Power demand prediction method and device based on fuzzy autoregressive distributed lag model

Abstract

The invention discloses a power demand prediction method and device based on a fuzzy autoregressive distributed lag model. The method comprises the steps that the historical total carbon emission amount of a target area in a target time period, historical carbon emission attribute data of all to-be-selected influence factors and a historical actual power demand value are acquired; decomposing the total historical carbon emission into a carbon emission decomposition relational expression among the historical carbon emission attribute data of the to-be-selected influence factors; on the basis of the carbon emission decomposition relational expression and historical carbon emission attribute data, determining the influence contribution amount of each to-be-selected influence factor; determining a power demand prediction influence factor based on the influence contribution amount; acquiring carbon emission power attribute data corresponding to the power demand prediction influence factors, and acquiring carbon-free power attribute data irrelevant to carbon emission; and inputting the carbon emission power attribute data and the carbon-free power attribute data into a fuzzy autoregressive distributed lag model for power demand prediction to obtain a power demand prediction result.

Description

Technical Field

[0001] The present invention relates to the field of smart grid technology, and in particular to a method and device for predicting power demand based on a fuzzy autoregressive distributed lag model. Background Art

[0002] As an important part of power grid planning and medium- and long-term transactions in the power market, power demand forecasting is an important basis for formulating unit investment and construction plans and power grid planning schemes. At this stage, with the proposal of the "dual carbon" goals, the uncertainty of future power demand has been further strengthened, and it will also have a profound impact on future changes in electricity consumption characteristics. The construction of power grids is characterized by high costs, long fundraising time, and long construction period. Therefore, the accuracy of forecasting future regional power demand under the background of "dual carbon" directly affects the operating costs of power grid companies. Accurately predicting future power demand and rationally planning the development of the power system are the key to ensuring the balance of power supply and demand, promoting energy transformation, and achieving the "dual carbon" goals.

[0003] At present, the influencing factors related to electricity demand forecasting in the context of dual carbon are usually determined manually and arbitrarily, and then electricity demand forecasting is carried out based on the influencing factors. However, due to the influence of human subjectivity, there may be cases where the influencing factors are determined incorrectly, resulting in low accuracy in electricity demand forecasting. Summary of the invention

[0004] The present invention provides a method and device for predicting electricity demand based on a fuzzy autoregressive distributed lag model, which mainly aims to improve the prediction accuracy of electricity demand under the dual-carbon background.

[0005] According to a first aspect of the present invention, there is provided a method for predicting power demand based on a fuzzy autoregressive distributed lag model, comprising:

[0006] Obtain the total historical carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, wherein each of the candidate influencing factors is a variety of influencing factors that affect carbon emissions in the power industry;

[0007] Decomposing the total historical carbon emissions into a relationship between the historical carbon emission attribute data of each of the selected influencing factors to obtain a carbon emission decomposition relationship;

[0008] Determine the contribution of each of the candidate influencing factors to the change in total carbon emissions based on the carbon emission decomposition relationship and the historical carbon emission attribute data;

[0009] Based on the impact contribution, determining the power demand forecast influencing factor related to carbon emissions from each of the candidate influencing factors;

[0010] Obtaining carbon emission power attribute data corresponding to power demand forecast influencing factors related to carbon emission, and obtaining carbon-free power attribute data of power demand forecast influencing factors unrelated to carbon emission;

[0011] The carbon emission electricity attribute data and the carbon-free electricity attribute data are input into a fuzzy autoregressive distributed lag model to perform electricity demand forecasting to obtain an electricity demand forecasting result.

[0012] According to a second aspect of the present invention, there is provided a power demand forecasting device based on a fuzzy autoregressive distributed lag model, comprising:

[0013] A first acquisition unit is used to acquire the total historical carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual power demand value, wherein each of the candidate influencing factors is a variety of influencing factors that affect carbon emissions in the power industry;

[0014] A decomposition unit, used for decomposing the total historical carbon emissions into a relationship between the historical carbon emission attribute data of each of the candidate influencing factors to obtain a carbon emission decomposition relationship;

[0015] A first determination unit, configured to determine the contribution of each of the candidate influencing factors to the change in the total carbon emissions based on the carbon emissions decomposition relationship and the historical carbon emissions attribute data;

[0016] A second determining unit is used to determine, based on the impact contribution, an electricity demand forecast influencing factor related to carbon emissions from among the candidate influencing factors;

[0017] A second acquisition unit is used to acquire carbon emission power attribute data corresponding to power demand forecast influencing factors related to carbon emission, and to acquire carbon-free power attribute data of power demand forecast influencing factors unrelated to carbon emission;

[0018] The prediction unit is used to input the carbon emission power attribute data and the carbon-free power attribute data into a fuzzy autoregressive distributed lag model to perform power demand prediction and obtain a power demand prediction result.

[0019] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for predicting power demand based on the fuzzy autoregressive distributed lag model.

[0020] According to a fourth aspect of the present invention, there is provided a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the above-mentioned method for predicting electricity demand based on the fuzzy autoregressive distributed lag model is implemented.

[0021] According to a method and device for predicting electricity demand based on a fuzzy autoregressive distributed lag model provided by the present invention, compared with the current method of arbitrarily determining the influencing factors related to electricity demand prediction under the dual-carbon background by artificial means, and then predicting electricity demand based on the influencing factors, the present invention decomposes the total historical carbon emissions into a carbon emission decomposition relationship between the historical carbon emission attribute data of each candidate influencing factor, and then determines the contribution of each candidate influencing factor to the change in the total carbon emissions according to the carbon emission decomposition relationship, and determines the electricity demand prediction influencing factors related to carbon emissions according to the contribution, and finally, according to the electricity demand prediction influencing factors, uses the fuzzy autoregressive distributed lag model to predict the electricity demand. Electricity demand is predicted, and the influencing factors related to electricity demand prediction are selected by the contribution of each candidate influencing factor to the change in total carbon emissions. This can improve the accuracy of determining the influencing factors related to electricity demand prediction under the dual carbon background, and thus improve the prediction accuracy of electricity demand under the dual carbon background. At the same time, electricity demand is predicted through the fuzzy autoregressive distributed lag model. The fuzzy autoregressive distributed lag model can handle the uncertainty factors in electricity demand prediction under the "dual carbon" background through fuzzy theory. The model can flexibly respond to complex situations such as changes in energy structure, and provide reliable support for electricity demand prediction under different low-carbon paths, thereby improving the accuracy of electricity demand prediction under the dual carbon background. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0023] Figure 1 A flow chart of a method for predicting power demand based on a fuzzy autoregressive distributed lag model provided by an embodiment of the present invention is shown;

[0024] Figure 2 Another flow chart of a method for predicting power demand based on a fuzzy autoregressive distributed lag model provided by an embodiment of the present invention is shown;

[0025] Figure 3 A schematic diagram of the structure of a power demand forecasting device based on a fuzzy autoregressive distributed lag model provided by an embodiment of the present invention is shown;

[0026] Figure 4 A schematic diagram of the structure of another power demand forecasting device based on a fuzzy autoregressive distributed lag model provided by an embodiment of the present invention is shown;

[0027] Figure 5 A schematic diagram of the physical structure of a computer device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0028] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0029] At present, the method of arbitrarily determining the influencing factors related to electricity demand forecasting under the dual carbon background and then forecasting electricity demand based on the influencing factors is affected by human subjectivity, which may lead to errors in determining the influencing factors, resulting in low accuracy in electricity demand forecasting.

[0030] In order to solve the above problems, an embodiment of the present invention provides a method for predicting power demand based on a fuzzy autoregressive distributed lag model, such as Figure 1 As shown, the method includes:

[0031] 101. Obtain the total historical carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, wherein each candidate influencing factor is a variety of influencing factors that affect carbon emissions in the power industry.

[0032] Among them, the target area is the area where electricity demand is predicted, and the target area can be based on provinces, cities, counties, etc. The target time period refers to each historical year in the target area; the total historical carbon emissions refer to the total carbon dioxide emissions of the target area during the target time period; each candidate influencing factor is an influencing factor related to carbon emissions that may affect electricity demand, including power supply structure influencing factors and energy structure influencing factors, and the power supply structure influencing factors and energy structure influencing factors are used as representatives of structural effect influencing factors; economic development influencing factors and population influencing factors are used as representatives of economic effect factors; carbon emission intensity influencing factors and energy intensity influencing factors are used as representatives of intensity effect factors. Table; historical carbon emission attribute data refers to the attribute data such as electricity consumption and power structure in the power structure influencing factors in each historical year; the historical carbon emission attribute data of energy structure influencing factors are the attribute data such as carbon emissions, electricity, total energy consumption, electrification rate, etc. generated by various energy sources in each historical year; the historical carbon emission attribute data of economic development influencing factors are the attribute data such as gross national product and per capita gross domestic product in each historical year, the historical carbon emission attribute data of population influencing factors are the attribute data such as population in each historical year, the historical carbon emission attribute data of carbon emission intensity influencing factors are the attribute data such as carbon emission coefficient in each historical year, and the historical carbon emission attribute data of energy intensity influencing factors are the attribute data such as energy intensity in each historical year.

[0033] Du Yu In the embodiment of the present invention, the power industry database records the total historical carbon emissions of each region at each historical time, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, and the above data can be obtained in the power industry database.

[0034] 102. Decompose the total historical carbon emissions into the relationship between the historical carbon emission attribute data of each candidate influencing factor to obtain the carbon emission decomposition relationship.

[0035] 103. Based on the carbon emission decomposition relationship and historical carbon emission attribute data, determine the contribution of each candidate influencing factor to the change in total carbon emissions.

[0036] 104. Based on the impact contribution, determine the electricity demand forecast influencing factors related to carbon emissions from among the candidate influencing factors.

[0037] For the embodiment of the present invention, the LDMI (Logarithmic Mean Divisia Index Method) decomposition method can be used to decompose the total historical carbon emissions into the form of the product of the historical carbon emission attribute data of each candidate influencing factor, so as to obtain the carbon emission decomposition relationship. Based on the principle of the LMDI decomposition method, a decomposition model suitable for determining the influencing factors is constructed. The decomposition model decomposes the total historical carbon emissions into the contribution of multiple candidate influencing factors. The formula of the LMDI decomposition method (carbon emission decomposition relationship) is used to calculate the contribution (influence contribution) of each candidate influencing factor to the change in the total carbon emissions. These influence contributions can be absolute values ​​or relative values ​​(such as contribution rates). After that, each influence contribution is analyzed to reveal the degree and direction of influence of each candidate influencing factor on the overall change of carbon emissions, and the causal relationship between these factors and electricity demand is studied through path analysis, so that in addition to the traditional electricity demand influencing factors, the electricity demand forecast influencing factors related to carbon emissions can also be determined. Therefore, by constructing a carbon emission decomposition relationship, and determining the influence contribution of each candidate influencing factor according to the carbon emission decomposition relationship, and finally selecting the electricity demand forecasting influencing factors related to carbon emissions according to the influence contribution, the time wasted in manually selecting influencing factors and the problem of incorrect selection can be avoided. Therefore, the embodiment of the present invention can improve the determination efficiency and accuracy of the influencing factors related to electricity demand forecasting, and further improve the prediction efficiency and prediction accuracy of electricity demand under the dual carbon background.

[0038] 105. Obtain carbon emission electricity attribute data corresponding to electricity demand forecast influencing factors related to carbon emissions, and obtain carbon-free electricity attribute data for electricity demand forecast influencing factors unrelated to carbon emissions.

[0039] Among them, the power demand forecasting influencing factors related to carbon emissions include at least one of the power structure influencing factors, energy structure influencing factors, economic development influencing factors, population influencing factors, carbon emission intensity influencing factors, and energy intensity influencing factors; the power demand forecasting influencing factors unrelated to carbon emissions include environmental influencing factors, etc.; the carbon emission power attribute data power demand preset area power consumption, power structure, carbon emissions, electricity, total energy consumption, electrification rate, gross national product, per capita gross domestic product, population, carbon emission coefficient, energy intensity Q, etc. generated by each energy source to produce electricity; carbon-free power attribute data includes weather data, temperature data, etc. For the embodiment of the present invention, to predict the power demand of a certain year, the carbon emission power attribute data corresponding to the power demand forecasting influencing factors related to carbon emissions in that year can be estimated, and the carbon-free power attribute data of the power demand forecasting influencing factors unrelated to carbon emissions can be obtained, and then based on the carbon emission power attribute data and the carbon-free power attribute data, the fuzzy autoregressive distributed lag model is used to predict the power demand, that is, to predict the required power consumption. Therefore, by using the fuzzy autoregressive distributed lag model to forecast electricity demand, we can flexibly respond to complex situations such as policy adjustments and changes in energy structure under the constraints of carbon emissions by introducing fuzzy theory and optimization algorithms, thereby providing reliable support for electricity demand forecasting under different low-carbon paths. The fuzzy autoregressive distributed lag model can effectively deal with the uncertainty factors in electricity demand forecasting under the background of "dual carbon", thereby improving the accuracy of electricity demand forecasting.

[0040] 106. Input the carbon emission electricity attribute data and the carbon-free electricity attribute data into the fuzzy autoregressive distributed lag model to perform electricity demand forecasting and obtain the electricity demand forecasting result.

[0041] Among them, the power demand forecasting influencing factors related to carbon emissions include at least one of the power structure influencing factors, energy structure influencing factors, economic development influencing factors, population influencing factors, carbon emission intensity influencing factors, and energy intensity influencing factors; the power demand forecasting influencing factors unrelated to carbon emissions include environmental influencing factors, etc.; the carbon emission power attribute data power demand preset area power consumption, power structure, carbon emissions, electricity, total energy consumption, electrification rate, gross national product, per capita gross domestic product, population, carbon emission coefficient, energy intensity Q, etc. generated by each energy source to produce electricity; carbon-free power attribute data includes weather data, temperature data, etc. For the embodiment of the present invention, to predict the power demand of a certain year, the carbon emission power attribute data corresponding to the power demand forecasting influencing factors related to carbon emissions in that year can be estimated, and the carbon-free power attribute data of the power demand forecasting influencing factors unrelated to carbon emissions can be obtained, and then based on the carbon emission power attribute data and the carbon-free power attribute data, the fuzzy autoregressive distributed lag model is used to predict the power demand, that is, to predict the required power consumption. Therefore, by using the fuzzy autoregressive distributed lag model to forecast electricity demand, we can flexibly respond to complex situations such as policy adjustments and changes in energy structure under the constraints of carbon emissions by introducing fuzzy theory and optimization algorithms, thereby providing reliable support for electricity demand forecasting under different low-carbon paths. The fuzzy autoregressive distributed lag model can effectively deal with the uncertainty factors in electricity demand forecasting under the background of "dual carbon", thereby improving the accuracy of electricity demand forecasting.

[0042] According to a method for predicting electricity demand based on a fuzzy autoregressive distributed lag model provided by the present invention, compared with the current method of manually and arbitrarily determining influencing factors related to electricity demand prediction under the dual-carbon background and then predicting electricity demand based on the influencing factors, the embodiment of the present invention decomposes the total historical carbon emissions into a carbon emission decomposition relationship between the historical carbon emission attribute data of each candidate influencing factor, and then determines the contribution of each candidate influencing factor to the change in the total carbon emissions according to the carbon emission decomposition relationship, and determines the electricity demand prediction influencing factors related to carbon emissions according to the contribution, and finally predicts the electricity demand based on the electricity demand prediction influencing factors using the fuzzy autoregressive distributed lag model. Power demand is predicted, and the influencing factors related to power demand forecast are selected by the contribution of each candidate influencing factor to the change in total carbon emissions. This can improve the accuracy of determining the influencing factors related to power demand forecast under the dual carbon background, and thus improve the prediction accuracy of power demand under the dual carbon background. At the same time, power demand is predicted by the fuzzy autoregressive distributed lag model. The fuzzy autoregressive distributed lag model can handle the uncertainty factors in power demand forecast under the "dual carbon" background through fuzzy theory. The model can flexibly respond to complex situations such as changes in energy structure, and provide reliable support for power demand forecast under different low-carbon paths, thereby improving the accuracy of power demand forecast under the dual carbon background.

[0043] Further, in order to better illustrate the above process of classifying data, as a refinement and extension of the above embodiment, the embodiment of the present invention provides another method for predicting power demand based on a fuzzy autoregressive distributed lag model, such as Figure 2 As shown, the method includes:

[0044] 201. Obtain the total historical carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, wherein each candidate influencing factor is a variety of influencing factors that affect carbon emissions in the power industry.

[0045] Specifically, the target area may be the same as or different from the area for which electricity demand forecasting is to be performed. The embodiment of the present invention does not limit the specific location of the target area. Within the target area, the total historical carbon emissions for multiple historical time periods, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value can be obtained. Then, the carbon emission decomposition relationship is determined based on the above-mentioned total historical carbon emissions and the historical carbon emission attribute data of each candidate influencing factor.

[0046] 202. Decompose the total historical carbon emissions into the relationship between the historical carbon emission attribute data of each candidate influencing factor to obtain the carbon emission decomposition relationship.

[0047] For the embodiment of the present invention, after obtaining the total historical carbon emissions of each year in the target area and the historical carbon emission attribute data of each candidate influencing factor, it is necessary to determine the carbon emission decomposition relationship based on the total historical carbon emissions and the historical carbon emission attribute data. Based on this, step 202 specifically includes: decomposing the total historical carbon emissions into the relationship between the historical carbon emission attribute data of each candidate influencing factor, and obtaining the carbon emission decomposition relationship as follows:

[0048]

[0049] Among them, H is the electricity consumption in each historical year, e i is the power structure, C i The carbon emissions generated by producing electricity from the i-th energy source in each historical year, E i is the electricity produced by the i-th energy source, E is the total energy consumption, D is the electrification rate, G is the gross national product, Y is the per capita gross domestic product, P is the population, and F i is the carbon emission coefficient, Q is the energy intensity, and C is the total historical carbon emissions, that is, the total historical carbon dioxide emissions.

[0050] 203. Based on the carbon emission decomposition relationship and historical carbon emission attribute data, determine the contribution of each candidate influencing factor to the change in total carbon emissions.

[0051] For the embodiment of the present invention, after determining the carbon emission decomposition relational expression, a carbon emission decomposition relational expression is also required to determine the contribution of each candidate influencing factor to the change in the total carbon emission. Based on this, step 203 specifically includes: using the logarithmic mean Dirichlet index method to decompose the carbon emission decomposition relational expression to obtain a carbon emission coefficient influence relational expression, and based on the carbon emission coefficient F of the i-th energy in year 0 i 0 and the carbon emission coefficient F of the i-th energy source in year T i T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T Carbon emission coefficient influence relationship, determine the contribution of carbon emission intensity influencing factors ΔC em ,in, The carbon emission decomposition relationship is decomposed by the logarithmic mean Dimitrov index method to obtain the power structure impact relationship, and the power structure of the i-th energy source in year 0 is calculated based on the power structure of the i-th energy source in year 0. and the power structure of the i-th energy source in year T Carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The power structure influence relationship is used to determine the contribution of the power structure influence factors ΔC es ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dirichlet index method to obtain the electrification rate impact equation, and based on the electrification rate D in year 0 0 and the electrification rate D in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of factors affecting energy structure ΔC int ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dimitrov index method to obtain the energy intensity impact equation, and based on the energy intensity Q 0 and the energy intensity Q in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of energy intensity influencing factors ΔC dp ,in, The carbon emission decomposition equation is decomposed by using the logarithmic mean Dimitrov index method to obtain the economic development impact equation, and based on the per capita gross domestic product Y in year 0 0 and the per capita GDP Y in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of economic development factors ΔC ey ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dimitrov index method to obtain the population impact equation, and based on the population size P in year 0 0 and the population size P in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of population influence factors ΔC P ,in,

[0052] Specifically, the LMDI decomposition method is used to decompose the carbon emission decomposition equation, that is, the items on the right side of the carbon emission decomposition equation can be expressed as the following formula:

[0053]

[0054]

[0055] Among them, ΔC represents the change in carbon emissions from year 0 to year T, C 0 represents the carbon emissions in year 0, C T represents the carbon emissions in year T, ΔC em Indicates the contribution of carbon emission coefficient, that is, the contribution of carbon emission intensity influencing factors, ΔC es Indicates the contribution of the factors affecting the power structure, ΔC int It represents the impact contribution of electrification rate, that is, the impact contribution of energy structure influencing factors, ΔC dq Indicates the contribution of factors affecting energy intensity, ΔC ey Indicates the contribution of factors affecting economic development, ΔC P Indicates the contribution of population influencing factors, F i 0 is the carbon emission coefficient of the i-th energy source in year 0, F i T is the carbon emission coefficient of the i-th energy source in year T, is the carbon emission of the i-th energy source in year 0, is the carbon emission of the i-th energy source in year T, is the power structure of the i-th energy source in year 0, is the power structure of the i-th energy source in year T, D 0 is the electrification rate in year 0, D T is the electrification rate in year T, Q 0 is the energy intensity in year 0, Q T is the energy intensity in year T, Y 0 is the per capita GDP in year 0, Y T is the per capita GDP in year T, P 0 is the population in year 0, P T is the population in year T. Therefore, by constructing a carbon emission decomposition relationship, and then determining the impact contribution of each candidate influencing factor according to the carbon emission decomposition relationship, the accuracy of determining the impact contribution can be improved, and then the accuracy of selecting factors that affect electricity demand forecasting under the dual carbon background can be improved.

[0056] 204. Based on the impact contribution, determine the electricity demand forecast influencing factors related to carbon emissions from among the candidate influencing factors.

[0057] For the embodiment of the present invention, after determining the impact contribution corresponding to each candidate influencing factor, it is also necessary to select the power demand forecast influencing factors related to the power demand forecast under the dual carbon background from each candidate influencing factor based on each impact contribution. Based on this, step 204 specifically includes: based on the impact contribution ΔC of the carbon emission intensity influencing factor em , the contribution of the factors affecting the power structure ΔC es , the contribution of factors affecting energy structure ΔC int , Contribution of factors affecting energy intensity ΔC dp , the contribution of economic development influencing factors ΔC ey , the contribution of population influencing factors ΔC P , determine the initial electricity demand forecast influencing factors related to carbon emissions from each of the candidate influencing factors; determine the influencing factor standard deviation σ of the historical carbon emission attribute data corresponding to each of the initial electricity demand forecast influencing factors xi , the standard deviation of the historical actual power demand value σ y , the i-th initial power demand forecast influencing factor X among the initial power demand forecast influencing factors i and the jth initial power demand forecast influencing factor X j The correlation coefficient C ij 、The influencing factor X of the i-th initial power demand forecast i The partial regression coefficient λ i ; Based on the standard deviation of the influencing factors σ xi , Standard deviation of demand value σ y , correlation coefficient C ij , partial regression coefficient λ i , determine the path coefficient P between each of the initial power demand forecast influencing factors and the historical actual power demand value i ,in, Based on the path coefficient Pi, the path coefficient judgment coefficient R of each of the initial power demand forecast influencing factors is determined. 2 ,in, C iy represents the influencing factor X of the i-th initial power demand forecast i The correlation coefficient between the initial electricity demand forecast value Y and the electricity demand forecast influencing factors related to carbon emissions is determined among the initial electricity demand forecast influencing factors based on the path coefficient judgment coefficient of each of the initial electricity demand forecast influencing factors.

[0058] Specifically, the impact contribution of carbon emission intensity influencing factors, the impact contribution of power structure influencing factors, the impact contribution of energy structure influencing factors, the impact contribution of energy intensity influencing factors, the impact contribution of economic development influencing factors, and the impact contribution of population influencing factors are used to determine the initial electricity demand forecast influencing factors related to carbon emissions among various candidate influencing factors, such as determining the target influencing factors whose impact contribution is greater than a preset threshold (the preset threshold is set according to actual demand) among various candidate influencing factors, and determining the target influencing factors as the initial electricity demand forecast influencing factors that affect the electricity demand forecast under the dual-carbon background constraints, and then determining the path coefficient between each initial electricity demand forecast influencing factor and the historical actual electricity demand value based on the historical carbon emission attribute data under the initial electricity demand forecast influencing factors. The specific calculation formula of the path coefficient is as follows:

[0059] P i =Z i +J i

[0060]

[0061] Among them, P i represents the path coefficient between the influencing factors of the i-th initial power demand forecast and the historical actual power demand value. If each initial power demand forecast influencing factor is X 1 , X 2 , …X i , …X k ,λ i Represents X i The partial regression coefficient of Represents X i The mean value, σ xi Represents X i The standard deviation of Y represents the power demand value. represents the mean of Y, σ y Represents the standard deviation of the power demand value Y, C ij Represents X i With X j The correlation coefficient between i For X i The direct path coefficient for Y indicates that when X i When one standard deviation unit is added, the standard deviation unit of change in Y is i The larger the absolute value of X is, the i The higher the relative importance of the change in Y, the greater the i When it is positive, it means X i The increase of will promote the increase of Y. i When negative, it means X i The increase of will inhibit the increase of Y, Ji Represents X i By X j Indirect impact on Y, J i Called X i The indirect path coefficient of X to Y i Through the indirect impact of other influencing factors on Y, P i Characterization Comprehensive consideration of direct and indirect effects of X i The effect on Y is called X i The comprehensive coefficient for Y.

[0062] Furthermore, based on the path coefficient P i , determine the path coefficient judgment coefficient R 2 , the path coefficient judgment coefficient is determined by the following formula:

[0063]

[0064] Among them, C iy represents the influencing factor X of the i-th initial power demand forecast i The correlation coefficient between the predicted power demand value Y and the path coefficient judgment coefficient R 2 It indicates the degree to which the selected influencing factors explain the electricity demand forecast. If the judgment coefficient is small, it means that the error is large or there are other important influencing factors that are not taken into account. Finally, according to the path coefficient judgment coefficient of each initial electricity demand forecast influencing factor, the electricity demand forecast influencing factor related to carbon emissions is determined in each initial electricity demand forecast influencing factor. For example, the target electricity demand forecast influencing factor with a path coefficient judgment coefficient greater than the preset coefficient threshold can be selected from each initial electricity demand forecast influencing factor, and finally the target electricity demand forecast influencing factor is determined as the electricity demand forecast influencing factor that affects the electricity demand forecast under the dual carbon constraint. Thus, the factors affected by carbon emissions are decomposed by the LMDI method, and then the causal relationship between these factors and electricity demand is studied by path analysis. In addition to the traditional electricity demand influencing factors, the influencing factors of electricity demand related to carbon emissions are screened out, so as to improve the accuracy of determining the factors that affect the electricity demand forecast under the dual carbon background. By predicting the electricity demand through accurately determined influencing factors, the accuracy of electricity demand prediction under the dual carbon background can be improved.

[0065] Further, after determining the influencing factors of power demand forecasting related to carbon emissions and power demand forecasting, it is necessary to optimize the energy structure in the influencing factors of power demand forecasting. Based on this, the method includes: determining the total energy demand balance condition in the power demand forecasting process, wherein the total energy demand balance condition is S i,t =W i,t +F i,t , S i,t The predicted energy supply of power supply energy in the energy structure influencing factors, W i,t is the amount of unused electricity generation energy in the ith energy source in the tth year, F i,t is the utilization of the i-th power generation energy in the t-th year, D t is the total energy demand planned in the electricity demand forecasting process, L t is the amount of clean energy consumed by the energy system in year t, i is the i-th energy in the energy structure influencing factors, n is the total amount of energy in the energy structure influencing factors, C t is the predicted total carbon emissions in the electricity demand forecasting process, C i,t is the carbon emission coefficient of the i-th energy source in year t; in order to achieve carbon peak and carbon neutrality in the future preset years, based on the predicted energy supply S i,t 、Total energy demand D t , predict total carbon emissions C t , the amount of energy used in the ith energy source in the tth year F i,t , determine the carbon emission constraints and non-negativity constraints in the power demand forecasting process, where the carbon emission constraint is C t ≤CS i,t , C.S. i,t is the upper limit of carbon emission constraint, and the non-negativity constraint condition is L t ,F i,t ≥0, Based on the carbon emission constraint and the non-negativity constraint, determine the carbon peak constraint and the carbon neutrality constraint; based on the total energy demand balance condition, the carbon peak constraint, and the carbon neutrality constraint, take the clean energy consumption L of the energy system in year t as t The minimum is the objective function, and the energy structure in the electricity demand forecast influencing factors is optimized to obtain the optimized energy structure; the carbon emission electricity attribute data corresponding to the electricity demand forecast influencing factors related to carbon emissions is obtained, including: obtaining the carbon emission electricity attribute data corresponding to the electricity demand forecast influencing factors after the energy structure is optimized.

[0066] Specifically, the main goal of the carbon pinch technology in optimizing the regional energy consumption structure of power demand forecast under carbon emission constraints is to calculate the minimum proportion of clean energy power generation under carbon emission and energy supply constraints by planning the carbon emissions generated by energy supply. The specific form of the optimization model is as follows:

[0067]

[0068] Among them, L t represents the amount of clean energy consumed by the energy system in year t, i is the i-th energy in the factors affecting the energy structure, W i,trepresents the amount of unused energy in the i-th energy source in year t, F i,t represents the utilization of the i-th energy in year t, S i,t represents the supply of the i-th energy in year t, C i,t represents the carbon emission coefficient of the i-th energy source in year t, C t represents the total carbon emissions in year t, D t represents the total energy demand in year t, CS i,t represents the upper limit of carbon emission constraint in year t, n is the total amount of energy in the energy structure influencing factors, S i,t =W i,t +F i,t ,

[0069] is the total energy demand balance condition, C t ≤CS i,t is the carbon emission constraint, L t ,F i,t ≥0 is a non-negativity constraint. Furthermore, under the premise of meeting carbon emission constraints, non-negativity constraints, and total energy demand balance, clean energy L t The objective function is to minimize the proportion of power generation. By allocating different types of energy consumption, the "dual carbon" critical scenario is explored, that is, the carbon peak constraint and carbon neutrality constraint are determined. For example, the "dual carbon" goals, namely, achieving carbon peak in 2030 and carbon neutrality in 2060, are included in the constraints of the model, and the carbon emission constraints and non-negative constraints are transformed into the following carbon peak constraints and carbon neutrality constraints:

[0070]

[0071] Among them, CL represents my country’s total carbon sink, t represents time, C t represents the total carbon emissions in year t, is the carbon peak constraint, C t | t＝2060 ≤CL is the carbon neutrality constraint. Finally, according to the total energy demand balance condition, carbon peak constraint condition, and carbon neutrality constraint condition, the clean energy consumption of the energy system in year t is L tTaking the minimum as the objective function, the energy structure in the factors affecting electricity demand forecast is optimized to obtain the optimized energy structure, and the carbon emission electricity attribute data corresponding to the factors affecting electricity demand forecast after the optimized energy structure are obtained. Then, according to the carbon emission electricity attribute data corresponding to the factors affecting electricity demand forecast after the optimized energy structure, the fuzzy autoregressive distributed lag model is used to predict electricity demand. In this way, the energy structure is optimized through the carbon peak constraints and carbon neutrality constraints, which can ensure that carbon emissions meet emission requirements while ensuring stable power supply. By exploring the dual-carbon critical scenario, it is possible to flexibly respond to complex situations such as policy adjustments and changes in energy structure in electricity demand forecasting, thereby improving the accuracy and flexibility of electricity demand forecasting.

[0072] 205. Obtain carbon emission electricity attribute data corresponding to electricity demand forecast influencing factors related to carbon emissions, and obtain carbon-free electricity attribute data of electricity demand forecast influencing factors unrelated to carbon emissions.

[0073] Furthermore, under the preset conditions of carbon peak and carbon neutrality constraints, the energy structure is optimized to obtain carbon emission electricity attribute data corresponding to the electricity demand forecast influencing factors after the optimized energy structure. The electricity demand is predicted by the carbon emission electricity attribute data to ensure that the electricity demand meets the dual carbon constraints. At the same time, in order to increase the accuracy of electricity demand prediction, it is also necessary to obtain carbon-free electricity attribute data of electricity demand forecast influencing factors that are not related to carbon emissions, such as temperature data, humidity data and weather data under environmental influencing factors.

[0074] 206. Determine a carbon emission power feature vector corresponding to the carbon emission power attribute data, and determine a carbon-free power feature vector corresponding to the carbon-free power attribute data.

[0075] 207. Multiply each element in the carbon emission power feature vector with the corresponding position element in the carbon-free power feature vector to obtain a fused power feature vector.

[0076] Specifically, the carbon emission power feature vector corresponding to the carbon emission power attribute data and the carbon-free power feature vector corresponding to the carbon-free power attribute data are determined by word embedding and other methods. If the carbon emission power feature vector is f(a 1 ,a 2 ,a 3 ), the carbon-free electricity characteristic vector is f(b 1 ,b 2 ,b 3 ), the carbon emission power feature vector and the carbon-free power feature vector are fused. The specific fusion process is to multiply the corresponding elements in the carbon emission power feature vector and the carbon-free power feature vector to obtain the fused feature vector f(a 1×b 1 ,a 2 ×b 2 ,a 3 ×b 3 ), and then the fused feature vector is input into the fuzzy autoregressive distributed lag model to predict the power demand, and the power demand prediction result is obtained. Therefore, by fusing the carbon emission power feature vector and the carbon-free power feature vector, more implicit features in the carbon emission power attribute data and the carbon-free power attribute data can be mined, which can make the features more complete, thereby improving the model analysis capability, that is, improving the prediction accuracy of power demand.

[0077] 208. The fused power feature vector is input into the fuzzy autoregressive distributed lag model to perform power demand forecasting and obtain the power demand forecasting result.

[0078] For the embodiment of the present invention, in order to improve the prediction accuracy of the fuzzy autoregressive distributed lag model, it is first necessary to construct a fuzzy autoregressive distributed lag model. Based on this, the method includes: determining an initial autoregressive distributed lag model, wherein the model function of the initial autoregressive distributed lag model is a 0 is a constant, a 1 is the regression coefficient at time t, is the time-lagged dependent variable (power demand forecast result) t-i The regression coefficient, β i,m is the time-lagged independent variable (carbon emission electricity attribute data and carbon-free electricity attribute data) x j,t-m The regression coefficient, ε t is the error term, i represents the i-th autocorrelation variable, j represents the j-th independent variable, k represents the number of factors affecting power demand forecasting, q represents the lag order of the time-lag independent variable, p represents the autocorrelation order of the time-lag dependent variable, and m represents the m-th time-lag independent variable; the membership function of the initial fuzzy number A~ is determined as Wherein, L() is the membership function, α is the center of the membership function, β is the expanded form of the membership function, and X represents the vector composed of the influencing factors of power demand forecasting; based on the initial fuzzy number A~, the regression coefficient of the initial autoregressive distributed lag model is modified to obtain the model function of the initial fuzzy autoregressive distributed lag model: Among them, X t =[1,t,y t-i ,x j,t-m ] t is the independent variable vector, represents the regression coefficient vector of fuzzy variables, is the power demand forecast result predicted by the initial fuzzy autoregressive distributed lag model; if the regression coefficient is a triangular fuzzy number, then determine the triangular fuzzy number The center and expansion type are (α l ,β l ); Based on the triangular fuzzy number The center and expansion type are (α l ,β l ) and the initial fuzzy autoregressive distributed lag model, determine the power demand forecast result y~ predicted by the initial fuzzy autoregressive distributed lag model t Membership in, X T represents the transpose of the vector consisting of the factors affecting the power demand forecast, X represents the vector consisting of the factors affecting the power demand forecast, Represents extended β l The transpose of To minimize As the objective function, the power demand forecast results Membership Take the confidence level h, if the power demand forecast result can be predicted when the membership degree is greater than h Then the constraint condition of the membership degree is obtained as follows: X t Represents the vector of the influencing factors of power demand forecast at time t; Based on the constraint condition of the membership degree, the first constraint condition of the initial fuzzy autoregressive distributed lag model is determined as The second constraint is represents the regression coefficient vector of fuzzy variables, β t represents the expanded form of the fuzzy variable regression coefficient vector; based on the first constraint condition and the second constraint condition, the triangular fuzzy number Extended beta l The minimum is the objective function, the fuzzy variable regression coefficients in the model function of the initial fuzzy autoregressive distributed lag model are optimized to obtain the initial fuzzy autoregressive distributed lag model after the optimized coefficients, and the initial fuzzy autoregressive distributed lag model after the optimized coefficients is determined as the fuzzy autoregressive distributed lag model.

[0079] Specifically, unlike static time series models that represent responses to current events, dynamic time series models use lagged variables to reflect the feedback of sequence variables over time. In power demand forecasting, many factors such as policy impacts cannot be simply described by static models, but their long-term effects must be considered. The autoregressive model, that is, a dynamic univariate model established using the lagged values ​​of the observed response variable, is usually expressed as follows:

[0080]

[0081] Among them, yt represents the observed value of the time series y at time point t, y t-n represents the observed value of the time series y at time point tn, n represents the lag order, is the autoregressive coefficient, ε t is the error term. The autoregressive model only relies on the lagged value of a single observed variable. If the power demand is only extrapolated based on the historical data of the autoregressive model, the impact of changes in policies and related external factors on power demand cannot be reflected. The autoregressive distributed lag model links the independent variable with the dependent variable and the lagged value. Based on this, the initial autoregressive distributed lag model is first determined. The initial autoregressive distributed lag model is a standard least squares regression model that includes independent variables and dependent variables. The model function of the initial autoregressive distributed lag model is as follows:

[0082]

[0083] Among them, a 0 is a constant, a 1 is the regression coefficient at time t, is the time-lagged dependent variable (power demand forecast result) t-i The regression coefficient, β i,m is the time-lagged independent variable (carbon emission electricity attribute data and carbon-free electricity attribute data) x j,t-m The regression coefficient, ε t is the error term, i represents the i-th autocorrelation variable, j represents the j-th independent variable, k represents the number of factors affecting power demand forecasting, q represents the lag order of the time-lagged independent variable, p represents the autocorrelation order of the time-lagged dependent variable, and m represents the m-th time-lagged independent variable.

[0084] Furthermore, since electricity demand and its influencing factors will be affected by various policy factors under the background of "dual carbon", and considering the uncertainty of the regression coefficient directly derived from historical data, the initial autoregressive distributed lag model is used to reflect the time lag effect of the variable, and the fuzzy number is introduced to reflect the uncertainty of the regression coefficient. The electricity demand forecasting method based on the fuzzy autoregressive distributed lag model is constructed to realize the electricity demand forecasting under the background of "dual carbon". Determine the fuzzy number The membership function of is shown in the following formula:

[0085]

[0086] Where L() is the membership function, α is the center of the membership function, β is the extended form of the membership function, and X represents the vector composed of the factors affecting power demand forecasting. If the membership function L satisfies the carbon emission constraint and the non-negativity constraint, then the fuzzy number is a symmetric fuzzy number, expressed as follows:

[0087] L(-x)=L(x)

[0088] L(0)=1

[0089] L(x 1 )≤L(x 2 ),0≤x 2 ≤x 1

[0090] Among them, x 1 represents the vector consisting of the first power demand forecast influencing factors, x 2 represents the vector composed of the first power demand forecast influencing factors, x represents the vector composed of the power demand forecast influencing factors, and further, the membership function of L is recorded as (α, β). When β = 0, the fuzzy relationship does not exist, and the fuzzy number That is, it becomes a real number. The specific establishment process of the fuzzy autoregressive distributed lag model is as follows:

[0091] ① Since there is a certain error in obtaining the regression coefficient through historical data, the regression coefficient is transformed into a fuzzy number to express the uncertainty of the coefficient. First, the regression coefficient of the initial autoregressive distributed lag model is expressed using fuzzy numbers to obtain the model function of the initial fuzzy autoregressive distributed lag model, as shown in the following formula:

[0092]

[0093] Among them, X t =[1,t,y t-i ,x j,t-m ] t represents the independent variable vector, Represents the regression coefficient vector of fuzzy variables, setting the regression coefficient to a triangular fuzzy number is the power demand forecast result predicted by the initial fuzzy autoregressive distributed lag model. Since the triangular fuzzy number has linear properties, the triangular fuzzy number can be calculated. The center and extension (α l ,β l ), and then the dependent variable y can be calculated. t Membership As shown in the following formula:

[0094]

[0095] Among them, X T represents the transpose of the vector consisting of the factors affecting the power demand forecast, X represents the vector consisting of the factors affecting the power demand forecast, Represents extended β l The transpose of .

[0096] ②Constructing the optimization model with minimum model uncertainty

[0097] From the model function of the initial fuzzy autoregressive distributed lag model, it can be seen that the error of the regression model prediction after the model is fuzzified is caused by the fuzziness of the regression coefficient. Therefore, the fuzzy independent variable regression coefficient vector A can be minimized by constructing an optimization model. t The fuzziness of the initial fuzzy number can be used to improve the prediction accuracy of the regression model. It can be seen from the membership function that when the expansion shape β is larger, the membership degree in a certain area (α-σ, α+σ) near the center value α of the membership function is also larger, indicating that The greater the probability that the actual value is at different points in (α-σ, α+σ), the greater the probability that the actual value is at different points in (α-σ, α+σ). Therefore, the extended form β represents the fuzziness of the fuzzy number. t is the objective function. The membership degree of takes the confidence level h, assuming that when the membership degree is greater than h, it can be observed that Combine dependent variables Membership The following formula is obtained:

[0098]

[0099] Among them, X t Represents the vector of factors affecting power demand forecast at time t. Further, simplifying the above formula, the first constraint and the second constraint of the initial fuzzy autoregressive distributed lag model can be obtained as follows:

[0100]

[0101] in, represents the regression coefficient vector of fuzzy variables, β t Represents the expanded form of the fuzzy variable regression coefficient vector. Further, according to the above constraints, a parameter estimation optimization model as shown below is constructed to obtain the model regression parameters that can minimize the model uncertainty, and then obtain the fuzzy autoregressive distributed lag model.

[0102]

[0103] Therefore, by introducing fuzzy numbers to reflect the uncertainty of regression coefficients, and constructing a fuzzy autoregressive distributed lag model based on fuzzy numbers, the constructed fuzzy autoregressive distributed lag model can reflect the time lag effect of variables when predicting power demand. Since there is a certain error in obtaining regression coefficients through historical data, transforming the regression coefficients into fuzzy numbers can express the uncertainty of the coefficients. At the same time, by constructing an optimization model to minimize the fuzziness of the regression coefficient vector of fuzzy independent variables, the prediction accuracy of the regression model can be improved, thereby improving the prediction accuracy of power demand.

[0104] Furthermore, after constructing the fuzzy autoregressive distributed lag model, the fused power feature vector is input into the fuzzy autoregressive distributed lag model for power demand forecasting, and the power demand forecasting result is obtained. Thus, the fuzziness of the fuzzy independent variable regression coefficient vector is minimized by constructing an optimization model. The model can handle the uncertainty and ambiguity in the prediction, improve the prediction accuracy, and adapt to the dynamic changes under the "dual carbon" policy.

[0105] According to another electricity demand forecasting method based on a fuzzy autoregressive distributed lag model provided by the present invention, compared with the prior art, the present invention decomposes the total historical carbon emissions into a carbon emission decomposition relationship between the historical carbon emission attribute data of each candidate influencing factor, and then determines the contribution of each candidate influencing factor to the change in the total carbon emissions according to the carbon emission decomposition relationship, and determines the electricity demand forecasting influencing factors related to carbon emissions according to the contribution, and finally predicts the electricity demand according to the electricity demand forecasting influencing factors using the fuzzy autoregressive distributed lag model, and selects the influencing factors related to the electricity demand forecasting according to the contribution of each candidate influencing factor to the change in the total carbon emissions, which can improve the determination accuracy of the influencing factors related to the electricity demand forecasting under the dual carbon background, and thus can improve the forecasting accuracy of the electricity demand under the dual carbon background, and at the same time, predicts the electricity demand through the fuzzy autoregressive distributed lag model, the fuzzy autoregressive distributed lag model can handle the uncertainty factors in the electricity demand forecasting under the "dual carbon" background through fuzzy theory, the model can flexibly respond to complex situations such as changes in energy structure, and provide reliable support for the electricity demand forecasting under different low-carbon paths, thereby improving the accuracy of the electricity demand forecasting under the dual carbon background.

[0106] Further, as Figure 1 In a specific implementation, the present invention provides a power demand forecasting device based on a fuzzy autoregressive distributed lag model, such as Figure 3 As shown, the device includes: a first acquisition unit 31, a decomposition unit 32, a first determination unit 33, a second determination unit 34, a second acquisition unit 35, and a prediction unit 36.

[0107] The first acquisition unit 31 can be used to obtain the historical total carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, wherein each of the candidate influencing factors is a variety of influencing factors that affect carbon emissions in the power industry.

[0108] The decomposition unit 32 may be used to decompose the total historical carbon emission into a relationship between the historical carbon emission attribute data of each of the candidate influencing factors to obtain a carbon emission decomposition relationship.

[0109] The first determination unit 33 may be configured to determine the contribution of each of the candidate influencing factors to the change in the total carbon emissions based on the carbon emissions decomposition equation and the historical carbon emissions attribute data.

[0110] The second determining unit 34 may be configured to determine, based on the impact contribution, an electricity demand forecasting influencing factor related to carbon emissions from among the candidate influencing factors.

[0111] The second acquisition unit 35 may be used to acquire carbon emission power attribute data corresponding to power demand forecast influencing factors related to carbon emission, and to acquire carbon-free power attribute data of power demand forecast influencing factors unrelated to carbon emission.

[0112] The prediction unit 36 ​​may be used to input the carbon emission power attribute data and the carbon-free power attribute data into a fuzzy autoregressive distributed lag model to perform power demand prediction and obtain a power demand prediction result.

[0113] In a specific application scenario, in order to determine the carbon emission decomposition relationship, the decomposition unit 32 can be specifically used to decompose the total historical carbon emission into the relationship between the historical carbon emission attribute data of each of the selected influencing factors, and obtain the carbon emission decomposition relationship as follows:

[0114] Among them, C is the total historical carbon emissions.

[0115] In a specific application scenario, in order to determine the contribution of each candidate influencing factor to the change in the total carbon emissions, the first determining unit 33 can be specifically used to decompose the carbon emission decomposition relationship using the logarithmic mean Dirichlet index method to obtain the carbon emission coefficient influence relationship, and based on the carbon emission coefficient F of the i-th energy in year 0 i 0 and the carbon emission coefficient F of the i-th energy source in year T i T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T Carbon emission coefficient influence relationship, determine the contribution of carbon emission intensity influencing factors ΔC em ,in, The carbon emission decomposition relationship is decomposed by the logarithmic mean Dimitrov index method to obtain the power structure impact relationship, and the power structure of the i-th energy source in year 0 is calculated based on the power structure of the i-th energy source in year 0. and the power structure of the i-th energy source in year T Carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The power structure influence relationship is used to determine the contribution of the power structure influence factors ΔC es ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dirichlet index method to obtain the electrification rate impact equation, and based on the electrification rate D in year 0 0 and the electrification rate D in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of factors affecting energy structure ΔC int ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dimitrov index method to obtain the energy intensity impact equation, and based on the energy intensity Q 0 and the energy intensity Q in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of energy intensity influencing factors ΔC dp ,in, The carbon emission decomposition equation is decomposed by using the logarithmic mean Dimitrov index method to obtain the economic development impact equation, and based on the per capita gross domestic product Y in year 0 0 and the per capita GDP Y in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of economic development factors ΔC ey ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dimitrov index method to obtain the population impact equation, and based on the population size P in year 0 0 and the population size P in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of population influence factors ΔC P ,in,

[0116] In a specific application scenario, in order to determine the influencing factors of power demand prediction related to carbon emissions, the second determining unit 34 can be specifically used to determine the influencing contribution ΔC based on the influencing factors of carbon emission intensity. em , the contribution of the factors affecting the power structure ΔC es, the contribution of factors affecting energy structure ΔC int , Contribution of factors affecting energy intensity ΔC dp , the contribution of economic development influencing factors ΔC ey , the contribution of population influencing factors ΔC P , determine the initial electricity demand forecast influencing factors related to carbon emissions from each of the candidate influencing factors; determine the influencing factor standard deviation σ of the historical carbon emission attribute data corresponding to each of the initial electricity demand forecast influencing factors xi , the standard deviation of the historical actual power demand value σ y , the i-th initial power demand forecast influencing factor X among the initial power demand forecast influencing factors i and the jth initial power demand forecast influencing factor X j The correlation coefficient C ij 、The influencing factor X of the i-th initial power demand forecast i The partial regression coefficient λ i ; Based on the standard deviation of the influencing factors σ xi , Standard deviation of demand value σ y , correlation coefficient C ij , partial regression coefficient λ i , determine the path coefficient P between each of the initial power demand forecast influencing factors and the historical actual power demand value i ,in, Based on the path coefficient Pi, the path coefficient judgment coefficient R of each of the initial power demand forecast influencing factors is determined. 2 ,in, C iy represents the influencing factor X of the i-th initial power demand forecast i The correlation coefficient between the initial electricity demand forecast value Y and the electricity demand forecast influencing factors related to carbon emissions is determined among the initial electricity demand forecast influencing factors based on the path coefficient judgment coefficient of each of the initial electricity demand forecast influencing factors.

[0117] In specific application scenarios, in order to optimize the energy structure of the factors affecting power demand forecasting, such as Figure 4 As shown, the device further comprises an optimization unit 37 .

[0118] The first determination unit 33 can also be used to determine the total energy demand balance condition in the power demand forecasting process, wherein the total energy demand balance condition is S i,t =W i,t +F i,t , S i,tThe predicted energy supply of power supply energy in the energy structure influencing factors, W i,t is the amount of unused electricity generation energy in the ith energy source in the tth year, F i,t is the utilization of the i-th power generation energy in the t-th year, D t is the total energy demand planned in the electricity demand forecasting process, L t is the amount of clean energy consumed by the energy system in year t, i is the i-th energy in the energy structure influencing factors, n is the total amount of energy in the energy structure influencing factors, C t is the predicted total carbon emissions in the electricity demand forecasting process, C i,t is the carbon emission coefficient of the i-th energy in the t-th year.

[0119] The first determination unit 33 may also be used to achieve carbon peak and carbon neutrality in a preset future period based on the predicted energy supply S i,t 、Total energy demand D t , predict total carbon emissions C t , the amount of energy used in the ith energy source in the tth year F i,t , determine the carbon emission constraints and non-negativity constraints in the power demand forecasting process, where the carbon emission constraint is C t ≤CS i,t , C.S. i,t is the upper limit of carbon emission constraint, and the non-negativity constraint condition is L t ,F i,t ≥0,

[0120] The first determination unit 33 can also be used to determine carbon peak constraints and carbon neutrality constraints based on the carbon emission constraints and the non-negativity constraints.

[0121] The optimization unit 37 can be used to calculate the clean energy consumption L of the energy system in the tth year based on the total energy demand balance condition, the carbon peak constraint condition, and the carbon neutrality constraint condition. t The minimum is the objective function, and the energy structure in the factors affecting power demand forecast is optimized to obtain the optimized energy structure.

[0122] The second acquisition unit 35 can be specifically used to acquire carbon emission power attribute data corresponding to the power demand forecast influencing factors after optimizing the energy structure.

[0123] In a specific application scenario, in order to construct a power demand prediction model, the device further includes a construction unit 38.

[0124] The construction unit 38 may be used to determine an initial autoregressive distributed lag model, wherein the model function of the initial autoregressive distributed lag model is: a 0 is a constant, a 1 is the regression coefficient at time t, is the time-lagged dependent variable (power demand forecast result) t-i The regression coefficient, β i,m is the time-lagged independent variable (carbon emission electricity attribute data and carbon-free electricity attribute data) x j,t-m The regression coefficient, ε t is the error term, i represents the i-th autocorrelation variable, j represents the j-th independent variable, k represents the number of factors affecting power demand forecasting, q represents the lag order of the time-lag independent variable, p represents the autocorrelation order of the time-lag dependent variable, and m represents the m-th time-lag independent variable; determine the initial fuzzy number The membership function is Wherein, L() is the membership function, α is the center of the membership function, β is the extended form of the membership function, and X represents the vector composed of the influencing factors of power demand forecasting; based on the initial fuzzy number The regression coefficient of the initial autoregressive distributed lag model is modified to obtain the model function of the initial fuzzy autoregressive distributed lag model: Among them, X t =[1,t,y t-i ,x j,t-m ] t is the independent variable vector, represents the regression coefficient vector of fuzzy variables, is the power demand forecast result predicted by the initial fuzzy autoregressive distributed lag model; if the regression coefficient is a triangular fuzzy number, then determine the triangular fuzzy number The center and expansion type are (α l ,β l ); Based on the triangular fuzzy number The center and expansion type are (α l ,β l ) and the initial fuzzy autoregressive distributed lag model, determine the power demand forecast result predicted by the initial fuzzy autoregressive distributed lag model Membership in, X T represents the transpose of the vector consisting of the factors affecting power demand forecast, Represents extended β l The transpose of To minimize As the objective function, the power demand forecast results Membership Take the confidence level h, if the power demand forecast result can be predicted when the membership degree is greater than h Then the constraint condition of the membership degree is obtained as follows: X t Represents the vector of the influencing factors of power demand forecast at time t; Based on the constraint condition of the membership degree, the first constraint condition of the initial fuzzy autoregressive distributed lag model is determined as The second constraint is represents the regression coefficient vector of fuzzy variables, β t represents the expanded form of the fuzzy variable regression coefficient vector; based on the first constraint condition and the second constraint condition, the triangular fuzzy number Extended beta l The minimum is the objective function, the fuzzy variable regression coefficients in the model function of the initial fuzzy autoregressive distributed lag model are optimized to obtain the initial fuzzy autoregressive distributed lag model after the optimized coefficients, and the initial fuzzy autoregressive distributed lag model after the optimized coefficients is determined as the fuzzy autoregressive distributed lag model.

[0125] In a specific application scenario, in order to perform power demand prediction, the prediction unit 36 ​​includes a determination module 361 , a multiplication module 362 , and a prediction module 363 .

[0126] The determination module 361 may be used to determine the carbon emission power feature vector corresponding to the carbon emission power attribute data, and determine the carbon-free power feature vector corresponding to the carbon-free power attribute data.

[0127] The multiplication module 362 may be used to multiply each element in the carbon emission power feature vector with the corresponding position element in the carbon-free power feature vector to obtain a fused power feature vector.

[0128] The prediction module 363 can be used to input the fused power feature vector into the fuzzy autoregressive distributed lag model to perform power demand prediction and obtain a power demand prediction result.

[0129] It should be noted that for other corresponding descriptions of the functional modules involved in the power demand forecasting device based on the fuzzy autoregressive distributed lag model provided in the embodiment of the present invention, reference can be made to Figure 1 The corresponding description of the method shown will not be repeated here.

[0130] Based on the above Figure 1The method shown, accordingly, an embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the following steps are implemented: obtaining the historical total carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, wherein each of the candidate influencing factors is a various influencing factors that affect carbon emissions in the power industry; decomposing the historical total carbon emissions into the relationship form between the historical carbon emission attribute data of each candidate influencing factor, and obtaining a carbon emission decomposition relationship; based on the carbon emission decomposition relationship and the historical carbon emission attribute data, determining the contribution of each candidate influencing factor to the change in the total carbon emissions; based on the contribution, determining the electricity demand forecast influencing factor related to carbon emissions among each of the candidate influencing factors; obtaining the carbon emission electricity attribute data corresponding to the electricity demand forecast influencing factor related to carbon emissions, and obtaining the carbon-free electricity attribute data of the electricity demand forecast influencing factor not related to carbon emissions; inputting the carbon emission electricity attribute data and the carbon-free electricity attribute data into the fuzzy autoregressive distributed lag model to perform electricity demand forecasting, and obtaining an electricity demand forecasting result.

[0131] Based on the above Figure 1 The method shown and Figure 3 The embodiment of the device shown in the figure, the embodiment of the present invention also provides a physical structure diagram of a computer device, such as Figure 5 As shown, the computer device includes: a processor 41, a memory 42, and a computer program stored in the memory 42 and executable on the processor, wherein the memory 42 and the processor 41 are both arranged on a bus 43, and when the processor 41 executes the program, the following steps are implemented: obtaining the historical total carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual power demand value, wherein each candidate influencing factor is a variety of influencing factors that affect carbon emissions in the power industry; decomposing the historical total carbon emissions into the relationship between the historical carbon emission attribute data of each candidate influencing factor, and obtaining the carbon emission decomposition relationship; based on the carbon emission decomposition relationship and the historical carbon emission attribute data, determine the impact contribution of each of the candidate influencing factors on the change of the total carbon emission; based on the impact contribution, determine the electricity demand forecast influencing factors related to carbon emissions among the candidate influencing factors; obtain the carbon emission electricity attribute data corresponding to the electricity demand forecast influencing factors related to carbon emissions, and obtain the carbon-free electricity attribute data of the electricity demand forecast influencing factors not related to carbon emissions; input the carbon emission electricity attribute data and the carbon-free electricity attribute data into the fuzzy autoregressive distributed lag model to perform electricity demand forecasting, and obtain the electricity demand forecasting result.

[0132] Through the technical scheme of the present invention, the present invention obtains the historical total carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, wherein each candidate influencing factor is a various influencing factor that affects carbon emissions in the power industry; and decomposes the historical total carbon emissions into the relationship form between the historical carbon emission attribute data of each candidate influencing factor to obtain a carbon emission decomposition relationship; at the same time, based on the carbon emission decomposition relationship and the historical carbon emission attribute data, the contribution of each candidate influencing factor to the change in the total carbon emissions is determined; and based on the contribution, the electricity demand forecast influencing factor related to carbon emissions is determined among each candidate influencing factor; then, the carbon emission electricity attribute data corresponding to the electricity demand forecast influencing factor related to carbon emissions is obtained, and the carbon-free electricity attribute data of the electricity demand forecast influencing factor not related to carbon emissions is obtained; finally, the carbon emission electricity attribute data and the carbon-free electricity attribute data are input into a fuzzy autoregressive distributed lag model to perform electricity demand forecasting, and obtain an electricity demand forecasting result. Therefore, by decomposing the total historical carbon emissions into the carbon emission decomposition relationship between the historical carbon emission attribute data of each candidate influencing factor, the contribution of each candidate influencing factor to the change in the total carbon emissions is determined according to the carbon emission decomposition relationship, and based on the contribution, the electricity demand forecasting influencing factors related to carbon emissions are determined. Finally, according to the electricity demand forecasting influencing factors, the electricity demand is predicted using the fuzzy autoregressive distributed lag model. The influencing factors related to the electricity demand forecast are selected based on the contribution of each candidate influencing factor to the change in the total carbon emissions. This can improve the accuracy of determining the influencing factors related to the electricity demand forecast under the dual carbon background, and thus improve the forecast accuracy of electricity demand under the dual carbon background. At the same time, the fuzzy autoregressive distributed lag model is used to predict electricity demand. The fuzzy autoregressive distributed lag model can handle the uncertainty factors in the electricity demand forecast under the "dual carbon" background through fuzzy theory. The model can flexibly respond to complex situations such as changes in the energy structure, and provide reliable support for electricity demand forecasting under different low-carbon paths, thereby improving the accuracy of electricity demand forecasting under the dual carbon background.

[0133] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0134] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting power demand based on fuzzy autoregressive distributed lag model, characterized in that: include: Obtain the total historical carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual electricity demand value, wherein each of the candidate influencing factors is a variety of influencing factors that affect carbon emissions in the power industry; Decomposing the total historical carbon emissions into a relationship between the historical carbon emission attribute data of each of the selected influencing factors to obtain a carbon emission decomposition relationship; Determine the contribution of each of the candidate influencing factors to the change in total carbon emissions based on the carbon emission decomposition relationship and the historical carbon emission attribute data; Based on the impact contribution, determining the power demand forecast influencing factor related to carbon emissions from each of the candidate influencing factors; Obtaining carbon emission power attribute data corresponding to power demand forecast influencing factors related to carbon emission, and obtaining carbon-free power attribute data of power demand forecast influencing factors unrelated to carbon emission; The carbon emission electricity attribute data and the carbon-free electricity attribute data are input into a fuzzy autoregressive distributed lag model to perform electricity demand forecasting to obtain an electricity demand forecasting result.

2. The method according to claim 1, characterized in that The influencing factors to be selected include power structure influencing factors, energy structure influencing factors, economic development influencing factors, population influencing factors, carbon emission intensity influencing factors, and energy intensity influencing factors. The historical carbon emission attribute data of the power structure influencing factors are the electricity consumption H, power structure e of each historical year, and the energy intensity influencing factors. i The historical carbon emission attribute data of the energy structure influencing factors is the carbon emission C generated by the electricity produced by the i-th energy source in each historical year. i , the amount of electricity produced by the i-th energy source E i , total energy consumption E, electrification rate D, the historical carbon emission attribute data of the economic development influencing factors are the gross national product G and per capita gross domestic product Y of each historical year, the historical carbon emission attribute data of the population influencing factors are the population P of each historical year, and the historical carbon emission attribute data of the carbon emission intensity influencing factors are the carbon emission coefficient F of each historical year. i ,The historical carbon emission attribute data of energy intensity influencing factors is the energy intensity Q of each historical year; Decomposing the total historical carbon emissions into the relationship between the historical carbon emission attribute data of each of the selected influencing factors to obtain a carbon emission decomposition relationship includes: The total historical carbon emissions are decomposed into the relationship between the historical carbon emission attribute data of each of the selected influencing factors, and the carbon emission decomposition relationship is obtained as follows: Among them, C is the total historical carbon emissions.

3. The method according to claim 2, characterized in that The determining, based on the carbon emission decomposition relation and the historical carbon emission attribute data, the contribution of each of the candidate influencing factors to the change in the total carbon emission amount includes: The carbon emission decomposition relationship is decomposed by the logarithmic mean Dirichlet index method to obtain the carbon emission coefficient influence relationship, and the carbon emission coefficient F of the i-th energy source in year 0 is calculated. i 0 and the carbon emission coefficient F of the i-th energy source in year T i T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T Carbon emission coefficient influence relationship, determine the contribution of carbon emission intensity influencing factors ΔC em ,in, The carbon emission decomposition relationship is decomposed by the logarithmic mean Dimitrov index method to obtain the power structure impact relationship, and the power structure of the i-th energy source in year 0 is calculated based on the power structure of the i-th energy source in year 0. and the power structure of the i-th energy source in year T Carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The power structure influence relationship is used to determine the influence contribution of the power structure influencing factors ΔC es ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dirichlet index method to obtain the electrification rate impact equation, and based on the electrification rate D in year 0 0 and the electrification rate D in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of factors affecting energy structure ΔC int ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dimitrov index method to obtain the energy intensity impact equation, and based on the energy intensity Q 0 and the energy intensity Q in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of energy intensity influencing factors ΔC dp ,in, The carbon emission decomposition equation is decomposed by using the logarithmic mean Dimitrov index method to obtain the economic development impact equation, and based on the per capita gross domestic product Y in year 0 0 and the per capita GDP Y in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of economic development factors ΔC ey ,in, The carbon emission decomposition equation is decomposed by the logarithmic mean Dimitrov index method to obtain the population impact equation, and based on the population size P in year 0 0 and the population size P in year T T , the carbon emissions of the i-th energy source in year 0 and the carbon emissions of the i-th energy source in year T The influence relationship of electrification rate is used to determine the contribution of population influence factors ΔC P ,in, 4. The method according to claim 1, characterized in that: The determining of the power demand forecast influencing factors related to carbon emissions from the selected influencing factors based on the influencing contribution amount includes: Contribution ΔC based on factors affecting carbon emission intensity em , the contribution of the factors affecting the power structure ΔC es , the contribution of factors affecting energy structure ΔC int , Contribution of factors affecting energy intensity ΔC dp , the contribution of economic development influencing factors ΔC ey , the contribution of population influencing factors ΔC P , determining the initial electricity demand forecast influencing factors related to carbon emissions among the candidate influencing factors; Determine the standard deviation σ of the influencing factors of the historical carbon emission attribute data corresponding to each of the influencing factors of the initial power demand forecast xi , the standard deviation of the historical actual power demand value σ y , the i-th initial power demand forecast influencing factor X among the initial power demand forecast influencing factors i and the jth initial power demand forecast influencing factor X j The correlation coefficient C ij 、The influencing factor X of the i-th initial power demand forecast i The partial regression coefficient λ i ; Based on the standard deviation of the influencing factors σ xi , Standard deviation of demand value σ y , correlation coefficient C ij , partial regression coefficient λ i , determine the path coefficient P between each of the initial power demand forecast influencing factors and the historical actual power demand value i ,in, Based on the path coefficient Pi, the path coefficient judgment coefficient R of each of the initial power demand forecast influencing factors is determined. 2 ,in, C iy represents the influencing factor X of the i-th initial power demand forecast i The correlation coefficient with the power demand forecast value Y; Based on the path coefficient judgment coefficient of each of the initial power demand forecast influencing factors, the power demand forecast influencing factor related to carbon emissions is determined among the initial power demand forecast influencing factors.

5. The method according to claim 1, characterized in that Before obtaining the carbon emission power attribute data corresponding to the power demand forecast influencing factors related to carbon emission, the method further includes: Determine the total energy demand balance condition in the power demand forecasting process, where the total energy demand balance condition is S i,t =W i,t +F i,t , S i,t The predicted energy supply of power supply energy in the energy structure influencing factors, W i,t is the amount of unused electricity generation energy in the ith energy source in the tth year, F i,t is the utilization of the i-th power generation energy in the t-th year, D t is the total energy demand planned in the electricity demand forecasting process, L t is the amount of clean energy consumed by the energy system in year t, i is the i-th energy in the energy structure influencing factors, n is the total amount of energy in the energy structure influencing factors, C t is the predicted total carbon emissions in the electricity demand forecasting process, C i,t is the carbon emission coefficient of the i-th energy source in year t; In order to achieve carbon peak and carbon neutrality within the preset years in the future, based on the predicted energy supply S i,t 、Total energy demand D t , predict total carbon emissions C t , the amount of energy used in the ith energy source in the tth year F i,t , determine the carbon emission constraints and non-negativity constraints in the power demand forecasting process, where the carbon emission constraint is C t ≤CS i,t , C.S. i,t is the upper limit of carbon emission constraint, and the non-negativity constraint condition is L t ,F i,t ≥0, Based on the carbon emission constraint condition and the non-negativity constraint condition, determine the carbon peak constraint condition and the carbon neutrality constraint condition; Based on the total energy demand balance condition, the carbon peak constraint condition, and the carbon neutrality constraint condition, the clean energy consumption of the energy system in year t is L t Minimum is the objective function, and the energy structure in the factors affecting power demand forecast is optimized to obtain the optimized energy structure; The obtaining of carbon emission power attribute data corresponding to power demand forecast influencing factors related to carbon emission includes: Obtain carbon emission electricity attribute data corresponding to the influencing factors of electricity demand forecast after optimizing the energy structure.

6. The method according to claim 1, characterized in that Before inputting the carbon emission power attribute data and the carbon-free power attribute data into the fuzzy autoregressive distributed lag model to perform power demand forecasting and obtain the power demand forecasting result, the method further includes: Determine an initial autoregressive distributed lag model, wherein the model function of the initial autoregressive distributed lag model is a0 is a constant, a1 is the regression coefficient at time t, is the time-lagged dependent variable (power demand forecast result) t-i The regression coefficient, β i,m is the time-lagged independent variable (carbon emission electricity attribute data and carbon-free electricity attribute data) x j,t-m The regression coefficient, ε t is the error term, i represents the i-th autocorrelation variable, j represents the j-th independent variable, k represents the number of factors affecting power demand forecasting, q represents the lag order of the time-lagged independent variable, p represents the autocorrelation order of the time-lagged dependent variable, and m represents the m-th time-lagged independent variable; Determine the initial fuzzy number The membership function is Wherein, L() is the membership function, α is the center of the membership function, β is the extended form of the membership function, and X represents the vector composed of the factors affecting power demand forecasting; Based on the initial fuzzy number The regression coefficient of the initial autoregressive distributed lag model is modified to obtain the model function of the initial fuzzy autoregressive distributed lag model: Among them, X t =[1,t,y t-i ,x j,t-m ] t is the independent variable vector, represents the regression coefficient vector of fuzzy variables, It is the power demand forecast result predicted by the initial fuzzy autoregressive distributed lag model; If the regression coefficient is a triangular fuzzy number, then determine the triangular fuzzy number The center and expansion type are (α l ,β l ); Based on the triangular fuzzy number The center and expansion type are (α l ,β l ) and the initial fuzzy autoregressive distributed lag model, determine the power demand forecast result predicted by the initial fuzzy autoregressive distributed lag model Membership in, X T represents the transpose of the vector consisting of the factors affecting the power demand forecast, X represents the vector consisting of the factors affecting the power demand forecast, Represents extended β l The transpose of Based on the membership To minimize As the objective function, the power demand forecast results Membership Take the confidence level h, if the power demand forecast result can be predicted when the membership degree is greater than h Then the constraint condition of the membership degree is obtained as follows: X t represents the vector of factors influencing power demand forecast at time t; Based on the constraint condition of the membership degree, the first constraint condition of the initial fuzzy autoregressive distributed lag model is determined as: The second constraint is represents the regression coefficient vector of fuzzy variables, β t Represents the expanded form of the regression coefficient vector of fuzzy variables; Based on the first constraint and the second constraint, a triangular fuzzy number Extended beta l The minimum is the objective function, the fuzzy variable regression coefficients in the model function of the initial fuzzy autoregressive distributed lag model are optimized to obtain the initial fuzzy autoregressive distributed lag model after the optimized coefficients, and the initial fuzzy autoregressive distributed lag model after the optimized coefficients is determined as the fuzzy autoregressive distributed lag model.

7. The method according to claim 1, characterized in that The step of inputting the carbon emission power attribute data and the carbon-free power attribute data into a fuzzy autoregressive distributed lag model to perform power demand forecasting to obtain a power demand forecasting result includes: Determine a carbon emission power feature vector corresponding to the carbon emission power attribute data, and determine a carbon-free power feature vector corresponding to the carbon-free power attribute data; Multiply each element in the carbon emission power feature vector by the corresponding position element in the carbon-free power feature vector to obtain a fused power feature vector; The fused power feature vector is input into the fuzzy autoregressive distributed lag model to perform power demand forecasting to obtain a power demand forecasting result.

8. A power demand forecasting device based on fuzzy autoregressive distributed lag model, characterized in that: include: A first acquisition unit is used to acquire the total historical carbon emissions of the target area within the target time period, the historical carbon emission attribute data of each candidate influencing factor, and the historical actual power demand value, wherein each of the candidate influencing factors is a variety of influencing factors that affect carbon emissions in the power industry; A decomposition unit, used for decomposing the total historical carbon emissions into a relationship between the historical carbon emission attribute data of each of the candidate influencing factors to obtain a carbon emission decomposition relationship; A first determination unit is used to determine the contribution of each of the candidate influencing factors to the change in the total carbon emissions based on the carbon emission decomposition relationship and the historical carbon emission attribute data; A second determining unit is used to determine, based on the impact contribution, an electricity demand forecast influencing factor related to carbon emissions from among the candidate influencing factors; A second acquisition unit is used to acquire carbon emission power attribute data corresponding to power demand forecast influencing factors related to carbon emission, and to acquire carbon-free power attribute data of power demand forecast influencing factors unrelated to carbon emission; The prediction unit is used to input the carbon emission power attribute data and the carbon-free power attribute data into a fuzzy autoregressive distributed lag model to perform power demand prediction and obtain a power demand prediction result.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images