Method for calculating carbon emissions of building life cycle based on multiple linear regression

Abstract

The present application relates to carbon emission processing technical field, especially in a kind of based on multiple linear regression's building full life cycle carbon emission calculation method.The present application includes the following steps: obtaining the carbon emission data of each stage of building by BIM model, adopts Latin hypercube sampling method and fuzzy clustering algorithm to extract key characteristic variable, such as material carbon emission coefficient, energy consumption intensity etc.;Carbon emission is modeled using multiple linear regression model, and carbon emission prediction model is constructed;Model parameters are optimized by differential evolution algorithm;Finally, the total amount of carbon emission is calculated, and the key influence factor of high carbon emission stage is analyzed by causal reasoning algorithm, optimization strategy is proposed and analysis report is generated, accurate prediction and effective optimization building full life cycle carbon emission are realized.The present application can not only accurately evaluate building full life cycle carbon emission, but also provide strong support for the optimization decision of carbon emission.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of carbon emission treatment, and in particular to a building full life cycle carbon emission calculation method based on multivariate linear regression. BACKGROUND

[0002] With the increasingly serious global climate change problem, the building industry, as one of the main sources of carbon emissions, has attracted much attention. The carbon emissions of the building full life cycle mainly include the stages of material production, construction, use and operation, and final demolition and waste treatment. In the building full life cycle, how to quantify carbon emissions and effectively propose emission reduction strategies has become an important research topic in the field of sustainable development. The current problems are as follows: in the prior art, the acquisition of building full life cycle carbon emission data usually relies on manual or discrete data collection methods, which easily leads to incomplete data and low efficiency, and lacks a unified standardized processing procedure, affecting the accuracy of subsequent analysis; traditional carbon emission calculation methods mostly rely on simple statistical models or single variable analysis, and it is difficult to fully consider the comprehensive influence of multi-dimensional factors such as materials, energy, equipment and construction process, so the prediction result is not accurate enough to meet the actual application requirements; the traditional method cannot comprehensively evaluate carbon emissions from the perspective of the building full life cycle, and lacks systematic carbon emission management means, which is difficult to meet the actual needs of the development of green building low carbonization. SUMMARY

[0003] To solve the above problems, the application provides a building full life cycle carbon emission calculation method based on multivariate linear regression, which solves the problems of how to comprehensively acquire carbon emission data in the building full life cycle, construct a multi-dimensional factor comprehensive analysis accurate prediction model, and propose effective low-carbon optimization strategies, so as to accurately evaluate the building full life cycle carbon emission and provide strong support for the optimization decision of carbon emission.

[0004] To achieve the above purpose, the technical scheme adopted by the application is:

[0005] The building full life cycle carbon emission calculation method based on multivariate linear regression comprises the following steps:

[0006] S1: acquiring carbon emission data of each stage in the building full life cycle through a BIM model;

[0007] S2: based on the carbon emission data, using a Latin hypercube sampling method for sampling, and simultaneously performing dynamic layering through a fuzzy clustering algorithm to extract key characteristic variables of building carbon emission; the key characteristic variables include material carbon emission coefficient, energy consumption intensity, equipment operation efficiency and construction process carbon emission characteristics;

[0008] S3: Based on the key feature variables, a multivariate linear regression model is used to model the carbon emissions of the building life cycle, and a carbon emission prediction model is constructed;

[0009] S4: The differential evolution algorithm is used to globally optimize the parameters of the carbon emission prediction model, and the dynamic adjustment of the model parameters is realized through iterative calculation;

[0010] S5: Calculate the total carbon emissions of the building life cycle, and analyze the causal relationship of the high-carbon emission stage in depth through causal reasoning algorithm, identify the influencing factors, and generate a carbon emission calculation and analysis report through the feedback mechanism.

[0011] Further, the carbon emission data includes design parameters, material types, equipment operation data, energy consumption data and construction process data.

[0012] Further, the step S2 includes the following steps:

[0013] Based on the BIM model, the carbon emission data is extracted, and is divided into material production stage, construction stage, operation and maintenance stage and demolition stage according to stage, and data groups are established respectively;

[0014] Latin hypercube sampling method is used to set constraint conditions to generate sampling points with uniform distribution characteristics in the data group;

[0015] Combined with fuzzy clustering algorithm, the carbon emission data is dynamically clustered according to the attribute similarity of the sampling points, forming a multi-layer grouping system with time, process and equipment type as dimensions;

[0016] Based on the multi-layer grouping system, a multi-modal fusion feature extraction algorithm is applied for analysis, and key feature variables are extracted;

[0017] The identified key feature variables are sorted by weight allocation.

[0018] Further, the formula of the Latin hypercube sampling method is as follows:

[0019]

[0020] Wherein, X ij represents the sampling result of the jth variable in the ith sample, that is, the sampling result of the carbon emission data; L j represents the minimum value of the carbon emission data; H j represents the maximum value of the carbon emission data; P ij represents the number of the jth variable in the discrete grouping after random arrangement; R ij represents a random number subject to uniform distribution; φ(R ij , σj represents an extended disturbance term, used to adjust the influence of random number R ij n represents the number of sampling points of carbon emission data; k j represents a periodic disturbance coefficient, used to introduce a periodic sinusoidal disturbance; represents a sinusoidal disturbance term, used to introduce nonlinear fluctuations on the sampling values of each variable.

[0021] Further, the construction process of the fuzzy clustering algorithm comprises the following steps:

[0022] Generate a fuzzy relation matrix based on the carbon emission data, and quantify the similarity between data through a membership function;

[0023] Initialize the clustering center, and use a dynamic hierarchical algorithm to assign data points;

[0024] Iteratively optimize the clustering center according to the objective function until the convergence condition is met;

[0025] Output the clustering result to form a multi-layer fuzzy clustering structure of building carbon emission data.

[0026] Further, the step S3 comprises the following steps:

[0027] Based on the key characteristic variables, establish a carbon emission prediction model using a multiple linear regression algorithm;

[0028] Use historical carbon emission data to fit the carbon emission prediction parameters by least squares method;

[0029] Use stepwise regression method to screen and optimize the weights of key characteristic variables;

[0030] Evaluate the fitting performance and prediction ability of the model through cross-validation, and finally output the optimized carbon emission prediction model.

[0031] Further, the formula of the carbon emission prediction model is as follows:

[0032]

[0033] Where E represents the total carbon emission of the building throughout its life cycle; β0represents the base value of the building's life cycle carbon emission; β i represents the linear regression coefficient of the i-th key characteristic variable; γ j represents the regression coefficient of the nonlinear term; δ k represents the regression coefficient of the interaction term; n represents the number of key characteristic variables; m represents the total number of square terms; p represents the total number of interaction terms; represents the linear contribution of the key characteristic variable to the total carbon emission; The squared terms of key characteristic variables represent the contribution of carbon emissions, reflecting the nonlinear relationship between the variables; This indicates the impact of the interaction between two variables on carbon emissions, used to model the synergistic effect between variables; ∈ indicates random error or carbon emission changes not explained by the model.

[0034] Furthermore, in step S4, the differential evolution algorithm specifically adopts a dual-objective optimization differential evolution algorithm, with carbon emission prediction error and computational efficiency as optimization objectives, and dynamically adjusts the model parameters.

[0035] Furthermore, in step S5, the causal inference algorithm specifically includes a method based on Bayesian networks and structural equation models. By establishing a causal network diagram of carbon emissions throughout the building's life cycle, and utilizing data-driven causal relationship inference technology, it identifies and quantifies key influencing factors and their causal paths in the high-carbon emission stage, and provides optimization strategy suggestions.

[0036] The beneficial effects of this invention are as follows:

[0037] This invention automates the acquisition of carbon emission data at each stage of a building's entire lifecycle through BIM models, improving the efficiency and accuracy of data collection and laying a solid foundation for subsequent modeling and analysis. By combining Latin hypercube sampling with fuzzy clustering algorithms, dynamic stratification and extraction of key carbon emission characteristic variables significantly reduce data redundancy, highlighting factors with significant impact on carbon emissions and improving model interpretability and analytical efficiency. A multiple linear regression model is used to model the carbon emissions throughout the entire lifecycle. By considering multidimensional factors such as materials, energy, equipment, and construction processes, a scientific and comprehensive carbon emission prediction model is constructed, providing accurate carbon emission estimates. Differential evolutionary algorithms are used to globally optimize model parameters, ensuring higher prediction accuracy and dynamic adaptability, thus enabling more flexible responses to the carbon emission characteristics of different building projects. Causal inference algorithms are employed to conduct in-depth analysis of high-carbon emission stages, identifying key influencing factors and providing targeted improvement measures for carbon reduction. Simultaneously, a feedback mechanism generates carbon emission calculation and analysis reports to help building design and construction teams optimize carbon emission management plans. Comprehensive and accurate carbon emission calculations and analyses can provide a scientific basis for carbon management throughout the entire building lifecycle, effectively promoting the construction industry toward low-carbon and sustainable development. Attached Figure Description

[0038] Fig. 1 This is a flowchart illustrating the building life-cycle carbon emission calculation method based on multiple linear regression, as presented in this invention.

[0039] Fig. 2 This is a flowchart illustrating step S2 provided in an embodiment of the present invention.

[0040] Fig. 3 is a flowchart of step S3 provided by an embodiment of the present application. DETAILED DESCRIPTION

[0041] Referring to Figs. 1-3 The present application relates to a building life cycle carbon emission calculation method based on multiple linear regression.

[0042] EMBODIMENT

[0043] The building life cycle carbon emission calculation method based on multiple linear regression comprises the following steps:

[0044] S1: Obtain carbon emission data of each stage in the building life cycle through the BIM model; the carbon emission data includes design parameters, material types, equipment operation data, energy consumption data, and construction process data.

[0045] Among them, the mainstream BIM modeling software (such as Revit, Archicad) is used to create a building model, which contains multi-dimensional information such as building design, structure, equipment, and construction. Import the design parameters of the building (such as building area, height, functional layout, etc.), construction materials (such as reinforced concrete, wood, glass, etc.), and construction process. Introduce the carbon emission coefficient database (such as the standard coefficient released by the Ministry of Ecology and Environment, the Ecoinvent database, etc.), and associate the carbon emission coefficients of building materials production, transportation, and construction stage to the material information in the BIM model. Through the simulation of energy consumption (such as electricity, gas, renewable energy) and equipment operation data (such as heating, ventilation and air conditioning, lighting system, elevator, etc.) in the building use stage, combined with the real data collected by the energy management system (EMS). According to the building life cycle stage (design, construction, operation, maintenance, and demolition), the data information related to carbon emission is extracted respectively to form the initial carbon emission data set.

[0046] Divide the building life cycle into the following stages and obtain the corresponding carbon emission data:

[0047] Production stage: involving carbon emissions of building material production, transportation and storage, including raw material mining, processing, manufacturing, etc.

[0048] Construction stage: use of mechanical equipment during construction, transportation of construction personnel, temporary facilities (such as scaffolding) at the construction site, and energy consumption.

[0049] Operation stage: energy consumption (electricity, gas, water, etc.) and equipment maintenance and operation during building use.

[0050] Maintenance stage: involving carbon emissions of regular repair, replacement of materials and equipment of the building.

[0051] Dismantling phase: disassembly, transportation and disposal (e.g. recycling or landfill) of waste materials after the end of the building's life.

[0052] S2: Based on the carbon emission data, Latin hypercube sampling method is used for sampling, and dynamic stratification is carried out through fuzzy clustering algorithm to extract key characteristic variables of building carbon emission; the key characteristic variables include material carbon emission coefficient, energy consumption intensity, equipment operation efficiency and construction process carbon emission characteristics;

[0053] The step S2 includes the following steps:

[0054] Based on the BIM model, the carbon emission data is extracted and divided into material production phase, construction phase, operation and maintenance phase and demolition phase according to the phase, and data groups are established respectively;

[0055] Latin hypercube sampling method is used to generate sampling points with uniform distribution characteristics in the data group by setting constraints;

[0056] It should be noted that in each data group, the variable range is defined (such as the material carbon emission coefficient range is 0.1-2.0 kgCO2 / kg), and the constraint condition is set (such as the upper and lower limits of the total amount of materials or the construction time). Latin hypercube sampling method (through tools such as MATLAB's lhsdesign or Python's pyDOE library) is used to generate sampling points with uniform distribution characteristics, ensuring coverage of variable range.

[0057] Constraint condition examples are as follows:

[0058] Material production phase: constraint the total carbon emission of each material to be less than a specified threshold;

[0059] Construction phase: constraint the total energy consumption within a certain range to meet the actual working condition;

[0060] Operation and maintenance phase: constraint the building service life to be 50 years.

[0061] The output sampling points contain multi-dimensional attribute data, ensuring representativeness and uniform distribution, such as: material types and their carbon emission coefficients; equipment operation efficiency data; energy consumption data of each process link.

[0062] Combined with fuzzy clustering algorithm, the carbon emission data is dynamically clustered according to the attribute similarity of the sampling points, forming a multi-layer grouping system with time, process and equipment type as dimensions;

[0063] Based on the multi-layer grouping system, multi-modal fusion feature extraction algorithm is applied for analysis to extract key characteristic variables;

[0064] The identified key feature variables are sorted by weight distribution.

[0065] The construction process of the fuzzy clustering algorithm includes the following steps:

[0066] A fuzzy relationship matrix is generated based on the carbon emission data, and the similarity between data is quantified by a membership function; each value of the matrix represents the degree of similarity between data points;

[0067] The clustering center is initialized, and a dynamic hierarchical algorithm is used to assign data points;

[0068] The clustering center is iteratively optimized according to the objective function until the convergence condition is met;

[0069] The clustering result is output, forming a multi-layer fuzzy clustering structure of building carbon emission data.

[0070] Further, the formula of the Latin hypercube sampling method is as follows:

[0071]

[0072] Where X ij represents the sampling result of the jth variable in the ith sample, i.e. the sampling result of the carbon emission data; L j represents the minimum value of the carbon emission data; H j represents the maximum value of the carbon emission data; P ij represents the number of the jth variable in the discrete grouping after random arrangement; R ij represents a random number subject to uniform distribution; φ(R ij , σ j ) represents an extended disturbance term, used to adjust the influence of random number R ij ; n represents the number of sampling points of carbon emission data; k j represents a periodic disturbance coefficient, used to introduce periodic sinusoidal disturbance; represents a sinusoidal disturbance term, used to introduce nonlinear fluctuations on the sampling value of each variable.

[0073] Specifically, the formula of φ(R ij , σ j ) is as follows:

[0074]

[0075] Where σ j represents the fluctuation factor of the jth variable, representing the dynamic characteristics of the variable.

[0076] S3: Based on the key characteristic variables, a multivariate linear regression model is used to model the carbon emissions of the building life cycle, and a carbon emission prediction model is constructed;

[0077] Wherein, the step S3 comprises the following steps:

[0078] Based on the key characteristic variables, a carbon emission prediction model using a multivariate linear regression algorithm is established;

[0079] It should be noted that the carbon emission prediction model takes the total carbon emissions of the whole life cycle as the dependent variable, and the key characteristic variables (such as material carbon emission coefficient, energy consumption intensity, equipment operation efficiency and construction process carbon emission characteristics) as the independent variables. The model assumes that the influence of each characteristic variable on the total carbon emissions is a linear relationship.

[0080] The extracted key characteristic variables and corresponding carbon emission historical data are sorted and formed into a data set required for modeling. The data set includes: input variables such as material carbon emission coefficient, energy consumption intensity, etc. Output variables: total carbon emissions of the whole life cycle. Standardize the data to convert different dimension data into dimensionless values to ensure that the characteristic variables are on the same scale. Build an initial multivariate linear regression model to represent the contribution of each key characteristic variable as a model parameter.

[0081] Using historical carbon emission data, the carbon emission prediction parameters are fitted by the least squares method;

[0082] Specifically, the carbon emission data set (such as the whole life cycle carbon emission records of multiple projects) is obtained from the historical data or actual measurement of the construction industry. The data samples need to cover a wide range of building types, sizes and life cycle characteristics to improve the universality of the model. Adopt the least squares method to adjust the regression coefficients to minimize the sum of squared errors between the predicted values and the actual values. Through iterative optimization, the model can best match the historical data. Output the initial regression coefficients of each characteristic variable as the influence weight of the total carbon emissions. Provide preliminary error evaluation results (such as mean square error) to judge the preliminary effect of fitting.

[0083] Stepwise regression method is used to screen and optimize the weight of key characteristic variables;

[0084] The fitting performance and prediction ability of the model are evaluated by cross-validation, and the optimized carbon emission prediction model is finally output.

[0085] Further, the formula of the carbon emission prediction model is as follows:

[0086]

[0087] where E represents the total carbon emissions in the building life cycle; β0represents the base value of the building life cycle carbon emissions, i.e. the predicted value when all variables take the value of zero; β i represents the linear regression coefficient of the i-th key feature variable, and the influence of the variable on carbon emissions per unit increase; γ j represents the regression coefficient of the nonlinear term; δ k represents the regression coefficient of the interaction term; n represents the number of key feature variables; m represents the total number of square terms; p represents the total number of interaction terms; represents the linear contribution of the key feature variable to the total carbon emissions; represents the contribution of the square term of the key feature variable to the carbon emissions, used to reflect the nonlinear relationship of the variable; represents the influence of the interaction between two variables on carbon emissions, used to model the synergistic effect between variables; ∈ represents random error or unexplained carbon emissions changes of the model.

[0088] S4: Adopting differential evolution algorithm to globally optimize the parameters of the carbon emission prediction model, and realizing dynamic adjustment of the model parameters through iterative calculation; the differential evolution algorithm specifically adopts a double-objective optimization differential evolution algorithm, taking carbon emission prediction error and calculation efficiency as optimization objectives, and dynamically adjusting the model parameters.

[0089] Specifically, a group of different model parameter combinations are randomly generated, which can be regarded as a group of candidate solutions, and each group of parameters includes key variables such as initial regression coefficients and intercepts of the model. Each parameter combination represents a possible model solution, and the population size (e.g. 50 groups of parameters) is set to ensure sufficient diversity of the algorithm. The parameter range is set according to the actual carbon emission data, for example, the weight range of material carbon emission coefficient is from low to high, and the value distribution of the initial parameters is uniform.

[0090] Each candidate parameter combination is run through the model to calculate the error of the model in predicting carbon emissions, as well as the efficiency of the model run (such as calculation time). Here, known data samples (such as the validation set) are used for evaluation to determine whether each group of parameters is reasonable. For example: if the prediction error is small and the calculation time is short, it means that this group of parameters is better than other combinations. If a group of parameters results in too large an error or too long a running time, this combination may be eliminated.

[0091] Based on the results of the last round of evaluation, new parameter combinations are generated. This step is called "mutation", that is, slightly adjusting the existing parameters to explore better solutions.

[0092] The range of parameter adjustment is dynamic, allowing a larger exploration range at the beginning, and gradually narrowing the range as the optimization iteration proceeds, to find more accurate parameters.

[0093] On the basis of the mutated parameters, new candidate solutions are generated by combining the advantages of different combinations through the "crossing" step. For example, one combination has a better material carbon emission weight, and another combination has a better energy consumption weight. The system will try to integrate these advantages. The parameter combination after crossing is evaluated again by the model, and the winners enter the next round of optimization, while the disadvantaged combinations are eliminated.

[0094] After each iteration, the algorithm generates better parameter combinations, continuously reducing prediction errors and improving computational efficiency.

[0095] The conditions for stopping the optimization process: the prediction error reaches the preset minimum standard, such as a relative error of less than 1%; the parameter variation amplitude tends to zero, indicating that the algorithm has found the optimal solution; the maximum number of iterations is reached, ensuring that the algorithm does not run indefinitely.

[0096] After the optimization is completed, the algorithm outputs the final optimal parameters, which will be used for the full life cycle prediction of building carbon emissions. The output includes the weight distribution of each key feature variable (such as the influence weight of the material carbon emission coefficient being 30% and the influence weight of energy consumption being 50%) and the overall performance indicators of the model (such as prediction error, computational efficiency, etc.).

[0097] S5: Calculate the total carbon emissions of the building's full life cycle and conduct in-depth analysis of the causal relationship of high-carbon emission stages through a causal reasoning algorithm to identify influencing factors and generate a carbon emission calculation and analysis report through a feedback mechanism. The causal reasoning algorithm specifically includes methods based on Bayesian networks and structural equation models, which establish a causal network diagram of building full life cycle carbon emissions, use data-driven causal relationship inference technology to identify and quantify key influencing factors and their causal paths in high-carbon emission stages, and provide optimization strategy suggestions.

[0098] In one embodiment, Bayesian networks are used to represent different stages of the building's full life cycle and their key feature variables as nodes. The relationships between nodes (causal paths) are defined, for example:

[0099] Material type → Material carbon emission coefficient → Production stage carbon emission.

[0100] Device efficiency → Energy consumption → Operation stage carbon emission.

[0101] Construction technology → Device usage time → Construction stage carbon emission.

[0102] Through statistical analysis of carbon emission data, direct and indirect relationships between variables are inferred. For example, the carbon emissions of high-energy consumption buildings during operation are significantly affected by the thermal performance of materials.

[0103] The Structural Equation Modeling (SEM) is used to analyze the strength of causal paths and quantify the contribution of each variable to carbon emissions. For example, it is found that the material carbon emission coefficient contributes 70% to the carbon emissions in the production phase. Based on the results of causal network analysis, the main influencing factors of high carbon emission stages are identified. For example:

[0104] Operation phase: Energy consumption intensity is the main influencing factor.

[0105] Construction phase: Long-term use of tower crane equipment is the key reason for high carbon emissions.

[0106] For each key factor, the impact degree is quantified, for example, a 10% improvement in equipment efficiency can reduce carbon emissions in the operation phase by 5%.

[0107] Based on the results of causal analysis, optimization strategies are proposed. For example:

[0108] Opt for low-carbon emission materials: Replace traditional high-carbon materials (such as ordinary cement) with green building materials (such as low-clinker cement).

[0109] Improve equipment operation efficiency: Upgrade HVAC systems or use more efficient renewable energy sources.

[0110] Optimize construction process: Reduce the operation time of tower crane and concrete mixer, and adopt low-carbon construction methods.

[0111] Quantify the potential carbon emission reduction effect of each optimization strategy. For example: replacing traditional lamps with LED lighting can reduce carbon emissions in the operation phase by 15%.

[0112] Cyclic feedback mechanism: The optimized strategy is reloaded and simulated through the BIM model to verify the actual carbon emission reduction effect of the optimization scheme. If the optimization effect does not meet the standard, the model parameters or strategy scheme can be adjusted again until the results meet the expectations.

[0113] In summary, the present application combines BIM models and carbon emission coefficient databases to comprehensively extract carbon emission data throughout the life cycle of a building, covering production, construction, operation, maintenance, and demolition stages, making carbon emission calculation more comprehensive and accurate. Latin hypercube sampling and fuzzy clustering algorithm are used to make data processing efficient and representative, greatly improving the accuracy of the prediction model. The differential evolution algorithm is used to globally optimize the parameters of the carbon emission prediction model, which can dynamically adjust the model parameters to reduce prediction error and improve calculation efficiency. Through iterative optimization, the optimal parameters of the whole life cycle carbon emission prediction model are output, making the prediction model more adaptable and universal.

[0114] The application introduces the causal reasoning algorithm of Bayesian network and structural equation model to identify the key influencing factors and causal paths of high carbon emission stages in the whole life cycle of buildings. Through causal analysis, the contribution rate of each variable to carbon emission can be quantified, providing a scientific basis for carbon emission reduction strategies. A carbon emission optimization feedback mechanism is established, and based on the results of causal analysis, operable emission reduction optimization strategies (such as optimizing low-carbon materials, optimizing construction process, improving equipment efficiency, etc.) are proposed. The optimized strategies are verified through BIM model, realizing a closed loop from data extraction, prediction, optimization to feedback, and improving the effectiveness and reliability of carbon emission reduction schemes.

[0115] The application is based on multiple linear regression, combined with nonlinear terms and interaction terms modeling, suitable for different types and scales of building projects, and has strong universality. Through phased data extraction and dynamic hierarchical analysis, it can flexibly cope with the carbon emission calculation requirements of different building life cycle characteristics. By accurately identifying and quantifying key influencing factors, it provides quantitative emission reduction effect evaluation. For example, improving equipment efficiency, selecting green building materials or optimizing construction process can significantly reduce the carbon emission in the whole life cycle of buildings, providing strong support for green building design and construction.

[0116] The above embodiments only describe the preferred embodiments of the application, and do not limit the scope of the application. Without departing from the design spirit of the application, various modifications and improvements of the technical solutions of the application made by ordinary engineering and technical personnel in the art shall fall within the protection scope determined by the claims of the application.

Claims

1. A method for calculating the carbon emissions of a building over its life cycle based on multiple linear regression, characterized in that, The method comprises the following steps: S1: Obtain carbon emission data of each stage in the whole life cycle of the building through a BIM model; S2: Based on the carbon emission data, use Latin hypercube sampling method for sampling, and use fuzzy clustering algorithm for dynamic layering to extract key characteristic variables of building carbon emission; the key characteristic variables include material carbon emission coefficient, energy consumption intensity, equipment operation efficiency and construction process carbon emission characteristics; S3: Based on the key characteristic variables, use a multiple linear regression model to model the carbon emission of the whole life cycle of the building, and build a carbon emission prediction model; S4: Use differential evolution algorithm to globally optimize the parameters of the carbon emission prediction model, and realize dynamic adjustment of the model parameters through iterative calculation; S5: Calculate the total carbon emission of the whole life cycle of the building, and use causal reasoning algorithm to analyze the causal relationship of the high-carbon emission stage in depth, identify the influencing factors, and generate a carbon emission calculation and analysis report through a feedback mechanism; The step S2 comprises the following steps: Based on the BIM model, the carbon emission data is extracted and divided into material production stage, construction stage, operation and maintenance stage and demolition stage according to stages, and data groups are established respectively; Using Latin hypercube sampling method, set constraint conditions to generate sampling points with uniform distribution characteristics in the data group; Combined with the fuzzy clustering algorithm, the carbon emission data is dynamically clustered according to the attribute similarity of the sampling points to form a multi-layer grouping system with time, process and equipment type as dimensions; Based on the multi-layer grouping system, a multi-modal fusion feature extraction algorithm is applied for analysis to extract key characteristic variables; For the identified key characteristic variables, weight allocation is used to sort the key characteristic variables; The formula of the Latin hypercube sampling method is as follows: ， wherein, represents the sampling result of the jth variable in the ith sample, i.e., the sampling result of the carbon emission data; represents the minimum value of the carbon emission data; represents the maximum value of the carbon emission data; represents the number of the jth variable in the discrete grouping after random arrangement; represents a random number subject to uniform distribution; represents an extended disturbance term, used to adjust the influence of the random number ; n represents the number of sampling points of the carbon emission data; represents a periodic disturbance coefficient, used to introduce periodic sinusoidal disturbance; represents a sinusoidal disturbance term, used to introduce nonlinear fluctuations on the sampling value of each variable.

2. The method of claim 1, wherein, The carbon emission data includes design parameters, material types, equipment operation data, energy consumption data and construction process data.

3. The method of claim 1, wherein, The construction process of the fuzzy clustering algorithm comprises the following steps: Based on the carbon emission data, a fuzzy relation matrix is generated, and the similarity between data is quantified by a membership function; Initialize the clustering center, and use a dynamic layering algorithm to distribute the data points; According to the objective function, iteratively optimize the clustering center until the convergence condition is met; Output the clustering result to form a multi-layer fuzzy clustering structure of building carbon emission data.

4. The method of claim 1, wherein, The step S3 comprises the following steps: Based on the key characteristic variables, a carbon emission prediction model using multiple linear regression algorithm is established; Using historical carbon emission data, the carbon emission prediction parameters are fitted by least squares method; Use stepwise regression method to screen and optimize the weight of the key characteristic variables; Through cross-validation, the fitting performance and prediction ability of the model are evaluated, and finally the optimized carbon emission prediction model is output.

5. The method of claim 4, wherein, The formula of the carbon emission prediction model is as follows: ， wherein, represents the total carbon emissions over the building life cycle; represents the base value of the building life cycle carbon emissions; represents the linear regression coefficient of the i-th key feature variable; represents the regression coefficient of the non-linear term; represents the regression coefficient of the interaction term; n represents the number of key feature variables; m represents the total number of square terms; p represents the total number of interaction terms; represents the linear contribution of the key feature variable to the total carbon emissions; represents the contribution of the square term of the key feature variable to the carbon emissions, used to reflect the non-linear relationship of the variable; represents the influence of the interaction between two variables on carbon emissions, used to model the synergistic effect between variables; represents the random error or the carbon emission variation not explained by the model.

6. The method of claim 1, wherein, In step S4, the differential evolution algorithm specifically uses a double-objective optimization differential evolution algorithm to dynamically adjust the model parameters, taking carbon emission prediction error and calculation efficiency as optimization objectives.

7. The method of claim 1, wherein, In step S5, the causal inference algorithm specifically includes a method based on Bayesian network and structural equation model, by establishing the causal network diagram of building life cycle carbon emissions, using data-driven causal relationship inference technology, identifying and quantifying the key influencing factors and their causal paths of high carbon emission stage, and providing optimization strategy suggestions.

Images