the global sensitivity analysis, key design parameters affecting the building performance indicators are identified.
[0018] Furthermore, the multi-objective optimization algorithm in step 6 is the non-dominated sorting genetic algorithm NSGA-II, and the Pareto optimal solution set is the set of solutions that balances greenhouse gas emissions, outdoor discomfort duration, and indoor discomfort duration.
[0019] By employing the above-described technology, the beneficial effects of the present invention compared to the prior art are as follows: 1) This invention uses the Urban Weather Generator (UWG) combined with measured data to perform localized correction on macro-meteorological data, generating high-precision local climate files. This reduces the prediction error of traditional typical meteorological year TMY data under the urban heat island effect and provides realistic boundary conditions for building performance simulation.
[0020] 2) This invention is based on a physical mechanism-driven urban canopy model, which quantifies the regulatory effect of urban morphology on microclimate and its feedback impact on building energy consumption and thermal comfort, and solves the problem of weak coupling between urban design and building performance analysis.
[0021] 3) This invention uses Latin hypercube sampling and machine learning surrogate model ANN / XGBoost to replace the energy-intensive repeated simulation calculations, and combines the NSGA-II algorithm to achieve fast multi-objective optimization. While ensuring accuracy, it reduces the computational cost by several orders of magnitude, meeting the needs of rapid decision-making in the early stages of design.
[0022] 4) This invention constructs the LCZ-building performance database and, combined with SHAP interpretability analysis, not only outputs the Pareto optimal solution set for greenhouse gas emissions and thermal comfort, but also clarifies the sensitivity and optimization direction of key design parameters under different local climate zones, providing a quantitative basis for climate-adaptive design. Attached Figure Description
[0023] Figure 1 A flowchart illustrating the LMUPO research framework provided in this embodiment of the invention; Figure 2 This is a flowchart of the urban morphology extraction process based on the CNN-MLP hybrid model in an embodiment of the present invention; Figure 3 This is a flowchart illustrating the UWG model architecture and local climate file generation verification process in an embodiment of the present invention. Figure 4This is a flowchart illustrating the execution of parameterized simulation and multi-scenario analysis in an embodiment of the present invention. Figure 5 This is a schematic diagram of the Pareto distribution of point-type buildings in an embodiment of the present invention; Figure 6 This is a schematic diagram of the Pareto distribution of the slab building in an embodiment of the present invention; Figure 7 This is a schematic diagram of the Pareto distribution of the U-shaped building in an embodiment of the present invention; Figure 8 This is a comparison chart of key design parameter optimization strategies under different LCZ types in the embodiments of the present invention. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0025] Conversely, this invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of the invention as defined in the claims. Furthermore, to provide a better understanding of the invention, certain specific details are described in detail below. However, those skilled in the art will fully understand the invention even without these detailed descriptions.
[0026] Example: A method for building performance analysis and optimization that integrates local climate zones. This example takes the central urban area of city T as an example, as detailed below: Step 1: Urban morphology extraction and classification based on LCZ
[0027] Reference Figure 1 and Figure 2 , Figure 1 This is a flowchart of the LMUPO research framework of the present invention, which shows the overall process from LCZ classification, local climate file generation, parametric simulation, multi-objective optimization to database construction; Figure 2 This is a flowchart of the urban morphology extraction and classification process based on the CNN-MLP hybrid model, which shows the process of fusing spatial features of remote sensing images and vector geometric features to output LCZ type.
[0028] First, the study area was defined as the central urban area of city T, and remote sensing images with a resolution of 1m and building vector data (including outlines and heights) were integrated.
[0029] To accurately identify building forms, a CNN-MLP hybrid model was constructed. Rasterized images of the building outline were input into the CNN module (based on the ResNet framework) to extract spatial structural features. Simultaneously, geometric features such as building area, number of corners, and aspect ratio were calculated from vector data and input into the MLP module. After feature fusion, classification results were output, determining that the target area mainly covers types such as LCZ1 (compact high-rise), LCZ2 (compact mid-rise), and LCZ3 (compact low-rise).
[0030] Step 2: Field Measurements and Local Climate File Generation
[0031] Reference Figure 3 The measured data input section on the left shows that from June 27 to July 3, 2025, AZ-8917SENTRY anemometers and AZ-87786 black ball thermometers were deployed in five typical LCZ clusters (AEs) in city T to collect measured meteorological data such as temperature, humidity, and wind speed at a height of 1.5m, as shown in Table 1: Table 1 Distribution of measured data collection points and correspondence of LCZ types
[0032] Simultaneously, macro-meteorological data were acquired: Typical Meteorological Year (TMY) data from LadybugTools and Observed Meteorological Year (AMY) data from Oikolab.
[0033] The TMY and AMY data mentioned above both originate from meteorological base stations on the outskirts of city A (i.e., Figure 3 The rural station model in the lower left corner represents the background climate field that is not significantly affected by the high-density urban morphology, and meets the consistency requirements of the UWG model for rural reference meteorological inputs.
[0034] Reference Figure 3 The data processing workflow for climate calculation and vertical diffusion in central rural areas utilizes the Dragonfly tool on the Grasshopper platform. Urban morphology parameters (building density, height, vegetation cover, etc.) extracted in step 1 are input, and the UWG model is run. TMY and AMY are used as rural background inputs respectively to generate corrected local climate and meteorological files: TMY+UWG and AMY+UWG.
[0035] Step 3: Data Validation and Filtering
[0036] Reference Figure 3 The right-hand urban boundary layer and canopy-building energy consumption output process compares and verifies the generated four types of meteorological datasets (TMY, AMY, TMY+UWG, AMY+UWG) with the measured data of the LCZ cluster within city A collected in step 2. The root mean square error (RMSE) and mean absolute percentage error (MAPE) are calculated.
[0037] Based on the error indices listed in Table 2, the AMY+UWG simulated values show the highest agreement with measured data in terms of air temperature (TA) (e.g., in the LCZ3 region, TA-RMSE = 2.28°C, MAPE = 6.52%). Therefore, AMY+UWG was selected as the target climate input data for subsequent simulations, thus resolving the homogenization bias of traditional TMY data. Table 2 Comparison of statistical parameters between simulated and measured data ; Here, RH represents relative humidity.
[0038] Step 4: Parametric Building Performance Simulation
[0039] Construct idealized building models encompassing point-type, slab-type, and U-type structures. Design parameters strictly adhere to Table 3, specifically including: building length (BL) 35-60m, floor height (SH) 2.7-3.4m, south-facing window-to-wall ratio (SWWR) 0.25-0.85, solar thermal gain coefficient (SHGC) 0.3-0.9, and green space spacing (TS) 4-10m, etc. Table 3 Idealized Model Parameter Settings
[0040] Based on the selected AMY+UWG climate file, large-scale parametric simulations were performed using Latin hypercube sampling on the OpenStudio / EnergyPlus platform. (Refer to...) Figure 4 This demonstrates the specific execution flow of the parametric simulation and multi-scenario analysis. It involves generating a large number of design parameter combinations through sampling and inputting them into a building energy consumption simulation engine. The system then calculates building carbon emissions and indoor / outdoor thermal comfort indices under different building types and local climate zones. Finally, it calculates greenhouse gas emissions (GHG), outdoor discomfort duration (PODH), and indoor discomfort duration (PIDH) under different parameter combinations.
[0041] Step 5: Establish a building performance proxy model
[0042] The parameter combinations generated in step 4 and their corresponding performance metrics are used to construct a training dataset. Considering that the target area may cover multiple local climate zone types (including but not limited to LCZ1, LCZ2, LCZ3, LCZ6, LCZ9, LCZ10, etc.), a multilayer feedforward neural network (ANN) is used to construct a surrogate model.
[0043] To quantitatively evaluate the prediction accuracy of the surrogate model, the following fitting evaluation index is introduced: Mean Absolute Error (MAE) represents the average of the absolute values of the differences between predicted and actual values, reflecting the average degree of deviation of the model's predictions. Mean Squared Error (MSE) represents the mean of the sum of squares of the differences between predicted and true values. It is more sensitive to larger errors, and the smaller the value, the stronger the model stability. The coefficient of determination (R²) characterizes the model's ability to explain sample variability. Its value ranges from negative infinity to 1. The closer R² is to 1, the better the model fits and the higher the reliability of the prediction.
[0044] The ANN model contains three hidden layers, each with 64 neurons. The ReLU activation function is used, and overfitting is controlled using early stopping. The accuracy evaluation results on the test set are shown in Table 4. These results demonstrate that the model's prediction accuracy meets the requirements for replacing high-energy-consuming simulations under various urban morphologies, including LCZ1, LCZ2, LCZ3, LCZ6, LCZ9, and LCZ10. For point-type buildings under LCZ3, GHG predicts R² = 0.984, and PODH predicts R² = 0.983; For slab-type buildings under LCZ2, GHG predicts R² = 0.974; For U-shaped buildings under LCZ1, GHG predicts R² = 0.987.
[0045] The above data proves that the accuracy of the surrogate model meets the requirements for replacing high-energy-consuming simulations: Table 4 Fitting Indices of Artificial Neural Network Models
[0046] Step 6: Global Sensitivity Analysis and Multi-Objective Optimization
[0047] Reference Figures 5-7 , Figure 5 This is a schematic diagram of the Pareto distribution of a point building in an embodiment of the present invention. It shows the Pareto optimal solution set distribution of the point building obtained by NSGA-II multi-objective optimization in three-dimensional space. The three coordinate axes correspond to greenhouse gas emissions, outdoor discomfort duration and indoor discomfort duration, respectively, showing the trade-off relationship between the three objectives and the non-dominated solution frontier. Figure 6 This is a schematic diagram of the Pareto distribution of a slab building in an embodiment of the present invention. It shows the Pareto optimal solution set distribution obtained by NSGA-II multi-objective optimization of the slab building, reflecting the non-dominated solution distribution characteristics between carbon emissions and indoor and outdoor thermal comfort. Figure 7This is a schematic diagram of the Pareto distribution of the U-shaped building in an embodiment of the present invention. It shows the Pareto optimal solution set distribution obtained by NSGA-II multi-objective optimization of the U-shaped building, reflecting the trade-off relationship of the three objectives under different combinations of design parameters.
[0048] First, based on the training dataset generated in step 4, a global sensitivity analysis was performed using the XGBoost algorithm combined with the SHAP (SHapley Additive exPlanations) analysis framework. By calculating the SHAP values of each input parameter, the main effects and interactive influences on the output performance indicators (GHG, PODH, PIDH) were quantified, and the solar thermal gain coefficient (SHGC), paving materials (GM), and greening configuration (TS) were identified as key design parameters affecting building performance.
[0049] Based on the key parameter paths identified in the sensitivity analysis above, and combined with the ANN surrogate model trained in step 5, a non-dominated sorting genetic algorithm with an elitist strategy (NSGA-II) is used for multi-objective optimization. The objective function is to minimize greenhouse gas emissions (GHG), outdoor discomfort duration (PODH), and indoor discomfort duration (PIDH), with constraints strictly following the parameter ranges set in Table 3 of step 4 (such as building height, window-to-wall ratio, etc.). The obtained Pareto optimal solution set shows the trade-offs among the three objectives; for example, the GHG of the LCZ3 point building corresponds to a specific PIDH value.
[0050] Step 7: Application of Results and Database Construction
[0051] Reference Figure 8 , Figure 8 This is a comparison chart of key design parameter optimization strategies under different LCZ types in the embodiments of the present invention. It shows the differences in recommended values for building height, tree spacing, paving materials, and solar heat gain coefficient for point-type, slab-type, and U-type building forms under different LCZ types, with the optimization goal of minimizing carbon emissions and improving outdoor or indoor thermal comfort.
[0052] Based on the Pareto optimal solution set obtained in step 6, and according to actual engineering design requirements (such as prioritizing the minimization of carbon emissions), representative compromise solutions are selected to construct a local climate zone-building performance database. Specific design parameter recommendations are provided for different LCZ types and building forms, as shown in Table 5. Table 5 provides specific design parameter recommendations for different LCZ types and building forms.
[0053] in, This indicates that, under this specific evaluation metric, this set of parameters is the optimal solution or Pareto optimal solution selected by a multi-objective optimization algorithm.
[0054] Designers can input the LCZ type of the project and the target building form, and then directly retrieve the verified parameter recommendation schemes from the table above.
[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for building performance analysis and optimization integrating local climate zones, characterized in that, Includes the following steps: Step 1: Based on the Local Climate Zone (LCZ) classification system, obtain and classify the urban morphology data of the target area to determine the local climate zone type of the target area; Step 2: Collect measured meteorological data of the target area and obtain macro-meteorological data corresponding to the target area; Step 3: Based on the Urban Weather Generator (UWG) model, local climate and meteorological files are generated by combining macro-meteorological data and urban morphology data. These files are then validated using measured meteorological data. The local climate and meteorological files with the highest degree of agreement with the measured meteorological data are selected as the target climate input data. Step 4: Construct a parametric idealized building model, set the range of values for building design parameters, conduct building performance simulation based on target climate input data, and obtain building performance indicators under different combinations of design parameters; Step 5: Construct a training dataset based on the combination of architectural design parameters and their corresponding architectural performance indicators, and train the architectural performance proxy model. Step 6: Based on the building performance proxy model, a multi-objective optimization algorithm is used to find the Pareto optimal solution set; Step 7: Construct an LCZ-building performance database based on the results of multi-scenario simulations, and provide parameter recommendations for building designs under the corresponding LCZ type based on the database.
2. The method for building performance analysis and optimization integrating local climate zones as described in claim 1, characterized in that, The specific process of obtaining and classifying the urban morphology data of the target area in step 1 is as follows: 1) Acquire remote sensing image data and building vector data of the target area; 2) Construct a CNN-MLP hybrid model, which includes a Convolutional Neural Network (CNN) module and a Multilayer Perceptron (MLP) module; 3) Use the CNN module to extract the spatial structural features of buildings from remote sensing image data, and use the MLP module to extract the geometric features of buildings from building vector data; 4) Integrate spatial structural features and geometric features, output building form classification results through the classification decision layer, and determine the local climate zone type of the target area based on the building form classification results.
3. The method for building performance analysis and optimization integrating local climate zones as described in claim 1, characterized in that, The macro-meteorological data includes typical meteorological year TMY data and measured meteorological year AMY data; the specific process of processing the macro-meteorological data and urban morphology data based on the Urban Weather Generator (UWG) model in step 3 to generate local climate and meteorological files is as follows: TMY and AMY data were used as rural reference meteorological inputs, and combined with parameters such as building density, building height, facade-to-site ratio, and vegetation coverage from urban morphology data, simulations were performed using the UWG model to generate TMY+UWG meteorological files and AMY+UWG meteorological files.
4. The method for building performance analysis and optimization integrating local climate zones as described in claim 1, characterized in that, The specific process of using the measured meteorological data to verify the local climate and meteorological files in step 3, and selecting local climate and meteorological files whose accuracy meets the preset requirements as the target climate input data, is as follows: 1) Calculate the error index between measured meteorological data and multiple candidate meteorological files; 2) Based on error indices, the meteorological file with the highest degree of agreement with the measured meteorological data is selected as the target climate input data. The error indices include root mean square error (RMSE) and mean absolute percentage error (MAPE).
5. The method for building performance analysis and optimization integrating local climate zones according to claim 1, characterized in that, The building performance indicators in step 4 include greenhouse gas emissions (GHG), outdoor discomfort duration (PODH), and indoor discomfort duration (PIDH). The outdoor discomfort duration (PODH) is calculated based on the Universal Thermal Climate Index (UTCI), and the indoor discomfort duration (PIDH) is calculated based on the Advanced Thermal Comfort Model (ACM). The greenhouse gas emissions (GHG) include carbon emissions during the building's operation phase.
6. The method for building performance analysis and optimization integrating local climate zones as described in claim 1, characterized in that, The specific process of constructing a training dataset based on the combination of building design parameters and the corresponding building performance indicators in step 5, and training the building performance proxy model based on the training dataset is as follows: 1) The Latin hypercube sampling method is used to sample architectural design parameters and generate multiple sets of architectural design parameter combinations; 2) Use a building energy consumption simulation engine to simulate the building performance of each combination of building design parameters, obtain the corresponding building performance indicators, and form a training dataset; 3) Based on the training dataset, train an artificial neural network (ANN) model or an extreme gradient boosting (XGBoost) model to obtain a building performance proxy model.
7. The method for building performance analysis and optimization integrating local climate zones as described in claim 6, characterized in that, After obtaining the Pareto optimal solution set in step 6, the following steps are also included: The SHAP interpretability analysis framework is used to analyze the building performance proxy model and quantify the global sensitivity and interactive effects of each building design parameter on the building performance index. Based on the results of the global sensitivity analysis, key design parameters affecting the building performance indicators are identified.
8. The method for building performance analysis and optimization integrating local climate zones according to claim 1, characterized in that, The multi-objective optimization algorithm in step 6 is the non-dominated sorting genetic algorithm NSGA-II, and the Pareto optimal solution set is the set of solutions that weighs the greenhouse gas emissions, outdoor discomfort duration, and indoor discomfort duration.