Road maintenance engineering carbon emission prediction method based on fuzzy polynomial neural network

Abstract

The invention provides a road maintenance project carbon emission prediction method based on a fuzzy polynomial neural network, and the method comprises the steps: collecting and calculating the carbon emission data of a plurality of completed road maintenance projects, calculating the carbon emission through employing a method combining a quota method and an emission coefficient method, and constructing a basic data set; based on the basic data set, identifying an emission heat source, matching engineering quantity parameters, and primarily screening and selecting key indexes to construct input data of a prediction model; performing nonlinear modeling and generalization ability mining through a fuzzy polynomial neural network model, and constructing and training a prediction model based on the input data; and predicting the carbon emission of the road maintenance project through the prediction model. According to the highway maintenance engineering carbon emission prediction method based on the fuzzy polynomial neural network, rapid and accurate prediction of carbon emission can be realized only based on a small amount of key information in the early stage of the highway maintenance engineering.

Description

Technical Field

[0001] The present invention belongs to the field of carbon emission, and in particular relates to a method for predicting carbon emission of highway maintenance projects based on fuzzy polynomial neural network. Background Art

[0002] At present, carbon emissions from highway maintenance are mostly quantified based on construction drawings and construction stages, but in the early stages of highway maintenance projects, there is a lack of accurate and specific bills of quantities, making it impossible to accurately calculate the carbon emissions of the projects. How to make a reasonable prediction of the carbon emissions of the projects based on a small amount of key information on highway maintenance projects is of great significance for the optimization and selection of maintenance plans, the control of carbon emission limits, and the formulation of carbon reduction plans.

[0003] Therefore, the present invention intends to propose a carbon emission prediction model for highway maintenance projects based on fuzzy polynomial neural networks, which can realize a rapid estimation of carbon emissions in the early stage of the project by using only indicators such as the number of kilometers of maintenance lanes, asphalt mixture usage, cement-stabilized gravel usage, and guardrail usage, thereby providing a basis for the comparison and selection of maintenance project plans and carbon reduction optimization. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks.

[0005] The technical solution of the present invention is as follows:

[0006] A method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks, the method comprising:

[0007] Collect and calculate carbon emission data of multiple completed highway maintenance projects, use a combination of quota method and emission coefficient method to calculate carbon emissions, and build a basic data set;

[0008] Identify emission heat sources, match engineering quantity parameters, and preliminarily screen and select key indicators based on the basic data set to construct input data for the prediction model;

[0009] Performing nonlinear modeling and generalization capability mining through a fuzzy polynomial neural network model, and building and training a prediction model based on the input data;

[0010] The prediction model is used to predict the carbon emissions of highway maintenance projects.

[0011] Furthermore, the carbon emissions are calculated as follows:

[0012]

[0013] Where: E is the carbon emission during highway maintenance and construction, i is the type of road construction material, Q iis the total consumption of the i-th road construction material, F i is the carbon emission factor of the production process of the i-th road construction material; D i is the average transportation distance of type i road construction material, f i is the carbon emission factor per unit weight of the i-th road construction material per unit transportation distance, j is the type of construction machinery, k is the energy type, T j is the total working time spent using the jth type of construction equipment, is the consumption of the kth energy by the jth construction machine in unit working time, f k is the carbon emission factor produced by the kth energy source.

[0014] Furthermore, the input data for building a prediction model based on the basic data set to identify the emission heat source, match the engineering quantity parameters, preliminarily screen and select key indicators are specifically:

[0015] Based on the basic data set, calculate and rank the carbon emissions of each road construction material to determine the main emission heat sources;

[0016] Based on the emission heat source identification results, track construction activities and match heat source materials to corresponding engineering quantity parameters;

[0017] Summarize the engineering quantity parameters corresponding to all emission heat sources in the basic data set, and preliminarily form a list of indicators that have a significant impact on carbon emissions and their frequency of occurrence;

[0018] According to the frequency of occurrence of indicators and the consumption or emission factors of various materials in maintenance projects, key indicators are finally selected as input data for the prediction model.

[0019] Furthermore, the key indicators are specifically:

[0020] Earth shoulder hardening, kerbstones, lime / lime fly ash stabilization, cement stabilization, asphalt concrete, asphalt stabilized macadam, cement concrete, markings, signboards, corrugated beam guardrails, concrete guardrails, lane kilometres and construction costs.

[0021] Furthermore, the nonlinear modeling and generalization capability mining are performed by using the fuzzy polynomial neural network model, and the prediction model is constructed and trained based on the input data specifically as follows:

[0022] Use principal component analysis to reduce the dimensionality of input data, extract principal components, and optimize data dimensions;

[0023] Perform fuzzy C-means clustering on the reduced-dimensional data to generate different fuzzy clusters;

[0024] Prediction is performed based on the fuzzy polynomial neural network model. By defining fuzzy rules, the input features are nonlinearly mapped through fuzzy sets to train the model;

[0025] Use ridge regression to regularize model parameters to prevent overfitting and improve the generalization ability of the model;

[0026] Generative adversarial networks are used for data enhancement to generate fake carbon emission data, which are then added to the training dataset, and multiple rounds of adversarial training are used to optimize the model.

[0027] Furthermore, the dimensionality reduction process of the input data by principal component analysis is specifically as follows:

[0028] Decentralize the data set;

[0029] Calculate the covariance matrix;

[0030] Find eigenvalues ​​and eigenvectors through eigenvalue decomposition;

[0031] Select the largest n eigenvectors to construct the dimension reduction matrix;

[0032] Multiply the original data by the dimensionality reduction matrix to obtain the reduced dimensionality dataset.

[0033] Furthermore, the fuzzy C-means clustering is performed on the reduced-dimensional data to generate different fuzzy clusters:

[0034] Set the initial number of clusters and fuzzy parameters, and use the Euclidean distance to calculate the membership of each data point to the cluster center;

[0035] The cluster centers are updated according to the membership function, and multiple rounds of iterative optimization are performed until the clusters converge.

[0036] Furthermore, prediction is performed based on the fuzzy polynomial neural network model. By defining fuzzy rules, the input features are nonlinearly mapped through fuzzy sets. The specific training model is:

[0037] Construct fuzzy rules and select input feature sets;

[0038] Perform fuzzy processing on each input feature and establish a fuzzy rule base;

[0039] Model the input data using a polynomial expression,

[0040] The mean square error is defined as the loss function, and the weights and biases are optimized through the back-propagation algorithm.

[0041] Furthermore, the generative adversarial network is used for data enhancement to generate fake carbon emission data and add it to the training data set. The multi-round adversarial training optimization model is as follows:

[0042] Use deep convolutional neural networks to build generators and discriminators, where the generator is used to generate fake carbon emission data and the discriminator is used to distinguish between real and fake data;

[0043] The real carbon emission data and fake data are mixed and input into the fuzzy polynomial neural network for training. After multiple rounds of adversarial training, the model is optimized to ensure that the generated data can improve the prediction accuracy.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] The carbon emission prediction model proposed in the present invention can quickly predict carbon emissions in the early stage of highway maintenance projects based on only a small amount of key information (such as maintenance mileage, asphalt mixture usage, etc.), without the need for a detailed bill of quantities, greatly reducing dependence on data integrity.

[0046] The present invention constructs a fuzzy polynomial neural network, can use fuzzy logic to process the ambiguity and complexity in highway maintenance data, and improves the prediction accuracy by combining polynomial expansion.

[0047] The principal component analysis and fuzzy C-means algorithm used in the model of the present invention can effectively process high-dimensional data and fuzzy data, enhancing the adaptability and robustness of the model. In addition, through ridge regression optimization and generative adversarial network data enhancement, the generalization ability and prediction accuracy of the model are further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the description and claims, serve to illustrate the embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0049] Figure 1 The carbon emission calculation process of highway maintenance engineering of the present invention is shown;

[0050] Figure 2 A schematic diagram of the process of the present invention is shown. DETAILED DESCRIPTION

[0051] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0052] The embodiment of the present invention provides a method for predicting carbon emissions of highway maintenance projects based on a fuzzy polynomial neural network, the method comprising:

[0053] Step 1: Establish a basic data set to collect and calculate carbon emission data of multiple completed highway maintenance projects.

[0054] In the process of calculating carbon emissions from highway maintenance project construction activities, the "emission factor method" is relatively simple and convenient to use in the material production, machinery construction, and transportation stages. It is only necessary to obtain the material or energy consumption and combine it with the corresponding emission factor to calculate the final emissions.

[0055] The present invention adopts a method combining the quota method and the carbon emission factor for calculation. The quota method is based on the current "Highway Engineering Budget Quota" (JTG / T 3832-2018) and "Highway Engineering Machinery Unit Cost Quota" (JTG / T 3833-2018) to determine the material consumption of each link of the maintenance plan, the energy consumption of construction machinery, and the fuel consumption of transportation vehicles. Combined with the carbon emission factors of materials and energy, the carbon emissions of various materials and energy are calculated and summed to finally obtain the total carbon emissions of the unit project. The carbon emission calculation process is as follows: Figure 1 shown.

[0056] The carbon emissions E during the road maintenance construction process is the sum of the carbon emissions in the material production stage, material transportation stage and on-site construction stage. The carbon emissions in each stage are calculated using a combination of the quota method and the emission coefficient method. The formula is as follows.

[0057]

[0058] Where: E is the carbon emission during highway maintenance and construction (tCO2e); i is the type of road construction material, i = 1, 2, 3, ..., n; Q i is the total consumption of the i-th road construction material (t); F i is the carbon emission factor of the production process of the i-th road construction material (tCO2e / t); D i is the average transportation distance of type i road construction material (km); f i is the carbon emission factor per unit weight of the i-th road construction material per unit transportation distance (tCO2e / (t·km)); j is the type of construction machinery, j=1,2,3,…,m; k is the type of energy, k=1,2,3,…,l; T j is the total working time consumed by using the jth type of construction equipment (h); is the consumption of the kth energy by the jth construction machine in unit working time (t); f k is the carbon emission factor produced by the kth energy source (tCO2e / (t·h).

[0059] Step 2: Screen key indicators of maintenance carbon emissions.

[0060] Identify emission heat sources. Based on the calculation results of the basic data set (20 maintenance projects), the carbon emissions generated by road construction materials in each maintenance project are ranked, and the top 15 road construction materials or those with cumulative emissions accounting for more than 90% of the total are found. According to statistics, they mainly include slaked lime, petroleum asphalt, cement, etc.

[0061] Matching engineering quantity parameters. Combined with the identification results of heat sources emitted by road construction materials, construction activities are tracked and engineering quantity parameters corresponding to heat source materials are matched, such as slaked lime corresponding to lime fly ash stabilized gravel in the engineering quantity list; petroleum asphalt corresponding to asphalt concrete in the engineering quantity list; cement corresponding to cement stabilized gravel in the engineering quantity list.

[0062] Preliminary screening of carbon emission impact indicators. Summarize the engineering quantity parameters corresponding to all emission heat sources in the basic data set, and preliminarily form a list of indicators that have a significant impact on carbon emissions and their frequency of occurrence.

[0063] Select carbon emission impact indicators. According to the frequency of occurrence of indicators, combined with the types of materials with large consumption or high emission factors in maintenance projects, the final key indicators are selected as input data for training the prediction model. These indicators include earth shoulder hardening (m3), curbstone (m), lime / lime fly ash stabilized type (m3), cement stabilized type (m3), asphalt concrete (m3), asphalt stabilized macadam (m3), cement concrete (m3), marking (m2), signboard (pieces), corrugated beam guardrail (m), concrete guardrail (m), lane kilometers (lane km), construction and installation costs (10,000 yuan) 13.

[0064] Table 1 List of carbon emission impact indicators (preliminary screening)

[0065]

[0066] Table 2 Highway maintenance project carbon emission data set (partial)

[0067]

[0068]

[0069] Step 3: Establish a maintenance carbon emission prediction model.

[0070] 1. Build a data preprocessing model, perform dimensionality reduction on carbon emission data through principal component analysis (PCA), reduce the dimensionality of high-dimensional carbon emission features, extract principal components, reduce redundant information, and optimize data dimensions.

[0071] 1.1 Select an appropriate number of principal components to ensure that the data after dimensionality reduction retains more than 95% of the total variance;

[0072] 1.2 Input the m-dimensional carbon emission dataset X that needs to be reduced to n-dimensional, and perform decentralization on X, that is, subtract the respective mean value from each feature.

[0073] 1.3 Calculate the covariance matrix

[0074] 1.4 Using the eigenvalue decomposition method to find the covariance matrix The eigenvalues ​​and eigenvectors of .

[0075] 1.5 Select eigenvalues ​​and use the largest n eigenvectors as row vectors to construct the matrix P.

[0076] 1.6 Multiply the original data by the constructed matrix to obtain the transformed reduced-dimensional matrix space and obtain the reduced-dimensional data set Y.

[0077] 2. Construct a clustering model based on the fuzzy C-means (FCM) algorithm to cluster the carbon emission data after dimensionality reduction and generate different fuzzy clusters to represent different carbon emission feature groups.

[0078] 2.1 Set the initial number of clusters and fuzzy parameters, and use the Euclidean distance to calculate the membership of each data point to the cluster center;

[0079] 2.2 Update the cluster center according to the membership function, and optimize through multiple rounds of iterations until the clusters converge.

[0080] The objective function and constraints of the fuzzy C-means algorithm are as follows:

[0081]

[0082] where u ij is the sample point x i With cluster center v j The membership degree, m is the fuzzy index (m>1), d ij is the sample point x i With cluster center v j The distance is Euclidean distance.

[0083] In order to minimize the objective function J, the Lagrange multiplier method is used on the objective function while satisfying the constraints to obtain the membership matrix U and cluster center v j .

[0084]

[0085] 3. Prediction is performed based on the fuzzy polynomial neural network (FPNN) model. By defining fuzzy rules, the input carbon emission characteristics are nonlinearly mapped through fuzzy sets to train the model.

[0086] 3.1 Construct fuzzy rules based on input carbon emission characteristics and use polynomial expressions to model the input data:

[0087] 3.2 Select the input carbon emission feature set, denoted as X = {x1, x2, ..., x n}, where x i Represents different categories of carbon emission factors, such as pavement engineering carbon emissions, traffic engineering carbon emissions, vehicle carbon emissions, etc.

[0088] 3.3 For each feature x i Perform fuzzy processing. Use membership function μ i (x i ) to map each input feature into a membership degree in a fuzzy set.

[0089] 3.4 Based on the fuzzified data, a fuzzy rule base is established based on each input feature. Through the fuzzy reasoning mechanism, the input fuzzy feature set is modeled through a polynomial function. The polynomial expression can be expressed as:

[0090]

[0091] Among them, a ij is the parameter to be optimized of the model, μ i (x i ) is the input feature x i The fuzzy membership value of , m is the highest degree of the polynomial.

[0092] 3.5 The weights and biases of the fuzzy polynomial neural network are optimized through the back-propagation algorithm, so that the model can accurately predict carbon emissions.

[0093] 4. Combined with ridge regression optimization, ridge regression is used to regularize the model parameters to prevent overfitting and improve the generalization ability of the model.

[0094] 5. Use generative adversarial networks (GANs) for data augmentation to generate fake carbon emission data and add them to the training dataset to enhance the robustness of the model.

[0095] 5.1 Use a deep convolutional neural network (DCNN) to build a generator and a discriminator, where the generator is used to generate fake carbon emission data and the discriminator is used to distinguish between real and fake data;

[0096] 5.2 The real carbon emission data and the fake data are mixed and input into the fuzzy polynomial neural network for training. After multiple rounds of adversarial training, the model is optimized to ensure that the generated data can improve the prediction accuracy.

[0097] Results of training performance of highway maintenance carbon emission prediction model

[0098] Compared with the ordinary linear regression prediction model:

[0099] ①Evaluation principle

[0100] The test data uses Mean MSE (mean of the mean square error) and STD MSE (standard deviation of the mean square error) to judge the prediction results. Mean MSE gives the overall prediction error level, and STD MSE gives the stability of the prediction error.

[0101] Mean MSE: The average of the mean square errors on multiple test sets. It indicates the average level of the overall prediction error of the model on different test sets. The smaller the Mean MSE, the better the overall prediction effect of the model.

[0102] formula:

[0103] Where N is the number of test sets, MSE i is the mean square error of the ith test set.

[0104] STD MSE: Standard deviation of the mean square error on multiple test sets. Indicates the degree of dispersion or fluctuation of the mean square error. The smaller the STD MSE, the more consistent the model's prediction error is on different test sets; the larger the STD MSE, the greater the fluctuation of the model's prediction error on different test sets.

[0105] formula:

[0106] Where N is the number of test sets, MSE i is the mean square error of the ith test set, and MeanMSE is the mean of the mean square errors.

[0107] ② Comparison of prediction results

[0108] The Mean MSE value of the ordinary linear regression result is about 0.25, and the STD MSE value is 0.044; the Mean MSE of the fuzzy polynomial neural network is 0.24, and the STD value is 0.029, which is better than the ordinary linear regression. This shows that the stability and accuracy of the prediction results of the fuzzy polynomial neural network are significantly improved, and the effect is obvious.

[0109] Example:

[0110] In order to verify the effectiveness of the carbon emission prediction model for highway maintenance projects, taking the S109 Jinlai Line (Outer Ring Line Jinlai Interchange-Anxing Road) maintenance project as an example, the carbon emission calculation method based on the emission coefficient method and the carbon emission prediction model were used to measure and predict the carbon emissions of its maintenance and construction activities.

[0111] The carbon emissions from material production, material transportation and on-site construction during the actual construction phase of the S109 Jinlai Line (Outer Ring Line Lailai Interchange-Anxing Road) maintenance project were calculated, and the results showed that the total carbon emissions of the project were 4205.12 tons. The carbon emissions prediction based on fuzzy polynomial neural network showed that the overall carbon emissions prediction value of the project was 3840.51 tons.

[0112] Table 3 Prediction results of S109 Jinlai Line

[0113]

[0114] Table 4 Comparison

[0115]

[0116] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks, characterized in that: The method comprises: Collect and calculate carbon emission data of multiple completed highway maintenance projects, use a combination of quota method and emission coefficient method to calculate carbon emissions, and build a basic data set; Identify emission heat sources, match engineering quantity parameters, and preliminarily screen and select key indicators based on the basic data set to construct input data for the prediction model; Performing nonlinear modeling and generalization capability mining through a fuzzy polynomial neural network model, and building and training a prediction model based on the input data; The prediction model is used to predict the carbon emissions of highway maintenance projects.

2. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks according to claim 1 is characterized in that: The calculation method of carbon emissions is: Where: E is the carbon emission during highway maintenance and construction, i is the type of road construction material, Q i is the total consumption of the i-th road construction material, F i is the carbon emission factor of the production process of the i-th road construction material; D i is the average transportation distance of type i road construction material, f i is the carbon emission factor per unit weight of the i-th road construction material per unit transportation distance, j is the type of construction machinery, k is the energy type, T j is the total working time spent using the jth type of construction equipment, is the consumption of the kth energy by the jth construction tool per unit working time, f k is the carbon emission factor produced by the kth energy source.

3. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks according to claim 1 is characterized in that: The input data for building a prediction model based on the basic data set to identify emission heat sources, match engineering quantity parameters, and preliminarily screen and select key indicators are as follows: Based on the basic data set, calculate and rank the carbon emissions of each road construction material to determine the main emission heat sources; Based on the emission heat source identification results, track construction activities and match heat source materials to corresponding engineering quantity parameters; Summarize the engineering quantity parameters corresponding to all emission heat sources in the basic data set, and preliminarily form a list of indicators that have a significant impact on carbon emissions and their frequency of occurrence; According to the frequency of occurrence of indicators and the consumption or emission factors of various materials in maintenance projects, key indicators are finally selected as input data for the prediction model.

4. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural network according to claim 3 is characterized in that: The key indicators are specifically: Earth shoulder hardening, kerbstones, lime / lime fly ash stabilization, cement stabilization, asphalt concrete, asphalt stabilized macadam, cement concrete, markings, signboards, corrugated beam guardrails, concrete guardrails, lane kilometres and construction costs.

5. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks according to claim 1 is characterized in that: The nonlinear modeling and generalization capability mining by using the fuzzy polynomial neural network model and the construction and training of the prediction model based on the input data are specifically as follows: Use principal component analysis to reduce the dimensionality of input data, extract principal components, and optimize data dimensions; Perform fuzzy C-means clustering on the reduced-dimensional data to generate different fuzzy clusters; Prediction is performed based on the fuzzy polynomial neural network model. By defining fuzzy rules, the input features are nonlinearly mapped through fuzzy sets to train the model; Use ridge regression to regularize model parameters to prevent overfitting and improve the generalization ability of the model; Generative adversarial networks are used for data enhancement to generate fake carbon emission data, which are then added to the training dataset, and multiple rounds of adversarial training are used to optimize the model.

6. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks according to claim 5 is characterized in that: The dimensionality reduction process of the input data by principal component analysis is specifically as follows: Decentralize the data set; Calculate the covariance matrix; Find eigenvalues ​​and eigenvectors through eigenvalue decomposition; Select the largest n eigenvectors to construct the dimension reduction matrix; Multiply the original data by the dimensionality reduction matrix to obtain the reduced dimensionality dataset.

7. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks according to claim 5 is characterized in that: The fuzzy C-means clustering is performed on the data after dimension reduction to generate different fuzzy clusters as follows: Set the initial number of clusters and fuzzy parameters, and use the Euclidean distance to calculate the membership of each data point to the cluster center; The cluster centers are updated according to the membership function, and multiple rounds of iterative optimization are performed until the clusters converge.

8. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks according to claim 5 is characterized in that: Prediction is performed based on the fuzzy polynomial neural network model. By defining fuzzy rules, the input features are nonlinearly mapped through fuzzy sets. The specific training model is as follows: Construct fuzzy rules and select input feature sets; Perform fuzzy processing on each input feature and establish a fuzzy rule base; Model the input data using a polynomial expression, The mean square error is defined as the loss function, and the weights and biases are optimized through the back-propagation algorithm.

9. The method for predicting carbon emissions from highway maintenance projects based on fuzzy polynomial neural networks according to claim 5 is characterized in that: Generative adversarial networks are used for data enhancement to generate fake carbon emission data and add them to the training data set. The multi-round adversarial training optimization model is as follows: Use deep convolutional neural networks to build generators and discriminators, where the generator is used to generate fake carbon emission data and the discriminator is used to distinguish between real and fake data; The real carbon emission data and fake data are mixed and input into the fuzzy polynomial neural network for training. After multiple rounds of adversarial training, the model is optimized to ensure that the generated data can improve the prediction accuracy.

Images