A method for calculating comprehensive carbon emissions of enterprises based on real-time energy consumption data

Abstract

The present invention relates to a method for calculating comprehensive carbon emissions of enterprises based on real-time energy consumption data. Basic energy consumption data, green new energy data and non-energy consumption data are collected to enrich the data sources, and compensation is performed to achieve reverse correction, further correcting the carbon emission calculation results of the enterprise, realizing the analysis of the enterprise's multi-energy consumption data, building a deep learning classification model, and combining the reverse correction calculation results to achieve reasonable and accurate calculation of the total comprehensive carbon emissions of the enterprise; it realizes consumption reduction early warning for high-carbon emission and over-carbon emission enterprises, changes the energy consumption behavior of enterprises, guides enterprises to accurately reduce carbon emissions, and guides their upgrading and transformation, which has great substantive significance, thereby ensuring that the total macro carbon emissions of the region operate within a reasonable range.

Description

Technical Field

[0001] The present invention belongs to the technical field of enterprise carbon emission calculation, and in particular relates to an enterprise comprehensive carbon emission calculation method based on real-time energy consumption data. Background Art

[0002] At present, China has clearly proposed the time point for achieving carbon neutrality, and carbon dioxide emissions will strive to reach a peak before 2030, and strive to achieve carbon neutrality before 2060; State Grid Corporation of China has also issued a carbon peak and carbon neutrality action plan, proposing to accelerate the construction of a clean, low-carbon, safe and efficient energy system, and continuously promote carbon emission reduction. It has clarified the main practices, research paths and action plans for promoting energy and power transformation, and helped achieve carbon peak and carbon neutrality from multiple aspects. However, the objective shortcomings of the existing enterprise carbon emission measurement technology are: 1) The data used for measurement are only data samples, with small scale, poor effectiveness, low reliability, and difficult to ensure accuracy, which affects the final measurement results; 2) The data source only contains basic energy consumption data, lacks other non-energy consumption data, and the data source is relatively single; 3) The current measurement method cannot achieve personalized and accurate measurement based on the characteristics of the enterprise itself, and cannot guide the management department to carry out scientific, reasonable and targeted carbon control strategies. Therefore, designing a comprehensive enterprise carbon emission measurement method based on real-time energy consumption data with a perfect collection scale, rich data sources, and suitable for enterprise characteristics has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a method for calculating comprehensive carbon emissions of an enterprise based on real-time energy consumption data.

[0004] The technical solution of the present invention is a method for calculating the comprehensive carbon emissions of enterprises based on real-time energy consumption data, which includes the following calculation steps: a first step, collecting enterprise carbon emission calculation data in real time on an enterprise basis, and the enterprise carbon emission calculation data includes basic energy consumption data, green new energy data and non-energy consumption data; a second step, performing preliminary preprocessing on the basic energy consumption data; a third step, performing data preprocessing on the basic energy consumption data through big data tools; a fourth step, using a CNN (convolutional neural network) classification model to classify the basic energy consumption data to obtain classification results; a fifth step, performing abnormal analysis and investigation on the classification results, and filtering and exporting data with unsatisfactory results or obvious deviations from historical values; a sixth step, numbering the enterprises, and establishing a carbon emission database corresponding to the classification results; a seventh step, performing reverse correction on the basic energy consumption data through green new energy data and non-energy consumption data; and an eighth step, exporting the data in the carbon emission database as the comprehensive carbon emission calculation results of the enterprise.

[0005] After adopting the above method, all enterprise data sets input into the classification model are collected in real time, so the data scale and timeliness are guaranteed. The calculation of enterprise carbon emissions is mainly based on electricity consumption data, supplemented by other multi-energy consumption data, so as to achieve accurate calculation of enterprise comprehensive carbon emissions. By using the CNN (convolutional neural network) classification model, the correlation between important data features can be retained to the greatest extent, and the calculation and classification of comprehensive carbon emissions can be accurately realized. By collecting basic energy consumption data, green new energy data and non-energy consumption data, the data source is enriched, and green new energy data and non-energy consumption data compensate for the basic energy consumption data to achieve reverse correction, and then The carbon emission calculation results of enterprises are corrected in one step; a carbon emission database is established by corresponding the enterprise numbering and classification results, realizing one enterprise, one data and one method, which is helpful to grasp the characteristics of the enterprise, more targeted and accurate calculation of the enterprise carbon emissions, and guide government departments to implement relevant policies; it realizes the analysis of the diversified energy consumption data of enterprises, builds a deep learning classification model, and combines the method of reverse correction of the calculation results to realize the reasonable and accurate calculation of the total comprehensive carbon emissions of enterprises; it realizes consumption reduction early warning for high-carbon emission and over-carbon emission enterprises, changes the energy consumption behavior of enterprises, guides enterprises to accurately reduce carbon emissions, and guides their upgrading and transformation, which is of great substantive significance, thereby ensuring that the total macro carbon emissions of the region operate within a reasonable range.

[0006] As a further improvement of the present invention, the basic energy consumption data includes enterprise electricity consumption data and comprehensive energy consumption data. The enterprise electricity consumption data is collected by detecting the low-voltage side of the transformer, including real-time data on the electricity consumption of the enterprise's production line; the comprehensive energy consumption data includes data on the enterprise's water, gas, heat, oil and coal consumption; the green new energy data collects new energy site information of hydropower, photovoltaics, energy storage and charging piles to generate a green new energy site list, which includes the name, voltage level, installed capacity, power content and address of the green new energy site; the non-energy consumption data includes uncontrollable energy consumption, energy consumption of related links and other data. The uncontrollable energy consumption includes gas consumption data directly discharged into the atmosphere by the enterprise during the production process and refrigeration process; the energy consumption of related links includes consumption data of enterprise raw material input, waste, logistics, travel and employee commuting, and the other data includes consumption data of the enterprise in the fields of transportation, housing and construction, and economic and information technology.

[0007] After adopting the above method, basic energy consumption data, green new energy data and non-energy consumption data involve multiple fields such as energy, industry, construction, transportation, and residents' lives, covering relevant data such as enterprise energy consumption, new energy projects, land use, and output value, so that the enterprise carbon emission calculation data is large in scale and highly reliable, ensuring the comprehensive calculation of the total amount of enterprise carbon emissions.

[0008] As a further improvement of the present invention, the data preprocessing includes manual modification processing, data normalization processing and data dimensionality reduction. The objects of the manual modification processing include multiple electricity user name data, missing data and outlier data. The multiple electricity user name data includes data of multiple user names and multiple records in an enterprise. The data normalization processing performs a linear transformation on the data so that the result value is mapped to between [0-1]; the data dimensionality reduction uses a PCA (principal component analysis) algorithm for dimensionality reduction.

[0009] After adopting the above method, the quality of basic energy consumption data is improved through data preprocessing, and the basic energy consumption data with different orders of magnitude and sizes are mapped to the same numerical range through normalization processing, which is convenient for comparing data from different enterprises. The main features of the basic energy consumption data are found through the PCA (principal component analysis) algorithm for analysis, while retaining as many data features as possible to make the data easier to classify.

[0010] As a further improvement of the present invention, the data normalization process utilizes the establishment of an artificial neural network, and the BatchNormalization function formula processes the data.

[0011] , where x (k) is the data neuron of the artificial neural network, E[x (k) ] is each batch of training data neurons x (k) The average value of Var[x (k) ] is each batch of data neurons x (k) degree of activation.

[0012] After adopting the above method, the normalization process is performed through the BatchNormalization function, which can reduce the influence of the distribution of the weight initialization value and make the network converge well. It can also speed up the convergence speed, reduce the computational complexity, increase the computational speed, improve the final data output accuracy, and increase the reliability of the measurement results.

[0013] As a further improvement of the present invention, the dimension reduction step of the PCA (principal component analysis) algorithm includes: step 1, organizing the enterprise carbon emission measurement data into a matrix X by column; step 2, performing zero mean on each row of the matrix X; step 3, calculating the covariance matrix C of the matrix X; step 4, calculating the eigenvalues ​​and corresponding eigenvectors of the covariance matrix C; step 5, arranging the eigenvectors into a new matrix by row from top to bottom according to the corresponding eigenvalue size, and taking the first k rows to form a matrix P; step 6, the matrix P is the data after the matrix X is reduced to k dimensions.

[0014] After adopting the above method, the high-dimensional data after normalization is reduced in dimensionality through the PCA (principal component analysis) algorithm, which can achieve more data features while compressing the data dimension, preventing overfitting and the occurrence of dimensionality disaster.

[0015] As a further improvement of the present invention, it also includes an enterprise carbon emission display platform, which is provided with an enterprise carbon emission evaluation module. The enterprise carbon emission evaluation module compares and evaluates the enterprise's comprehensive carbon emission calculation results with the relevant national enterprise carbon emission standards. The comparative evaluation includes high carbon emissions, low carbon emissions and normal carbon emissions.

[0016] After adopting the above method, by establishing a corporate carbon emissions display platform, it is convenient to conduct visual supervision and early warning of corporate carbon emissions. At the same time, it is convenient for companies to intuitively understand their own carbon emissions based on comparative evaluation, which helps guide companies to reduce carbon emissions. DETAILED DESCRIPTION

[0017] A method for calculating comprehensive carbon emissions of enterprises based on real-time energy consumption data includes the following calculation steps: a first step, collecting enterprise carbon emission calculation data in real time based on the enterprise as a unit, and the enterprise carbon emission calculation data includes basic energy consumption data, green new energy data and non-energy consumption data; a second step, performing preliminary preprocessing on the basic energy consumption data; a third step, performing data preprocessing on the basic energy consumption data through big data tools; a fourth step, using a CNN (convolutional neural network) classification model to classify the basic energy consumption data to obtain classification results; a fifth step, performing abnormal analysis and investigation on the classification results, and filtering and exporting data with unsatisfactory results or obvious deviations from historical values; a sixth step, numbering the enterprises, and establishing a carbon emission database corresponding to the classification results; a seventh step, performing reverse correction on the basic energy consumption data through green new energy data and non-energy consumption data; and an eighth step, exporting the data in the carbon emission database as the comprehensive carbon emission calculation results of the enterprise.

[0018] By collecting all enterprise data sets input into the classification model in real time, the scale and timeliness of the data are guaranteed. The calculation of enterprise carbon emissions is mainly based on electricity consumption data, supplemented by other multi-energy consumption data, to achieve accurate calculation of enterprise comprehensive carbon emissions. By using the CNN (convolutional neural network) classification model, the correlation between important data features can be retained to the greatest extent, and the calculation and classification of comprehensive carbon emissions can be accurately realized. By collecting basic energy consumption data, green new energy data and non-energy consumption data, the data source is enriched, and green new energy data and non-energy consumption data compensate for the basic energy consumption data to achieve reverse correction, and further correction. The carbon emission calculation results of enterprises; after the enterprise is numbered and the corresponding carbon emission database is established with the classification results, one enterprise one data and one method are realized, which helps to grasp the characteristics of the enterprise, more targeted and accurate calculation of the enterprise carbon emissions, and guide government departments to implement relevant policies; it realizes the analysis of the diversified energy consumption data of enterprises, builds a deep learning classification model, and combines the method of reverse correction of the calculation results to realize the reasonable and accurate calculation of the total comprehensive carbon emissions of enterprises; it realizes consumption reduction early warning for high-carbon emission and over-carbon emission enterprises, changes the energy consumption behavior of enterprises, guides enterprises to accurately reduce carbon emissions, and guides their upgrading and transformation, which has great substantive significance, thereby ensuring that the total macro carbon emissions of the region operate within a reasonable range.

[0019] The basic energy consumption data includes enterprise electricity consumption data and comprehensive energy consumption data. The enterprise electricity consumption data is collected by detecting the low-voltage side of the transformer, including real-time data on the electricity consumption of the enterprise's production line; the comprehensive energy consumption data includes data on the enterprise's water, gas, heat, oil and coal consumption; the green new energy data collects new energy site information of hydropower, photovoltaics, energy storage and charging piles to generate a list of green new energy sites. The list of green new energy sites includes the name, voltage level, installed capacity, power content and address of the green new energy site; the non-energy consumption data includes uncontrollable energy consumption, energy consumption of related links and other data. The uncontrollable energy consumption includes gas consumption data directly discharged into the atmosphere by the enterprise during the production process and refrigeration process; the energy consumption of related links includes consumption data of the enterprise's raw material input, waste, logistics, travel and employee commuting, and the other data includes the enterprise's consumption data in the fields of transportation, housing and construction, and economic and information technology.

[0020] Through basic energy consumption data, green new energy data and non-energy consumption data involving energy, industry, construction, transportation, residents' lives and other fields, covering corporate energy consumption, new energy projects, land use, output value and other related data, the company's carbon emissions calculation data is large in scale and highly reliable, ensuring the comprehensive calculation of the company's total carbon emissions.

[0021] The data preprocessing includes manual modification processing, data normalization processing and data dimensionality reduction. The objects of the manual modification processing include multiple electricity user name data, missing data and outlier data. The multiple electricity user name data includes data of multiple user names and multiple records in an enterprise. The data normalization processing performs a linear transformation on the data so that the result value is mapped to between [0-1]; the data dimensionality reduction uses the PCA (principal component analysis) algorithm for dimensionality reduction.

[0022] The quality of basic energy consumption data is improved through data preprocessing. Basic energy consumption data with different orders of magnitude and sizes are mapped to the same numerical range through normalization processing to facilitate comparison of data from different enterprises. The main features of basic energy consumption data are found through the PCA (principal component analysis) algorithm for analysis, while retaining as many data features as possible to make the data easier to classify.

[0023] The data normalization process utilizes the establishment of an artificial neural network.

[0024] BatchNormalization function formula processes data, normalized data

[0025] , where x (k) is the data neuron of the artificial neural network, E[x (k) ] is each batch of training data neurons x (k) The average value of Var[x (k) ] is each batch of data neurons x (k) degree of activation.

[0026] Normalization through the BatchNormalization function can reduce the impact of the distribution of weight initialization values, allowing the network to converge well; it can also speed up convergence, reduce computational complexity, increase computational speed, improve the accuracy of the final data output, and increase the reliability of measurement results.

[0027] The dimensionality reduction steps of the PCA (principal component analysis) algorithm include: step 1, organizing the enterprise carbon emission measurement data into a matrix X by column; step 2, performing zero mean on each row of the matrix X; step 3, calculating the covariance matrix C of the matrix X; step 4, calculating the eigenvalues ​​and corresponding eigenvectors of the covariance matrix C; step 5, arranging the eigenvectors into a new matrix by row from top to bottom according to the corresponding eigenvalue size, and taking the first k rows to form a matrix P; step 6, the matrix P is the data after the matrix X is reduced to k dimensions.

[0028] By using the PCA (principal component analysis) algorithm to reduce the dimensionality of high-dimensional data after normalization, it is possible to retain more data features while compressing the data dimension, thereby preventing overfitting and the occurrence of dimensionality disaster.

[0029] It also includes an enterprise carbon emissions display platform, which has an enterprise carbon emissions evaluation module. The enterprise carbon emissions evaluation module compares and evaluates the enterprise's comprehensive carbon emissions calculation results with the relevant national enterprise carbon emissions standards. The comparative evaluation includes high carbon emissions, low carbon emissions and normal carbon emissions.

[0030] By establishing a corporate carbon emissions display platform, it is easier to conduct visual supervision and early warning of corporate carbon emissions. It is also easier for companies to intuitively understand their own carbon emissions based on comparative evaluation, which helps guide companies to reduce carbon emissions.

Claims

1. A method for calculating comprehensive carbon emissions of an enterprise based on real-time energy consumption data, characterized by: The method includes the following calculation steps: first, collecting enterprise carbon emission calculation data in real time based on the enterprise as a unit. The enterprise carbon emission calculation data includes basic energy consumption data, green new energy data and non-energy consumption data; The second step is to pre-process the basic energy consumption data. The third step is to pre-process the basic energy consumption data using big data tools. The fourth step is to classify the basic energy consumption data using a convolutional neural network classification model to obtain classification results. The fifth step is to conduct an abnormal analysis on the classification results and filter out data with unsatisfactory results or data that obviously deviates from historical values. The sixth step is to number the enterprises and establish a carbon emission database corresponding to the classification results. Step 7: Reversely correct the basic energy consumption data through green new energy data and non-energy consumption data; Step 8: Export the data in the carbon emission database as the result of the enterprise's comprehensive carbon emission calculation; the reverse correction compensates the basic energy consumption data through green new energy data and non-energy consumption data; the basic energy consumption data includes enterprise electricity consumption data and comprehensive energy consumption data. The enterprise electricity consumption data is collected by detecting the low-voltage side of the transformer, including real-time data on the electricity consumption of the enterprise's production line; the comprehensive energy consumption data includes data on the enterprise's water, gas, heat, oil and coal consumption; the green new energy data collects new energy site information of hydropower, photovoltaics, energy storage and charging piles to generate a list of green new energy sites, which includes the name, voltage level, installed capacity, power content and address of the green new energy site; the non-energy consumption data includes uncontrollable energy consumption, energy consumption in related links and other data. The uncontrollable energy consumption includes gas consumption data directly discharged into the atmosphere by the enterprise during the production process and refrigeration process; the energy consumption in related links includes consumption data of raw material input, waste, logistics, travel and employee commuting of the enterprise, and the other data includes consumption data of the enterprise in the fields of transportation, housing and construction, and economics and information technology.

2. According to claim 1, a method for calculating comprehensive carbon emissions of an enterprise based on real-time energy consumption data is characterized by: The data preprocessing includes manual modification processing, data normalization processing and data dimensionality reduction. The objects of the manual modification processing include multiple electricity user name data, missing data and outlier data. The multiple electricity user name data includes data of multiple user names and multiple records in an enterprise. The data normalization processing performs a linear transformation on the data so that the result value is mapped to between [0-1]; the data dimensionality reduction uses a principal component analysis algorithm to perform dimensionality reduction.

3. According to claim 2, a method for calculating comprehensive carbon emissions of an enterprise based on real-time energy consumption data is characterized by: The dimensionality reduction steps of the principal component analysis algorithm include: step 1, organizing the enterprise carbon emission measurement data into a matrix X by column; step 2, performing zero mean on each row of the matrix X; step 3, calculating the covariance matrix C of the matrix X; step 4, calculating the eigenvalues ​​and corresponding eigenvectors of the covariance matrix C; step 5, arranging the eigenvectors into a new matrix by row from top to bottom according to the corresponding eigenvalues, and taking the first k rows to form a matrix P; step 6, the matrix P is the data after the matrix X is reduced to k dimensions.

4. According to claim 1, a method for calculating comprehensive carbon emissions of an enterprise based on real-time energy consumption data is characterized by: It also includes an enterprise carbon emission evaluation module, which compares and evaluates the enterprise's comprehensive carbon emission calculation results with the relevant national enterprise carbon emission standards, and the comparative evaluation includes high carbon emissions, low carbon emissions and normal carbon emissions.