A method for predicting long-term carbon emissions in energy-consuming areas

Abstract

The present invention provides a method for predicting long-term carbon emissions in an energy-consuming area, which collects macro data and micro data of the energy-consuming area, and performs the following steps: first establish a macro prediction model, then establish a micro prediction model, and then establish a combined prediction model, establish an optimal solution model for two weights in the combined prediction model, iteratively solve the optimal solution model to obtain the optimal two weight values, and correct the combined prediction model to obtain the final combined prediction model, and substitute the final combined prediction model into MATLAB software for iterative solution to obtain the predicted value of long-term carbon emissions in the energy-consuming area. Since the present invention reasonably combines the macro prediction model and the micro prediction model, and at the same time adopts a special model structure for the weights of the macro prediction model and the micro prediction model, it is possible to iteratively solve the optimal weights to respectively exert the effects of the macro prediction model and the micro prediction model, so that the results of the long-term carbon emissions prediction of the energy-consuming area are more accurate.

Description

Technical Field

[0001] The present invention specifically relates to a method for predicting long-term carbon emissions in an energy-consuming area, and belongs to the technical field of carbon emission prediction. Background Art

[0002] The scientific formulation of carbon emission reduction plans is to obtain future carbon emission trends through long-term carbon emission forecasts and then formulate carbon emission reduction plans. Therefore, whether the long-term carbon emission forecast is accurate and timely has become the core of the scientific formulation of carbon emission reduction plans. In the prior art, long-term carbon emission forecasts are mainly achieved through two models, namely macro forecast models and micro forecast models. Among them, the macro forecast model collects macro data of energy-consuming areas (annual changes in regional permanent population, annual changes in regional permanent population, and annual changes in energy intensity, etc.) to predict the long-term carbon emissions of energy-consuming areas; the micro forecast model collects micro data of energy-consuming areas (daily electricity and energy consumption of all industries in the energy-consuming area, etc.), first predicts the short-term carbon emissions of the energy-consuming area, and finally fits the short-term carbon emission forecast results to obtain the long-term carbon emission forecast results of the energy-consuming area. Although the macro prediction model in the existing technology can reveal the long-term trend of carbon emissions in energy-consuming areas, it needs to be fitted based on the official data released in the past, which has a lag and cannot make timely long-term carbon emission predictions for energy-consuming areas. Although the micro prediction model can perform detailed simulations of energy-consuming areas and make timely and accurate predictions of short-term carbon emissions in energy-consuming areas, the micro model prediction is not accurate enough in predicting long-term carbon emissions in energy-consuming areas due to the large amount of engineering work and high fitting difficulty in long-term predictions. Summary of the invention

[0003] The technical problem to be solved by the present invention is: how to reasonably combine the macro prediction model and the micro prediction model in the long-term carbon emission prediction to make the long-term carbon emission prediction results of the energy consumption area more accurate.

[0004] The technical solution proposed in the present invention is: a method for predicting long-term carbon emissions in an energy-consuming area, collecting macro data and micro data of the energy-consuming area, the macro data including historical annual carbon emission data, annual resident population change, annual per capita GDP change, annual energy intensity change, annual interactive carbon emission change in different energy-consuming areas, annual energy intensity change and annual energy structure change, the micro data including daily electricity and energy consumption data of all industries in the energy-consuming area, and performing the following steps:

[0005] Step 1: Establish a macro prediction model for the energy consumption area, as shown in the following formula (1):

[0006]

[0007] In formula (1), is the long-term carbon emission forecast value predicted by the macro forecast model; h is the correction coefficient, σ, and τ are respectively the traditional prediction deviation, policy impact and the carbon emission control willingness of residents in the energy consumption area, k1, k2 and k3 are respectively the above σ, and the empirical correction coefficient of τ; ΔP is the annual change in the permanent population of the energy-consuming area; ΔA is the annual change in per capita GDP of the energy-consuming area; ΔT is the annual change in energy intensity of the energy-consuming area; ΔF is the annual change in carbon emissions of different energy-consuming areas; ΔS is the annual change in industrial structure of the energy-consuming area; ΔE is the annual change in energy structure of the energy-consuming area; a is the carbon emission data of the energy-consuming area in the previous year;

[0008] Step 2: Establish a micro-prediction model for the energy consumption area, as shown in the following formula (2):

[0009]

[0010] In formula (2), is the long-term carbon emission forecast value predicted by the micro-forecast model; v is the time for the micro-forecast model to make long-term carbon emission forecasts; Z t is the total predicted carbon emissions of all industries in the energy-consuming area on day t, ρ i is the energy consumption carbon emission factor of the ith industry; E it is the standard coal consumption of the ith industry on the tth day; X in is the electricity and energy consumption data of the ith industry converted into standard coal consumption; γ i is the connection weight between the input layer and the output layer of the i-th industry; β ij is the connection weight between the hidden layer and the output layer of the i-th industry; η i is the incentive function of the ith industry; b ij is the threshold vector of the i-th industry;

[0011] Step 3: Establish a combined prediction model for the energy consumption area, as shown in the following formula (3):

[0012]

[0013] In formula (3), and are the combined weights of the macro prediction model and the micro prediction model respectively;

[0014] Establishing weights and The optimal combination solution model is shown in formula (4):

[0015]

[0016] In formula (4), λ is the penalty factor, CE m-w , They are the historical forecast carbon emissions of w years forward when the combined forecast model, macro forecast model and micro forecast model are used to forecast long-term carbon emissions, Q m-w is the actual carbon emissions of w years ahead when making long-term carbon emissions forecasts; UC is the uncertainty information entropy;

[0017] Step 4: Iterate the optimal combination solution model and calculate the optimal weight after the solution is completed. and Substituting into the combined prediction model, the final combined prediction model is obtained, as shown in the following formula (5):

[0018]

[0019] Step 5: Substitute the final combined prediction model into MATLAB software for iterative solution, and output the long-term carbon emission prediction results of the energy consumption area after the solution is completed.

[0020] Furthermore, the specific iterative solution process in step 4 is as follows:

[0021] Step 4.1: Set the weight scale N and the maximum number of iterations T termax and the number of combined prediction models, etc., to generate N corresponding random weight pairs in the 2-dimensional [0,1] interval, and normalize the weights in each weight pair so that the sum of their weights is 1;

[0022] Step 4.2: Substitute the N corresponding random weight pairs into the optimal combination solution model and calculate N and UC, calculate N fitness S using formula (6), get N fitness S, arrange the N fitness S from large to small and assign the values ​​of the first three fitness S after arrangement to the fitness α_score, β_score, δ_score of populations α, β, δ and the position vector of the corresponding population respectively

[0023]

[0024] Step 4.3: Substitute the X calculated in step 4.2 α , X β and X δ Substituting into the following equations (7) and (8) we can get the position vectors of other positions of the population:

[0025] D α =C1⊙X α -X

[0026] D β =C2⊙X β -X

[0027] D δ =C3⊙X δ -X (7)

[0028] X1=X α -A1⊙D α

[0029] X2=X β -A2⊙D β

[0030] X3=X δ -A3⊙D δ

[0031] X(t+1) = (X1+X2+X3) / 3 (8)

[0032] In formulas (8) and (9), D α , D β , D δ are the position vector after position update, A i and C i is the coordination coefficient vector; X is the position vector of the current population; X(t+1) is the position vector of the updated population;

[0033] Step 4.4: Use the following formula (9) to determine whether the iteration calculation is completed. If the following formula (9) is satisfied, the iteration is considered to be completed and the position vector X is output. α The corresponding weight pair and The value of is taken as the optimal weight and If the following formula (10) is not satisfied, repeat steps 4.1 to 4.3 until formula (9) is satisfied.

[0034] t≤T termax (9).

[0035] Beneficial effects of the present invention: Since the present invention reasonably combines the macro prediction model and the micro prediction model, and adopts a special model structure for the weights of the macro prediction model and the micro prediction model, it is possible to iteratively solve the optimal weights to respectively bring out the effects of the macro prediction model and the micro prediction model, thereby making the results of long-term carbon emission prediction of energy-consuming areas more accurate. DETAILED DESCRIPTION

[0036] The following is a further description of a method for predicting long-term carbon emissions in an energy-consuming area according to the present invention in conjunction with a specific embodiment.

[0037] The present invention provides a method for predicting long-term carbon emissions in an energy-consuming area, collects macro data and micro data of the energy-consuming area, wherein the macro data includes historical annual carbon emission data, annual resident population change, annual per capita GDP change, annual energy intensity change, annual interactive carbon emission change of different energy-consuming areas, annual energy intensity change and annual energy structure change, and the micro data includes daily electricity and energy consumption data of all industries in the energy-consuming area, and performs the following steps:

[0038] Step 1: Establish a macro prediction model for energy consumption areas, as shown in the following formula (1):

[0039]

[0040] In formula (1), is the long-term carbon emission forecast value predicted by the macro forecast model; h is the correction coefficient, σ, and τ are respectively the traditional prediction deviation, policy impact and the carbon emission control willingness of residents in the energy consumption area, k1, k2 and k3 are respectively the traditional prediction deviation σ, policy impact and the empirical correction coefficient of the residents' carbon emission control will τ in the energy-consuming area; ΔP is the annual change in the permanent population of the energy-consuming area; ΔA is the annual change in per capita GDP in the energy-consuming area; ΔT is the annual change in energy intensity in the energy-consuming area; ΔF is the annual interactive carbon emission change in different energy-consuming areas; ΔS is the annual change in the industrial structure of the energy-consuming area; ΔE is the annual change in the energy structure of the energy-consuming area; a is the carbon emission data of the energy-consuming area in the previous year.

[0041] Step 2: Establish a micro-prediction model for energy consumption areas, as shown in the following formula (2):

[0042]

[0043] In formula (2), is the long-term carbon emission forecast value predicted by the micro-forecast model; v is the time for the micro-forecast model to make long-term carbon emission forecasts; Z t is the total predicted carbon emissions of all industries in the energy-consuming area on day t, ρ i is the energy consumption carbon emission factor of the ith industry; E it is the standard coal consumption of the ith industry on the tth day; X in is the electricity and energy consumption data of the ith industry converted into standard coal consumption; γ i is the connection weight between the input layer and the output layer of the i-th industry; β ij is the connection weight between the hidden layer and the output layer of the i-th industry; η i is the incentive function of the ith industry; b ijis the threshold vector of the i-th industry;

[0044] Step 3: Establish a combined prediction model for energy consumption areas, as shown in the following formula (3):

[0045]

[0046] In formula (3), and are the combined weights of the macro forecast model and the micro forecast model respectively;

[0047] Establishing weights and The optimal combination solution model is shown in formula (4):

[0048]

[0049] In formula (4), λ is the penalty factor, CE m-w , They are the historical carbon emissions predicted w years ago when making long-term carbon emission forecasts, Q m-w is the actual carbon emissions of w years forward when making long-term carbon emissions forecasts;

[0050] Step 4: Weight and The optimal combination solution model is iteratively solved. The specific iterative solution process is as follows:

[0051] Step 4.1: Set the weight scale N and the maximum number of iterations T termax and the number of combined prediction models, etc., to generate N corresponding random weight pairs in the 2-dimensional [0,1] interval, and normalize the weights in each weight pair so that the sum of their weights is 1;

[0052] Step 4.2: Substitute the N corresponding random weight pairs into the optimal combination solution model and calculate the N and UC, calculate N fitness S using formula (6), get N fitness S, arrange the N fitness S from large to small and assign the values ​​of the first three fitness S after arrangement to the fitness α_score, β_score, δ_score of populations α, β, δ and the position vector of the corresponding population respectively

[0053]

[0054] Step 4.3: Substitute the X calculated in step 4.2 into α , X β and X δ Substituting into the following equations (7) and (8) we can get the position vectors of other positions of the population:

[0055] D α =C1⊙X α -X

[0056] D β =C2⊙X β -X

[0057] D δ =C3⊙X δ -X (7),

[0058] X1=X α -A1⊙D α

[0059] X2=X β -A2⊙D β

[0060] X3=X δ -A3⊙D δ

[0061] X(t+1) = (X1+X2+X3) / 3 (8),

[0062] In formulas (8) and (9), D α , D β , D δ They are the position vector after position update, A i and C i is the coordination coefficient vector; X is the position vector of the current population; X(t+1) is the position vector of the updated population;

[0063] Step 4.4: Use the following formula (9) to determine whether the iteration calculation is completed. If the following formula (9) is satisfied, the iteration is considered to be completed and the position vector X is output. α The corresponding weight pair and The value of is taken as the optimal weight and If the following formula (10) is not satisfied, repeat steps 4.1 to 4.3 until formula (9) is satisfied.

[0064] t≤T termax (9);

[0065] After the solution is completed, the optimal weight will be obtained and Substituting into the combined prediction model, the final combined prediction model is obtained, as shown in the following formula (5):

[0066]

[0067] Step 5: Substitute the final combined prediction model into the MATLAB software for iterative solution. After the solution is completed, the long-term carbon emission prediction results of the energy consumption area will be output.

[0068] The following is a long-term carbon emission forecast for a certain energy consumption area. The macro data of the energy consumption area is shown in Table 1 below, and the micro data is shown in Table 2 below.

[0069]

[0070] Table 1

[0071]

[0072]

[0073] Table 2

[0074] First, follow step 4 in this method to adjust the weights and The optimal combination solution model is iteratively solved to obtain the optimal weight and are 0.63 and 0.37 respectively, so the final combined prediction model of this energy consumption area is obtained, as shown in the following formula (10):

[0075]

[0076] The macro data in Table 1, the micro data in Table 2 and the final combined prediction model in formula (10) are substituted into Matlab software for solution, and the long-term carbon emission prediction results of this energy consumption area are obtained as shown in Table 3 below. The root mean square error (RMSE) comparison table of the long-term carbon emission prediction results obtained by this method and the macro prediction model and micro prediction model is shown in Table 4 below.

[0077]

[0078] Table 3

[0079]

[0080] Table 4

[0081] It can be clearly seen from Table 4 that the long-term carbon emission prediction results obtained by this method are more accurate.

Claims

1. A method for predicting long-term carbon emissions in energy-consuming areas, characterized by: Collect macro data and micro data of the energy consumption area, the macro data includes historical annual carbon emission data, annual permanent population change, annual per capita GDP change, annual energy intensity change, annual interactive carbon emission change of different energy consumption areas, and annual energy structure change, the micro data includes daily electricity and energy consumption data of all industries in the energy consumption area, and perform the following steps: Step 1: Establish a macro prediction model for the energy consumption area, as shown in the following formula (1): In formula (1), is the long-term carbon emission forecast value predicted by the macro forecast model; h is the correction coefficient, σ, and τ are respectively the traditional prediction deviation, policy impact and the carbon emission control willingness of residents in the energy consumption area, k1, k2 and k3 are respectively the above σ, and the empirical correction coefficient of τ; ΔP is the annual change in the permanent population of the energy-consuming area; ΔA is the annual change in per capita GDP of the energy-consuming area; ΔT is the annual change in energy intensity of the energy-consuming area; ΔF is the annual change in interactive carbon emissions of different energy-consuming areas; ΔS is the annual change in industrial structure of the energy-consuming area; ΔE is the annual change in energy structure of the energy-consuming area; a is the carbon emission data of the energy-consuming area in the previous year; Step 2: Establish a micro-prediction model for the energy consumption area, as shown in the following formula (2): In formula (2), is the long-term carbon emission forecast value predicted by the micro-forecast model; v is the time for the micro-forecast model to make long-term carbon emission forecasts; Z t is the total predicted carbon emissions of all industries in the energy-consuming area on day t, ρ i is the energy consumption carbon emission factor of the ith industry; E it is the standard coal consumption of the ith industry on the tth day; X in is the electricity and energy consumption data of the ith industry converted into standard coal consumption; γ i is the connection weight between the input layer and the output layer of the i-th industry; β ij is the connection weight between the hidden layer and the output layer of the i-th industry; η i is the incentive function of the ith industry; b ij is the threshold vector of the i-th industry; Step 3: Establish a combined prediction model for the energy consumption area, as shown in the following formula (3): In formula (3), and are the combined weights of the macro prediction model and the micro prediction model respectively; Establishing weights and The optimal combination solution model is shown in formula (4): In formula (4), λ is the penalty factor, CE m-w , They are the historical forecast carbon emissions of w years forward when the combined forecast model, macro forecast model and micro forecast model are used to forecast long-term carbon emissions, Q m-w is the actual carbon emissions of w years forward when making long-term carbon emissions forecasts; UC is uncertainty information entropy; Step 4: Iterate the optimal combination solution model and calculate the optimal weight after the solution is completed. and Substituting into the combined prediction model, the final combined prediction model is obtained, as shown in the following formula (5): Step 5: Substitute the final combined prediction model into MATLAB software for iterative solution, and output the long-term carbon emission prediction results of the energy consumption area after the solution is completed.

2. The method for predicting long-term carbon emissions in energy-consuming areas according to claim 1 is characterized by: The specific iterative solution process in step 4 is as follows: Step 4.1: Set the weight scale N and the maximum number of iterations T termax And the number parameters of the combined prediction model are used to generate N corresponding random weight pairs in the 2-dimensional [0,1] interval, and the weights in each weight pair are normalized so that the sum of their weights is 1; Step 4.2: Substitute the N corresponding random weight pairs into the optimal combination solution model and calculate N and UC, calculate N fitness S using formula (6), get N fitness S, arrange the N fitness S from large to small and assign the values ​​of the first three fitness S after arrangement to the fitness α_score, β_score, δ_score of populations α, β, δ and the position vector of the corresponding population respectively Step 4.3: Substitute the X calculated in step 4.2 α , X β and X δ Substituting into the following equations (7) and (8) we can get the position vectors of other positions of the population: D α =C1⊙X α -X D β =C2⊙X β -X D δ =C3⊙X δ -X (7) In formulas (7) and (8), D α , D β , D δ are the position vectors after the position update, C1, C2 and C3 are the first coordination coefficient vectors of populations α, β and δ respectively; A1, A2 and A3 are the second coordination coefficient vectors of populations α, β and δ respectively; X is the position vector of the current population; X(o+1) is the position vector of the updated population; Step 4.4: If the number of iterations of the population reaches the maximum number of iterations T termax , then the iterative calculation of the population ends, and the position vector X is output at this time α The corresponding weight pair and As the optimal weight and If the number of iterations of the population does not reach the maximum number of iterations T termax , then repeat steps 4.1 to 4.3 until the number of iterations of the population reaches the maximum number of iterations T termax until.

Images