Carbon emission near real-time calculation method based on electricity-energy-carbon correlation model

Abstract

The application discloses a kind of carbon emission near real-time calculation methods based on electric-energy-carbon correlation model, comprising the following steps: using the carbon emission related power user clustering of SCV-LV, after the user group data of clustering is handled by EEMD, as the input variable of PSO-LSTM model;Using the regularity component, trend component and random component of user class power time series data obtained by EEMD power data time sequence feature;Based on regularity component, trend component and random component respectively carry out industrial production process carbon emission estimation, purchase indirect carbon emission estimation of implicit power, and fossil energy consumption carbon emission estimation.The application makes full use of historical information, fully and timely reflects the change of carbon emission, real-time monitoring of high-emission industry carbon emission influence, reduce the pollution to environment.

Description

Technical Field

[0001] This invention relates to the field of carbon emissions, and in particular to a near real-time carbon emission calculation method based on an electricity-energy-carbon correlation model. Background Technology

[0002] With the development of society and the economy, energy utilization and environmental pollution have become increasingly important issues. Carbon emission accounting is an important foundational task.

[0003] Current carbon emission accounting methods are mainly divided into production-side carbon emission accounting methods, which estimate carbon emissions by using energy activity level data and parameters such as carbon content and carbon oxidation rate; and consumption-side carbon emission accounting methods, which mainly refer to obtaining the carbon emission footprint of products through life cycle assessment methods and estimating the carbon emissions of consumer products throughout their entire life cycle.

[0004] Due to the lag in the release of basic statistical data, current carbon emission accounting methods have the following drawbacks:

[0005] 1) It relies on annual energy statistics and cannot distinguish subtle industry differences; 2) It lacks accurate response to real-time events; 3) The low frequency and poor real-time performance of monitoring make it more in demand for carbon emission data that is more reliable and has lower latency. Summary of the Invention

[0006] This invention provides a near real-time carbon emission calculation method based on an electricity-energy-carbon correlation model. This invention fully utilizes historical information to comprehensively and promptly reflect changes in carbon emissions, monitors the impact of carbon emissions from high-emission industries in real time, and mitigates environmental pollution. See the description below for details:

[0007] A near-real-time carbon emission calculation method based on an electricity-energy-carbon correlation model, the method comprising the following steps:

[0008] Carbon emission-related electricity user clustering using SCV-LV, and the clustered user group data processed by EEMD, are used as input variables for the PSO-LSTM model;

[0009] Use the time-series characteristics of EEMD's electricity consumption data to obtain the regularity, trend, and random components of user-class electricity consumption time-series data;

[0010] Carbon emissions from industrial production processes, indirect carbon emissions from purchased electricity, and carbon emissions from fossil fuel consumption are estimated based on regularity components, trend components, and random components, respectively.

[0011] The carbon emission estimation for the industrial production process is achieved by using an autoregressive distributed lag model to make short-term predictions of industrial product output.

[0012] yt,γ =a0+a1t+φ1y t-1,γ +φ2y t-2,γ …φ p y t-p,γ +β0x t,γ +β1x t-1 …β q x t-q +u t,γ

[0013] In the formula, y t,γ φ1 represents the output of major industrial products; a0 represents the basic output of major industrial products; a1t represents the rate of change of time on product output; φ1y t-1,γ +φ2y t-2,γ …φ p y t-p,γ This represents the combined impact of historical product output on current product output; β0x t,γ +β1x t-1 …β q x t-q This indicates the combined impact of current and historical electricity consumption on current product output; t,γ This represents the corresponding random variable.

[0014] The indirect carbon emissions implied by the purchased electricity are estimated as follows: the carbon flow of the input node is equal to the carbon flow of the output node. The carbon emission factor of the power grid in each region is calculated by solving the carbon flow balance equation.

[0015]

[0016] In the formula, EF el,m E represents the regional electricity carbon emission factor for the current month. eg,m This indicates the carbon emissions from electricity generation in the current month; ps m This indicates the monthly electricity consumption in the area. This represents the sum of electricity transferred out of province i; This indicates the carbon emissions implied by the electricity transmitted to province i; EF j This represents the average carbon emission factor of the power grid in province j.

[0017] Furthermore, the carbon emission estimation of fossil energy consumption includes: carbon emission estimation based on the PSO-LSTM algorithm;

[0018] By comprehensively considering the seasonal characteristics of "electricity-energy" at the same time point in previous years, the accuracy of corresponding predictions can be improved.

[0019] F i,m =PSOLSTM θ (P regular,m ,P trend,m ,P random,m,PO s,m )

[0020] In the formula, F i,m This indicates the estimated consumption of different fossil fuels in the current month; PO s,m PSOLSTM represents monthly product output data for different industries. θ This represents the PSO-LSTM algorithm program.

[0021] The beneficial effects of the technical solution provided by this invention are:

[0022] (1) Compared with the existing technology that simply uses energy statistics data for accounting, this invention combines the powerful prediction capabilities of PSO-LSTM (long short-term memory network optimized by particle swarm optimization algorithm) and constructs an "electricity-energy-carbon" correlation model based on electricity consumption characteristic data, carbon emission data, and industry activity level related data. This can effectively utilize user historical data information to achieve the purpose of accurate near real-time estimation of energy carbon emissions.

[0023] (2) Compared with the clustering methods in the prior art, the present invention uses a supervised clustering method to conduct targeted clustering analysis on a large number of electricity users in different industries, find user classes that play a key role in estimating carbon emissions in the industry, and use the EEMD (Ensemble Empirical Mode Decomposition) method to extract their electricity consumption characteristic data, laying the foundation for subsequent model construction.

[0024] Therefore, this invention can make full use of existing historical data and information. The carbon emission estimation method can help understand the real-time changes of high-emission industries in the region, support the formulation of effective carbon emission reduction decision-making recommendations, monitor the impact of carbon emissions from high-emission industries in real time, and reduce environmental pollution. Attached Figure Description

[0025] Figure 1 A flowchart of a near real-time carbon emission calculation method based on an electricity-energy-carbon correlation model;

[0026] Figure 2 Here is a flowchart of the EEMD algorithm proposed in this invention;

[0027] Figure 3 This is a schematic diagram of the training and prediction process of the PSO-LSTM network.

[0028] Figure 4 Construct a map for "electricity-energy-carbon";

[0029] Figure 5 The following are example diagrams illustrating carbon emissions from the steel, chemical, building materials, and paper industries, respectively, using actual data from Tianjin as an example of the method of this invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below.

[0031] To address the problems existing in the background technology, the present invention's embodiments have revealed that carbon emissions from fossil fuel combustion and carbon emissions from purchased electricity are the main components of carbon emissions. Electricity has a strong correlation with both, and can reflect carbon emissions in a timely and accurate manner. Furthermore, electricity data possesses characteristics such as strong real-time performance, high data quality, and consistent statistical standards; it is easy to acquire and has a wide collection scope, providing massive amounts of data for carbon emission estimation and analysis.

[0032] The "electricity-energy-carbon" correlation model constructed in this embodiment of the invention fully considers the correlation between different user classes and industry carbon emissions, as well as the impact of production activity levels on carbon emissions. The supervised clustering method used can effectively select the user class that makes the model estimate the best, and the PSO-LSTM method can effectively identify and utilize the temporal characteristics of the data to improve the estimation accuracy and reliability of the model.

[0033] Example

[0034] A near-real-time carbon emission calculation method based on an electricity-energy-carbon correlation model, comprising the following steps:

[0035] 101: Carbon emission-related electricity users were clustered using SCV-LV (Supervised Clustering Method Around Latent Variables). The clustered user group data was then processed by EEMD and used as the input variable for the PSO-LSTM model.

[0036] This step utilizes hierarchical clustering to iteratively extract a series of variable groups by maximizing the within-group correlation of the extracted variable groups and maximizing the contribution of the variable groups to the response variable. The clustering criterion is to maximize the global identity of the variable groups while controlling for the loss of local identity based on the response variable.

[0037] The goal of clustering is to obtain clustering information between variables, improve the predictive ability of the model, and enhance its interpretability. The approach is as follows: First, cluster the response variable y with all explanatory variables {x1, x2, x3, ..., x...}. p The data are grouped together and then subjected to hierarchical clustering. After obtaining the multi-level clustering results, the optimal clustering level is determined, i.e., the best single-level clustering result that meets the requirements is selected. The specific steps are as follows:

[0038] 1) Initialization;

[0039] The response variable y and all explanatory variables {x1, x2, x3, ..., x} are used to define the response variable y and all explanatory variables {x1, x2, x3, ..., x}. p} put together, that is, U={y,x1,x2,x3,…,x p}, x p This indicates that there are a total of p explanatory variables.

[0040] 2) Hierarchical clustering;

[0041] The initial p classes need to be clustered layer by layer until all p+1 variables belong to the same class. In the process of moving from one layer to the next (e.g., from p+1 classes to p classes), we have... There are several possibilities, and we need to choose the one that minimizes the decrease in global identity. Assume the current layer has K variable classes, denoted as G1, G2, G3, ..., G... K Global identity is then defined as the latent variable c obtained by each group in that layer. k The sum of correlations between each variable in each corresponding group, where the latent variable is a linear combination c of the variables contained in the corresponding variable group. k =G k v k v k This is the vector of combination coefficients.

[0042] The expression for the global identity criterion T is:

[0043]

[0044] In the formula: |G K |For variable group G K The number of variables contained in it, x j Let j be the j-th explanatory variable.

[0045] It is easy to prove that the latent variable c obtained by maximizing criterion T k It is exactly G K The first principal component obtained from the variables is thus equivalent to maximizing the first eigenvalue of the covariance matrix of each variable group. λ1 is the first eigenvalue. Consider any two variable classes G in the current aggregation layer. A c * / j and G B Then G A With G B During the process of merging into the same category, the decrease in identity is:

[0046]

[0047] Therefore, the division from the current clustering layer to the next clustering layer is based on merging the two clusters corresponding to the minimum value of Δ into one cluster. This reflects the aforementioned approach of maximizing the intra-group correlation of the extracted variable groups and maximizing the contribution of the variable groups to the response variable.

[0048] 3) Determine the optimal polymerization level.

[0049] After obtaining the clustering results for each layer, to determine the optimal clustering level, we need to select the layer with the smallest change in local identity. Consider the variable group containing y in the (l-1)th layer and the lth layer, and denote the corresponding latent variable as c. l-1 and c l The local identity criterion is then defined as:

[0050]

[0051] In the formula: n is the sample size; c * / j For variables in x j The latent variable corresponding to the variable group before merging with y (at level l) (at level l-1); j|al-1,j=1 indicates that x j Belongs to G l-1 The value of j; j|al-1,j=0 means that x j Not belonging to G l-1 The value of j; jalj = 1 means that x j Belongs to G l The value of j.

[0052] Therefore, the rate of change of local identity between two adjacent layers is:

[0053]

[0054] in, For the local identity of the (l-1)th layer, This represents the local identity of the l-th layer.

[0055] It can be seen that the local identity criterion reflects the loss ratio of identity between two adjacent aggregation levels. Given a loss threshold, the optimal aggregation level is defined as: l * =maxl(Γ l ≤γ), where γ is usually set to 5%.

[0056] In this embodiment of the invention, the response variable is determined to be the monthly carbon emissions of the industry, y, and the explanatory variables are {x1, x2, x3, ..., x...}. p This represents monthly electricity consumption data for industry users.

[0057] 102: Use the time-series characteristics of EEMD's electricity consumption data to obtain the regularity, trend, and random components of user-class electricity consumption time-series data;

[0058] Due to the influence of industry scale, seasonal electricity consumption habits, and random fluctuations in electricity consumption, electricity consumption behavior curves vary significantly across industries. Based on the user clustering described above, it is necessary to extract time-series features from user electricity consumption data that is strongly correlated with carbon emissions. This embodiment of the invention utilizes the EEMD method to obtain the overall trend characteristics, seasonal characteristics, and random fluctuation characteristics of electricity consumption curves for various industries.

[0059] This step specifically includes:

[0060] 1) Add white noise;

[0061] Using the EEMD method, different types of white noise need to be added to the initial time series to transform the electricity consumption time series data x k,t It is decomposed into a series of intrinsic mode functions (IMF) components.

[0062]

[0063] In the formula, imf k,i This represents the i-th intrinsic mode component obtained by decomposing the electricity consumption time-series data of the k-th user class using EEMD; r k,n It is the residual component after decomposing the original time series data, where n is the sample size.

[0064] 2) Determine the components.

[0065] EEMD decomposes time series data by connecting local maxima and minima with cubic spline curves to form upper and lower envelopes, and then taking the mean m. j-1 Then from the initial sequence x k,t Subtracting the average of these envelopes from the mean yields the remaining sequence r. k,i Therefore, when calculating each IMF process, we have:

[0066] r k,i =r k,i-1 -imf k,i (6)

[0067] In the formula, when i = 1, r k,0 =x k,t .

[0068] Iterate continuously on the remaining sequence r k,i Extract the (i+1)th IMF until the remaining sequence r k,n Since further decomposition is not possible, the mean of the IMF obtained by adding different white noises is finally calculated. Finally, the regularity, trend, and random components of the user-class electricity consumption time-series data are obtained, P. regular,m ,P trend,m ,P random,m .

[0069]

[0070] In the formula, EEMD represents the EEMD algorithm program, P m This represents the monthly electricity consumption of the user group. That is, using the EEMD algorithm described above, the user group's electricity consumption data is decomposed into three types of time-series variables, P... regular,m ,P trend,m ,P random,m The characteristic is that the sum of the three types of variables equals the electricity consumption data.

[0071] 103: Estimation of carbon emissions from industrial production processes;

[0072] 1) Carbon emission estimation;

[0073] Carbon emissions from industrial production processes mainly refer to the intermediate process carbon emissions generated during the production of industrial products due to the use of catalysts, reactants, etc. According to the IPCC greenhouse gas emissions inventory, carbon emissions from industrial production processes can be calculated based on the corresponding industrial product output data and carbon emission factors.

[0074] Taking the steel industry as an example:

[0075] E Process,steel =∑PO steel ×EF steel (8)

[0076] In the formula, E Process,steel This indicates carbon emissions from the industrial production process in the steel industry; PO steel This indicates the output of major industrial products in the steel industry, such as steel and pig iron; EF steel This indicates the carbon emission factor of the industrial production process corresponding to the product.

[0077] 2) Industrial product output forecast.

[0078] It should be noted that industrial product output data also exhibits a certain lag (typically 1 to 2 months). Therefore, an autoregressive distributed lag model (ARDL) is used to make short-term forecasts of industrial product output.

[0079] y t,γ =a0+a1t+φ1y t-1,γ +φ2y t-2,γ …φ p y t-p,γ +β0x t,γ +β1x t-1 …β q x t-q +u t,γ (9)

[0080] In the formula, yt,γ φ1 represents the output of major industrial products; a0 represents the basic output of major industrial products; a1t represents the rate of change of time on product output; φ1y t-1,γ +φ2y t-2,γ …φ p y t-p,γ This represents the combined impact of historical product output on current product output (autoregressive component); β0x t,γ +β1x t-1 …β q x t-q This represents the combined impact of current and historical electricity consumption on current product output (the lagged effect portion); u t,γ This represents the corresponding random variable.

[0081] 104: Estimates of indirect carbon emissions implied by electricity purchases;

[0082] This step includes:

[0083] 1) Carbon emission estimation;

[0084] Carbon emissions from electricity production primarily originate from the generation side, but the responsibility for these emissions rests with the consumer side. Therefore, when calculating industry energy carbon emissions, the implicit carbon emissions from purchased electricity must be considered.

[0085] E el,m =P m ×EF el,m (10)

[0086] In the formula, E el,m This indicates the carbon emissions inherent in electricity; P m Indicates monthly electricity consumption; EF el,m This indicates the carbon emission factor of electricity in the region for the current month.

[0087] 2) Calculation of carbon emission factors.

[0088] This invention employs a method for calculating the electricity carbon emission factor that considers carbon flow. That is, the carbon flow at the input node equals the carbon flow at the output node. By solving the carbon flow balance equation, the electricity carbon emission factor for each regional power grid can be calculated.

[0089]

[0090] In the formula, EF el,m E represents the regional electricity carbon emission factor for the current month. eg,m This indicates the carbon emissions from electricity generation in the current month; ps m This indicates the monthly electricity consumption in the area. This represents the sum of electricity transferred out of province i; This indicates the carbon emissions implied by the electricity transmitted to province i; EF jThis represents the average carbon emission factor of the power grid in province j.

[0091] 105: Estimation of carbon emissions from fossil fuel consumption;

[0092] This step includes:

[0093] 1) Training dataset setup;

[0094] This invention employs historical fossil fuel consumption, historical electricity consumption characteristics, and historical product output data as training sets. An electricity-energy correlation model is constructed to estimate real-time industry energy consumption data using real-time electricity and output data. Real-time carbon emissions are estimated using the emission factor method.

[0095] 2) Calculate historical carbon emissions using the emission factor method;

[0096] The monthly fossil fuel carbon emissions were calculated using the emission factor method, specifically as follows:

[0097] E i,m =∑F i,m ×EF i (12)

[0098]

[0099] In the formula, E i,m This indicates the carbon emissions from the consumption of different fossil fuels in that month; EF i Indicates the carbon emission factor of different fossil fuels; CC i OF represents the carbon content per unit calorific value of the i-th fossil fuel; i This represents the carbon oxidation rate of the i-th fossil fuel; F represents the ratio of the number of CO2 atoms to the number of C atoms. i,m This indicates the consumption of different fossil fuels in a given month.

[0100] 3) Carbon emission estimation based on PSO-LSTM algorithm.

[0101] The model training uses the PSO-LSTM algorithm, which can comprehensively consider the seasonal characteristics of "electricity-energy" at the same time point in previous years, thereby improving the corresponding prediction accuracy.

[0102] F i,m =PSOLSTM θ (P regular,m ,P trend,m ,P random,m ,PO s,m (14)

[0103] In the formula, F i,m This indicates the estimated consumption of different fossil fuels in the current month; PO s,mPSOLSTM represents monthly product output data for different industries. θ This represents the PSO-LSTM algorithm program.

[0104] The specific process is as follows: Figure 4 As shown.

[0105] Example 2

[0106] The feasibility of the scheme in Example 1 is verified below with specific examples, as detailed in the following description:

[0107] Based on actual data from Tianjin, simulation analysis was conducted on the above embodiments, and the beneficial effects obtained from the embodiments of the present invention are as follows: Figure 5 As shown, the monthly carbon emissions of Tianjin were estimated using the model proposed in this embodiment of the invention. The error rate between the estimated value and the actual value remained at approximately 3%, and the estimation capability remained good even when carbon emission data fluctuated drastically. It can be considered that the estimated values ​​obtained using this embodiment of the invention are sufficiently reasonable, practical, and informative. This helps in analyzing the impact of many unforeseen events on carbon emissions.

[0108] In summary, this invention provides a near-real-time carbon emission estimation method based on supervised clustering and an "electricity-energy-carbon" correlation model. This invention fully utilizes historical information to reflect changes in carbon emissions in a timely manner, monitors the impact of carbon emissions from high-emission industries in real time, and takes corresponding energy-saving and emission-reduction measures to transform high-carbon processes, improve the utilization rate of clean energy, and avoid environmental pollution. For example, the steel industry can carry out short-range steelmaking transformation, i.e., electric arc furnace steelmaking.

[0109] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A near-real-time carbon emission calculation method based on an electricity-energy-carbon correlation model, characterized in that, The method includes the following steps: Carbon emission-related electricity user clustering using SCV-LV, and the clustered user group data processed by EEMD, are used as input variables for the PSO-LSTM model; Use the time-series characteristics of EEMD's electricity consumption data to obtain the regularity, trend, and random components of user-class electricity consumption time-series data; Carbon emissions from industrial production processes, indirect carbon emissions from purchased electricity, and carbon emissions from fossil fuel consumption are estimated based on regularity components, trend components, and random components, respectively. The carbon emission estimation for the industrial production process is achieved by using an autoregressive distributed lag model to make short-term predictions of industrial product output. ； In the formula, Indicates the output of major industrial products; This indicates the basic output of major industrial products; Indicates the rate of impact of time on product output; This indicates the combined impact of historical product output on current product output; This indicates the combined impact of current and historical electricity consumption on current product output; This represents the corresponding random variable; The indirect carbon emissions implied by the purchased electricity are estimated as follows: the carbon flow of the input node is equal to the carbon flow of the output node. The carbon emission factor of the power grid in each region is calculated by solving the carbon flow balance equation. ； In the formula, This indicates the regional electricity carbon emission factor for the current month; This indicates the carbon emissions from electricity generation in the current month; This indicates the monthly electricity consumption in the area. This represents the sum of electricity transferred out of province i; This indicates the carbon emissions implied by the electricity transmitted to province i; This represents the average carbon emission factor of the power grid in province j. The carbon emission estimation of fossil energy consumption includes: carbon emission estimation based on the PSO-LSTM algorithm; By comprehensively considering the seasonal characteristics of electricity and energy at the same time points in previous years, the accuracy of corresponding predictions can be improved. ； In the formula, This represents the estimated consumption of different fossil fuels in the current month; This represents monthly product output data for different industries. This represents the PSO-LSTM algorithm program; Represents the regularity components of user-class electricity consumption time-series data; Indicates a trend component; Represents a random component.

Images