Highway Carbon Emission Analysis Method, Storage Medium and Computer Terminal Device Based on Dynamic Bayesian Network

Abstract

The technical problem to be solved by the present invention is to provide a highway carbon emission analysis method, a storage medium and a computer terminal device based on a dynamic Bayesian network, which can realize the dynamic analysis and probabilistic evaluation of highway carbon emissions during the operation period. The highway carbon emission analysis method based on a dynamic Bayesian network includes establishing a Kaya identity formula according to the key factors affecting the total highway carbon emissions in the research target area; taking the key factors as hidden variables, and taking the primary variables and secondary variables affecting each hidden variable as observed variables to establish a dynamic Bayesian network model; using the dynamic Bayesian network model to simulate the states of the hidden variables to generate posterior distribution samples of the hidden variables at different time points; using the LMDI decomposition method to calculate the contribution degree of each hidden variable in each hidden variable sample, and statistically analyzing the contribution degrees of all samples to generate the probability distribution of the contribution degrees of each hidden variable. The present invention helps to formulate more scientific and accurate dynamic emission reduction strategies.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission analysis, and specifically to a method for analyzing highway carbon emissions based on a dynamic Bayesian network, a storage medium, and a computer terminal device. Background Art

[0002] With the continuous expansion of the highway network scale, the carbon emission problem during the operation period of highways has become increasingly prominent, bringing great pressure to the environment. Identifying the main contributing factors of highway carbon emissions is of great significance for studying carbon emission reduction throughout the life cycle of highways in the target area.

[0003] Currently, the research on highway carbon emissions is mostly based on static models, which are difficult to comprehensively and dynamically analyze the influencing factors of carbon emissions and their changing laws. Existing research mostly uses static models for factor decomposition of carbon emissions. For example, the combination of the traditional Kaya identity and the LMDI decomposition method can identify the driving factors of carbon emissions, but it is difficult to describe the impact mechanism of time-varying characteristics such as highway mileage expansion and dynamic changes in traffic volume on carbon emissions. Although some scholars have tried to introduce Bayesian networks to handle uncertainties, there are still significant defects in the construction of variable systems. There are three major bottlenecks in the current technology: First, static models cannot reflect the dynamic evolution law of carbon emission driving factors during the life cycle of highways; second, the research on the dynamic correlation mechanism of latent variables is insufficient; finally, the multi-level conduction mechanism of carbon emission influencing factors has not been clarified, and it is difficult to establish a quantitative response relationship between macro policies and micro operations. Although existing research has constructed a highway carbon emission model based on Kaya-LMDI, it does not consider the cumulative effect of road network dynamic expansion on carbon emissions; for another example, although the proposed Bayesian network method introduces traffic volume variables, it does not achieve dynamic reasoning for multiple time steps. These defects result in the lack of time series prediction ability and risk quantification assessment function in existing methods when formulating dynamic emission reduction strategies, and it is difficult to meet the urgent need for precise control of highway carbon emissions during the operation period under the "dual carbon" goal. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for analyzing highway carbon emissions based on a dynamic Bayesian network, a storage medium, and a computer terminal device that can achieve dynamic analysis and probabilistic evaluation of highway carbon emissions during the operation period.

[0005] The technical solution adopted by the present invention to solve its technical problems is: A method for analyzing highway carbon emissions based on a dynamic Bayesian network, including the following steps:

[0006] Establish a Kaya identity formula according to the key factors affecting the total carbon emissions of highways in the target research area, and the key factors include the total mileage of highways in the target research area , traffic volume , the total energy consumption of each vehicle and the carbon emission intensity , where the carbon emission intensity is the carbon emissions generated per unit of energy consumption, and the Kaya identity formula is:

[0007] , where is the total highway carbon emissions in the research target area;

[0008] Taking the key factors as latent variables, and taking the first-level variables and second-level variables that affect each latent variable as observed variables, and establishing a dynamic Bayesian network model according to the dependence relationship and conditional probability distribution between variables, where the first-level variables directly affect the corresponding latent variables, and the second-level variables indirectly affect the corresponding latent variables by affecting the first-level variables;

[0009] Using the dynamic Bayesian network model to simulate the states of latent variables and generate posterior distribution samples of latent variables at different time points;

[0010] Using the LMDI decomposition method to calculate the contribution of each latent variable in each latent variable sample from time step t to t + 1 to the change in the total highway carbon emissions in the research target area, and statistically analyzing the contributions of all samples to generate the probability distribution of the contributions of each latent variable, and then obtaining the interval graph of the percentage of the contribution of each latent variable to the change in the total highway carbon emissions in the research target area under the set confidence level.

[0011] Furthermore, the total mileage , traffic flow , the total energy consumption of each vehicle , carbon emission intensity from time step t to t + 1 in the research target area contribute to the total highway carbon emissions are respectively , , , ;

[0012]

[0013]

[0014]

[0015]

[0016] Among them: is the total highway mileage in the research target area at time t + 1, is the total mileage of expressways in the research target area at time t, is the traffic volume of expressways in the research target area at time t + 1, is the traffic volume of expressways in the research target area at time t, is the total energy consumption per vehicle of expressways in the research target area at time t + 1, is the total energy consumption per vehicle of expressways in the research target area at time t, is the carbon emission intensity of expressways in the research target area at time t + 1, is the carbon emission intensity of expressways in the research target area at time t, is the logarithmic mean weight, , is the total carbon emission of expressways in the research target area at time t + 1, is the total carbon emission of expressways in the research target area at time t.

[0017] Furthermore, the first-level variables of the total mileage of expressways in the research target area include the bridge-tunnel ratio and the number of lanes; the first-level variables of the traffic volume of expressways in the research target area include the regional population density and the level of regional economic activities; the first-level variables of the total energy consumption per vehicle of expressways in the research target area include the vehicle type distribution and the temperature; the first-level variables of the carbon emission intensity of expressways in the research target area include the energy structure and vehicle technology.

[0018] Furthermore, the secondary variables include the regional population quantity and the regional GDP.

[0019] Furthermore, the Metropolis-Hastings algorithm is used to generate samples of the posterior distribution of the latent variables.

[0020] Furthermore, the confidence level is 95% confidence level.

[0021] The present invention also provides a computer terminal device, including:

[0022] one or more processors;

[0023] a memory; and

[0024] one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by

[0025] The one or more processors execute the program, and when the program is executed by the processor, the above-mentioned method for analyzing highway carbon emissions based on a dynamic Bayesian network is implemented.

[0026] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for analyzing highway carbon emissions based on a dynamic Bayesian network are implemented.

[0027] The beneficial effects of the present invention are as follows: The method for analyzing highway carbon emissions based on a dynamic Bayesian network, the storage medium and the computer terminal device of the present invention take the total highway mileage, traffic volume, total energy consumption per vehicle, and carbon emission intensity in the research target area as hidden variables, and screen their direct influencing factors as primary variables, and further screen their interval influencing factors as secondary variables to establish a dynamic Bayesian network model to determine the distribution law of the hidden variables, and then simulate the states of the hidden variables and the total carbon emissions at different times, and calculate the contribution probability distribution of each hidden variable to carbon emissions in combination with the LMDI decomposition method, so as to realize the dynamic analysis of highway carbon emission factors. The present invention can more accurately understand the influence degree of each hidden variable (key factor) on the carbon emissions of highways in different periods in the research target area, and is helpful for formulating dynamic emission reduction strategies more scientifically and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flowchart of the present invention;

[0029] Figure 2 is the total mileage of highways in Chengdu ( ); the contribution percentage CS to the change in total carbon emissions and the interval graph of the 95% probability density;

[0030] Figure 3 is the traffic volume of highways in Chengdu ( ) the contribution percentage to the change in total carbon emissions and the interval graph of the 95% probability density;

[0031] Figure 4 is the total energy consumption per vehicle of highways in Chengdu ( ) the contribution percentage to the change in total carbon emissions and the interval graph of the 95% probability density;

[0032] Figure 5 is the carbon emission intensity of highways in Chengdu ( ) the contribution percentage to the change in total carbon emissions and the interval graph of the 95% probability density. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0034] As Figure 1 shown, a method for analyzing highway carbon emissions based on a dynamic Bayesian network of the present invention includes the following steps:

[0035] S1. Establish a Kaya identity formula according to the key factors affecting the total highway carbon emissions in the research target area. The key factors include the total mileage of the highways in the research target area , traffic volume , total energy consumption per vehicle and carbon emission intensity . The carbon emission intensity is the carbon emissions generated per unit of energy consumption. The Kaya identity formula is:

[0036] , where is the total highway carbon emissions in the research target area;

[0037] S2. Using the key factors as hidden variables, using the primary variables and secondary variables that affect each hidden variable as observed variables, and establishing a dynamic Bayesian network model according to the dependence relationship and conditional probability distribution between variables. Among them, the primary variables directly affect the corresponding hidden variables, and the secondary variables indirectly affect the corresponding hidden variables by affecting the primary variables;

[0038] S3. Use the dynamic Bayesian network model to simulate the hidden variable states and generate posterior distribution samples of the hidden variables at different time points;

[0039] S4. Use the LMDI (Logarithmic Mean Divisia Index) decomposition method to calculate the contribution degree of each hidden variable from time step t to t + 1 in each hidden variable sample to the change of the total highway carbon emissions in the research target area, and count the contribution degrees of all samples to generate a probability distribution of the contribution degrees, and then obtain an interval graph of the percentage of the contribution degree of each hidden variable to the change of the total highway carbon emissions in the research target area under the set confidence level.

[0040] In the present invention, the calculation method of the traffic volume is the ratio of the number of motor vehicles N to the total mileage of the highway , that is .

[0041] In the embodiment of the present invention, considering the data collectability, the primary variables of the total mileage of the highways in the research target area include the bridge-tunnel ratio ( ) and the number of lanes ( ); The traffic flow of expressways within the research target area The first-level variables of include regional population density (Pop) and regional economic activity level (Eco); The total energy consumption per vehicle on the expressways within the research target area The first-level variables of include vehicle type distribution ( ), temperature (Tmp), etc. The carbon emission intensity of expressways within the research target area The first-level variables of include energy structure ( ), vehicle technology (Vte). The bridge-tunnel ratio ( ), number of lanes ( ) data can be obtained from road construction plans and transportation departments respectively. Regional population density (Pop) data can be obtained from population censuses. Regional economic activity level (Eco) data can be obtained from economic statistics. Vehicle type distribution ( ) data can be obtained from traffic flow monitoring. Temperature (Tmp) data can be obtained from meteorological stations. Energy structure ( ) data can be obtained from energy departments. Vehicle technology (Vte) data can be obtained from vehicle manufacturers.

[0042] In some embodiments, only some of the above factors are also used as the first-level variables. Of course, other factors can also be added according to needs.

[0043] Generally, macro factors are screened as second-level variables, such as: policy factors (such as environmental protection policies, traffic planning policies, etc., affecting first-level variables such as energy structure ( ), number of lanes ( )), regional population (Pop), regional GDP (Gdp), etc. Regional population not only affects traffic flow ( ), but may also affect the total highway mileage ( ) to meet traffic demand, and its data can be obtained from population censuses. Regional economic level affects traffic flow ( ) and vehicle technology (Vte). Economic development may lead to more freight demand and more advanced vehicle technology, and its data can be obtained from economic statistics.

[0044] Additive decomposition is preferably used in the present invention. Specifically, the total mileage of expressways, traffic flow , total energy consumption per vehicle , carbon emission intensity The contribution degrees of the total carbon emissions of expressways within the research target area from adjacent time steps t to t + 1 are respectively , , , ;

[0045]

[0046]

[0047]

[0048]

[0049] Wherein: is the total mileage of expressways in the research target area at time t + 1, is the total mileage of expressways in the research target area at time t, is the traffic volume of expressways in the research target area at time t + 1, is the traffic volume of expressways in the research target area at time t, is the total energy consumption per vehicle on expressways in the research target area at time t + 1, is the total energy consumption per vehicle on expressways in the research target area at time t, is the carbon emission intensity of expressways in the research target area at time t + 1, is the carbon emission intensity of expressways in the research target area at time t, is the logarithmic mean weight, , is the total carbon emission of expressways in the research target area at time t + 1, is the total carbon emission of expressways in the research target area at time t.

[0050] The present invention takes the total mileage of expressways, traffic volume, total energy consumption per vehicle, and carbon emission intensity in the research target area as latent variables, screens their direct influencing factors as first-level variables, and further screens their spacing influencing factors as second-level variables to establish a dynamic Bayesian network model to determine the distribution law of latent variables. Then, it simulates the states of latent variables and total carbon emissions at different times, and combines the LMDI decomposition method to calculate the contribution probability distribution of each latent variable to carbon emissions, so as to realize the dynamic analysis of expressway carbon emission factors. The present invention can more accurately understand the influence degree of each latent variable (key factor) on the carbon emissions of expressways in the research target area at different times, and helps to formulate more scientific and accurate dynamic emission reduction strategies.

[0051] The establishment process of the dynamic Bayesian network model is specifically as follows:

[0052] (1) Structure learning

[0053] Initialize the network structure, create an empty Bayesian network, add latent variables, primary variables, and secondary variables, and add preliminary edges based on expertise and possible relationships between variables;

[0054] Add observation data and input the collected data set into the Bayesian network. The data should include the observation values of all variables, such as total highway mileage, traffic volume, energy consumption per vehicle, carbon emission intensity, bridge-tunnel ratio, number of lanes, regional population, GDP, etc., to ensure the integrity and accuracy of the data and handle missing values and outliers;

[0055] Scoring and optimization: take the preliminary network structure created in the previous step as the current structure, perform local operations on the current network structure, such as adding an edge, deleting an edge, or reversing the direction of an edge, and calculate the BIC value for the network structure after each operation. The smaller the BIC value, the better the model. Select the operation with the smallest BIC value, update the current structure, and repeat the above process until no operation can further reduce the BIC value, that is, reach the local optimal solution. The formula of Bayesian Information Criterion (BIC) is:

[0056]

[0057] Where: n is the number of data samples; k is the number of parameters of the model; L is the maximum likelihood estimate of the model, which is obtained by calling the statsmodels library function.

[0058] Model validation and adjustment: randomly divide the data set into K subsets of similar size (usually K is 5 or 10). For each subset, use it as a test set and the remaining K-1 subsets as training sets. Train the Bayesian network model on the training set and validate it on the test set. Record the performance indicators of the model (such as prediction accuracy). Repeat the above process K times, selecting a different subset as the test set each time. Finally, average the results of K validations to obtain a comprehensive performance evaluation of the model. According to the results of cross-validation, adjust the model structure. If the model performs poorly during the validation process, the structure can be manually adjusted, such as adding or deleting certain edges to better conform to the actual causal relationship.

[0059] Scoring and optimization as well as model verification and adjustment are repeated until the comprehensive performance evaluation of the model achieves satisfactory results, which indicates that the structure of the Bayesian network can better reflect the dependency and causal relationship between variables, thus providing a reliable model basis for subsequent carbon emission analysis.

[0060] (2) Parameter Learning

[0061] Initialize the parameters, and use the existing historical data to assign initial values to the conditional probability parameters in the Bayesian network. These initial values will serve as the starting point for the EM algorithm. For each variable, estimate its initial conditional probability based on the observed frequencies in the historical data. For example, for the relationship between the total highway mileage ( ), the bridge-tunnel ratio ( ), the number of lanes ( ), etc., which are first-level variables, initialize the corresponding conditional probability parameter X according to the distribution of the total highway mileage under different bridge-tunnel ratios and numbers of lanes in the historical data;

[0062]

[0063] Among them, is the variable, is its parent node variable, count represents the number of occurrences in the data, is the conditional probability of the variable when the parent node takes a specific value ;

[0064] Execute the E-step (expectation step), and use the current parameter estimates to calculate the expected value Q of the latent variable. In the case of discrete variables, this involves calculating the expected number of occurrences of each latent variable in different states based on the existing conditional probabilities and observed data. For example, for the latent variable of carbon emission intensity ( ), combined with the conditional probabilities of its parent node variables (such as the energy structure , the first-level variables such as vehicle technology Vtech, etc.) and the observed data, calculate the expected value of in different possible states, and these expected values will serve as the basis for subsequent parameter updates.

[0065]

[0066] Among them, represents the posterior probability distribution of the latent variable in the t-th iteration, is the observed data, is the parameter estimate of the current iteration, represents the conditional posterior probability of the latent variable under the current parameter , calculated through Bayesian quantification;

[0067] Perform M steps (maximization steps) to re-estimate the conditional probability parameters in the Bayesian network based on the expected values calculated in the E step. For discrete variables, this is typically achieved by maximizing the likelihood function. Specifically, based on the expected values of the latent variables and the observed data obtained in the E step, recalculate the conditional probability of each variable to maximize the likelihood of the data under the current parameters. For example, update the conditional probability parameters between first-level variables such as the total highway mileage ( ) and the bridge and tunnel ratio ( ), the number of lanes ( ) etc., to better reflect the dependencies in the data.

[0068]

[0069] Among them, is the updated parameter estimate value;

[0070] Calculate the parameter change amount and compare the difference between the parameters estimated in the current iteration round and the parameters in the previous round. If the parameter change amount is less than the preset convergence threshold, it is considered that the EM algorithm has converged and the iteration can be stopped; otherwise, continue to the next step. The setting of the convergence threshold needs to be determined according to the specific problem and data characteristics. Usually, a smaller threshold means a more accurate estimate, but it may also increase the computational cost.

[0071]

[0072] Among them, represents the norm, usually the Euclidean norm;

[0073] Repeat the above steps until the convergence condition or the maximum number of iterations is reached. This cyclic process continuously alternates between the E step and the M step to optimize the parameters, gradually improving the model's fit to the data. Each iteration makes the parameters closer to the true values, thus continuously improving the conditional probability estimation in the Bayesian network.

[0074] Verify the model performance. Use methods such as cross-validation to evaluate the performance of the Bayesian network model learned by the EM algorithm. Divide the dataset into a training set and a validation set. Use the training set for parameter learning, and then evaluate indicators such as the prediction accuracy rate and likelihood value of the model on the validation set to ensure that the model has good generalization ability and stability. If the model performs poorly on the validation set, it may be necessary to adjust the model structure or re-initialize the parameters, and then execute the EM algorithm again for parameter learning. Save the final parameter estimation results. When the EM algorithm converges and the model performance verification passes, save the conditional probability parameters in the finally estimated Bayesian network to complete the establishment of the dynamic Bayesian network model.

[0075] Specifically, algorithms such as Expectation Maximization and Particle Filtering can be used for latent variable state simulation. In the embodiments of the present invention, MCMC simulation (Markov Chain Monte Carlo simulation) is adopted, and a sample sequence of the posterior distribution of the latent variable is generated through the Metropolis-Hastings / Gibbs algorithm. Initialize the conditional probability parameters in the Bayesian network according to the existing historical data. Set the initial state of the MCMC simulation, including the initial values of the latent variables. Select a distribution , for generating new candidate states.

[0076] During the iterative process, for each iteration step t, generate a candidate state, starting from the current state , and use to generate a new candidate state . Calculate the acceptance probability: Calculate the acceptance probability , and its formula is:

[0077]

[0078] where is the target distribution, is the proposal distribution.

[0079] Accept or reject the candidate state, generate a uniformly distributed random number u ∼ Uniform(0,1). If u ≤ , then accept the candidate state , and set . Otherwise, reject the candidate state , and set . In this way, the target distribution is clarified.

[0080] After obtaining the target distribution, each sampling is based on the conditional probability relationship of the Bayesian network:

[0081]

[0082] where represents the set of parent nodes of node (such as "Regional GDP → Vehicle ownership → "). Set the burn-in period to K iterations to eliminate the influence of the initial value, and then perform K effective samplings. In the embodiments of the present invention, the number of iterations is 1000, that is, K = 1000, to generate a set of latent variable samples for year T and year T + 1:

[0083]

[0084] This set of latent variable samples satisfies the Bayesian network constructed in the previous step.

[0085] Specifically, based on the additive form of the LMDI decomposition, the carbon emission change , , is the total highway carbon emissions in the research target area at time T + 1, , is the total highway carbon emissions in the research target area at time T, and is decomposed into the sum of the contributions of each factor: , where the contribution of each factor for all samples is calculated as:

[0086] The total highway mileage of all samples contribution

[0087]

[0088] Among them, is the k-th value of , is the total highway carbon emissions in the research target area at time t, , is the total highway carbon emissions in the research target area at time 0; is the total mileage the k-th value at time T + 1; is the total mileage the k-th value at time T; The traffic volume of all samples contribution , the total energy consumption per vehicle contribution , and the carbon emission intensity contribution are calculated in the same way, and no specific examples are given here. In the numerator, is the carbon emission change of a single simulation, and the denominator is the logarithmic difference.

[0089] This formula ensures no residual decomposition through the Weighted Logarithmic Mean Divisia Index (LMDI-I).

[0090] When generating the contribution probability distribution, for each simulation (k = 1 to 1000), calculate the proportion of the contribution of each factor:

[0091]

[0092] Among them, is the change in the total mileage the k-th value, is the change in the total carbon emissions C at the k-th value.

[0093] The same applies to the remaining factors. Summarize the results of 1000 simulations to generate the probability distribution of the contribution of each factor.

[0094] The confidence level can be set as needed, such as: 99%, 95%, 90%, etc. In the present invention, a 95% confidence level is preferably selected, which can provide a relatively high certainty and balance the precision and the interval width.

[0095] The interval graph of the contribution percentage of each latent variable to the change in the total carbon emissions of the expressways in the research target area at a 95% confidence level is specifically as follows:

[0096] Use Kernel Density Estimation (KDE) to fit the distribution curve and calculate the 95% confidence interval (such as " The contribution is 15% - 30%"). Select a Gaussian kernel that is symmetric, smooth, and suitable for most continuous distributions as the kernel function

[0097]

[0098] Among them, K(u): kernel function, u: input variable, representing the distance between the data point and the center of the kernel function,

[0099] The bandwidth h is the core parameter of KDE, which controls the smoothness of the density curve.

[0100]

[0101] Among them, : sample standard deviation, : sample interquartile range, : sample size. Select the optimal h by minimizing the integrated squared error (MISE). For each position x, calculate its density estimate value :

[0102]

[0103] Generate 1000 equally spaced grid points based on the sample minimum and maximum values. Calculate the density point by point: for each grid point x, accumulate the contributions of all sample points of the kernel function to integrate the density curve and find the interval covering 95% of the probability density.

[0104] Such as Figures 2 to 5 , which is the percentage of the contribution of the total mileage of the expressways in Chengdu City ( ), traffic flow ( ), total energy consumption per vehicle ( ), and carbon emission intensity ( ) to the change in the total carbon emissions and the interval graph of the 95% probability density obtained by the method of the present invention. It can be seen from the graph that the confidence interval is overall located in the positive interval and has a wide range, indicating that highway growth needs to be preferentially controlled. The confidence interval spans positive and negative values, indicating a large uncertainty in the growth of traffic flow. The increase in the number of retained vehicles after 2020 and the emergence of new energy vehicles after 2022 have caused the contribution of this variable to show inflection points multiple times.

[0105] The present invention also provides a computer terminal device, including:

[0106] One or more processors;

[0107] A memory; and

[0108] One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by

[0109] The one or more processors. When the program is executed by the processor, it implements the above-mentioned method for analyzing highway carbon emissions based on a dynamic Bayesian network.

[0110] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the above-mentioned method for analyzing highway carbon emissions based on a dynamic Bayesian network.

Claims

1. A method for analyzing highway carbon emissions based on a dynamic Bayesian network, characterized in that, Including the following steps: Establish a Kaya identity formula based on the key factors affecting the total highway carbon emissions in the research target area, where the key factors include the total mileage of highways in the research target area , traffic volume , total energy consumption per vehicle and carbon emission intensity , where the carbon emission intensity is the carbon emissions generated per unit of energy consumption, and the Kaya identity formula is: , where is the total highway carbon emissions in the research target area; Taking the key factors as latent variables, taking the first-level variables and second-level variables that affect each latent variable as observed variables, and establishing a dynamic Bayesian network model according to the dependency relationships and conditional probability distributions among the variables, wherein the first-level variables directly affect the corresponding latent variables, and the second-level variables indirectly affect the corresponding latent variables by affecting the first-level variables; Using the dynamic Bayesian network model to simulate the states of the latent variables and generating posterior distribution samples of the latent variables at different time points; Using the LMDI decomposition method to calculate the contribution of each latent variable in each latent variable sample from adjacent time steps t to t + 1 to the change in the total highway carbon emissions in the research target area, and statistically analyzing the contributions of all samples to generate a probability distribution of the contributions of each latent variable, and then obtaining an interval graph of the percentage of the contribution of each latent variable to the change in the total highway carbon emissions in the research target area under a set confidence level.

2. The method for analyzing highway carbon emissions based on a dynamic Bayesian network according to claim 1, wherein The total mileage of highways within the research target area , traffic volume , total energy consumption per vehicle , carbon emission intensity The contribution degrees of the total carbon emissions of highways within the research target area from adjacent time steps t to t + 1 are respectively , , , ; Wherein: is the total mileage of expressways in the research target area at time t + 1, is the total mileage of expressways in the research target area at time t, is the traffic flow of expressways in the research target area at time t + 1, is the traffic flow of expressways in the research target area at time t, is the total energy consumption per vehicle of expressways in the research target area at time t + 1, is the total energy consumption per vehicle of expressways in the research target area at time t, is the carbon emission intensity of expressways in the research target area at time t + 1, is the carbon emission intensity of expressways in the research target area at time t, is the logarithmic mean weight, , is the total carbon emissions of expressways in the research target area at time t + 1, is the total carbon emissions of expressways in the research target area at time t.

3. The method for analyzing highway carbon emissions based on a dynamic Bayesian network according to claim 1, wherein Total mileage of expressways within the research target area The first-level variables of include the bridge-tunnel ratio and the number of lanes; Traffic volume of expressways within the research target area The first-level variables of include the regional population density and the level of regional economic activities; Total energy consumption per vehicle on expressways within the research target area The first-level variables of include the vehicle type distribution and the temperature; Carbon emission intensity of expressways within the research target area The first-level variables of include the energy structure and vehicle technology.

4. The method for analyzing highway carbon emissions based on a dynamic Bayesian network according to claim 1, wherein, The second-level variables include the regional population quantity and the regional GDP.

5. The method for analyzing highway carbon emissions based on a dynamic Bayesian network according to claim 1, wherein Using the Metropolis-Hastings algorithm to generate the posterior distribution samples of the latent variables.

6. The method for analyzing highway carbon emissions based on a dynamic Bayesian network according to claim 1, characterized in that The confidence level is 95% confidence level.

7. A computer terminal device, characterized in that, Including: One or more processors; A memory; And One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the program is executed by the processor, it implements the method for analyzing highway carbon emissions based on a dynamic Bayesian network according to any one of claims 1-6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for analyzing highway carbon emissions based on a dynamic Bayesian network according to any one of claims 1-6.

Images