Carbon emission prediction method and related apparatus

Abstract

The present invention is applicable to the technical field of carbon emission monitoring. Provided are a carbon emission prediction method and a related apparatus. The method comprises: acquiring total carbon emission and total power consumption of each industry; calculating a correlation index of the total carbon emission and the total power consumption, so as to screen a target monitored industry; acquiring carbon emission and power consumption of the target monitored industry; on the basis of the carbon emission and the power consumption, calculating an electricity-to-carbon emission transfer coefficient based on the target monitored industry; on the basis of the target monitored industry, matching a target monitored enterprise, and on the basis of randomness, periodicity and climate factors, constructing a power consumption prediction model of the target monitored enterprise; acquiring actual power consumption of the target monitored enterprise and inputting same into the power consumption prediction model for training; matching an outputted predicted power consumption value with the electricity-to-carbon emission transfer coefficient; and generating a predicted carbon emission value. On the basis of the linear relationship between carbon emission and power consumption, the present invention uses power consumption prediction to estimate predicted carbon emission values, thus effectively improving the efficiency and accuracy of carbon emission prediction.

Description

A method and related apparatus for predicting carbon emissions Technical Field

[0001] This invention relates to the field of carbon emission monitoring technology, and in particular to a method and related apparatus for predicting carbon emissions. Background Technology

[0002] Currently, long-term carbon emission monitoring and assessment can provide a scientific basis for low-carbon development. Governments and enterprises can formulate timely and reasonable carbon reduction policies to help enterprises better cope with carbon market challenges and improve their competitiveness. By accurately monitoring and predicting carbon emissions, we can better adapt to climate change and lead green technology innovation and carbon market transformation.

[0003] In existing technologies, carbon emission monitoring technologies are mainly divided into two types: accounting methods and real-time monitoring methods. Accounting methods do not directly measure CO2 emissions but estimate CO2 emissions caused by specific activities based on emission factor methods or material balance methods. Real-time monitoring methods, also known as direct measurement methods, utilize continuous emission monitoring systems (CEMS) to monitor the concentration and flow rate of CO2 emitted from emission sources in real time to obtain instantaneous and continuous CO2 emission data. Real-time monitoring methods determine the total emissions by measuring the flow rate, velocity, and concentration of emitted gases using monitoring instruments or continuous metering equipment. This is divided into manual monitoring and continuous flue gas emission monitoring. The former involves temporary measurements using on-site equipment, while the latter focuses on continuous monitoring of facilities such as thermal power plants to accurately calculate the concentration and emission of carbon dioxide in flue gas. Because carbon emissions exhibit a certain degree of fluctuation, estimations using accounting methods can lead to errors in the predicted values. While real-time monitoring methods have relatively high prediction accuracy, their monitoring efficiency is low when dealing with large amounts of data.

[0004] Summary of the Invention

[0005] This application provides a method and related apparatus for predicting carbon emissions, which addresses the problem of low efficiency in carbon emission monitoring.

[0006] The first aspect of this application provides a method for predicting carbon emissions, including:

[0007] Obtain the total carbon emissions and total electricity consumption of each industry within the first preset time period;

[0008] The target monitoring industries are screened by calculating the correlation index between the total carbon emissions and the total electricity consumption. The correlation index represents the linear correlation between the total carbon emissions and the total electricity consumption.

[0009] The carbon emissions and electricity consumption of the target monitoring industry are obtained within a second preset time period, and the electrocarbon transmission coefficient based on the carbon emissions and electricity consumption is calculated, wherein the electrocarbon transmission coefficient is the carbon emission data expressed in terms of the electricity consumption data of the target monitoring industry.

[0010] A power consumption prediction model for the target monitoring industry is constructed based on randomness, periodicity, and climatic factors.

[0011] The electricity consumption of the target monitoring industry within a third preset time period is obtained and input into the electricity consumption prediction model for training.

[0012] The predicted electricity consumption of the target monitored industry within a target preset time period, output from the electricity consumption prediction model, is matched with the electrocarbon conductivity coefficient.

[0013] Generate a predicted value for the carbon emissions of the target monitoring industry within the target preset time period.

[0014] Furthermore, the step of screening target monitoring industries by calculating the correlation index between total carbon emissions and total electricity consumption includes:

[0015] A carbon-electricity correlation index based on the total carbon emissions and total electricity consumption is constructed using the grey relational analysis method.

[0016] Target monitoring industries are selected based on the characteristics of the aforementioned electrocarbon correlation index.

[0017] Furthermore, the derivation process of the expression for the electrocarbon correlation index is as follows: EM T,s =(em) 1,s ,em 2,s ,……,em T,s EC T,s =(ec 1,s ,ec 2,s ,……,ec T,s );

[0018] Where: EM T,s For the carbon emissions of industry s in year T, em 1,s For the first year's carbon emissions of industry s, em 2,s For the second year's carbon emissions of industry s, em T,s For year T, the carbon emissions of the industry; EC T,s For the electricity consumption data of industry s in year T, ec 1,s For the first year's electricity consumption of the S industry, EC 2,s For the electricity consumption of industry S in the second year, EC T,sLet be the electricity consumption of industry s in year T; y be the standardized series; and ρ be the resolution coefficient. The grey relational coefficient is the coefficient between two sequences.

[0019] Furthermore, the calculation of the electrocarbon transmission coefficient based on the carbon emissions and electricity consumption for the target monitoring industry includes:

[0020] The mathematical model for the multiple linear relationship between carbon emissions and electricity consumption is as follows:

[0021] In the formula: em t,s For industry s, the carbon emissions over time t; ec t,s X represents the electricity consumption of industry s at time t; t,K,s For variables specific to different industries; c0 is the constant term of the model; μ t,s It is an exogenous random perturbation; it follows a normal distribution with an expected value of 0; β k For regression coefficients, coef s The electrical conductivity coefficient of carbon;

[0022] When the industry is set to use no electricity at all, i.e., EC t,s When μ = 0, its carbon emissions also tend to close to 0; the constant term in the baseline model equals 0; μ t,s Exogenous random shocks do not affect the estimated variable ec t,s With variable em t,s If the coefficients are unbiased, then the simplified model is: em t,s =coef s ×ec t,s

[0023] Based on the simplified model, the electrocarbon conductivity coefficient is calculated as the ratio of the mean carbon emission sequence to the mean electricity consumption sequence, expressed as:

[0024] In the formula: coef s em is the electrocarbon conductivity coefficient. t,s For industry s, the carbon emissions over time t; ec t,s The electricity consumption of industry s at time t; The average carbon emissions of industry s in year T; Let be the average electricity consumption series of industry s in year T.

[0025] Furthermore, the electricity consumption prediction model for the target monitoring industry, constructed based on randomness, periodicity, and climatic factors, includes:

[0026] Construct a time-based dummy variable model based on randomness;

[0027] An autoregressive moving average model is constructed based on periodicity;

[0028] A hot and cold diurnal regression model was constructed based on climatic factors;

[0029] The prediction sequences of the time dummy variable model, the autoregressive moving average model, and the hot and cold daily regression model are fitted to generate the electricity consumption prediction model for the target prediction industry.

[0030] Furthermore, the expression for the time dummy variable model is:

[0031] In the formula: ec t,n The periodicity of electricity consumption over time; variable day i The number of days in a week, hour j This represents the number of hours in a day; to avoid perfect collinearity among variables, 6 dummy variables for days and 23 for hours are used. and These are the estimated coefficients for the day and hour variables, respectively; For random interference items;

[0032] The expression for the autoregressive moving average model is: ec t,n =a0+a1ec t―1,n +a2ec t―1,n +…+a p ec t―p,n +ε t +ε t―1 +…+ε t―q ;

[0033] In the formula: ec t,n Let be the electricity consumption sequence of enterprise n at time t; α, p, q are the parameters to be estimated in the model; ε t , ε t―1 ……,ε t―q Used to describe random perturbations in a sequence;

[0034] The expression for the hot and cold daily regression model is:

[0035] Heating temperature for several days:

[0036] Cooling degree days ec t,n =δ1×HDD t +δ2×CDD t +δ3×X t +ε t ;

[0037] In the formula: T iFor real-time temperature; T a The low temperature threshold; T b High temperature threshold; ec t,n For enterprise n, the electricity consumption sequence at time t; HDD t and CDD t These are temperature variables; X t Other control variables related to object n; ε t For random disturbance terms;

[0038] Using the electricity consumption fitted sequences from the time dummy variable model, autoregressive moving average model, and hot / cold daily regression model as explanatory variables, regression analysis was performed with the target electricity consumption to obtain the expression for the overall prediction model:

[0039] In the formula: and These are prediction sequences representing the periodicity, randomness, and climatic influence of electricity forecasts, respectively; ec t,n For electricity consumption prediction sequences; γ and These are the explanatory coefficients and the estimated values, respectively. This is a composite prediction sequence.

[0040] Furthermore, the expression for matching the predicted electricity consumption value of the target monitored industry within the target preset time period, output from the electricity consumption prediction model, with the electrocarbon conductivity coefficient is as follows:

[0041] In the formula: For the predicted carbon emissions sequence; coef s The electrical conductivity coefficient of carbon; This is a composite prediction sequence for electricity consumption.

[0042] Furthermore, generating the predicted carbon emissions of the target monitoring industry within the target preset time period includes:

[0043] Obtain the actual daily carbon emissions of the target monitored industry;

[0044] Randomly select several predicted daily carbon emission data and corresponding actual daily carbon emission data within a fourth preset time period, and calculate the carbon emission prediction sequence and actual carbon emission sequence based on the predicted daily carbon emission data and actual daily carbon emission data.

[0045] The difference between the predicted carbon emission sequence and the actual carbon emission sequence is used to determine whether there is an error in the electricity consumption prediction model. If so, the parameters and variables are dynamically corrected.

[0046] If not, then generate a predicted carbon emission value for the target monitoring industry within the target preset time period.

[0047] Furthermore, the expression for the actual daily carbon emissions is as follows:

[0048] In the formula: em day,n For actual daily carbon emissions; fossil day,n,f For different fossil energy sources, including various industries and n, daily; ef f Carbon emission factor; cem day,n This refers to cement consumption; ef cem For unit emission factor; ec day,n Daily electricity consumption; ce r The carbon emission factor for electricity consumption.

[0049] A second aspect of this application provides a carbon emission prediction device, comprising:

[0050] The acquisition unit is used to acquire the total carbon emissions and total electricity consumption of each industry within a first preset time period;

[0051] A screening unit is used to screen target monitoring industries by calculating the correlation index between the total carbon emissions and the total electricity consumption, wherein the correlation index represents the linear correlation between the total carbon emissions and the total electricity consumption.

[0052] The calculation unit is used to obtain the carbon emissions and electricity consumption of the target monitoring industry within a second preset time period, and to calculate the electrocarbon transmission coefficient based on the carbon emissions and electricity consumption of the target monitoring industry. The electrocarbon transmission coefficient represents the carbon emissions data expressed by the electricity consumption data of the target monitoring industry.

[0053] The construction unit is used to construct a power consumption prediction model for the target monitored industry based on randomness, periodicity, and climatic factors.

[0054] The input unit is used to obtain the electricity consumption of the target monitoring industry within a third preset time period and input it into the electricity consumption prediction model for training.

[0055] The output unit is used to match the predicted electricity consumption value of the target monitored industry within a target preset time period, which is output from the electricity consumption prediction model, with the electrocarbon conductivity coefficient.

[0056] The generation unit is used to generate the predicted carbon emissions of the target monitoring industry within the target preset time period.

[0057] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0058] 1. By constructing an industry-specific correlation index for electricity and carbon emissions, suitable industries for monitoring can be efficiently selected. The index is built based on grey relational analysis and correlation analysis, and comprehensively considers the horizontal characteristics and historical trends of industry electricity and carbon emissions data, thereby ensuring more accurate selection of monitoring industries and improving data accuracy.

[0059] 2. Based on the close relationship between a company's production activities and its electricity consumption, electricity consumption is used as an indicator for monitoring corporate carbon emissions. By establishing a transmission coefficient between electricity and carbon emissions, the transmission coefficients of monitored companies and their corresponding industries were matched, enabling real-time carbon emission monitoring of the monitored entities. This process significantly improves the flexibility and real-time performance of the monitoring system and effectively enhances the accuracy of the monitoring results.

[0060] 3. This invention introduces a time-series composite forecasting model, fully considering various factors such as periodicity, randomness, and climate, thus making short-term emission high-frequency forecasts more comprehensive. This model not only encompasses the establishment of time dummy variable models, time-series moving average autoregressive models, and HDD-CDD models, but also comprehensively forecasts electricity consumption from multiple perspectives, ensuring the comprehensiveness and accuracy of the forecasts. This effectively improves the accuracy of electricity consumption forecasts while simultaneously increasing the accuracy of carbon emission forecasts.

[0061] 4. This invention scientifically evaluates the errors between the monitoring and prediction models using paired sample mean tests and bias probability assessment methods to monitor and predict errors, and dynamically adjusts the model's settings based on this assessment. This method ensures the timeliness and accuracy of the model parameters, further improving the robustness and data accuracy of the entire system. Attached Figure Description

[0062] Figure 1 is a schematic flowchart of an embodiment of a carbon emission prediction method provided by the present invention;

[0063] Figure 2 is a schematic flowchart of another embodiment of a carbon emission prediction method provided by the present invention.

[0064] Figure 3 is a schematic diagram of the industry identification standard for electrical carbon correlation in this invention;

[0065] Figure 4 is a schematic flowchart of another embodiment of a carbon emission prediction method provided by the present invention.

[0066] Figure 5 is a schematic flowchart of another embodiment of a carbon emission prediction method provided by the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] The carbon emission prediction method in this embodiment aims to improve the efficiency and accuracy of carbon emission prediction. This method can be implemented in a system, on a server, or on a terminal; no specific limitation is made.

[0069] Example 1

[0070] Please refer to Figure 1. An embodiment of a carbon emission prediction method includes the following steps:

[0071] S11. Obtain the total carbon emissions and total electricity consumption of each industry within the first preset time period;

[0072] In this embodiment, the first preset time refers to the period within which data is collected, and the specific time is not limited. First, the total carbon emissions of various industries by energy category are collected. The carbon emissions of each industry consist of three parts: carbon emissions from fossil fuel use, industrial processes, and electricity consumption. Carbon emissions from fossil fuel use and industrial processes are considered direct carbon emissions; carbon emissions from electricity consumption are considered indirect carbon emissions. Summing up the direct and indirect carbon emissions of an industry yields the total emissions of industry s in year t, as expressed below:

[0073] In the formula: em t,s Let be the total carbon emissions of industry s in year t; Direct carbon emissions; This refers to indirect carbon emissions.

[0074] Specifically, direct carbon emissions include carbon emissions from the combustion of fossil fuels in production processes and carbon emissions from industrial processes, primarily cement production. The data for each industry and its fossil fuel sources in year t is as follows: t,s,f Its carbon emission factor ef f Carbon emission factor ef obtained through interactive calculation f The lower heating value v of the f-th fuel f Carbon content per unit calorific value f f Oxidation rate o f This is related to the C-CO2 conversion coefficient of 44 / 12. Industrial process carbon emissions are determined by cement consumption (cem). t,s and unit emission factor ef cem The direct carbon emissions from electricity generated by industry s in year t are obtained by multiplication. The expression is as follows:

[0075] Indirect carbon emissions from electricity are determined by the industry's total electricity consumption over a corresponding period (ec). t,s The carbon emission factor ce of electricity consumption in the region r in year t t,r Multiply to obtain the indirect carbon emissions from electricity in industry s in year t. The expression is as follows:

[0076] The physical quantities in the formula have been expressed as described above.

[0077] In this embodiment, the total electricity consumption of each industry can be collected from the electricity system database or obtained from other systems. Electricity consumption is used as the basic data. The method of acquisition is not limited here, but data preprocessing is required based on the acquired initial data. The method of preprocessing is also not limited.

[0078] S12. Target monitoring industries are screened by calculating the correlation index between total carbon emissions and total electricity consumption. The correlation index represents the linear correlation between total carbon emissions and total electricity consumption.

[0079] In this embodiment, the grey relational analysis method is first used to analyze the linear relationship between the industry's historical total carbon emissions and total electricity consumption. Then, the Pearson correlation test is used to calculate the linear correlation between total carbon emissions and total electricity consumption to characterize the degree of correlation. The purpose of screening target monitoring industries by calculating the correlation index between total carbon emissions and total electricity consumption is to reduce the interference of unnecessary data. Based on the linear correlation between carbon emissions and electricity consumption, carbon emissions can be predicted by calculating electricity consumption.

[0080] S13. Obtain the carbon emissions and electricity consumption of the target monitoring industry within a second preset time period, and calculate the electrocarbon transmission coefficient based on the carbon emissions and electricity consumption of the target monitoring industry. The electrocarbon transmission coefficient is the carbon emission data expressed in terms of the electricity consumption data of the target monitoring industry.

[0081] After collecting the carbon emissions and electricity consumption of the target industry over a certain period of time, the electrocarbon transmission coefficient of the target industry is calculated. The electrocarbon transmission coefficient is expressed as follows: the carbon emission data is represented by the industry's electricity consumption data. In this way, the carbon emission data can be estimated by calculating the electricity consumption data. The acquisition and calculation based on electricity consumption data is faster and simpler than directly calculating carbon emissions, which can improve the efficiency of the final prediction.

[0082] S14. Match target monitoring enterprises with target monitoring industries, and construct electricity consumption prediction models for target monitoring enterprises based on randomness, periodicity and climate factors;

[0083] Based on the industry's carbon emission transmission coefficient calculated in step S13, the electricity consumption data of monitored enterprises is used as a proxy variable to generate enterprise carbon emission sequences. It is necessary to match the industry category to which the enterprises belong, as each industry has a different carbon emission transmission coefficient. Therefore, this invention uses enterprise carbon emissions to reflect the industry's carbon emission situation.

[0084] First, short-term electricity consumption of the target monitored enterprises is predicted using a time-based dummy variable model, an autoregressive moving average model (ARMA), and a hot-cold daily regression model (HDD-CDD). These models are used to fit the periodicity, randomness, and climatic influence of high-frequency electricity consumption, respectively. Finally, a composite electricity consumption prediction model is formed by fitting the prediction sequences from these regression models. The overall prediction model is then fitted again based on the models for randomness, periodicity, and climatic influence, resulting in higher accuracy of the electricity consumption prediction values.

[0085] S15. Obtain the electricity consumption of the target monitored enterprise within the third preset time period and input it into the electricity consumption prediction model for training;

[0086] In this embodiment, after determining the overall electricity consumption prediction model, the actual electricity consumption of the target monitored enterprise within a certain period is collected. The method for collecting electricity consumption data is similar to that in step S11 and will not be described in detail here. Furthermore, by inputting the data into the prediction model, the predicted electricity consumption value within a certain prediction period can be obtained.

[0087] S16. Match the predicted electricity consumption of the target monitored enterprise within the target preset time period output from the electricity consumption prediction model with the carbon conductivity coefficient.

[0088] In this embodiment, the electrocarbon conductivity coefficient (coef) of industry s is constructed. s Using the high-frequency electricity consumption data of the monitored enterprise n, ec t,n High-frequency carbon emission sequences from enterprise point sources are generated as proxy variables. First, the industry category *s* to which enterprise *n* belongs is matched, as each industry has a different electrocarbon conductivity coefficient *coef*. s Secondly, the corresponding industry's electrocarbon conductivity coefficient (coef) s Electricity consumption data of enterprises t,n Multiply to obtain the enterprise-level carbon emission sequence em t,n The expression for the electrocarbon conductivity coefficient is:

[0089] The predicted value of carbon emissions can be estimated from the predicted value of electricity consumption obtained from the above expression.

[0090] S17. Generate the predicted carbon emissions of the target monitoring enterprise within the target preset time period.

[0091] In this embodiment, the high-frequency carbon emission data of enterprises exhibits high-frequency fluctuation characteristics as it changes with the enterprise's electricity consumption data. Therefore, after determining the high-frequency carbon emission amount based on the carbon conduction coefficient, since emission activities are related to enterprise production activities and have periodicity over time, a measurement model including time dummy variables is constructed to fit the high-frequency carbon emission em. t,n The expression for how it changes over time is as follows: day j =1·{t∈i}; hour j =1·{t=j};

[0092] In the formula, the variable day i Indicates the number of days in a week, hour j This represents the number of hours in a day; to avoid perfect collinearity among variables, 6 dummy variables for days and 23 for hours are used. and Let represent the estimated coefficients of the day and hour variables, respectively. For random interference items; in addition, 1·{…} is an indicator function, which takes the value of 1 if the condition in the compound parentheses is met, and takes the value of 0 otherwise.

[0093] Based on the above optimization of carbon emission predictions, the flexibility and real-time performance of carbon emission monitoring have been improved, and the monitoring results have become more accurate.

[0094] Example 2

[0095] Please refer to Figure 2. The method of screening target monitoring industries by calculating the correlation index between total carbon emissions and total electricity consumption in this invention also includes the following steps:

[0096] S121. Construct an electricity-carbon correlation index based on total carbon emissions and total electricity consumption using the grey relational analysis method;

[0097] S122. Select target monitoring industries based on the characteristics of the electrocarbon correlation index;

[0098] Specifically, the derivation process of the expression for the electrocarbon correlation index is as follows: EM T,s =(em) 1,s ,em 2,s ,……,em T,s EC T,s =(ec 1,s ,ec 2,s ,……,ec T,s );

[0099] Where: EM T,s For the carbon emissions of industry s in year T, em 1,s For the first year's carbon emissions of industry s, em 2,s For the second year's carbon emissions of industry s, em T,s For year T, the carbon emissions of the industry; EC T,s For the electricity consumption data of industry s in year T, ec 1,s For the first year's electricity consumption of the S industry, EC 2,s For the electricity consumption of industry S in the second year, EC T,s Let be the electricity consumption of industry s in year T; y be the standardized series; and ρ be the resolution coefficient. The grey relational coefficient is the coefficient between two sequences.

[0100] Simultaneously, Pearson correlation test was used to calculate and obtain EM. T,s With EC T,s The linear correlation between them is used to characterize the degree of association. The mean of the carbon emission series is calculated. and electricity consumption sequence After taking the mean, the correlation coefficient can be further calculated. The expression is as follows:

[0101] grey Correlation coefficient and Pearson correlation coefficient Both are positively correlated with the degree of correlation between sequences. In this embodiment, the above two indicators are used as correlation indicators for electricity data and carbon emission data.

[0102] Based on the time-varying trend and magnitude of the grey relational index, industries with high correlation to carbon emissions and potential feasible monitoring industries are identified. For correlation trend identification, a difference sequence Δξ of the correlation is constructed. T,s Used to determine, Δξ T,s A value greater than 0 indicates that the correlation between electricity and carbon emissions in industry s shows an increasing trend between two adjacent time periods, suggesting potential cost-monitoring targets; while Δξ T,s A value less than 0 indicates that the correlation between electricity and carbon emissions in industry s decreases over a short period, suggesting a trend of decoupling between electricity and carbon emissions, which is detrimental to the application of electricity-carbon monitoring technology. The expression is as follows:

[0103] Δξ T,s =ξ T,s ―ξ T―1,s ;Δξ T,s >0 increases with adjacent time intervals; Δξ T,s <0 indicates a decrease in adjacent time intervals;

[0104] Based on the distribution range of correlation indicators, a threshold is set to screen industries with high electricity-carbon correlation. Two types of correlation indicators are constructed based on grey correlation and Pearson correlation coefficient, with a distribution range of 0 to 1. The closer to 1, the stronger the correlation between electricity and carbon in the industry; 0.8 to 1 indicates extremely strong correlation between sequences; 0.6 to 0.8 indicates strong correlation; 0.4 to 0.6 indicates moderate correlation; and below 0.4 indicates weak correlation between sequences. In this embodiment, θ = 0.7 is set as the threshold to determine whether the carbon emissions and electricity consumption of an industry have a strong correlation. Combining the trend identification of correlation, this invention identifies feasible monitoring objects with high electricity-carbon correlation based on two-dimensional criteria, as shown in Figure 3.

[0105] Example 3

[0106] The present invention calculates the electrocarbon conductivity coefficient based on carbon emissions and electricity consumption for the target monitoring industry, and also includes the following steps:

[0107] A multiple linear equation is constructed to correlate industry carbon emissions with, but not limited to, electricity consumption, to assess the linear relationship between changes in electricity consumption and total carbon emissions. The total carbon emissions of industry *s* are the explanatory variable in the econometric model. In the baseline mathematical model specification, the regression model variables only include electricity consumption. Variable X t,K,s c0 is used as a dynamic adjustment variable for the model, and is set to 0 in the baseline case; based on the error assessment of the model in subsequent steps, the variable X is further adjusted. t,K,s Adjustments and corrections are made to c0 to improve the model's fit.

[0108] The mathematical model for the multiple linear relationship between carbon emissions and electricity consumption is as follows:

[0109] In the formula: em t,s For industry s, the carbon emissions over time t; ec t,s X represents the electricity consumption of industry s at time t; t,K,s For variables specific to different industries; c0 is the constant term of the model; μ t,s It is an exogenous random perturbation; it follows a normal distribution with an expected value of 0; β k For regression coefficients, coef s The electrical conductivity coefficient of carbon;

[0110] When the industry is set to use no electricity at all, i.e., EC t,s When μ = 0, its carbon emissions also tend to close to 0; the constant term in the baseline model equals 0; μ t,s Exogenous random shocks do not affect the estimated variable ec t,s With variable em t,s The coefficients are unbiased at time t; therefore, the simplified model is: em t,s =coef s ×ect,s

[0111] Based on the simplified model, the electrocarbon conductivity coefficient is calculated as the ratio of the mean carbon emission sequence to the mean electricity consumption sequence, expressed as:

[0112] In the formula: coef s em is the electrocarbon conductivity coefficient. t,s For industry s, the carbon emissions over time t; ec t,s The electricity consumption of industry s at time t; The average carbon emissions of industry s in year T; Let be the average electricity consumption series of industry s in year T.

[0113] Electrocarbon conductivity coefficient (coef) s This represents the total carbon emissions per unit of electricity used by the monitored industry, including direct carbon emissions. and indirect carbon emissions The carbon conductivity coefficient is different from the carbon emission coefficient (ce) of electricity consumption. t,r The latter only represents the carbon emissions generated by electricity generation per unit of electricity consumed.

[0114] Example 4

[0115] Please refer to Figure 4. The electricity consumption prediction model for the target monitoring enterprise constructed in this invention based on randomness, periodicity, and climatic factors also includes the following steps:

[0116] S141. Constructing a time-based dummy variable model based on randomness;

[0117] A time-based dummy model is constructed to explain the periodic variations in high-frequency electricity consumption. A multiple regression model is established using time-based dummy variables to describe the electricity consumption (ec) t,n Periodic characteristics that change over time.

[0118] The expression for the time dummy variable model is:

[0119] In the formula: ec t,n The periodicity of electricity consumption over time; variable day i The number of days in a week, hour j This represents the number of hours in a day; to avoid perfect collinearity among variables, 6 dummy variables for days and 23 for hours are used. and These are the estimated coefficients for the day and hour variables, respectively; This is a random interference item.

[0120] Replace the parameters in the time dummy variable model with Substituting the explanatory variables back into the data allows us to obtain the fitted sequence of the electrical energy of the monitored object n. and The expression is as follows:

[0121] The physical quantities in the formula have the same meaning as those in the expression of the time dummy variable model mentioned above.

[0122] S142. Construct an autoregressive moving average model based on periodicity;

[0123] An ARMA(p,q) model is constructed to explain the randomness of high-frequency electricity. Real-time electricity is correlated with electricity in the previous stage; therefore, the electricity level at time t (ec) is constructed. t,n Compared to the previous battery level (EC) t―1,n , ec t―1,n , ...,ec t―p,n The model is a regression model. Simultaneously, the model is affected by a previous random shock q; p and q represent the optimal lag order, and the parameters to be estimated in the model are α, p, and q. The time accuracy of the model's prediction depends on the time scale of the data t, and is consistent with the time scale of the electricity data. That is, the expression for the autoregressive moving average model is: ec t,n =a0+a1ec t―1,n +a2ec t―1,n +…+a p ec t―p,n +ε t +ε t―1 +…+ε t―q ;

[0124] In the formula: ec t,n Let be the electricity consumption sequence of enterprise n at time t; α, p, q are the parameters to be estimated in the model; ε t , ε t―1 ……,ε t―q Used to describe random perturbations in a sequence.

[0125] Obtain the estimated values ​​of the parameters to be estimated in the ARMA model. Then, by substituting the explanatory variables back, the fitted electricity sequence can be obtained, as shown in the following expression:

[0126] The physical quantities in the formula have the same meaning as those in the expression of the autoregressive moving average model mentioned above.

[0127] S143. Construct a hot-cold diurnal regression model based on climatic factors;

[0128] A HDD-CDD model is constructed to explain the variations in electricity consumption caused by climatic factors. Heating demand days (HDD) and cooling demand days (CDD) are used, with the former representing a fixed amount of heating demand and the latter representing a fixed amount of cooling demand. The expression for the hot / cold day regression model is as follows:

[0129] Heating temperature for several days:

[0130] Cooling degree days ec t,n =δ1×HDD t +δ2×CDD t +δ3×X t +ε t ;

[0131] In the formula: T i For real-time temperature; T a The low temperature threshold; T b High temperature threshold; ec t,n For enterprise n, the electricity consumption sequence at time t; HDD t and CDD t These are temperature variables; X t Other control variables related to object n; ε t This is a random disturbance term.

[0132] The model measures the electricity consumption (ec) of monitored enterprise n. t,n With time-varying HDD t CDD t Correlation between indicators; X t This represents other control variables concerning firm n. The coefficients are estimated based on the HDD-CDD model. Substituting back the explanatory variables yields the fitted sequence of electricity. The expression is as follows:

[0133] The physical quantities in the formula have the same meaning as those in the expression of the above-mentioned hot-cold daily regression model.

[0134] S144. Fit the prediction sequences of the time dummy variable model, the autoregressive moving average model, and the hot and cold daily regression model to generate the electricity consumption prediction model for the target enterprise.

[0135] Predicted sequence and These models respectively characterize the periodicity, randomness, and climatic influence of electricity generation forecasts. Combining the fitted sequences from the three highly predictive models can further improve the accuracy of electricity generation forecasts. The three fitted sequences are then used as explanatory variables in a regression analysis with actual electricity consumption to obtain estimates of the explanatory coefficients γ for each variable. The composite fitting sequence can be further obtained. The expression is as follows:

[0136] In the formula: and These are prediction sequences representing the periodicity, randomness, and climatic influence of electricity forecasts, respectively; ec t,n For electricity consumption prediction sequences; γ and These are the explanatory coefficients and the estimated values, respectively. This is a composite prediction sequence.

[0137] Finally, the composite prediction sequence The electrocarbon conductivity coefficient (coef) of the monitored object n s By matching and interacting, a high-frequency prediction sequence of short-term carbon emissions can be obtained, as shown in the following expression:

[0138] In the formula: For the predicted carbon emissions sequence; coef s The electrical conductivity coefficient of carbon; This is a composite prediction sequence for electricity consumption.

[0139] Example 5

[0140] Please refer to Figure 5. The method for generating the predicted carbon emissions of the target monitoring enterprise within the target preset time period in this invention also includes the following steps:

[0141] S171. Obtain the actual daily carbon emissions of the target monitored enterprises;

[0142] S172. Randomly select several predicted daily carbon emission data and corresponding actual daily carbon emission data within a fourth preset time period, and calculate the carbon emission prediction sequence and actual carbon emission sequence based on the daily predicted carbon emission data and actual daily carbon emission data.

[0143] S173. Determine whether there is an error in the electricity consumption prediction model based on the difference between the carbon emission prediction sequence and the actual carbon emission sequence;

[0144] S174. When there are errors in the electricity consumption prediction model, the parameters and variables are dynamically corrected.

[0145] S175. When there is no error in the electricity consumption prediction model, the predicted carbon emissions of the target monitoring enterprise within the target preset time period are generated.

[0146] A method for assessing the error of a monitoring model is established based on paired-samples t-tests. First, high-frequency monitoring data of monitoring object n are aggregated. To daily level And calculate the actual daily carbon emission level of object n, as shown in the following expression:

[0147] In the formula: em day,n For actual daily carbon emissions; fossil day,n,f For different fossil energy sources, including various industries and n, daily; ef f Carbon emission factor; cem day,n This refers to cement consumption; ef cem For unit emission factor; ec day,n Daily electricity consumption; ce r The carbon emission factor for electricity consumption.

[0148] An emission monitoring sequence is constructed by randomly selecting N daily monitoring data points and corresponding actual emission data within a sample period. and actual emission sequence EM day,n And perform a paired sample mean t-test on the two sequences, as shown in the following expression:

[0149] In the formula: D day To monitor the difference between the sequence and the actual sequence; the mean of the difference sequence is calculated. and variance s D The t-statistic can then be obtained. The p-value corresponding to the t-statistic distribution can be used to determine whether there is a significant difference between the monitored sequence and the actual sequence. This invention sets the threshold to 0.1. If the p-value is below 0.1, it is considered that there is a significant difference between the monitored sequence and the actual sequence, and the model will return to step 21 for dynamic adjustment of parameters and variables.

[0150] An error assessment method for establishing an electricity consumption prediction model based on sequence trend deviation assessment. First, a deviation assessment range is constructed based on the actual electricity consumption sequence of monitored enterprise n. Thresholds a1, a2, a3, a4 (a1 < a2 < 1 < a3 < a4) are set to define the threshold range, as shown in the following expression:

[0151] Among them, if the predicted power consumption satisfy If the predicted electricity volume is high, then the prediction accuracy is considered high; satisfy If the predicted electricity is tolerable, then the prediction error is considered acceptable; if the predicted electricity is... or The prediction error is considered large. Since electricity consumption is a continuous time series, relying on the electricity consumption at a single point in time is insufficient to determine the model's prediction accuracy. Therefore, this invention uses the probability distribution of the full-time sample to determine the probability that the predicted electricity consumption sequence exceeds a critical tolerable range, thereby identifying the model's prediction error. The expression is as follows:

[0152] When probability P a and P b Simultaneously satisfy P a >90% and P b If P > 95%, the model is considered to have high prediction accuracy, and the carbon emission prediction calculation in the above steps can be continued; if P a >90% and P b If any condition is violated in more than 95% of cases, return to the above steps to dynamically correct the composite prediction model.

[0153] In this embodiment, the paired sample mean test method and the deviation probability assessment method for monitoring and prediction errors are used to evaluate the error between the monitoring model and the prediction model, and the model setting parameters are dynamically adjusted based on this, which effectively improves the accuracy of the prediction model.

[0154] Example 6

[0155] One embodiment of a carbon emission prediction device includes the following steps:

[0156] The total carbon emissions and total electricity consumption acquisition unit 101 is used to acquire the total carbon emissions and total electricity consumption of each industry within a first preset time period.

[0157] The target monitoring industry screening unit 102 is used to screen target monitoring industries by calculating the correlation index between total carbon emissions and total electricity consumption. The correlation index represents the linear correlation between total carbon emissions and total electricity consumption.

[0158] The electrocarbon conduction coefficient calculation unit 103 is used to obtain the carbon emissions and electricity consumption of the target monitoring industry within a second preset time period, and calculate the electrocarbon conduction coefficient based on the carbon emissions and electricity consumption of the target monitoring industry. The electrocarbon conduction coefficient represents the carbon emissions data using the electricity consumption data of the target monitoring industry.

[0159] Electricity consumption prediction model building unit 104 is used to match target monitoring enterprises according to the target monitoring industry, and to build electricity consumption prediction models for target monitoring enterprises based on randomness, periodicity and climate factors.

[0160] The electricity consumption input unit 105 of the target monitored enterprise is used to acquire the electricity consumption of the target monitored enterprise within a third preset time period and input it into the electricity consumption prediction model for training.

[0161] The target electricity consumption prediction output unit 106 is used to match the electricity consumption prediction value of the target monitored enterprise within the target preset time period output from the electricity consumption prediction model with the carbon conductivity coefficient.

[0162] The carbon emission prediction generation unit 107 is used to generate the predicted carbon emission values ​​of the target monitored enterprises within the target preset time period.

[0163] The specific functions and uses of the units in this embodiment are similar to the steps in the aforementioned embodiments one to six, and will not be repeated here.

[0164] It is understood that those skilled in the art can combine various implementation methods in the above embodiments under the guidance of the above examples to obtain technical solutions with multiple implementation methods.

[0165] 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, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting carbon emissions, characterized in that, include: Obtain the total carbon emissions and total electricity consumption of each industry within the first preset time period; The target monitoring industries are screened by calculating the correlation index between the total carbon emissions and the total electricity consumption. The correlation index represents the linear correlation between the total carbon emissions and the total electricity consumption. The carbon emissions and electricity consumption of the target monitoring industry are obtained within a second preset time period, and the electrocarbon transmission coefficient based on the carbon emissions and electricity consumption is calculated, wherein the electrocarbon transmission coefficient is the carbon emission data expressed in terms of the electricity consumption data of the target monitoring industry. Based on the target monitoring industry, target monitoring enterprises are matched, and a power consumption prediction model for the target monitoring enterprises is constructed based on randomness, periodicity, and climate factors. The electricity consumption of the target monitored enterprise within a third preset time period is obtained and input into the electricity consumption prediction model for training; The predicted electricity consumption of the target monitored enterprise within the target preset time period, output from the electricity consumption prediction model, is matched with the electrocarbon conductivity coefficient. Generate a predicted value for the carbon emissions of the target monitored enterprise within the target preset time period.

2. The carbon emission prediction method according to claim 1, characterized in that, The method of screening target monitoring industries by calculating the correlation index between total carbon emissions and total electricity consumption includes: A carbon-electricity correlation index based on the total carbon emissions and total electricity consumption is constructed using the grey relational analysis method. Target monitoring industries are selected based on the characteristics of the aforementioned electrocarbon correlation index.

3. The carbon emission prediction method according to claim 2, characterized in that, The derivation process of the expression for the electrocarbon correlation index is as follows: EM T,s =(em) 1,s ,em 2,s ,……,em T,s EC T,s =(ec 1,s ,ec 2,s ,……,ec T,s ); Where: EM T,s For the carbon emissions of industry s in year T, em 1,s For the first year's carbon emissions of industry s, em 2,s For the second year's carbon emissions of industry s, em T,s For year T, the carbon emissions of the industry; EC T,s For the electricity consumption data of industry s in year T, ec 1,s For the first year's electricity consumption of the S industry, EC 2,s For the electricity consumption of industry S in the second year, EC T,s Let be the electricity consumption of industry s in year T; y be the standardized series; and ρ be the resolution coefficient. The grey relational coefficient is the coefficient between two sequences.

4. The carbon emission prediction method according to claim 1, characterized in that, The calculation of the electrocarbon transmission coefficient based on the carbon emissions and electricity consumption for the target monitoring industry includes: The mathematical model for the multiple linear relationship between carbon emissions and electricity consumption is as follows: In the formula: em t,s For industry s, the carbon emissions over time t; ec t,s X represents the electricity consumption of industry s at time t; t,K,s For variables specific to different industries; c0 is the constant term of the model; μ t,s It is an exogenous random perturbation; it follows a normal distribution with an expected value of 0; β k For regression coefficients, coef s The electrical conductivity coefficient of carbon; When the industry is set to use no electricity at all, i.e., EC t,s When μ = 0, its carbon emissions also tend to close to 0; the constant term in the baseline model equals 0; μ t,s Exogenous random shocks do not affect the estimated variable ec t,s With variable em t,s If the coefficients are unbiased, then the simplified model is: Based on the simplified model, the electrocarbon conductivity coefficient is calculated as the ratio of the mean carbon emission sequence to the mean electricity consumption sequence, expressed as: In the formula: coef s em is the electrocarbon conductivity coefficient. t,s For industry s, the carbon emissions over time t; ec t,s The electricity consumption of industry s at time t; The average carbon emissions of industry s in year T; Let be the average electricity consumption series of industry s in year T.

5. The carbon emission prediction method according to claim 1, characterized in that, The electricity consumption prediction model for the target monitored enterprises, constructed based on randomness, periodicity, and climatic factors, includes: Construct a time-based dummy variable model based on randomness; An autoregressive moving average model is constructed based on periodicity; A hot and cold diurnal regression model was constructed based on climatic factors; The prediction sequences of the time dummy variable model, the autoregressive moving average model, and the hot and cold daily regression model are fitted to generate the electricity consumption prediction model for the target enterprise.

6. The carbon emission prediction method according to claim 5, characterized in that, The expression for the time dummy variable model is: In the formula: ec t,n This refers to the periodic characteristics of electricity consumption over time. variable day i The number of days in a week, hour j This represents the number of hours in a day; to avoid perfect collinearity among variables, 6 dummy variables for days and 23 for hours are used. and These are the estimated coefficients for the day and hour variables, respectively; For random interference items; The expression for the autoregressive moving average model is: at t,n <a0+a1ec t―1,n +a2ec t―1,n +…+a p at t―p,n +ε t +ε t―1 +…+ε t―q 4 In the formula: ec t,n Let be the electricity consumption sequence of enterprise n at time t; α, p, q are the parameters to be estimated in the model; ε t , ε t―1 ……,ε t―q Used to describe random perturbations in a sequence; The expression for the hot and cold daily regression model is: Heating temperature for several days: Refrigeration for several days at t,n (δ1×HDD t +δ2×CDD t +δ3×X t +ε t 4 In the formula: T i For real-time temperature; T a The low temperature threshold; T b High temperature threshold; ec t,n For enterprise n, the electricity consumption sequence at time t; HDD t and CDD t These are temperature variables; X t Other control variables related to object n; ε t For random disturbance terms; Using the electricity consumption fitted sequences from the time dummy variable model, autoregressive moving average model, and hot / cold daily regression model as explanatory variables, regression analysis was performed with the target electricity consumption to obtain the expression for the overall prediction model: In the formula: and These are prediction sequences representing the periodicity, randomness, and climatic influence of electricity forecasts, respectively; ec t,n For electricity consumption prediction sequences; γ and These are the explanatory coefficients and the estimated values, respectively. This is a composite prediction sequence.

7. The carbon emission prediction method according to claim 1, characterized in that, The expression for matching the predicted electricity consumption value of the target monitored enterprise within the target preset time period, output from the electricity consumption prediction model, with the electrocarbon conductivity coefficient is as follows: In the formula: For the predicted carbon emissions sequence; coef s The electrical conductivity coefficient of carbon; This is a composite prediction sequence for electricity consumption.

8. The carbon emission prediction method according to claim 1, characterized in that, The process of generating the predicted carbon emissions of the target monitoring enterprise within the target preset time period includes: Obtain the actual daily carbon emissions of the target monitored enterprise; Randomly select several predicted daily carbon emission data and corresponding actual daily carbon emission data within a fourth preset time period, and calculate the carbon emission prediction sequence and actual carbon emission sequence based on the predicted daily carbon emission data and actual daily carbon emission data. The difference between the predicted carbon emission sequence and the actual carbon emission sequence is used to determine whether the electricity consumption prediction model has any errors. If the result is poor, then the parameters and variables will be dynamically adjusted. If not, then generate a predicted carbon emission value for the target monitored enterprise within the target preset time period.

9. The carbon emission prediction method according to claim 8, characterized in that, The expression for the actual daily carbon emissions is: In the formula: em day,n For the actual daily carbon emissions of each industry (n); fossil day,n,f For different fossil energy sources, including various industries and n, daily; ef f Carbon emission factor; cem day,n This refers to cement consumption; ef cem Unit emission factor; ec day,n Daily electricity consumption; ce r The carbon emission factor for electricity consumption.

10. A carbon emission prediction device, characterized in that, include: The carbon emission and electricity consumption acquisition unit is used to acquire the total carbon emission and electricity consumption of each industry within a first preset time period. The target monitoring industry screening unit is used to screen target monitoring industries by calculating the correlation index between the total carbon emissions and the total electricity consumption, wherein the correlation index represents the linear correlation between the total carbon emissions and the total electricity consumption. The electrocarbon conductivity calculation unit is used to obtain the carbon emissions and electricity consumption of the target monitoring industry within a second preset time period, and to calculate the electrocarbon conductivity based on the carbon emissions and electricity consumption of the target monitoring industry. The electrocarbon conductivity represents the carbon emissions data expressed by the electricity consumption data of the target monitoring industry. The electricity consumption prediction model building unit matches target monitoring enterprises according to the target monitoring industry and is used to build an electricity consumption prediction model for the target monitoring enterprises based on randomness, periodicity and climate factors. The electricity consumption input unit of the target monitored enterprise is used to obtain the electricity consumption of the target monitored enterprise within a third preset time period and input it into the electricity consumption prediction model for training. The target electricity consumption prediction output unit is used to match the electricity consumption prediction value of the target monitored enterprise within the target preset time period output from the electricity consumption prediction model with the electrocarbon conductivity coefficient. A carbon emission prediction generation unit is used to generate a carbon emission prediction value for the target monitoring enterprise within the target preset time period.

Images