Parameter uncertainty-considered explainable building reconstruction rapid optimization method

Abstract

The invention relates to an interpretable building reconstruction rapid optimization method considering parameter uncertainty. The method comprises the following steps: S1, collecting information and carrying out data processing on original data; s2, building a building simulation model and calibrating weather data; s3, uncertain parameter sensitivity is analyzed, and a simulation model is calibrated; s4, building reconstruction measures are set, and reconstruction multi-objective optimization is carried out; s5, training a rapid prediction model and explaining a black box model decision mechanism; the method is based on an optimization algorithm, a prediction algorithm and a simulation kernel, and aims to solve the technical problems of uncertainty, simulation speed and interpretability in building reconstruction multi-objective optimization.

Description

Technical Field

[0001] This invention relates to an optimization method, specifically to a rapid optimization method for interpretable building renovation that takes into account parameter uncertainties, belonging to the field of urban building energy conservation technology. Background Technology

[0002] Cities consume over 60% of the world's primary energy and emit over 70% of greenhouse gases. Of this, newly constructed buildings account for only 1-3% of the existing building stock annually, meaning the majority of energy demand still comes from existing structures. Renovating existing buildings is of great significance for reducing energy consumption and improving residents' quality of life. Adopting active, passive, and renewable energy-based building renovation measures can effectively improve the performance of existing buildings. Decisions regarding building renovation schemes require extensive simulation using simulation models to assess their performance improvement potential.

[0003] However, uncertainties in building geometry, equipment parameters, and user behavior parameters lead to discrepancies between simulation results and actual building performance, directly impacting the effectiveness of building renovation schemes. Furthermore, the improvement of building performance through renovation measures and their combinations requires extensive performance simulation experiments, consuming significant computational resources and time. In addition, the complexity of physical simulation model calculations and the ambiguity of the decision-making mechanisms within data-driven models make it difficult to understand the complex relationship between renovation measures and building performance. Therefore, a new solution is urgently needed to address this technical problem. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing an interpretable rapid optimization method for building renovation that considers parameter uncertainty. This method is based on optimization algorithms, prediction algorithms, and simulation kernels, and aims to solve the technical challenges of uncertainty, simulation speed, and interpretability in multi-objective optimization of building renovation.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: a rapid optimization method for interpretable building renovation considering parameter uncertainties, characterized in that the method includes the following steps:

[0006] S1: Collect information and process the raw data;

[0007] S2: Build building simulation models and calibrate weather data;

[0008] S3: Sensitivity analysis of uncertain parameters and calibration of simulation model;

[0009] S4: Set building renovation measures and perform multi-objective optimization of the renovation;

[0010] S5: Train a fast prediction model and explain the decision-making mechanism of the black-box model.

[0011] S1 involves collecting information and processing the raw data. This step provides high-quality building energy consumption information and indoor thermal environment data for model calibration and multi-objective optimization during building renovation, significantly improving the accuracy of the simulation model and reducing uncertainties in building renovation assessments. Specifically:

[0012] S1.1: Collect data on building geometry, building envelope performance, and equipment performance through architectural design drawings, GIS platforms, and field surveys;

[0013] S1.2: Obtain building performance data, including electricity consumption and indoor thermal environment data, through energy management platforms, sensors, and field surveys;

[0014] S1.3: Outliers in building energy consumption and indoor thermal environment data are identified using the Z-score and interquartile range methods, and these outliers are treated as missing values. The relevant calculation formulas are as follows:

[0015]

[0016] If |Z|>3, i.e. (the 3σ rule), then it is considered an outlier, where x is the data point, μ is the mean of the data, and σ is the standard deviation of the data.

[0017] IQR = Q3 - Q1

[0018] If x is located in [Q1-1.5IQR, Q3+1.5IQR], it is considered an outlier, where Q1 is the 25th percentile and Q3 is the 75th percentile.

[0019] S1.4: Missing values ​​are filled using Lagrange interpolation to ensure the completeness of building performance data. The relevant calculation formula is as follows:

[0020]

[0021] Where L(x) is the interpolation result, x i y i Let n represent the independent (time point) and dependent variable values ​​of the data sample points, where n is the degree of the polynomial, equal to the data sample size minus 1, and l is the value of the dependent variable. i (x) is the i-th basis function, ensuring that the polynomial is in x. i The value y is taken at the location i The values ​​are set to 0 for other sample points. This step provides high-quality building energy consumption information and indoor thermal environment data for model calibration and multi-objective optimization during the building renovation process, which can significantly improve the accuracy of the simulation model and reduce the uncertainty in building renovation assessment.

[0022] S2 involves constructing a building simulation model and calibrating it using weather data. Weather data is a crucial uncertainty parameter affecting building performance. This step, based on measured microclimate data from meteorological stations and microclimate simulation calibration weather data, improves the accuracy of the building simulation model. Details are as follows:

[0023] S2.1: Based on building information collected from multiple sources, a building physical model is constructed using the Rhinocero and Grasshopper platforms;

[0024] S2.2: Based on the thermal performance of the building envelope, user behavior, building functions, and equipment performance information, a building simulation model is constructed using Ladybug plugins.

[0025] S2.3: Based on microclimate data collected from small weather stations, the microclimate is simulated using Dragonfly and the Urban Weather Generator (UWG), and the EPW format file of typical year climate data is modified.

[0026] S2.4: Build a building simulation model by writing a program on the Grasshopper platform, and call the Energyplus simulation core to perform performance simulation on the optimization goals of the building renovation, including building energy consumption and indoor thermal comfort duration.

[0027] S3 involves sensitivity analysis and calibration of the simulation model for uncertain parameters. Uncertain parameters in building simulation models can cause discrepancies between simulation results and actual performance, directly impacting building renovation decisions. This step identifies key uncertain parameters affecting building performance through sensitivity analysis, reducing the computational resources and time required for subsequent model calibration. This step further calibrates the key uncertain parameters using a single-objective optimization algorithm, quickly obtaining a large number of simulation models that meet accuracy requirements. This step improves the accuracy of simulation results and the speed of the optimization process through sensitivity analysis and model calibration. Details are as follows:

[0028] S3.1: Select the uncertain parameters that affect the optimization objective (such as thermal performance parameters of the building envelope, various time schedules, temperature setpoints, etc.), and determine the probability distribution (such as normal distribution, uniform distribution) and interval range of each parameter;

[0029] S3.2: Generate combined samples with uncertain parameters using Latin hypercube sampling (LHS), and obtain the performance simulation results of each combined sample using a building simulation model. The relevant calculation formulas are as follows:

[0030]

[0031] Where, x ij The j-th sample value of the i-th parameter, u j The numbers are uniformly random numbers in the range [0, 1]. The inverse cumulative distribution function of the parameter;

[0032] S3.3: Calculate the Sobol main effect index to identify the key parameters affecting building performance among the uncertain parameters, and distinguish between main effects and interaction effects. The relevant calculation formula is as follows:

[0033]

[0034] Among them, S i It is the main effect index of the input parameter, which measures the proportion of the parameter's contribution to the output variance. Its value ranges from [0, 1]. i It is the variance of the conditional expectation of the output, i.e., x i The variance component, interpreted separately, is V, which represents the total variance of the model output. It is parameter x i The total effect index, which includes the main effects and all interaction effects with other parameters, ranges from [0, 1]. It is the expected value of the output conditional variance when other parameters are fixed, reflecting x. i The contribution of their interactions to the total variance;

[0035] S3.4: Set the calibration range and step size for the identified key uncertainty parameters, and select the evaluation index for the simulation uncertainty of the evaluation model;

[0036] S3.5: Using Opssum to call a single-objective optimization algorithm, a program is written to calibrate the simulation model. Uncertain parameters in building simulation models cause discrepancies between simulation results and actual performance, directly impacting building renovation decisions. This step identifies key uncertain parameters affecting building performance through sensitivity analysis, reducing the computational resources and time required for subsequent model calibration. This step further calibrates the key uncertain parameters using a single-objective optimization algorithm, quickly obtaining a large number of simulation models that meet accuracy requirements. This step improves the accuracy of simulation results and the speed of the optimization process through sensitivity analysis and model calibration.

[0037] Step S4 involves setting building renovation measures and performing multi-objective optimization. This step utilizes a high-precision building simulation model to conduct multi-objective optimization of the building renovation, yielding a large amount of simulation result data. The Pareto solution set and its visualization results can be used for building renovation decisions, improving the interpretability of the decision results. Details are as follows:

[0038] S4.1: Set the building renovation measures and their renovation levels. Renovation measures include adding a thermal insulation layer to the building envelope, replacing high-performance windows, replacing high-performance HVAC systems, adding roof photovoltaic systems, etc. Further set the levels of different renovation measures and obtain the input parameters corresponding to each renovation level of different renovation measures.

[0039] S4.2: Select the optimization objectives for the building renovation, use the Grasshopper platform to call the simulation model to evaluate the optimization effect after the renovation, and connect to the Wallacei plugin. Optimization objectives include building energy consumption, indoor thermal comfort duration, and net present value;

[0040] S4.3: Set the modification level of the modification measures through the gene pool battery and connect it to the Wallacei plugin;

[0041] S4.4: A multi-objective optimization genetic algorithm is invoked via the Wallacei plugin to perform multi-objective optimization of building renovation, using renovation measures as genes. This step conducts multi-objective optimization of building renovation based on a high-precision building simulation model, yielding a large amount of simulation result data. The Pareto solution set and its visualization results can be used for building renovation decision-making, improving the interpretability of the decision results.

[0042] S4.5: Save large amounts of data generated during multi-objective optimization using the TT toolbox plugin.

[0043] S4.6: Visualize data using tools such as Origin and Python, analyze the Pareto solution set distribution characteristics, evaluate the performance improvement effect of different combinations of modification measures, and make modification decisions based on actual needs.

[0044] S5 involves training a fast prediction model and interpreting the black-box model's decision-making mechanism. Building simulation models suffer from drawbacks such as consuming significant computational resources and having difficulty interpreting simulation results. This step, by training a fast prediction model and using SHAP to interpret the black-box model, can clearly explain the complex relationship between modification measures and optimization objectives, improving the scientific rigor and interpretability of modification decisions. Specifically:

[0045] S5.1: Construct a training database based on the large amount of data generated by the multi-objective optimization process, including building form factors, user behavior plans, renovation measures, and optimization objectives;

[0046] S5.2: Use prediction algorithms by writing code in Python, including MLR, MLP, KNN, DT, RF, and XGBoost algorithm models;

[0047] S5.3: Using building form factors, user behavior plans, and renovation measures as inputs, and building renovation goals as prediction targets, the algorithm hyperparameters are adjusted through grid retrieval and the K-fold method.

[0048] S5.4: Calculate Shapley values ​​using Shapely additive exPlanations by writing Python code. The relevant calculation formula is as follows:

[0049]

[0050] Where, φ i (f) is the SHAP value of feature i, N is the set of all features, S is the subset of features that does not contain feature i, and f(S) is the model prediction output given the feature subset S.

[0051] S5.5: Use charts and visualization tools to explain the complex and multifaceted relationships between transformation measures and optimization goals.

[0052] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the aforementioned rapid optimization method for interpretable building renovation considering parameter uncertainties.

[0053] A computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the aforementioned rapid optimization method for interpretable building renovation considering parameter uncertainties.

[0054] Compared to existing technologies, this invention offers the following advantages: This method calibrates 13 input parameters using a single-objective optimization algorithm, resulting in a large number of simulation models conforming to ASHRAE Guideline 14. The uncertainty assessment index |NMBE| for all building simulation models can be reduced to the 0.01% level, far below the recommended value of 5%. Compared to manual calibration, this method significantly reduces the uncertainty of simulation models, improving the speed and accuracy of model calibration. Furthermore, this method constructs a database using the large amount of data generated during multi-objective optimization, trains a fast prediction model, and employs SHAP analysis to explain the decision-making mechanism of the black-box model, interpreting the complex relationships between modification measures and optimization objectives, thus enhancing the credibility of optimization results. Building simulation models suffer from drawbacks such as consuming significant computational resources and having difficulty interpreting simulation results; this step, by training a fast prediction model and using SHAP to interpret the black-box model, clearly explains the complex relationships between modification measures and optimization objectives, improving the scientific rigor and interpretability of modification decisions. Attached Figure Description

[0055] Figure 1 This is a flowchart of an interpretable rapid optimization method for building renovation that considers parameter uncertainties, according to the present invention.

[0056] Figure 2 These are the calibration results of the simulation model in the embodiments of the present invention.

[0057] Figure 3 This is an uncertainty analysis diagram of an embodiment of the present invention.

[0058] Figure 4This is a visual example of the optimized global interpretation results in an embodiment of the present invention.

[0059] Figure 5 This is a visual example of the optimized local interpretation results in an embodiment of the present invention. Detailed Implementation

[0060] To enhance understanding of the present invention, the embodiments will be described in detail below with reference to the accompanying drawings.

[0061] Example 1: See Figure 1-5 A rapid optimization method for interpretable building renovation considering parameter uncertainties, the method comprising the following steps:

[0062] S1: Collect information and process the raw data;

[0063] S2: Build building simulation models and calibrate weather data;

[0064] S3: Sensitivity analysis of uncertain parameters and calibration of simulation model;

[0065] S4: Set building renovation measures and perform multi-objective optimization of the renovation;

[0066] S5: Train a fast prediction model and explain the decision-making mechanism of the black-box model.

[0067] S1 involves collecting information and processing the raw data, as detailed below:

[0068] S1.1: Collect data on building geometry, building envelope performance, and equipment performance through architectural design drawings, GIS platforms, field surveys, and other methods;

[0069] S1.2: Obtain building electricity consumption information through energy management platforms, sensors, and field surveys;

[0070] S1.3: Outliers in building energy consumption information are identified using the Z-score and interquartile range methods, and these outliers are treated as missing values. The relevant calculation formulas are as follows:

[0071]

[0072] If |Z| > 3, i.e. (the 3σ rule), then it is considered an outlier. Here, x is the data point, μ is the mean of the data, and σ is the standard deviation of the data.

[0073] IQR = Q3 - Q1

[0074] If x is located in the range [Q1-1.5IQR, Q3+1.5IQR], it is considered an outlier. Here, Q1 is the 25th percentile and Q3 is the 75th percentile.

[0075] S1.4: Missing values ​​are filled using Lagrange interpolation to ensure the completeness of building performance data. The relevant calculation formulas are as follows:

[0076]

[0077] Where L(x) is the interpolation result, x i y i Let n represent the independent (time point) and dependent variable values ​​of the data sample points, where n is the degree of the polynomial, equal to the data sample size minus 1, and l is the value of the dependent variable. i (x) is the i-th basis function, ensuring that the polynomial is in x. i The value y is taken at the location i , take 0 for other sample points.

[0078] S2, construct the building simulation model and calibrate the weather data, as detailed below:

[0079] S2.1: Based on building information collected from multiple sources, a building physical model is constructed using the Rhinocero and Grasshopper platforms;

[0080] S2.2: Based on information such as the thermal performance of the building envelope, user behavior, building functions, and equipment performance, a building simulation model is constructed using Ladybug plugins.

[0081] S2.3: Based on microclimate data collected from small weather stations, microclimates are simulated using Dragonfly and Urban Weather Generator (UWG), and the EPW format files of typical year climate data are modified;

[0082] S2.4: Write programs using the Grasshopper platform to call the Energyplus simulation core to simulate the optimization goals of building renovation, including but not limited to building energy consumption, indoor thermal comfort duration, etc.

[0083] S3, Sensitivity analysis of uncertain parameters and calibration of the simulation model, as detailed below:

[0084] S3.1: Select the uncertain parameters that affect the optimization objective (such as thermal performance parameters of the building envelope, various time schedules, temperature setpoints, etc.), and determine the probability distribution (such as normal distribution, uniform distribution) and interval range of each parameter;

[0085] S3.2: Generate combined samples with uncertain parameters using Latin hypercube sampling (LHS), and obtain the performance simulation results of each combined sample using a building simulation model. The relevant calculation formulas are as follows:

[0086]

[0087] Where, x ij The j-th sample value of the i-th parameter, u j The numbers are uniformly random numbers in the range [0, 1]. The inverse cumulative distribution function of the parameter;

[0088] S3.3: Calculate the Sobol main effect index to identify the key parameters affecting building performance among the uncertain parameters, and distinguish between main effects and interaction effects. The relevant calculation formulas are as follows:

[0089]

[0090] Among them, S i This is the main effect index of the input parameter, measuring the proportion of the parameter's contribution to the output variance, with a value range of [0, 1]. i It is the variance of the conditional expectation of the output, i.e., x i The variance component is explained separately. V is the total variance of the model output. It is parameter x i The total effect index, which includes its main effect and all interaction effects with other parameters, ranges from [0, 1]. It is the expected value of the output conditional variance when other parameters are fixed, reflecting x. i The contribution of their interactions to the total variance;

[0091] S3.4: Set the calibration range and step size for the identified key uncertainty parameters, and select an evaluation index for the simulation uncertainty of the evaluation model. In this example, |NMBE| of Energy Usage Intensity (EUI) is selected as the model uncertainty evaluation index. Table 1 shows the range of variation for the uncertainty parameters set in this example, and the program is written to calibrate the simulation model.

[0092] Table 1

[0093] Calibration parameters unit Parameter range Step length Lighting schedule ％ 20-95 1 Personnel Schedule ％ 20-95 1 Equipment schedule ％ 20-95 1 Ventilation schedule ％ 20-100 1 Hot water schedule ％ 20-83 1 Heating temperature set point ℃ 18-22,16-20,or 13-17 0.1 Cooling temperature set point ℃ 24-28or 26-30 0.1 Wall U-value W / (m2·K) 0.85-0.95 0.2 Roof U-value W / (m2·K) 0.66-0.71 0.1 Window U-value W / (m2·K) 3-3.25 0.05 Window SHGC value - 0.43-0.68 0.05 EER - 1.9–2.2 0.05 COP - 0.75-1 0.05

[0094] The relevant and algorithmic formulas are as follows:

[0095]

[0096] EUI s It is used to calibrate the building's energy consumption simulation values, EUI. m It is a measurement of building energy consumption for calibration.

[0097] S3.5: Use Opossum to call the RBFOpt single-objective optimization algorithm and write a program to calibrate the simulation model. Figure 2 The results for this embodiment are the model calibration results for 15 buildings using |NMBE| as the model uncertainty assessment index. Figure 2The 5% figure represents the recommended value for annual building energy consumption calibration according to ASHRAE Guideline 14. Numerous simulation models conform to the recommended values ​​for each building.

[0098] S4, set building renovation measures and perform multi-objective optimization of the renovation, as detailed below:

[0099] S4.1: Table 2 defines the building renovation measures and their levels. Renovation measures include adding insulation to the building envelope, replacing windows with high-performance ones, replacing HVAC systems with high-performance ones, and adding rooftop photovoltaic systems. Further define the levels for different renovation measures and obtain the corresponding input parameters for each level.

[0100] Table 2

[0101]

[0102] S4.2: Select the optimization objectives for the building renovation, use the Grasshopper platform to call the simulation model to evaluate the optimization effect after the renovation, and connect it to the Wallacei plugin. In this example, Energy Usage Intensity (EUI), Thermal Comfort Duration (TCD), and Net Present Value (NPV) are selected as the optimization objectives for the building renovation. The relevant calculation formulas are as follows:

[0103]

[0104] Where EU refers to the building's total annual energy consumption, PE is the building's rooftop photovoltaic energy generation, and S is the building's total floor area. t For the Cth t Net cash flow for the year, r is the discount rate, C0 is the initial investment cost, n is the project duration, p i q is the unit price of the i-th material. i It is the quantity of the i-th material. This represents the proportion of material costs in the total cost. The example references data from the People's Bank of China and related literature, setting the discount rate r to 2%, and the material cost... The percentage is set to 0.6, and n is set to 20 years;

[0105] S4.3: Set the modification level of the modification measures through the gene pool battery and connect it to the Wallacei plugin;

[0106] S4.4: A multi-objective optimization genetic algorithm is invoked via the Wallacei plugin to perform multi-objective optimization of building modifications, using modification measures as genes. The relevant data expression formulas for multi-objective optimization are as follows:

[0107] minF=[f(EUI),-f(CT),-f(NPV)] T

[0108] f(EUI) = f1(Wall) U Roof U Wd U Wd SHGC Wall R Roof R EER, COP, PV)

[0109] f(CT) = f2(Wall) U Roof U Wd U Wd SHGC Wall R Roof R EER, COP, PV)

[0110] f(NPV) = f3(Wall) U Roof U Wd U Wd SHGC Wall R Roof R EER, COP, PV)

[0111] Where minF is the objective function vector, and f(EUI), f(CT), and f(NPV) are the objective functions that need to be optimized simultaneously. U Roof U Wd U Wd SHGC Wall R Roof R EER, COP, and PV are used to optimize genes (parameters for building renovation measures) for multiple objectives.

[0112] S4.5: Save large amounts of data generated during multi-objective optimization using the TT toolbox plugin.

[0113] S4.6: Visualize data using tools such as Origin and Python, analyze the Pareto solution set distribution characteristics, evaluate the performance improvement effect of different combinations of modification measures, and make modification decisions based on actual needs. Figure 3 This provides a visualization of the feasible solutions obtained after multi-objective optimization for each building.

[0114] S5, Training a Fast Prediction Model and Explaining the Black-Box Model Decision-Making Mechanism, as detailed below:

[0115] S5.1: Construct a training database based on the large amount of data generated by the multi-objective optimization process, including building form factors, user behavior plans, renovation measures, and optimization objectives;

[0116] S5.2: Use prediction algorithms by writing code in Python, including but not limited to MLR, MLP, KNN, DT, RF, XGBoost and other algorithm models;

[0117] S5.3: Using building form factors, user behavior plans, and renovation measures as inputs, and building renovation goals as prediction targets, the algorithm's hyperparameters are adjusted through grid retrieval and the K-fold method. Table 3 shows the model's prediction performance index R. 2 ,

[0118] Table 3

[0119]

[0120] The relevant calculation formulas are as follows:

[0121]

[0122] Among them, y pre，i y act，i Let y be the i-th predicted value and the actual value of the model, respectively. act，i The average of the actual values, where n represents the sample size;

[0123] S5.4: Calculate the Shapley value using Shapely additive exPlanations by writing Python code. The relevant calculation formula is as follows:

[0124]

[0125] Where, φ i (f) is the SHAP value of feature i, N is the set of all features, S is the subset of features that does not contain feature i, and f(S) is the model prediction output given the feature subset S.

[0126] S5.5: Use charts and visualization tools to explain the complex and multifaceted relationships between transformation measures and optimization goals. Figure 4 and Figure 5 Visualizing Shapley values ​​can be used to explain the importance of different modification measures and their complex and multifaceted relationship with optimization objectives.

[0127] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made based on the above technical solutions all fall within the scope of protection of the claims of the present invention.

Claims

1. A rapid optimization method for interpretable building renovation considering parameter uncertainty, characterized in that, The method includes the following steps: S1: Collect information and process the raw data; S2: Build building simulation models and calibrate weather data; S3: Sensitivity analysis of uncertain parameters and calibration of simulation model; S4: Set building renovation measures and carry out renovation and optimization; S5: Train a fast prediction model and explain the decision-making mechanism of the black-box model.

2. The rapid optimization method for interpretable building renovation considering parameter uncertainty according to claim 1, characterized in that, S1, collect information and process the raw data, as follows: S1.1: Collect data on building geometry, building envelope performance, and equipment performance through architectural design drawings, GIS platforms, and field surveys; S1.2: Obtain building performance data, including electricity consumption and indoor thermal environment data, through energy management platforms, sensors, and field survey methods; S1.3: Outliers in building electricity consumption and indoor thermal environment data are identified using the Z-score and interquartile range methods, and these outliers are treated as missing values. The relevant calculation formulas are as follows: If |Z| > 3 (the 3σ rule), then it is considered an outlier, where x is the data point, μ is the mean of the data, and σ is the standard deviation of the data. IQR = Q3 - Q1 If x is located in [Q1-1.5IQR, Q3+1.5IQR], it is considered an outlier, where Q1 is the 25th percentile and Q3 is the 75th percentile. S1.4: Missing values ​​are filled using Lagrange interpolation to ensure the completeness of building performance data. The relevant calculation formula is as follows: Where L(x) is the interpolation result, x i y i Let n represent the independent (time point) and dependent variable values ​​of the data sample points, where n is the degree of the polynomial, equal to the data sample size minus 1, and l is the value of the dependent variable. i (x) is the i-th basis function, ensuring that the polynomial is in x. i The value y is taken at the location i , take 0 for other sample points.

3. The rapid optimization method for interpretable building renovation considering parameter uncertainty according to claim 2, characterized in that, S2, construct the building simulation model and calibrate the weather data, as detailed below: S2.1: Based on building information collected from multiple sources, a building physical model is constructed using the Rhinocero and Grasshopper platforms; S2.2: Based on the thermal performance of the building envelope, user behavior, building functions, and equipment performance information, a building simulation model is constructed using Ladybug plugins. S2.3: Based on microclimate data collected from small weather stations, the microclimate is simulated using Dragonfly and the Urban Weather Generator (UWG), and the EPW format file of typical year climate data is modified. S2.4: Build a building simulation model by writing a program on the Grasshopper platform, and call the Energyplus simulation core to perform performance simulation on the optimization goals of the building renovation, including building energy consumption and indoor thermal comfort duration.

4. The rapid optimization method for interpretable building renovation considering parameter uncertainty according to claim 3, characterized in that, S3, Sensitivity analysis of uncertain parameters and calibration of the simulation model, as detailed below: S3.1: Select the uncertain parameters that affect the optimization objective, and determine the probability distribution and range of each parameter; S3.2: Generate combined samples with uncertain parameters using Latin hypercube sampling (LHS), and obtain the performance simulation results of each combined sample using a building simulation model. The relevant calculation formulas are as follows: Where, x ij The j-th sample value of the i-th parameter, u j The numbers are uniformly random numbers in the range [0, 1]. The inverse cumulative distribution function of the parameter; S3.3: Calculate the Sobol main effect index, identify the key parameters affecting building performance among the uncertain parameters, and distinguish between main effects and interaction effects. The relevant calculation formula is as follows: Among them, S i It is the main effect index of the input parameter, which measures the proportion of the parameter's contribution to the output variance. Its value ranges from [0, 1]. i It is the variance of the conditional expectation of the output, i.e., x i The variance component, interpreted separately, is V, which represents the total variance of the model output. It is parameter x i The total effect index, which includes the main effects and all interaction effects with other parameters, ranges from [0, 1]. It is the expected value of the output conditional variance when other parameters are fixed, reflecting x. i The contribution of their interactions to the total variance; S3.4: Set the calibration range and step size for the identified key uncertainty parameters, and select the evaluation index for the simulation uncertainty of the evaluation model; S3.5: Call the single-objective optimization algorithm through Opossum, and use the model uncertainty evaluation index as the optimization objective of the single-objective optimization algorithm to write a program to calibrate the simulation model.

5. The rapid optimization method for interpretable building renovation considering parameter uncertainty according to claim 2, characterized in that, S4, set building renovation measures and perform multi-objective optimization of the renovation, as detailed below: S4.1: Set the building renovation measures and their renovation levels. The renovation measures include adding a thermal insulation layer to the building envelope, replacing high-performance windows, replacing high-performance HVAC systems, and adding roof photovoltaic systems. Further set the levels of different renovation measures and obtain the input parameters corresponding to each renovation level of different renovation measures. S4.2: Select the optimization objectives for building renovation, use the Grasshopper platform to call the simulation model to evaluate the optimization effect after renovation, and connect to the Wallacei plugin. Optimization objectives include building energy consumption, indoor thermal comfort duration, and net present value. S4.3: Set the modification level of the modification measures through the gene pool battery and connect it to the Wallacei plugin; S4.4: A multi-objective optimization genetic algorithm is invoked via the Wallacei plugin to perform multi-objective optimization of building modifications, using modification measures as genes. S4.5: Save the large amount of data generated during the multi-objective optimization process using the TT toolbox plugin. S4.6: Visualize the data using tools such as Origin and Python, analyze the Pareto solution set distribution characteristics, evaluate the performance improvement effect of different combinations of modification measures, and make modification decisions based on actual needs.

6. The rapid optimization method for interpretable building renovation considering parameter uncertainty according to claim 2, characterized in that, S5, Training a Fast Prediction Model and Explaining the Black-Box Model Decision-Making Mechanism, as detailed below: S5.1: Construct a training database based on the large amount of data generated by the multi-objective optimization process, including building form factors, user behavior plans, renovation measures, and optimization objectives; S5.2: Use prediction algorithms by writing code in Python, including MLR, MLP, KNN, DT, RF, and XGBoost algorithm models; S5.3: Using building form factors, user behavior plans, and renovation measures as inputs, and building renovation goals as prediction targets, the algorithm hyperparameters are adjusted through grid retrieval and the K-fold method. S5.4: Calculate Shapley values ​​using Shapely additive exPlanations by writing Python code. The relevant calculation formula is as follows: Where, φ i (f) is the SHAP value of feature i, N is the set of all features, S is the subset of features that does not contain feature i, and f(S) is the model prediction output given the feature subset S. S5.5: Use charts and visualization tools to explain the complex and multifaceted relationships between transformation measures and optimization goals.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the interpretable rapid optimization method for building renovation that takes into account parameter uncertainties, as described in any one of claims 1 to 6.

8. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by a processor, the computer instructions implement the interpretable rapid optimization method for building renovation that takes into account parameter uncertainties as described in any one of claims 1-6.

Images