Enterprise carbon emission data verification method based on electric energy carbon model

Abstract

The invention provides an enterprise carbon emission data verification method based on an electric energy carbon model. The enterprise carbon emission data verification method is characterized by comprising the following steps: step 1, establishing the electric energy carbon model; 2, executing a carbon emission data verification program; 3, comprehensive judgment and error report positioning are carried out; according to the novel enterprise carbon emission data verification method, an electric energy carbon model is firstly established for predicting carbon emission, then a set of detailed verification program is designed, multi-dimensional verification is carried out in combination with a model prediction value and a value reported by an enterprise based on a'material accounting method ', finally possible error report reasons are positioned, and verification efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission accounting and verification, and in particular to a method for verifying enterprise carbon emission data based on an electric energy carbon model. Background Art

[0002] There are currently two methods for measuring carbon emissions data, one is the "material accounting method" based on the principle of carbon element balance, and the other is the "flue gas monitoring method" based on direct flue gas monitoring. The "material accounting method" mainly calculates carbon emissions by monitoring the consumption of carbon-containing raw materials and fuels in the production process of the enterprise, combined with the corresponding emission coefficient. The "flue gas monitoring method" uses relevant equipment to monitor parameters such as flue gas flow and carbon dioxide concentration in the flue gas at the emission port to determine carbon emissions. In comparison, the "material accounting method" is more mature in application and is the only method adopted by the national carbon market.

[0003] Accurate and reliable carbon emission data is the basis for enterprises to fulfill their obligations in the national carbon market, and it involves the economic interests of enterprises. In order to verify the carbon emission data of enterprises, deter enterprises from falsifying data, and ensure the quality of carbon emission data, the national carbon market has taken the following measures: 1) The national carbon market has set up a carbon emission verification procedure, and a third-party agency will review the annual carbon emission data reported by enterprises. Usually, the original bills and production reports of enterprises will be checked, and on-site inspections and interviews will be conducted; 2) The national carbon market requires enterprises to keep evidence of key parameters and supporting materials for carbon emission accounting on a monthly basis, which will be reviewed regularly by the competent authorities, identify abnormal data and issue timely warnings; 3) The national carbon market will organize special supervision and assistance on the quality of carbon emission data reports from time to time, conduct on-site spot checks on some enterprises that have been verified by third-party agencies but still have the risk of data falsification, and verify carbon emission data again.

[0004] In addition to the above work required by the national carbon market, some other methods have also emerged to verify the carbon emission data reported by enterprises based on the "material accounting method". The first type is to compare the carbon emission data obtained by the "flue gas monitoring method" with the results of the "material accounting method"; the second type is to compare the carbon emission data predicted by mathematical models such as "electricity carbon" and "electricity carbon" with the results of the "material accounting method".

[0005] The third-party carbon emission verification and special supervision and assistance for carbon emission data reporting quality adopted by the national carbon market cannot timely detect problems in corporate carbon emission reporting, and require a lot of manpower and material resources, which is not suitable as a regular carbon emission data verification measure. The monthly evidence review mechanism adopted by the national carbon market lacks professional methodology, can only detect very obvious misreporting, and is inefficient.

[0006] The installation and construction cost of the "flue gas monitoring method" is high, and the current penetration rate is very low. It is difficult to become a universal carbon emission data verification technology in the short term. Various mathematical models such as "electricity carbon" and "electricity carbon" only compare the differences in carbon emission results. The verification procedures are not rigorous enough, resulting in low accuracy of the verification conclusions, and the original data indicators that are misreported cannot be located. Summary of the invention

[0007] In order to make up for the deficiencies of the prior art, the embodiment of the present application proposes a method for verifying enterprise carbon emission data based on an electric energy carbon model to solve the problems existing in the prior art.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] A new method for verifying corporate carbon emissions data first establishes an electric energy carbon model to predict carbon emissions, then designs a detailed verification procedure, combines the model prediction values ​​with the values ​​reported by the enterprise based on the "material accounting method" to perform multi-dimensional verification, and finally locates possible causes of misreporting, thereby improving verification efficiency and accuracy.

[0010] The specific process is as follows:

[0011] 1. Establishing an Electricity Carbon Model

[0012] 1. Data Collection

[0013] Collect historical production data of a large number of enterprises, including detailed electricity consumption data of different production links (such as raw material preparation, clinker calcination, cement grinding, etc. of cement enterprises), production equipment parameters (power, operating time, load rate, etc.), carbon-containing raw materials and fuel usage, etc., and obtain corresponding carbon emission data measured by the "material accounting method".

[0014] 2. Screening of key independent variables

[0015] Analyze the intrinsic relationship between electricity and carbon emissions in the production process of enterprises, and determine the impact mechanism of electricity on carbon emissions in various links. From the collected data, preliminarily select key indicators that are closely related to corporate carbon emissions and can be reflected by electricity data as alternative variables.

[0016] The correlation analysis method in statistics is used to measure the correlation between each alternative variable and carbon emissions. For each alternative variable in the enterprise production data and carbon emission data, the Pearson correlation coefficient between them is calculated. A correlation threshold r is set. th For each alternative variable X, if its correlation coefficient (r XY ) is greater than r th , then the independent variable is selected as the key independent variable.

[0017] 3. Regression model construction

[0018] The collected enterprise data is divided into a training set and a validation set in a ratio of 6:4. The training set is used to train the model, and the validation set is used for model evaluation and optimization.

[0019] Support vector machine regression (SVR), random forest model, full subset regression model, ridge regression, Lasso model, and neural network algorithm were used for regression modeling. The independent variable data of the training set was used as the model input, and carbon emissions were used as the model output to train the model.

[0020] After training, the goodness of fit (R 2 ), Adjusted R 2 ), mean square error (MSE), and mean absolute error (MAE).

[0021] In order to comprehensively compare the impact of different indicators on the model, the evaluation indicators of each model are standardized. Let R 2 for Adjusted R 2 for MSE value is MSE i , MAE value is MAE i The normalization formula is as follows:

[0022] For R 2 and the adjusted R 2 (larger values ​​are better):

[0023]

[0024] Where max(R 2 ) and min(R 2 ) is the R 2 The maximum and minimum values ​​of max(Adjusted R 2 ) and min(Adjusted R 2 ) is the adjusted R in all models 2 The maximum and minimum values ​​of .

[0025] For MSE and MAE (smaller values ​​are better):

[0026]

[0027] Among them, max(MSE) and min(MSE) are the maximum and minimum values ​​of MSE in all models, and max(MAE) and min(MAE) are the maximum and minimum values ​​of MAE in all models.

[0028] Assign weights to each evaluation indicator, R 2 The weight is 0.3, and the adjusted R 2 is 0.3, MSE is 0.2, and MAE is 0.2. Then calculate the comprehensive score of each model i :

[0029]

[0030] The models are sorted according to the calculated comprehensive scores, and the model with the highest comprehensive score is the optimal model and is used as the final selected regression model.

[0031] (II) Implementation of carbon emission data verification procedures

[0032] 1. Data preparation

[0033] Using the above-mentioned established electric energy carbon model, the production-related data (i.e., model independent variable data) of the enterprise's current production cycle (such as monthly or daily) is input into the model to obtain the carbon emission forecast data for that cycle.

[0034] Obtain the carbon emission data calculated and reported by the enterprise based on the "material accounting method" during the same production cycle.

[0035] 2. Carbon emissions verification and misstatement risk assessment

[0036] (1) Difference analysis

[0037] Calculate the difference ratio Δ between the carbon emission value predicted by the electric energy carbon model and the carbon emission value reported by the "material accounting method".

[0038] The mean μ and standard deviation σ are calculated based on the statistical distribution of the difference ratio between the predicted value and the reported value of the electric energy carbon model in the training set and the validation set.

[0039] If |Δ|≤μ+2σ, the verification conclusion of this step is low risk (normal range), indicating that the difference between the predicted carbon emission value and the reported value is within a reasonable range and the data credibility is high;

[0040] If μ+2σ<|Δ|<μ+3σ, the conclusion of this step verification is medium risk (suspicious range), which means that the difference exceeds the normal small fluctuation range, but has not reached the level of serious deviation, and further inspection is required;

[0041] If |Δ|≥μ+3σ, the conclusion of this step verification is high risk (abnormal range), indicating that the difference is significant, and the reported carbon emission values ​​may have major problems and a high possibility of misreporting.

[0042] (2) Outlier Check

[0043] Calculate the mean μ and standard deviation σ based on the statistical distribution of the independent variables in the training set and the validation set. Determine the theoretical maximum value T of the above indicators based on physical laws, expert knowledge, and national regulations. max and minimum value T min .

[0044] Relative and proportional indicators in the model's independent variables are selected for outlier verification.

[0045] If the independent variable X∈[ux-2σx,ux+2σx], the verification conclusion of this step is low risk (normal range);

[0046] If the independent variable X∈(ux-3σx, ux-2σx]∪[ux+2σx, ux+3σx), the verification conclusion of this step is medium risk (suspicious range);

[0047] If the independent variable X∈[T min ,ux-3σx]∪[ux+3σx,T max ], then the verification conclusion of this step is high risk (abnormal range);

[0048] If the independent variable X∈(-∞, T min )∪(T max , +∞), then the verification conclusion of this step is to confirm the misstatement.

[0049] (3)Consistency check

[0050] Based on the enterprise's production process knowledge and data analysis methods, identify the independent variable combinations with stable quantitative relationships as consistency verification objects. For each group of selected independent variables X1, X2...X m …X n , a quantitative relationship model is established based on process principles and historical statistical data, X n =f(X1, X2…X m ). X1, X2…X in the data to be verified m Substitute the calculated X in function f n Predicted value Calculate predicted values The actual data to be verified The difference ratio D measures the degree of deviation from the consistency of the independent variable data to be verified.

[0051]

[0052] If D≤10%, the verification conclusion of this step is low risk (logical consistency);

[0053] If 10%<D<20%, ​​the verification conclusion of this step is medium risk (slight inconsistency);

[0054] If D ≥ 20%, the verification conclusion of this step is high risk (serious inconsistency)

[0055] (4) Stability verification

[0056] Based on the statistical distribution of the historical data of the year-on-year fluctuation range V of the independent variable in the training set and the validation set, its mean μ and standard deviation σ are calculated.

[0057] Relative and proportional indicators in the model's independent variables are selected for stability verification.

[0058] If |V|≤uv+2σv, the verification conclusion of this step is low risk (stable);

[0059] If μv+2σv<|V|<μv+3σv, the verification conclusion of this step is medium risk (mild fluctuation);

[0060] If |V|≥μv+3σv, the verification conclusion of this step is high risk (large fluctuations).

[0061] (III) Comprehensive judgment and misstatement location

[0062] The results of the four steps of comprehensive difference analysis, outlier check, consistency check and stability check are integrated together as the input for comprehensive judgment. Each step outputs its own risk level (low risk, medium risk, high risk and misstatement).

[0063] The difference analysis directly compares the carbon emission values ​​predicted by the electric energy carbon model with the carbon emission values ​​reported by the "material accounting method". It is a direct comparison of the accuracy of carbon emission data, and its results have a key impact on the final conclusion. Therefore, the difference analysis is given the highest priority in the overall comprehensive judgment.

[0064] The following comprehensive judgment rules are designed, including four judgment conclusions: misstatement, high risk, medium risk and low risk. Among them:

[0065] Misstatement means that the verification procedure has clearly found errors in the values ​​reported by the enterprise; high risk means a negative conclusion, indicating that after a comprehensive verification of the carbon emission data, it was found that the data has a significant risk of misstatement; medium risk means a reserved conclusion, indicating that the carbon emission data is basically reliable as a whole, but there are some aspects that need further verification; low risk means an unreserved conclusion, indicating that no factors that may lead to misstatement have been found, and the quality and reliability of the carbon emission data is high.

[0066] Table 1: Comprehensive judgment rules

[0067]

[0068] When the comprehensive judgment conclusion is medium risk, high risk and misstatement, the independent variables related to suspected misstatements determined in each medium and high risk verification step are located, and the independent variable name, verification step, suspected misstatement performance (such as difference value, abnormal value, inconsistent relationship, unstable fluctuation) and possible misstatement cause analysis are output.

[0069] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0070] (1) Reduce the cost of verification:

[0071] The existing third-party verification procedures in the national carbon market require third-party organizations to invest a lot of manpower to conduct on-site surveys, check original invoices and production reports, interviews, etc., which is costly. However, this application is mainly based on the company's own production data and electric energy carbon model for verification, and there is no need for a third-party organization to conduct such complex on-site work. Taking a medium-sized cement company as an example, traditional third-party verification may cost tens of thousands of yuan each time, but using this method, after the model is established, the main cost of the company is data collection and model operation, which greatly saves expenses. For the method of comparing and verifying the "flue gas monitoring method" with the "material accounting method", additional installation and maintenance of flue gas monitoring equipment is required, and the equipment procurement, installation and commissioning, and subsequent operation and maintenance costs are relatively high. The electric energy carbon model of this application is established based on the relevant parameters of the company's existing production data, without the need for additional hardware equipment investment, reducing the verification cost.

[0072] (2) Causes of positioning misreporting:

[0073] Existing national carbon market measures (such as third-party verification, review by competent authorities, etc.) often find it difficult to quickly and accurately locate specific causes after discovering data anomalies, and require further in-depth investigation of each link in the company's production, which is a cumbersome and time-consuming process. Other verification methods (such as the comparison between the "flue gas monitoring method" and the "material accounting method") cannot effectively point out which link in the production process caused the carbon emission data error after discovering the difference. This application uses a multi-step verification procedure (outlier verification, consistency verification, stability verification, etc.) to accurately locate possible causes of misreporting based on the abnormal conditions of the independent variables in each step. For example, if an abnormal power consumption of a specific device is found in the outlier verification, and the power consumption fluctuation of the device is abnormal in the stability verification, it can be clearly pointed out that there may be problems in the production links related to the equipment, such as equipment failure, unreasonable operating parameters, etc.

[0074] (3) Improve verification efficiency:

[0075] When enterprises use this application method to conduct monthly carbon emission data self-inspection, they only need to input the production data into the electric energy carbon model and run the verification program to complete the verification in a short time. Compared with the traditional method of waiting for annual verification by a third-party agency, the efficiency is greatly improved. Enterprises can find problems and make adjustments in a timely manner to avoid affecting carbon market compliance due to data problems. When the competent authorities review a large number of corporate carbon emission data, they can use this method to quickly screen out companies that may have data problems. Through the automated operation of the model and verification program, there is no need to conduct on-site spot checks and detailed manual verification of each enterprise as in traditional special supervision and assistance. It improves regulatory efficiency and ensures the timeliness and effectiveness of carbon market data review. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 Schematic diagram of the enterprise carbon emission data verification method based on the electric energy carbon model. DETAILED DESCRIPTION

[0077] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] This embodiment takes a cement production enterprise as an example to explain in detail the enterprise carbon emission data verification method based on the electric energy carbon model of the present invention.

[0079] 1. Establishing an electric energy carbon model:

[0080] 1. Data collection phase:

[0081] For cement companies, collect their historical production data for the past five years (the time can be adjusted according to the availability of actual data). This includes the power consumption data (power, operating time, etc.) of the ball mill in the raw material preparation process, the power consumption and operating parameters of the rotary kiln in the clinker calcination process, and the power usage of the mill in the cement grinding process. At the same time, record the usage of carbon-containing raw materials (such as limestone, clay, etc.) and fuels (such as coal, natural gas, etc.) in each link. And obtain the carbon emission data calculated according to the "material accounting method" corresponding to these production data. These data will form the basis for subsequent model establishment and analysis.

[0082] 2. Screening of key independent variables:

[0083] In-depth analysis of the intrinsic relationship between these indicators and carbon emissions in the cement production process. The comprehensive energy consumption per unit of clinker directly reflects the energy utilization efficiency in the production process and is closely related to carbon emissions. The higher the energy consumption, the higher the carbon emissions. The proportion of self-generated electricity affects the company's dependence on external electricity purchases, which in turn affects the electricity-related carbon emissions in the carbon emission calculation. Clinker output is an important reflection of production scale, and its changes will affect the overall carbon emission level. Coal consumption is one of the important sources of carbon emissions, which is directly related to the carbon emissions during the combustion process.

[0084] The Pearson correlation coefficient between each alternative variable and carbon emission data is calculated using the correlation analysis method. Set the correlation threshold r th =0.6. After calculation, assuming that the correlation coefficient between unit clinker comprehensive energy consumption and carbon emissions is r XY1 =0.7, the correlation coefficient between the proportion of self-generated electricity and carbon emissions r XY2 =-0.75 (negative correlation means that the higher the proportion of self-generation, the lower the carbon emissions may be). The correlation coefficient between clinker production and carbon emissions is r XY2 =0.65, the correlation coefficient between coal consumption and carbon emissions r XY2 =0.85, and the absolute values ​​of these correlation coefficients are greater than 0.6, thus selecting the comprehensive energy consumption per unit clinker, the proportion of self-generated electricity, the clinker output, and the coal consumption as the key independent variables.

[0085] 3. Regression model construction and optimization

[0086] The collected enterprise data are divided into training set and validation set in a ratio of 6:4. The training set is used to train six models: support vector machine regression (SVR), random forest model, full subset regression model, ridge regression, Lasso model, and neural network algorithm. The key independent variable data such as unit clinker comprehensive energy consumption, self-generation ratio, clinker output, coal consumption, etc. are used as model input, and carbon emission data is used as output to train the model.

[0087] After training, the goodness of fit (R 2 ), Adjusted R 2 ), mean square error (MSE), and mean absolute error (MAE).

[0088] The evaluation indicators of each model are standardized according to the standardized formula. Assume that R 2 for Adjusted R 2 for MSE value is MSE i , MAE value is MAE i .

[0089] For R 2 and the adjusted R 2 (larger values ​​are better):

[0090]

[0091] For MSE and MAE (smaller values ​​are better):

[0092]

[0093] Assign a weight to each evaluation indicator, R 2 The weight is 0.3, and the adjusted R 2 is 0.3, MSE is 0.2, and MAE is 0.2. Then calculate the comprehensive score of each model i :

[0094]

[0095] After calculation, assuming that the neural network algorithm has the highest comprehensive score, it will be used as the final selected regression model for subsequent carbon emission prediction.

[0096] 2. Execute the carbon emission data verification procedure:

[0097] 1. Data preparation:

[0098] After the monthly production is completed, the cement enterprise's production-related data (i.e., model independent variable data) such as unit clinker comprehensive energy consumption, self-generated electricity ratio, clinker output, and coal consumption for that month are input into the established electric energy carbon model to obtain the carbon emission forecast data for that month. At the same time, the carbon emission data calculated and reported by the enterprise in the same month based on the "material accounting method" is obtained for verification.

[0099] 2. Carbon emissions verification and misstatement risk assessment:

[0100] (1) Difference analysis:

[0101] The difference ratio between the carbon emission value predicted by the calculated electric energy carbon model and the carbon emission value reported by the “material accounting method” is Δ=-0.11.

[0102] According to the statistical distribution of the difference ratio between the predicted value and the reported value of the electric energy carbon model in the training set and the validation set, the mean u=0.06 and σ=0.02 are calculated.

[0103] Since |Δ|=0.11>u+2σ=0.06+2×0.02=0.1 and |Δ|=0.11<u+3σ=0.06+3×0.02=0.12, the verification conclusion of this step is medium risk (suspicious range).

[0104] (2) Outlier verification:

[0105] For the comprehensive energy consumption per unit clinker, the average value u is calculated based on historical data. x =105 and standard deviation σ x = 10. According to physical laws and expert knowledge, the theoretical maximum value of comprehensive energy consumption per unit clinker is determined. max =150, minimum value Tmin=70.

[0106] The comprehensive energy consumption per unit clinker of the enterprise is 100∈[85, 125], and the verification conclusion is low risk (normal range).

[0107] The same outlier check operation was performed on the independent variables such as self-generation ratio, clinker output, and coal consumption, and the corresponding mean, standard deviation, theoretical maximum value, and minimum value were determined according to their historical data distribution and physical meaning. The check conclusions were all low risk.

[0108] (3)Consistency check:

[0109] Based on cement production process knowledge, a combination of independent variables with a stable quantitative relationship is identified. It is determined that there is a certain relationship between coal consumption, clinker output and comprehensive energy consumption per unit clinker. Coal consumption X1 and clinker output X2 are selected as consistency verification objects, and a quantitative relationship model X3 = f(X1, X2) is established based on historical statistical data (assuming that X3 is a related indicator of comprehensive energy consumption per unit clinker).

[0110] Substitute X1 and X2 in the data to be verified into function f to calculate the predicted value of X3 Calculate the predicted value and the actual data to be verified The difference ratio The verification conclusion is high risk (serious inconsistency).

[0111] (4) Stability check:

[0112] Select relative indicators such as the proportion of self-generation and ratio indicators for stability verification. Calculate its mean μ based on the statistical distribution of the historical data of the year-on-year fluctuation range of the self-generation ratio in the training set and the validation set. v = 0.04 and standard deviation σ v =0.01.

[0113] Since |V|=0.03≤μ v +2σ v =0.04+2×0.01=0.06, the verification conclusion is low risk (stable).

[0114] The same stability check operation was performed on the selected independent variables such as unit clinker comprehensive energy consumption, clinker output, and coal consumption, and the corresponding fluctuation range mean and standard deviation were calculated based on their historical data. The check conclusions were all low risk.

[0115] 3. Comprehensive judgment and misstatement location:

[0116] 1. Application of comprehensive judgment rules:

[0117] The results of the four steps of comprehensive difference analysis, outlier check, consistency check and stability check are as follows: The difference analysis result is medium risk, the outlier check result is low risk, the consistency check result is high risk, and the stability check result is low risk.

[0118] Since the difference analysis has the highest priority, combined with other verification results, according to the comprehensive judgment rules, this situation can be judged as high risk. This shows that after comprehensive verification, there is a significant risk of misstatement in carbon emission data, which requires enterprises to focus on and conduct in-depth analysis of the reasons.

[0119] 2. Misstatement Location:

[0120] Because the consistency check result is high risk, the consistency check process is reviewed. For the verification of the above-mentioned coal consumption X1, clinker output X2 and unit clinker comprehensive energy consumption X3 related indicators, it is found that D ≥ 20%, indicating that there is a serious inconsistency in the relationship between this group of independent variables.

[0121] Further analysis may be due to production process adjustments (such as the use of new coal formulas or production equipment upgrades that affect the relationship between clinker output and energy consumption), which makes the theoretical relationship no longer applicable, or the data on coal consumption or clinker output during data collection is inaccurate, which affects the calculation of the comprehensive energy consumption per unit of clinker and causes the consistency check to fail.

[0122] Through the above specific implementation methods, cement enterprises can use the method of the present invention to effectively verify carbon emission data, timely discover and locate possible misreporting, improve the accuracy and reliability of carbon emission data, and reduce verification costs and improve verification efficiency. In practical applications, enterprises can appropriately adjust various parameters and thresholds according to their own production characteristics and data conditions to adapt to different production environments and data quality requirements.

[0123] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

[0124] In addition, it should be understood that although this specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment have also been appropriately combined to form other implementation modes that are easy for those skilled in the art to understand.

Claims

1. A method for verifying enterprise carbon emission data based on an electric energy carbon model, characterized in that: The following steps are involved: Step 1: Establish an electric energy carbon model; Step 2: Execute the carbon emission data verification procedure; Step 3: Comprehensive judgment and misstatement location.

2. According to the method for verifying enterprise carbon emission data based on electric energy carbon model in claim 1, it is characterized in that: The step 1 specifically includes: data collection, key independent variable screening and regression model construction.

3. According to claim 2, a method for verifying enterprise carbon emission data based on electric energy carbon model is characterized in that: The data collection specifically involves: collecting historical production data of a large number of enterprises, including different production links, detailed electricity consumption data, production equipment parameters, carbon-containing raw materials and fuel usage information, and obtaining corresponding carbon emission data measured by the material accounting method.

4. According to claim 2, a method for verifying enterprise carbon emission data based on electric energy carbon model is characterized in that: The key independent variable screening is specifically as follows: analyzing the intrinsic relationship between electricity and carbon emissions in the production process of the enterprise, and determining the impact mechanism of electricity on carbon emissions in various links; preliminarily screening out key indicators related to the carbon emissions of the enterprise and which can be reflected by electricity data from the collected data as alternative independent variables, and using the correlation analysis method in statistics to measure the degree of correlation between each alternative independent variable and carbon emissions; for each alternative independent variable in the enterprise production data and carbon emission data, calculating the Pearson correlation coefficient between them; setting a correlation threshold r th ; For each alternative variable X, if its correlation coefficient with carbon emission data Y is r XY The absolute value is greater than r th , then the independent variable is selected as the key independent variable.

5. The method for verifying enterprise carbon emission data based on electric energy carbon model according to claim 2 is characterized in that: The regression model construction is specifically as follows: the collected enterprise data is divided into a training set and a validation set according to a ratio of 6:4; the training set is used for training the model, and the validation set is used for model evaluation and optimization; Support vector machine regression SVR, random forest model, full subset regression model, ridge regression, Lasso model, and neural network algorithm were used for regression modeling. The independent variable data of the training set was used as the model input, and carbon emissions were used as the model output to train the model. After training, the goodness of fit R of each model is calculated on the validation set. 2 , Adjusted R 2 , mean square error MSE, mean absolute error MAE; In order to comprehensively compare the impact of different indicators on the model, the evaluation indicators of each model are standardized. Let R 2 for Adjusted R 2 for MSE value is MSE i , MAE value is MAE i , the standardized formula is as follows: For R 2 and the adjusted R 2 : Where max(R 2 ) and min(R 2 ) is the R 2 The maximum and minimum values ​​of max(AdjustedR 2 ) and min(AdjustedR 2 ) is the adjusted R in all models 2 The maximum and minimum values ​​of ; For MSE and MAE: Among them, max(MSE) and min(MSE) are the maximum and minimum values ​​of MSE in all models, and max(MAE) and min(MAE) are the maximum and minimum values ​​of MAE in all models; Assign a weight to each evaluation indicator, R 2 The weight is 0.3, and the adjusted R 2 is 0.3, MSE is 0.2, MAE is 0.2, and then the comprehensive score of each model is calculated. i : The models are sorted according to the calculated comprehensive scores, and the model with the highest comprehensive score is the optimal model and is used as the final selected regression model.

6. The method for verifying enterprise carbon emission data based on electric energy carbon model according to claim 1 is characterized in that: The step 2 specifically includes: data accuracy and carbon emission verification and misstatement risk assessment.

7. The method for verifying enterprise carbon emission data based on electric energy carbon model according to claim 6 is characterized in that: The data preparation specifically includes: using the above-mentioned established electric energy carbon model, inputting the production-related data of the enterprise's current production cycle into the model, obtaining the carbon emission forecast data of the cycle, and obtaining the carbon emission data calculated and reported by the enterprise based on the "material accounting method" in the same production cycle.

8. The method for verifying enterprise carbon emission data based on electric energy carbon model according to claim 6 is characterized in that: The carbon emission verification and misstatement risk assessment specifically include: Ⅰ. Difference analysis: Calculate the difference ratio Δ between the carbon emission value predicted by the electric energy carbon model and the carbon emission value reported by the "material accounting method"; calculate its mean μ and standard deviation σ based on the statistical distribution of the difference ratio between the predicted value and the reported value of the electric energy carbon model in the training set and the validation set; if |Δ|≤μ+2σ, the verification conclusion of this step is low risk, indicating that the difference between the predicted carbon emission value and the reported value is within a reasonable range and the data credibility is high; if μ+2σ<|Δ|<μ+3σ, the verification conclusion of this step is medium risk, which means that the difference exceeds the normal small fluctuation range, but has not reached the level of serious deviation and needs further inspection; if |Δ|≥μ+3σ, the verification conclusion of this step is high risk, indicating that the difference is significant, and the reported carbon emission value may have major problems and has a high possibility of misreporting; II. Outlier verification: Calculate the mean μ and standard deviation σ based on the statistical distribution of the independent variables in the training set and the validation set; determine the theoretical maximum value T of the above indicators based on physical laws, expert knowledge, and national regulations. max and minimum value T min ; Select relative and proportional indicators in the model independent variables for outlier verification; if the independent variable X∈[ux-2σx, ux+2σx], the verification conclusion of this step is low risk; if the independent variable X∈(ux-3σx, ux-2σx]∪[ux+2σx, ux+3σx), the verification conclusion of this step is medium risk; if the independent variable X∈[T min ,ux-3σx]∪[ux+3σx,T max ], The verification conclusion of this step is high risk; if the independent variable X∈(-∞, T min )∪(T max , +∞), The verification conclusion of this step is to confirm the misstatement; III. Consistency check: Based on the enterprise's production process knowledge and data analysis methods, identify the independent variable combinations with stable quantitative relationships as the consistency check objects; for each group of selected independent variables X1, X2...X m …X n , a quantitative relationship model is established based on process principles and historical statistical data, X n =f(X1, X2…X m ); X1, X2…X m Substitute the calculated X in function f n Predicted value Calculate predicted values The actual data to be verified The difference ratio D measures the degree of deviation from the consistency of the independent variable data to be verified; If D≤10%, the verification conclusion of this step is low risk; If 10%<D<20%, ​​the verification conclusion of this step is medium risk; If D ≥ 20%, the verification conclusion of this step is high risk; IV. Stability check: According to the statistical distribution of the historical data of the year-on-year fluctuation range V of the independent variable in the training set and the validation set, calculate its mean μ and standard deviation σ; select relative indicators and proportional indicators in the model independent variables for stability check; if |V|≤uv+2σv, the verification conclusion of this step is low risk; if μv+2σv<|V|<μv+3σv, the verification conclusion of this step is medium risk; if |V|≥μv+3σv, the verification conclusion of this step is high risk.

9. The method for verifying enterprise carbon emission data based on electric energy carbon model according to claim 1 is characterized in that: Step 3 is specifically: comprehensive difference analysis, outlier check, consistency check and stability check. Each step outputs its own risk level, which includes low risk, medium risk, high risk and misstatement. These results are integrated together as input for comprehensive judgment.

Images