A machine learning based method for predicting carbon emissions in bridge construction phases

Abstract

The application discloses a bridge construction stage carbon emission prediction method based on machine learning, first, a carbon emission calculation model is constructed based on a life cycle assessment method; then, activity data of the bridge construction stage carbon emission are acquired, and engineering variables influencing the carbon emission are extracted to construct a variable space; then, pre-processing is carried out, and the variable space is divided into a training sample set, a candidate sample set and a test sample set; then, the training sample set is used to construct a Kriging model, a sequential sampling strategy is introduced, and an information selection criterion is combined to dynamically optimize and train the Kriging model; after training, the test sample set is used to evaluate the Kriging model to obtain a bridge carbon emission prediction model; and finally, the bridge carbon emission prediction model is used for carbon emission prediction in the bridge construction stage. Through construction of a multi-source data feature space facing the bridge construction stage and a sequential sampling-ISC-Kriging model, high-precision, strong-generalization and extensible prediction of the bridge construction stage carbon emission are realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of bridge engineering carbon emission prediction, and particularly relates to a bridge construction stage carbon emission prediction method based on machine learning. BACKGROUND

[0002] As a key component in the transportation infrastructure system, the carbon emission problem of bridges is attracting more and more attention. Among them, bridge engineering has the characteristics of large resource consumption, long construction period and complex structure, and produces a large amount of greenhouse gas emissions in the whole life cycle, especially in the construction stage. Studies have shown that the carbon emissions in the construction stage of the bridge can account for more than 60% of the total emissions in the whole life cycle, mainly including the use of high-carbon building materials such as cement and steel, direct emissions generated by the operation of construction equipment and machinery, and indirect emission links such as transportation and processing.

[0003] At present, the evaluation method of bridge carbon emission mainly takes life cycle assessment (LCA) as the core framework, supplemented by inventory analysis and carbon factor method to account for the emissions of various activity processes; such method has the advantages of mature theoretical system and strong traceability, and has been widely used in the carbon assessment of bridges and other civil engineering facilities. However, with the expansion of engineering scale and the refinement of carbon management demand, the traditional LCA method gradually exposes the following technical bottlenecks; on the one hand, the prediction ability is insufficient: LCA emphasizes more on post-analysis, and lacks the forward-looking prediction ability of carbon emissions under different design schemes in the future, which is difficult to meet the needs of early control and optimization of carbon emissions; on the other hand, the nonlinear modeling ability is weak: the carbon emission of bridge is affected by structure size, material strength, construction technology, geographical environment and other multi-dimensional factors, and the variables often show strong nonlinear and interactive coupling characteristics, so the traditional statistical model is difficult to effectively fit and express; on the other hand, the data utilization rate is low: the current LCA method mostly uses static input, and cannot fully tap the data rules and knowledge potential in the existing engineering projects, resulting in that a large amount of historical data cannot be efficiently reused.

[0004] In recent years, with the development of artificial intelligence technology, especially machine learning algorithms, its advantages in engineering prediction modeling have gradually emerged. Machine learning can effectively capture the implicit rules between high-dimensional and nonlinear data, improve the generalization ability and prediction accuracy of the model, and has achieved good application results in the fields of building carbon emission prediction and structural performance evaluation. Some existing technologies have attempted to introduce machine learning into infrastructure carbon emission modeling, such as the invention patent with the application number 2024113086288, "Highway bridge carbon emission prediction method based on random forest", which applies random forest to bridge carbon emission prediction to provide optimization guidance for energy saving and carbon reduction in bridge construction; and the invention patent with the application number 2024115158454 and the invention name "Building implicit carbon emission prediction method and device based on machine learning", which uses machine learning algorithms to construct the correlation between influencing factors such as project characteristics, construction characteristics and management level and construction consumption, and quickly and conveniently carries out carbon emission quantitative prediction in the preliminary design stage of the building. Although the existing technology has achieved certain results, there are problems such as unsystematic model feature selection, uneven training sample distribution, and insufficient engineering capability suitable for complex bridge construction scenarios.

[0005] Therefore, it is urgent to construct a bridge construction stage carbon emission prediction method that integrates multi-source variable input, nonlinear machine learning modeling and engineering practice adaptation mechanism, to break through the limitations of current carbon assessment tools in precision, efficiency and universality, and realize scientific decision support for low-carbon construction of bridges. Based on this background, the present application proposes a bridge construction stage carbon emission prediction method based on machine learning, which improves the carbon emission modeling precision and practicality in complex engineering environment by constructing an efficient and generalizable prediction model, and provides key technical support for green and low-carbon construction of bridge engineering. SUMMARY

[0006] The main purpose of the present application is to overcome the problems of insufficient prediction accuracy, poor variable adaptability, weak nonlinear fitting ability and poor engineering universality of existing bridge carbon emission assessment and prediction methods, and to propose a bridge construction stage carbon emission prediction method based on machine learning, which constructs a multi-source data feature space for the bridge construction stage, constructs a Kriging model (sequential sampling-ISC-Kriging model) combined with sample filling criteria by introducing a sequential sampling strategy, and realizes high-precision, strong generalization and scalable prediction of carbon emission in the bridge construction stage.

[0007] In order to achieve the above purpose, the first purpose of the present application is to provide a bridge construction stage carbon emission prediction method based on machine learning, comprising the following steps:

[0008] construct a carbon emission calculation model based on the life cycle assessment method, the carbon emission calculation model including a carbon emission calculation model of carbon emission generated by building material consumption, a carbon emission calculation model of carbon emission generated by building material transportation, a carbon emission calculation model of carbon emission generated by construction machinery equipment, and a carbon emission calculation model of carbon emission generated by energy consumption;

[0009] obtain activity data of carbon emission in the bridge construction stage based on the carbon emission calculation model, and extract engineering variables affecting carbon emission to construct a variable space;

[0010] preprocess variable data in the variable space, and divide the variable data into a training sample set, a candidate sample set and a test sample set;

[0011] construct a Kriging model using the training sample set, introduce a sequential sampling strategy combined with an information selection criterion to add the candidate sample set to the training sample set to dynamically optimize and train the Kriging model, and evaluate the Kriging model using the test sample set after training to obtain a bridge carbon emission prediction model;

[0012] use the bridge carbon emission prediction model to predict carbon emission in the bridge construction stage.

[0013] As a preferred technical solution, based on the inventory analysis method in the life cycle assessment method, carbon emission generated by building material consumption, carbon emission generated by building material transportation, carbon emission generated by construction machinery equipment and carbon emission generated by energy consumption are classified according to the sources of carbon emission in the bridge construction stage;

[0014] construct a carbon emission calculation model for each type of carbon emission respectively; wherein:

[0015] the carbon emission calculation model of carbon emission generated by building material consumption is:

[0016] ,

[0017] in the formula, C 1 is the carbon emission amount generated by building material consumption in the bridge construction stage, V i is the carbon emission factor of the first i building material, Q i1 is the amount of the first i building material in the bridge construction stage;

[0018] the carbon emission calculation model of carbon emission generated by building material transportation is:

[0019] ,

[0020] in the formula, C2 represents the carbon emissions generated during the transportation of building materials during the bridge construction phase. η ij To adopt the first j The first unit mass of the transportation method i Carbon emissions per unit distance for a type of building material. T ij1 For building materials, the first j The transport distance of each mode of transport;

[0021] The carbon emission calculation model for the carbon emissions generated by the construction machinery and equipment is as follows:

[0022] ,

[0023] In the formula, C 3 represents the carbon emissions generated by construction machinery and equipment during the bridge construction phase. M i For the first i Carbon emission factors of various construction machinery S i1 For the first i The number of shifts for various types of construction machinery;

[0024] The carbon emission calculation model for the carbon emissions generated by the energy consumption is as follows:

[0025] ,

[0026] In the formula, C 4 represents the carbon emissions generated by energy consumption during the bridge construction phase. E i For the first i Carbon emission factors of various energy sources E i1 For the first i The amount of energy consumed.

[0027] As a preferred technical solution, the construction of the variable space specifically involves:

[0028] The correlation coefficient of Pearson was used to conduct correlation analysis on the engineering variables, and the correlation coefficients between each engineering variable and carbon emissions were analyzed and a heat map was drawn.

[0029] Analysis of variance was performed on the categorical variables in each engineering variable to test their significance.

[0030] Based on correlation coefficients and / or variances, the engineering variables are divided into three levels, and the top m engineering variables of each level are selected as key variables to construct the variable space.

[0031] As a preferred technical solution, the preprocessing includes:

[0032] The outliers in the variable data are removed, and the missing values are filled in;

[0033] The category variables in the variable data are converted into virtual variable encoding or classification variables by using one-hot encoding or MATLAB categorical format.

[0034] The numerical characteristic variables in the variable data are normalized.

[0035] As a preferred technical solution, the Kriging model is dynamically optimized and trained, specifically:

[0036] The Kriging model is constructed, the number of iterations and the termination condition are set, and the training is performed on the training sample set;

[0037] The trained Kriging model is used to predict the candidate sample set, and the predicted value and mean square error of each candidate sample are calculated;

[0038] The ISC value of each candidate sample is calculated based on the information selection criterion, and the candidate sample corresponding to the minimum ISC value is selected as the new training sample to be added to the training sample set, and the candidate sample is removed from the candidate sample set;

[0039] The training sample set is updated, and the trained Kriging model is optimized and trained;

[0040] The iteration is continuously performed until the candidate sample set is empty, and the training is terminated.

[0041] As a preferred technical solution, the Kriging model is evaluated by using mean square error and determination coefficient.

[0042] The second object is to provide a bridge construction phase carbon emission prediction system based on machine learning, which is applied to the bridge construction phase carbon emission prediction method, including a carbon emission calculation module, a variable construction module, a data processing module, a model training module, and a carbon emission prediction module.

[0043] The carbon emission calculation module is used to obtain activity data of carbon emissions in the bridge construction phase, and to construct a carbon emission calculation model based on the life cycle assessment method.

[0044] The variable construction module is used to extract engineering variables affecting carbon emissions based on the activity data of carbon emissions in the bridge construction phase, and to construct a variable space.

[0045] The data processing module is used to preprocess the variable data in the variable space, and to divide it into a training sample set, a candidate sample set, and a test sample set.

[0046] The model training module is configured to use a training sample set to construct a Kriging model, introduce a sequential sampling strategy combined with an information selection criterion to add a candidate sample set to the training sample set to dynamically optimize and train the Kriging model, and use a test sample set to evaluate the Kriging model after training to obtain an optimal Kriging model.

[0047] The carbon emission prediction module is configured to use the optimal Kriging model to predict carbon emissions in the bridge construction stage.

[0048] A third object is to provide an electronic device comprising:

[0049] at least one processor; and a memory connected to the at least one processor in communication; wherein

[0050] The memory stores computer program instructions executable by the at least one processor, and the computer program instructions are executed by the at least one processor to enable the at least one processor to perform the bridge construction stage carbon emission prediction method.

[0051] A fourth object is to provide a computer-readable storage medium storing a program, when the program is executed by a processor, the bridge construction stage carbon emission prediction method is implemented.

[0052] A fifth object is to provide a computer program product comprising a computer program or instructions, when the computer program or instructions are executed by a processor, the bridge construction stage carbon emission prediction method is implemented.

[0053] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0054] 1. The nonlinear prediction accuracy is significantly improved:

[0055] The sequential sampling-ISC-Kriging model proposed in the present application dynamically identifies the prediction uncertainty area by introducing an information selection criterion (Information Selection Criterion, ISC) in the prediction process, and optimizes the training sample distribution using a sequential sampling strategy, thereby effectively improving the response capability of the model to complex nonlinear input variables. The engineering verification results show that the prediction error of the model in multiple bridge type scenarios is stably controlled within 5%, which is significantly better than the prediction performance of the existing Kriging model.

[0056] 2. Support multi-dimensional high-complexity input variable modeling:

[0057] The application establishes a multi-dimensional input feature space covering key factors of bridge construction, including but not limited to nine influencing factors such as bridge length, bridge width, structure type, main material type, construction method, transportation distance, component size, beam height, and construction equipment type; the modeling structure can comprehensively depict the influence mechanism of carbon emission, effectively solve the limitations of existing methods in terms of multi-variable adaptability and modeling accuracy, and is suitable for bridge carbon emission modeling requirements under various construction conditions.

[0058] 3. Strong model generalization ability and low sample dependence:

[0059] To address the problem of limited number and uneven quality of actual engineering data samples, the application introduces a sequential sampling strategy combined with the ISC mechanism to actively identify high-uncertainty areas during model training, gradually supplement sample data with information gain value, and realize optimal allocation of sample resources. This mechanism significantly reduces the dependence of the model on the number of high-quality training samples, enhances the robustness and generalization ability of the model under small sample conditions, and is particularly suitable for data scarcity problems in regional bridge projects or new bridge scenarios.

[0060] 4. Good engineering adaptability and generalization:

[0061] The method proposed in the application has been modeled and verified based on the measured and investigated data of 105 medium and small span bridges in Guangdong Province, covering typical structure forms such as prestressed hollow slab beam, concrete box beam, steel box beam, and steel plate beam. The model shows good adaptability under different bridge types, geographical regions, and construction modes, and has the potential for universality and promotion across bridge types and projects, providing technical support for carbon emission prediction for bridge construction projects in multiple regions.

[0062] 5. Support for low-carbon decision optimization:

[0063] Compared with traditional evaluation methods that can only provide post-hoc emission analysis, the method proposed in the application has the ability of prospective prediction and decision support, and can provide quantitative basis for material selection, construction method evaluation, and transportation path planning in the bridge design stage. In addition, the model can be integrated with building information model (BIM) or bridge design optimization platform to realize multi-objective design optimization under the constraint of carbon emission target, and provide strong data support and intelligent algorithm guarantee for low-carbon management of bridge engineering throughout the life cycle. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0065] Figure 1 The figure is a whole flow chart of the bridge construction stage carbon emission prediction method based on machine learning in the embodiment of the present application.

[0066] Figure 2 The figure is a prediction effect diagram of two models in a low-dimensional variable space in the embodiment of the present application.

[0067] Figure 3 The figure is a prediction effect diagram of two models in a medium-dimensional variable space in the embodiment of the present application.

[0068] Figure 4 The figure is a prediction effect diagram of two models in a high-dimensional variable space in the embodiment of the present application.

[0069] Figure 5 The figure is a structure schematic diagram of the bridge construction stage carbon emission prediction system based on machine learning in the embodiment of the present application.

[0070] Figure 6 The figure is a structure schematic diagram of the electronic device in the embodiment of the present application. DETAILED DESCRIPTION

[0071] In order to enable persons skilled in the art to better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative labor are within the scope of protection of the present application.

[0072] In the present application, "embodiment" means that the specific features, structures or characteristics described in combination with the embodiment can be contained in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily mean the same embodiment, nor is it an independent or alternative embodiment to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in the present application can be combined with other embodiments.

[0073] As shown in Figure 1 The embodiment provides a bridge construction stage carbon emission prediction method based on machine learning, which comprises the following steps:

[0074] S1, a carbon emission calculation model is constructed based on a life cycle assessment method, including a carbon emission calculation model of building material consumption, a carbon emission calculation model of building material transportation, a carbon emission calculation model of construction machinery and equipment, and a carbon emission calculation model of energy consumption.

[0075] Further, the present application is based on the inventory analysis method in the life cycle assessment method (LCA), and is classified into carbon emissions generated by building material consumption, carbon emissions generated by building material transportation, carbon emissions generated by construction machinery and equipment, and carbon emissions generated by energy consumption according to the sources of carbon emissions in the bridge construction stage; and a carbon emission calculation model is constructed for each type of carbon emission. Among them:

[0076] The carbon emission calculation model of carbon emissions generated by building material consumption is:

[0077] ,

[0078] In the formula, C 1 is the carbon emission amount generated by building material consumption in the bridge construction stage, and the unit is: kilogram of carbon dioxide equivalent (kgCO2e); V i is the carbon emission factor of the i th building material, and the unit can be: kilogram of carbon dioxide equivalent per ton (kgCO2e / t), kilogram of carbon dioxide equivalent per square (kgCO2e / m 2 ), kilogram of carbon dioxide equivalent per cubic (kgCO2e / m 3 ); Q i1 is the amount of the i th building material in the bridge construction stage, and the unit can be: ton (t), square (m 2 ), cubic (m 3 ).

[0079] The carbon emission calculation model of carbon emissions generated by building material transportation is:

[0080] ,

[0081] In the formula, C 2 is the carbon emission amount generated by building material transportation in the bridge construction stage, and the unit is: kilogram of carbon dioxide equivalent (kgCO2e); η ij is the carbon emission amount of the j th building material per unit mass per unit distance transported by the i th transportation mode, and the unit is: kilogram of carbon dioxide equivalent per ton per kilometer (kgCO2e / t·km); T ij1 is the transportation distance of the building material by the j th transportation mode, and the unit is: kilometer (km).

[0082] The carbon emission calculation model of carbon emissions generated by construction machinery and equipment is:

[0083] ,

[0084] In the formula, C 3 is the carbon emission of construction machinery and equipment in the bridge construction stage, in units of kilogram of carbon dioxide equivalent (kgCO2e); M i is the carbon emission factor of the first i kind of construction machinery, in units of kilogram of carbon dioxide equivalent per shift (kgCO2e / shift); S i1 is the number of shifts of the first i kind of construction machinery, in units of shifts.

[0085] The carbon emission calculation model of carbon emission generated by energy consumption is:

[0086] ,

[0087] In the formula, C 4 is the carbon emission of energy consumption in the bridge construction stage, in units of kilogram of carbon dioxide equivalent (kgCO2e); E i is the carbon emission factor of the first i kind of energy, which can be in units of kilogram of carbon dioxide equivalent per kilowatt-hour (kgCO2e / kWh), kilogram of carbon dioxide equivalent per ton (kgCO2e / t), kilogram of carbon dioxide equivalent per liter (kgCO2e / L), kilogram of carbon dioxide equivalent per cubic meter (kgCO2e / m 3 ); E i1 is the consumption of the first i kind of energy, which can be in units of kilowatt-hour (kWh), ton (t), liter (L), or cubic meter (m 3 ).

[0088] S2, based on the carbon emission calculation model, obtaining activity data of carbon emission in the bridge construction stage, and extracting engineering variables affecting carbon emission to construct a variable space.

[0089] Further, before constructing the bridge carbon emission prediction model, reasonably selecting the prediction index is the key to ensuring the accuracy of the model. The influencing factors of carbon emission in the bridge construction stage are complex, involving multiple aspects such as the functional positioning of the bridge, the structural form, the geometric parameters, the material selection, the construction method, and the transportation conditions. The present application performs correlation analysis on the engineering variables and carbon emission in the bridge construction stage, uses Pearson correlation coefficient analysis and variance analysis to select prediction variables to construct a variable space, so that the variables can more comprehensively describe the characteristics of the bridge and affect the carbon emission level at different levels; specifically:

[0090] S2.1, first, the Pearson correlation coefficient is used to analyze the correlation of the engineering variables, the correlation coefficient of each engineering variable and carbon emission is analyzed and a heat map is drawn.

[0091] S2.2, the variance analysis test significance of the category variables in each engineering variable.

[0092] S2.3, based on the correlation coefficient and / or variance, each engineering variable is divided into three levels, and the first m engineering variables in each level are selected as the key variables to construct the variable space.

[0093] In this embodiment, the data of 105 medium and small span bridges in multiple cities in Guangdong Province is collected, and the activity data such as the structure size, construction method, main material type, transportation distance of the bridge construction stage is obtained; after Pearson correlation coefficient and variance analysis, it is divided into three levels, including extremely significant level, significant level and less significant level. For the extremely significant level, if the p value of the engineering variable of the non-category variable is much smaller than 0.05, it is divided into this level, if the p value of the engineering variable of the category variable is much smaller than 0.05 and the F value is relatively high, it is divided into this level; for the significant level, if the p value of the engineering variable of the non-category variable is less than 0.05, it is divided into this level, if the p value of the engineering variable of the category variable is less than 0.05 and the F value is relatively low, it is divided into this level; for the less significant level, the engineering variables that do not meet the above two levels are divided into this level. Finally, the first 3 engineering variables in each level are selected as the key variables, and a variable space containing 9 types of variables such as bridge length, bridge width, beam height, hole number, bridge type, main material type (such as concrete, steel), construction method (precast or cast-in-place), transportation distance and bridge function (pedestrian bridge, vehicle and pedestrian bridge, etc.) is constructed, which is used for subsequent bridge carbon emission prediction model prediction; these variables can comprehensively describe the characteristics of the bridge and affect the carbon emission level at different levels, and the selection of these variables is based on the significant influence on carbon emission, the availability of data and the practical significance of engineering.

[0094] S3, the variable data in the variable space is preprocessed and divided into training sample set, candidate sample set and test sample set.

[0095] Further, in order to ensure the reliability and accuracy of the data, the data needs to be preprocessed, including:

[0096] S3.1, the outliers in the variable data are removed, and the missing values are filled to ensure the integrity of the data.

[0097] S3.2, the category variable in the variable data (such as bridge function, bridge type, main material type, construction method) is converted into virtual variable coding or classification variable by using one-hot coding or MATLAB categorical format.

[0098] S3.3, the numerical characteristic variable in each variable data (such as hole number, beam height, bridge width, bridge length, transportation distance) is normalized to eliminate the influence between different dimensions, reduce the influence of scale difference on model parameter estimation, and improve the stability and convergence of the model.

[0099] In this embodiment, the data of 105 medium and small span bridges collected after variable screening is input into MATLAB for data preprocessing operation, and the preprocessed data is randomly divided into a training sample set containing 50 samples, a candidate sample set containing 35 samples and a test sample set containing 20 samples.

[0100] S4, the Kriging model is constructed using the training sample set, the candidate sample set is added to the training sample set to dynamically optimize and train the Kriging model by introducing the sequential sampling strategy combined with the information selection criterion; after training, the Kriging model is evaluated using the test sample set to obtain a bridge carbon emission prediction model.

[0101] Further, the traditional Kriging model has problems of insufficient prediction accuracy, poor adaptability to high-dimensional nonlinear variables, dependence on static samples, weak generalization ability and the like in bridge construction stage carbon emission prediction, and is difficult to meet the modeling demand in complex engineering scenarios. The sequential sampling-ISC-Kriging model proposed in the present application effectively improves the adaptability and prediction accuracy of the model to complex variable space by introducing the information selection criterion (Information Selection Criterion, ISC) and the sequential sampling strategy, while maintaining the advantages of Kriging space modeling, and is especially suitable for bridge carbon emission prediction scenarios with limited samples and highly coupled variables; the dynamic optimization training process of the sequential sampling-ISC-Kriging model is as follows:

[0102] S4.1, the Kriging model is constructed and the number of iterations and termination conditions are set, and the training is performed on the training sample set.

[0103] S4.2, the trained Kriging model is used to predict the candidate sample set, and the predicted value and mean square error of each candidate sample are calculated.

[0104] S4.3, calculate the ISC value of each candidate sample based on the information selection criterion (ISC), and select the candidate sample corresponding to the minimum ISC value as the new training sample to be added to the training sample set, and remove the candidate sample from the candidate sample set.

[0105] S4.4, update the training sample set and optimize the training of the Kriging model;

[0106] S4.5, continuously iterate until the candidate sample set is empty to terminate the training.

[0107] Further, after the optimization training is completed, the test sample set is used to evaluate the Kriging model after the optimization training, and the mean square error (MSE) and the determination coefficient (R²) are used as the evaluation indexes.

[0108] In this embodiment, the variable space is divided into three types of combinations to form a multi-dimensional high-complexity input variable, including: low dimension (3 variables, including bridge length, main material type and bridge width three influence variables), medium dimension (5 variables, including bridge length, main material type, bridge width, beam height and bridge type five influence variables) and high dimension (9 variables, including bridge length, main material type, bridge width, beam height, bridge type, construction method, hole number, transportation distance and bridge function nine influence variables); the above three models are trained under different dimensions, the model performance is compared, and the model accuracy is evaluated by using the MSE (mean square error), R² (determination coefficient) and other indexes; the results are shown in the following table 1:

[0109] Table 1 Performance index data of three models

[0110]

[0111] As can be seen from Table 1, under the low-dimensional, medium-dimensional and high-dimensional variable space, the performance of the sequential sampling-ISC-Kriging model is better than that of the traditional Kriging model in all indexes, and the specific performance is as follows: under the low-dimensional variable space (bridge length, main material type, bridge width), the MSE of the sequential sampling-ISC-Kriging model is 0.1837, which is reduced by about 24.4% compared with the Kriging model; R<2> is increased to 0.7414, which shows that even under the condition of limited number of variables, the model also has stronger fitting ability. In the medium-dimensional variable space (newly added beam height, bridge type), the advantage of the model is further shown, the MSE of the sequential sampling-ISC-Kriging model is 0.0778, which is only 41.2% of the Kriging model, and R<2> is significantly increased to 0.8904, close to the high fitting state. In the high-dimensional variable space (containing nine key variables), the MSE of the sequential sampling-ISC-Kriging model is further reduced to 0.0680, and R<2> reaches 0.9200, which shows that the model has excellent learning ability and global expression ability when dealing with complex variable relationships. Therefore, with the increase of the dimension of the variable, the Kriging model gradually cannot cope with the non-linear fitting ability and residual control, while the sequential sampling-ISC-Kriging model always maintains excellent prediction performance after introducing the information selection criterion and the sequential sampling strategy, which reflects good multi-dimensional variable adaptability, model stability and generalization ability. The results verify that the modeling framework proposed in the application can significantly alleviate the problems of underfitting and weak generalization of the traditional Kriging model in the high-dimensional variable space, and is particularly suitable for bridge engineering scenarios with complex carbon emission influencing factors and strong multi-variable interaction.

[0112] S5, the bridge carbon emission prediction model is used for carbon emission prediction in the bridge construction stage.

[0113] Further verify the practicability and prediction accuracy of the model, that is, the prediction accuracy of the sequential sampling-ISC-Kriging model, compare the prediction results of the Kriging model, and thus illustrate the prediction effect of the bridge carbon emission prediction model in the application. Select a certain cross-creek bridge in Foshan City, Guangdong Province as an actual application case, take the measured unit volume carbon emission value (665.93 kgCO2e / m<3>) as a reference, compare the prediction results and error rates of the two models under three types of variable combinations, and the results are shown in Table 2, Figure 2 、 3 、4 as follows:

[0114] Table 2 Prediction results and errors of three types of models

[0115]

[0116] From Table 2, it can be seen that under the low-dimensional combination of variable space, the prediction error of the Kriging model is 22.31%, which deviates far from the actual value; the prediction error of the sequential sampling-ISC-Kriging model is 15.37%, the error is obviously reduced, and it shows more robust reasoning ability under the condition of missing information. And from Figure 2 It can be seen that although the prediction results of the Kriging model are relatively concentrated, there are obvious discrete points, especially in the low value area and the high value area, which is not ideal; there is obvious deviation between the predicted value and the true value, and the data points are relatively dense but lack obvious linear trend. The effect diagram of the sequential sampling-ISC-Kriging model proposed in the application shows that the data points are relatively dense and concentrated, and show good linear consistency; especially in the densely populated data interval, the prediction effect is significantly enhanced, and can better fit the true value distribution.

[0117] In the variable space dimensional combination, the Kriging model prediction value in Table 2 is significantly lower than the measured value, with an error of 16.91%, which shows that its fitting effect does not significantly improve after the variable increases, and from Figure 3 It can be seen that the data points of the Kriging model are distributed in a certain aggregation, but there are still discrete points, especially in the high and low prediction value areas; there is a certain deviation between the predicted value and the true value, although the performance is improved compared with the low-dimensional combination of variable space, but in the complex bridge carbon emission prediction task, it still shows local underfitting phenomenon. The prediction error of the sequential sampling-ISC-Kriging model is reduced to 7.18%, which is close to the actual value, and the accuracy is greatly improved, Figure 3 It is also shown in the above that the data points are more dense and concentrated, and show good linear consistency, especially in the densely populated data interval, which can accurately fit the true value distribution, so as to verify its high adaptability to the coupling relationship between typical engineering variables.

[0118] In the high-dimensional combination of variable space, Table 2 reflects that the prediction error of the Kriging model is 12.54%, which has certain systematic deviation; from Figure 4It can be seen from Table 6 that there is still a certain deviation between the predicted value of the Kriging model and the true value, especially in the area with more data, which shows a certain degree of discreteness; although the data points are more densely distributed than the variable space low dimension and variable space dimension combination, there are still some abnormal values deviating, which shows that although the variable space high dimension combination introduces more characteristic variables (such as construction method, hole number, transportation distance and bridge function), the fitting ability of Kriging model to nonlinear change is still insufficient; therefore the fitting ability of the model in the high value area and the low value area is still weak, especially in the comprehensive modeling of complex bridge characteristics, it is difficult to accurately capture the coupling relationship between variables. The error of the sequential sampling-ISC-Kriging model is further compressed to 5.65%, which maintains high precision prediction ability under the condition of high dimensional data, and shows excellent nonlinear feature recognition and global fitting ability. And from Table 7, it can be seen that the data points are highly concentrated, showing good linear consistency, especially in the data intensive area and the high value area, which shows excellent prediction ability; the fitting of the predicted value and the true value is very high, and there is almost no obvious deviation; the significant improvement of this effect is mainly due to the application of active learning strategy (sequential sampling-ISC method), which effectively extracts the characteristic correlation between complex variables in the process of continuously optimizing sample selection, significantly improves the generalization ability and prediction accuracy of the model. Figure 4

[0119] In summary, the prediction error of the sequential sampling-ISC-Kriging model under the combination of three types of variables is significantly lower than that of the Kriging model, and the error gradually decreases with the increase of the number of variables, which shows good variable adaptability and model expansibility, and can provide reliable carbon emission prediction basis for actual bridge projects; its high precision prediction ability also provides forward-looking technical support for subsequent low-carbon design, construction process optimization, material selection, etc.

[0120] It should be noted that for the foregoing method embodiments, in order to facilitate description, they are all described as a series of action combinations, but those skilled in the art should know that the present application is not limited by the order of the described actions, because according to the present application, certain steps can be performed in other orders or simultaneously.

[0121] Example 2

[0122] ​Based on the same idea as the machine learning-based bridge construction phase carbon emission prediction method in the above embodiment, the present application also provides a machine learning-based bridge construction phase carbon emission prediction system, which can be used to execute the machine learning-based bridge construction phase carbon emission prediction method described above. For the convenience of description, in the structural schematic diagram of the embodiment of the machine learning-based bridge construction phase carbon emission prediction system, only the parts related to the embodiment of the present application are shown, and those skilled in the art can understand that the illustrated structure does not constitute a limitation on the device, which can include more or fewer components than the illustrated, or combine certain components, or different component arrangements.

[0123] As shown in Figure 5 Another embodiment of the present application provides a machine learning-based bridge construction phase carbon emission prediction system, which includes a carbon emission calculation module, a variable construction module, a data processing module, a model training module, and a carbon emission prediction module.

[0124] The carbon emission calculation module is used to obtain activity data of carbon emission in the bridge construction phase, and to construct a carbon emission calculation model based on the life cycle assessment method.

[0125] The variable construction module is used to extract engineering variables affecting carbon emission based on the activity data of carbon emission in the bridge construction phase, and to construct a variable space.

[0126] The data processing module is used to pre-process each variable data in the variable space, and to divide it into a training sample set, a candidate sample set, and a test sample set.

[0127] The model training module is used to construct a Kriging model using the training sample set, to introduce a sequential sampling strategy combined with a sample filling criterion to add the candidate sample set to the training sample set to dynamically optimize and train the Kriging model, and to use the test sample set to evaluate the Kriging model after training to obtain an optimal Kriging model.

[0128] The carbon emission prediction module is used to use the optimal Kriging model to predict carbon emission in the bridge construction phase.

[0129] It should be noted that the machine learning-based bridge construction phase carbon emission prediction method system of the present application corresponds one-to-one to the machine learning-based bridge construction phase carbon emission prediction method of the present application, and the technical features and advantages described in the embodiment of the machine learning-based bridge construction phase carbon emission prediction method are applicable to the embodiment of the machine learning-based bridge construction phase carbon emission prediction system, and the specific content can be referred to the description in the method embodiment of the present application. Therefore, it is declared here.

[0130] Furthermore, in the above embodiments of the machine learning-based carbon emission prediction system for bridge construction, the logical division of each program module is merely illustrative. In practical applications, the above functions can be assigned to different program modules as needed, for example, for the sake of corresponding hardware configuration requirements or the convenience of software implementation. That is, the internal structure of the machine learning-based carbon emission prediction system for bridge construction can be divided into different program modules to complete all or part of the functions described above.

[0131] Example 3

[0132] like Figure 6 As shown, in one embodiment, an electronic device is provided for implementing the above-described method for predicting carbon emissions during the bridge construction phase. The electronic device may include a first processor, a first memory, and a bus, and may also include a computer program stored in the first memory and executable on the first processor, such as a carbon emission prediction program for the bridge construction phase.

[0133] The first memory includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, disk, optical disk, etc. In some embodiments, the first memory can be an internal storage unit of an electronic device, such as a portable hard drive. In other embodiments, the first memory can be an external storage device of the electronic device, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc. Furthermore, the first memory can include both internal and external storage units of the electronic device. The first memory can be used not only to store application software and various types of data installed on the electronic device, such as the code for a carbon emission prediction program during bridge construction, but also to temporarily store data that has been output or will be output.

[0134] The first processor in some embodiments can be composed of integrated circuits, for example, can be composed of a single packaged integrated circuit, or can be composed of multiple packaged integrated circuits of the same function or different functions, including one or more combinations of central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The first processor is the control unit of the electronic device, which connects various components of the entire electronic device through various interfaces and lines, and executes various functions and processes data of the electronic device by running or executing programs or modules stored in the first memory (such as the bridge construction stage carbon emission prediction program), and calling data stored in the first memory.

[0135] Figure 6 Only the electronic device with components is shown, and those skilled in the art can understand that, Figure 6 The structure shown does not constitute a limitation on the electronic device, and can include fewer or more components than shown, or combine certain components, or different component arrangements.

[0136] The bridge construction stage carbon emission prediction program stored in the first memory in the electronic device is a combination of multiple instructions, which, when running in the first processor, can achieve:

[0137] A carbon emission calculation model is constructed based on a life cycle assessment method, and the carbon emission calculation model includes a carbon emission calculation model of building material consumption, a carbon emission calculation model of building material transportation, a carbon emission calculation model of construction machinery and equipment, and a carbon emission calculation model of energy consumption;

[0138] Activity data of carbon emissions in the bridge construction stage are obtained based on the carbon emission calculation model, and engineering variables affecting carbon emissions are extracted to construct a variable space;

[0139] The variable data in the variable space are preprocessed and divided into a training sample set, a candidate sample set, and a test sample set;

[0140] A Kriging model is constructed using the training sample set, a sequential sampling strategy is introduced, and a sample filling criterion is combined to add the candidate sample set to the training sample set to dynamically optimize and train the Kriging model; after training, the Kriging model is evaluated using the test sample set to obtain a bridge carbon emission prediction model;

[0141] The bridge carbon emission prediction model is used for carbon emission prediction in the bridge construction stage.

[0142] Further, the modules / units of the electronic device, if implemented in the form of software function units and sold or used as independent products, can be stored in a nonvolatile computer-readable storage medium. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM).

[0143] Those of ordinary skill in the related art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments of each method. In the embodiments provided in the present application, any reference to memory, storage, database or other medium can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0144] The technical features of the above embodiments can be combined in any way. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not contradict, they should be considered within the scope of the present application.

[0145] The above embodiments are the preferred embodiments of the present application, but the embodiments of the present application are not limited to the above embodiments, and any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present application should be considered as equivalent replacement modes and should be included in the protection scope of the present application.

Claims

1. A machine learning-based method for predicting carbon emissions during bridge construction, characterized in that, Includes the following steps: A carbon emission calculation model is constructed based on the life cycle assessment method. The carbon emission calculation model includes a carbon emission calculation model for the consumption of building materials, a carbon emission calculation model for the transportation of building materials, a carbon emission calculation model for the construction machinery and equipment, and a carbon emission calculation model for the energy consumption. Based on a carbon emission calculation model, activity data on carbon emissions during the bridge construction phase are obtained, and engineering variables affecting carbon emissions are extracted to construct a variable space. The activity data includes the structural dimensions, construction methods, main material types, and transportation distances during the bridge construction phase. The data of each variable in the variable space are preprocessed and divided into training sample set, candidate sample set and test sample set; The Kriging model is constructed using the training sample set. A sequential sampling strategy is introduced, combined with the information selection criterion, to add the candidate sample set to the training sample set for dynamic optimization training of the Kriging model. After training, the Kriging model was evaluated using a test sample set to obtain a bridge carbon emission prediction model. The bridge carbon emission prediction model was used to predict carbon emissions during the bridge construction phase. The dynamic optimization training of the Kriging model specifically involves: Build a Kriging model and set the number of iterations and termination conditions, and train it on the training sample set; The trained Kriging model is used to predict the candidate sample set, and the predicted value and mean squared error of each candidate sample are calculated. The ISC value of each candidate sample is calculated based on the information selection criterion, and the candidate sample with the smallest ISC value is selected as a new training sample and added to the training sample set. The candidate sample is then removed from the candidate sample set. Update the training sample set and optimize the trained Kriging model; The training process continues iteratively until the candidate sample set is empty, at which point the training terminates.

2. The method for predicting carbon emissions during the bridge construction phase according to claim 1, characterized in that, Based on the inventory analysis method in the life cycle assessment approach, carbon emissions are classified according to the sources of carbon emissions during the bridge construction stage, into carbon emissions from the consumption of building materials, carbon emissions from the transportation of building materials, carbon emissions from construction machinery and equipment, and carbon emissions from energy consumption. For each type of carbon emission, a carbon emission calculation model is constructed; where: The carbon emission calculation model for the carbon emissions generated by the consumption of building materials is as follows: ， In the formula, C 1 represents the carbon emissions generated by the consumption of building materials during the bridge construction phase. V i For the first i Carbon emission factors of various building materials Q i1 For the bridge construction phase i The amount of various building materials used; The carbon emission calculation model for the carbon emissions generated from the transportation of building materials is as follows: ， In the formula, C 2 represents the carbon emissions generated during the transportation of building materials during the bridge construction phase. η ij To adopt the first j The first unit mass of the transportation method i Carbon emissions per unit distance for a type of building material. T ij1 For building materials, the first j The transport distance of each mode of transport; The carbon emission calculation model for the carbon emissions generated by the construction machinery and equipment is as follows: ， In the formula, C 3 represents the carbon emissions generated by construction machinery and equipment during the bridge construction phase. M i For the first i Carbon emission factors of various construction machinery S i1 For the first i The number of shifts for various types of construction machinery; The carbon emission calculation model for the carbon emissions generated by the energy consumption is as follows: ， In the formula, C 4 represents the carbon emissions generated by energy consumption during the bridge construction phase. E i For the first i Carbon emission factors of various energy sources N i1 For the first i The amount of energy consumed.

3. The method for predicting carbon emissions during the bridge construction phase according to claim 1, characterized in that, The construction of the variable space specifically refers to: The correlation coefficient of Pearson was used to conduct correlation analysis on the engineering variables, and the correlation coefficients between each engineering variable and carbon emissions were analyzed and a heat map was drawn. Analysis of variance was performed on the categorical variables in each engineering variable to test their significance. Based on correlation coefficients and / or variances, the engineering variables are divided into three levels, and the top m engineering variables of each level are selected as key variables to construct the variable space.

4. The method for predicting carbon emissions during the bridge construction phase according to claim 1, characterized in that, The preprocessing includes: Outliers were removed and missing values ​​were filled in the data of each variable; Convert categorical variables in the variable data into dummy variable codes or categorical variables using one-hot encoding or MATLAB categorical format; Normalize the numerical characteristic variables in each variable's data.

5. The method for predicting carbon emissions during the bridge construction phase according to claim 1, characterized in that, The Kriging model is evaluated using mean squared error and coefficient of determination.

6. A machine learning-based carbon emission prediction system for bridge construction, characterized in that, The carbon emission prediction method for the bridge construction stage applied to any one of claims 1-5 includes a carbon emission calculation module, a variable construction module, a data processing module, a model training module, and a carbon emission prediction module; The carbon emission calculation module is used to acquire activity data on carbon emissions during the bridge construction phase and to construct a carbon emission calculation model based on the life cycle assessment method. The variable construction module is used to extract engineering variables affecting carbon emissions based on activity data of carbon emissions during the bridge construction phase, and to construct a variable space; The data processing module is used to preprocess the data of each variable in the variable space and divide it into training sample set, candidate sample set and test sample set. The model training module is used to construct a Kriging model using a training sample set. It introduces a sequential sampling strategy combined with information selection criteria to add candidate sample sets to the training sample set and dynamically optimize the Kriging model. After training, the Kriging model is evaluated using a test sample set to obtain the optimal Kriging model. The carbon emission prediction module is used to use the optimal Kriging model to predict carbon emissions during the bridge construction phase. The dynamic optimization training of the Kriging model specifically involves: Build a Kriging model and set the number of iterations and termination conditions, and train it on the training sample set; The trained Kriging model is used to predict the candidate sample set, and the predicted value and mean squared error of each candidate sample are calculated. The ISC value of each candidate sample is calculated based on the information selection criterion, and the candidate sample with the smallest ISC value is selected as a new training sample and added to the training sample set. The candidate sample is then removed from the candidate sample set. Update the training sample set and optimize the trained Kriging model; The training process continues iteratively until the candidate sample set is empty, at which point the training terminates.

7. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to said at least one processor; wherein, The memory stores computer program instructions that can be executed by the at least one processor to enable the at least one processor to perform the bridge construction phase carbon emission prediction method as described in any one of claims 1-5.

8. A computer-readable storage medium storing a program, characterized in that, When the program is executed by the processor, it implements the carbon emission prediction method for the bridge construction phase as described in any one of claims 1-5.

9. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the processor, they implement the carbon emission prediction method for the bridge construction phase as described in any one of claims 1-5.

Images