Flat glass industry carbon emission prediction method based on electric-carbon correlation

Abstract

The present application relates to a kind of flat glass industry carbon emission prediction method based on electric carbon correlation, comprising the following steps: step S1: obtaining flat glass industry carbon emission correlation quantity, and data preprocessing is carried out;Step S2: based on SVR regression prediction, obtain the correlation between electricity and carbon emission;Step S3: according to the carbon emission correlation quantity after preprocessing, based on SVR regression prediction, establish the correlation between electricity and correlation quantity, and the correlation between correlation quantity and carbon emission except electricity;Step S4: based on the correlation obtained in step S2, build electric carbon model;Step S5: input the data to be measured into electric carbon model, obtain predicted carbon emission.The present application establishes the correlation model between electricity consumption and carbon emission by SVR and historical data, realizes the purpose of predicting carbon emission by electricity consumption data in flat glass industry.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data fitting and electric carbon model establishment, and relates to a flat glass industry carbon emission prediction method based on an electric carbon correlation relationship. BACKGROUND

[0002] With the continuous acceleration of urbanization in China, the flat glass industry closely related to the construction and real estate industry has also developed rapidly. At present, the number and output of flat glass enterprises in China rank first in the world. The flat glass industry is a typical high energy consumption and high emission industry, and under the wave of the current low-carbon economy, the development of the flat glass industry is facing severe challenges.

[0003] Support vector regression (SVR) is a machine learning method based on the idea of support vector machine, which uses statistical learning theory for regression calculation. This method is suitable for limited sample research, can obtain global optimal solution in theory, and the complexity of the calculation process is independent of the sample dimension, which can achieve optimal effect in function approximation, regression prediction, etc., and can effectively solve the problem of small data samples in the flat glass industry.

[0004] In order to achieve the goal of "carbon peak and carbon neutralization" and realize sustainable economic development, carbon emission prediction and modeling technology will become a hot research topic. At present, there are few related researches on industrial carbon emission prediction, and the establishment of electric carbon model in the flat glass industry not only promotes industrial development, but also has great economic and environmental protection significance. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a flat glass industry carbon emission prediction method based on an electric carbon correlation relationship, which establishes an electricity consumption and carbon emission correlation relationship model through SVR and historical data to achieve the purpose of predicting carbon emissions through electricity consumption data in the flat glass industry.

[0006] To achieve the above purpose, the present application adopts the following technical solutions:

[0007] A flat glass industry carbon emission prediction method based on an electric carbon correlation relationship, comprising the following steps:

[0008] Step S1: Obtain the flat glass industry carbon emission correlation quantity and perform data preprocessing;

[0009] Step S2: Obtain the electric carbon emission correlation relationship based on SVR regression prediction;

[0010] Step S3: Establish the electric correlation quantity and the correlation quantity to the electric carbon emission correlation relationship based on SVR regression prediction according to the preprocessed carbon emission correlation quantity;

[0011] Step S4: based on the correlation relationship obtained in step S2, an electric-carbon model is constructed;

[0012] Step S5: input the to-be-tested data into the electric-carbon model to obtain the predicted carbon emission.

[0013] Further, the step S1 specifically comprises:

[0014] Step S11: selecting the flat glass output as the flat glass industry carbon emission correlation quantity;

[0015] Step S12: taking the unit flat glass output comprehensive power consumption and the unit flat glass output comprehensive energy consumption as the standards to preprocess the data, calculating the unit flat glass output comprehensive power consumption and the unit flat glass output comprehensive energy consumption, and the calculation formula is as follows:

[0016]

[0017] The value range of the unit flat glass output comprehensive power consumption and the unit flat glass output comprehensive energy consumption is determined, and the data not within the value range is taken as singular data, which is excluded when the correlation relationship model is established.

[0018] Further, the step S2 specifically comprises:

[0019] Step S21: obtaining N groups of monthly electricity consumption, flat glass output and carbon emission data of a certain flat glass enterprise, and excluding singular data according to the value range of the unit flat glass output comprehensive power consumption and the unit flat glass output comprehensive energy consumption to form a sample set containing n groups of data;

[0020] Step S22: selecting m samples in the sample set to form a training sample set, and the remaining n-m samples form a test sample set; selecting a kernel function to process the training sample set, constructing an SVR model; setting the initial value of ε, and applying grid search and cross-validation method to determine the values of parameters C and γ, and finally obtaining the optimal parameters ε ∗ , C ∗ and γ ∗ ;

[0021] Step S23: simulating the data of the training sample set to obtain the optimal solution of the model and the regression function f(x), substituting all the data of the training sample set and the test sample set into the function to output the fitting values, and performing linear regression on the fitting results and the true values, and judging the learning and generalization ability of the model according to the correlation coefficient R 2 ; if the learning and generalization ability of the model is poor, return to step S22 until the optimal solution is obtained, and the electric-carbon emission correlation relationship is obtained.

[0022] Further, the step S3 is specifically:

[0023] Step S31: establish the power consumption to the correlation quantity correlation according to steps S22 to S23;

[0024] Step S32: establish the flat glass production to the de-electricity carbon emission correlation according to steps S22 to S23.

[0025] Further, the step S4 is specifically:

[0026] The electricity to correlation quantity function is substituted into the flat glass production to de-electricity carbon emission function, and the power consumption to de-electricity carbon emission correlation established by the flat glass production is obtained;

[0027] The electricity to carbon emission correlation is added to the power consumption to de-electricity carbon emission correlation, and the power consumption to carbon emission correlation is obtained, that is, the model to be solved.

[0028] Compared with the prior art, the present application has the following beneficial effects:

[0029] 1. The present application establishes a power consumption and carbon emission correlation model by SVR and historical data, realizes the purpose of predicting carbon emission of the flat glass industry through power consumption data, and realizes the purpose of predicting carbon emission of the flat glass industry through power consumption data.

[0030] 2. The flat glass industry electricity-carbon correlation modeling method based on SVR can obtain an electricity-carbon emission mathematical model, and simultaneously solve the problems of carbon emission prediction and calculation of the flat glass industry, and has an important role in carbon emission reduction of the flat glass industry. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is a method flowchart of the present application;

[0032] Figure 2 is a flat glass industry electricity-carbon correlation model based on SVR in an embodiment of the present application. DETAILED DESCRIPTION

[0033] The present application will be further described below in combination with the drawings and embodiments.

[0034] Please refer to Figure 1 The present application provides a flat glass industry carbon emission prediction method based on electricity-carbon correlation, comprising the following steps:

[0035] Step S1: obtaining flat glass industry carbon emission correlation quantity and performing data preprocessing;

[0036] Step S11: selecting flat glass production as the flat glass industry carbon emission correlation quantity;

[0037] Step S12: Select a typical enterprise in the flat glass industry and compile its carbon verification report to obtain historical data such as the enterprise's electricity consumption and carbon emissions.

[0038] Referring to the energy efficiency indicators of major products in the flat glass industry in the National Industrial Energy Efficiency Guidelines, the comprehensive electricity consumption and comprehensive energy consumption per unit of flat glass production were determined as the standards for data preprocessing. The comprehensive electricity consumption and comprehensive energy consumption per unit of flat glass production are calculated using the following formulas:

[0039]

[0040]

[0041] The range of comprehensive power consumption per unit of flat glass production is determined to be [8,12], and the range of comprehensive energy consumption per unit of flat glass production is determined to be [19,22]. Data outside this range are considered as outliers and are removed when establishing the correlation model.

[0042] Step S2: Obtain the correlation between electricity-to-electricity carbon emissions based on SVR regression prediction;

[0043] Step 2.1: Obtain 12 sets of monthly electricity consumption, related energy consumption and carbon emission data of a certain flat glass enterprise, and remove outlier data according to the range of values ​​of comprehensive electricity consumption per unit of flat glass production and comprehensive energy consumption per unit of flat glass production in Step 1.2, forming a sample set containing 8 sets of data;

[0044] Step 2.2: Select 5 samples from the sample set to form the training sample set, and the remaining 3 samples to form the test sample set. Select a kernel function to process the training sample set and construct the SVR model. Set the initial value of ε to 0.01, and use lattice search and cross-validation to determine the values ​​of parameters C and γ, finally obtaining the optimal parameter C. ∗ 5.89, γ ∗ It is 2.8;

[0045] Step 2.3: Simulate the training sample set data to obtain the optimal solution of the model and the regression function f(x). Substitute all the data from the training and test sample sets into the function to output the fitted values, and perform linear regression between the fitted results and the true values, based on the correlation coefficient R. 2 The results determine the model's learning and generalization ability. If the model's learning and generalization ability is poor, it is necessary to return to step 2.2 until the optimal solution is obtained, thus obtaining the correlation between electricity and carbon emissions.

[0046] Step S3: Based on the preprocessed carbon emission correlation, establish the correlation between electricity and correlation, and between correlation and carbon emissions excluding electricity, based on SVR regression prediction;

[0047] Step 3.1: Establish the correlation between electricity consumption and flat glass production as described in steps 2.2 to 2.3, with the optimal parameter C. ∗ γ is 0.01. ∗ It is 3;

[0048] Step 3.2: Establish the correlation between flat glass production and carbon emissions (excluding electricity) according to steps 2.2 to 2.3, with the optimal parameter C. ∗ It is 7.5, γ ∗ It is 3;

[0049] Step S4: Based on the correlation obtained in step S2, construct the electrocarbon model;

[0050] Step 4.1: Substitute the electricity consumption to flat glass production function obtained in Step 3 into the flat glass production to carbon emissions (excluding electricity) function to obtain the correlation between electricity consumption and carbon emissions (excluding electricity) established through flat glass production.

[0051] Step 4.2: Add the electricity-to-electricity carbon emission correlation obtained in Step 2 to the electricity consumption-to-non-electricity carbon emission correlation obtained in Step 4.1 to obtain the electricity consumption-to-carbon emission correlation, which is the desired model. Figure 2 As shown.

[0052] Step S5: Input the data to be measured into the electric carbon model to obtain the predicted carbon emissions.

[0053] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should be included in the scope of the present invention.

Claims

1. A method for predicting carbon emissions of flat glass industry based on electric-carbon correlation, characterized in that, The method comprises the following steps: Step S1: obtaining carbon emission related quantities of the flat glass industry and performing data preprocessing; Step S2: obtaining an electricity-to-electricity carbon emission quantity correlation based on SVR regression prediction; Step S3: establishing an electricity-to-related quantity correlation and a related quantity-to-except-electricity carbon emission quantity correlation based on SVR regression prediction according to the preprocessed carbon emission related quantities; Step S4: constructing an electricity carbon model based on the correlation obtained in step S2; Step S5: inputting the to-be-tested data into the electricity carbon model to obtain predicted carbon emission quantities; The step S1 is specifically: Step S11: selecting flat glass production as the carbon emission related quantity of the flat glass industry; Step S12: preprocessing the data by taking the unit flat glass production comprehensive electricity consumption and the unit flat glass production comprehensive energy consumption as the standards, calculating the unit flat glass production comprehensive electricity consumption and the unit flat glass production comprehensive energy consumption, and the calculation formula is: determining the value range of the unit flat glass production comprehensive electricity consumption and the unit flat glass production comprehensive energy consumption, and taking the data not in the value range as singular data, which is excluded when the correlation model is established; The step S4 is specifically: substituting the electricity-to-related quantity function into the flat glass production-to-except-electricity carbon emission quantity function to obtain the electricity quantity-to-except-electricity carbon emission quantity correlation established by the flat glass production; adding the electricity-to-electricity carbon emission quantity correlation and the electricity quantity-to-except-electricity carbon emission quantity correlation to obtain the electricity quantity-to-carbon emission quantity correlation, which is the model to be solved.

2. The method for predicting carbon emissions of the flat glass industry based on the electrical-carbon correlation relationship according to claim 1, characterized in that, The step S2 is specifically: Step S21: obtaining N groups of monthly electricity consumption, flat glass production and carbon emission quantity data of a certain flat glass enterprise and excluding singular data according to the value range of the unit flat glass production comprehensive electricity consumption and the unit flat glass production comprehensive energy consumption to form a sample set containing n groups of data; Step S22: selecting m samples in the sample set to form a training sample set, and the remaining n-m samples to form a test sample set; selecting a kernel function to process the training sample set, constructing an SVR model; setting an initial value of ε, applying a grid search and cross-validation method to determine the values of parameters C and γ, and finally obtaining optimal parameters ε * , C * , and γ * ; Step S23: Simulate the training sample set data to obtain the model optimal solution and the regression function f(x), and substitute all data of the training sample set and the test sample set into the function to output the fitting value, and perform linear regression on the fitting result and the true value, and determine the learning and generalization ability of the model according to the correlation coefficient R 2 The result determines the learning and generalization ability of the model; if the learning and generalization ability of the model is poor, return to step S22 until the optimal solution is obtained, and the correlation between the electricity-to-electricity carbon emission is obtained. 3.The method of claim 2, wherein, The step S3 is specifically: Step S31: establishing the electricity quantity-to-related quantity correlation according to steps S22 to S23; Step S32: establishing the flat glass production-to-except-electricity carbon emission quantity correlation according to steps S22 to S23.

Images