Clean energy yield prediction method and system

Abstract

The invention discloses a clean energy yield prediction method and system, and relates to the technical field of environmental protection engineering. The STIRPAT variable coefficient model is provided and is used for accurately predicting the clean energy yield of a region. The model aims at calculating and evaluating the clean energy yield of each region and analyzing the influence of different energy activities, policies and economic development on clean energy production. According to the method, the advantages of the integrated STIRPAT model and the variable coefficient model are combined, and the influence mechanism of the clean energy yield can be effectively established. Through adoption of a self-adaptive optimizer method of dynamic element optimization, the unknown quantity of the model is efficiently estimated, and the accuracy of a prediction result is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental protection engineering, and particularly to a method and system for predicting the output of clean energy. Background Art

[0002] With the increasingly severe global climate change problem, promoting the development and application of clean energy has become the core goal for countries around the world to achieve sustainable development. Clean energy not only helps to reduce the dependence on fossil energy, but also can significantly reduce greenhouse gas emissions, and is an important means to achieve the green and low-carbon transformation and protect the ecological environment.

[0003] As an important energy resource-rich area in China, the Yellow River Basin is rich in clean energy resources such as hydropower, wind energy, and solar energy, and has great development potential. The Yellow River Basin covers many places, and there are significant differences in aspects such as per capita GDP, energy structure, and natural conditions in the covered areas. Especially, the distribution of hydropower, wind energy, and solar energy resources shows obvious regional differences. How to rationally allocate clean energy resources in such a diverse region and give full play to their maximum utilization potential has become an urgent problem to be solved.

[0004] In the production process of clean energy, multiple factors will affect the energy output and utilization efficiency. For example, climate change has a particularly significant impact on wind energy and solar energy. Factors such as sunshine duration, wind speed change, and precipitation in different regions directly determine the output of clean energy. In addition, the construction of regional power transmission and power transmission facilities, the policy support of local governments, and the degree of advancement of scientific and technological innovation will all affect the actual development progress of clean energy. Therefore, how to accurately predict the output of clean energy in this region and evaluate the impact of different natural factors and policy factors on energy production has become the key to promoting the sustainable development of clean energy in the Yellow River Basin.

[0005] Currently, the existing methods for predicting clean energy mainly rely on traditional statistical analysis and empirical models. However, these methods have certain limitations when facing the changing climate, complex geographical environment, and different per capita GDP in the Yellow River Basin. Especially, the regional differences and dynamic changes make it difficult for traditional models to accurately predict energy output. With the development of technologies such as big data and artificial intelligence, new prediction methods based on machine learning and data mining have gradually been proposed and applied to the prediction of clean energy. However, how to combine the actual situation of the Yellow River Basin, through multi-dimensional and multi-level data analysis and modeling, to form a more accurate clean energy prediction system is still an urgent problem to be solved. Summary of the Invention

[0006] To overcome the problem in the above-mentioned existing technologies of how to combine the actual situation of the Yellow River Basin, through multi-dimensional and multi-level data analysis and modeling, to form a more accurate clean energy prediction system, the main object of the present invention is to provide a method and system for predicting clean energy production.

[0007] To achieve the above object, the present invention adopts the following technical solutions. A method for predicting clean energy production includes:

[0008] Obtain multiple indicators related to clean energy and determine the factors affecting the indicators. The indicators include historical clean energy production, environmental protection expenditure, and per capita GDP. The factors affecting the indicators include technological level, urbanization level, and export volume;

[0009] Construct a linearized STIRPAT model by combining multiple indicators, obtain the logarithmic function of the STIRPAT model, and construct a STIRPAT variable coefficient model based on the factors affecting the indicators;

[0010] Construct a loss function using the logarithmic function, and adopt dynamic meta-optimization based on an adaptive optimizer until the loss function reaches the upper limit of the number of iterations, and obtain the coefficients of the determined STIRPAT variable coefficient model;

[0011] Obtain the environmental protection expenditure, per capita GDP, technological level, urbanization level, export volume, and historical clean energy production of the area to be predicted in the prediction year, input the STIRPAT variable coefficient model determined by the coefficients, and obtain the prediction result of the clean energy production in the area to be predicted.

[0012] The linear STIRPAT model is expressed by the following formula:

[0013]

[0014] Where, I t represents the environmental index representing historical clean energy production ENP; a represents the constant term, A t represents the GDP representing per capita GDP, P t represents the urbanization level UR representing the population size; T t represents the technological progress RD representing the technological level, e t represents the error term;

[0015] The construction of the STIRPAT variable coefficient model is expressed by the following formula:

[0016] LENP it =La + β 1 LGDP it +β 2 LENGP it +β 3(LENGP × GDP) it + β 4 LRD it + β 5 LUR it + β 6 LEX it + Zα(u it ) + ε it

[0017] Wherein, L() represents the logarithmic function; ENP it represents the observed value of historical clean energy production in the i-th individual at the t-th period and the t-th period; GDP it represents the observed value of economic development in the i-th individual at the t-th period and the t-th period; ENGP it represents the observed value of environmental protection expenditure in the i-th individual at the t-th period; (ENGP × GDP) it represents the observed value of the cross-term of environmental protection expenditure and per capita GDP in the i-th individual at the t-th period; RD it represents the observed value of scientific and technological progress in the i-th individual at the t-th period; UR it represents the observed value of the urbanization level in the i-th individual at the t-th period; EX it represents the observed value of the export volume in the i-th individual at the t-th period; u it represents the observed value of the energy consumption structure in the i-th individual at the t-th period, β i (i = 1, 2, 3, 4, 5, 6) represents the coefficient of the independent variable; Z = (GDP it ', LENGP it ', RD it ', UR it ', EX it '), α(·) represents the non-parametric influence mechanism of each variable on NEP, ε it represents the error term, and i = 1, …, N, t = 1, …, T represent the number of individuals and time respectively.

[0018] The obtained STIRPAT variable coefficient model for coefficient determination includes the following steps:

[0019] The initial coefficient of the linearized STIRPAT model is the initial weight of multiple indicators. Set the hyperparameters and coefficients of the linearized STIRPAT model, including the initial learning rate, time window size, and dynamic feedback mechanism;

[0020] Based on the linear STIRPAT model structure, substitute the initialized coefficient into the STIRPAT variable coefficient model to construct the initial STIRPAT variable coefficient model;

[0021] Based on the clean energy output as the prediction target and the actual observed data, a loss function is constructed using a logarithmic function; the formula is as follows:

[0022]

[0023] Dynamic meta-optimization is adopted until the upper limit of the number of iterations of the loss function is reached, and the coefficients of the determined STIRPAT variable coefficient model are obtained;

[0024] The determined coefficients are substituted into the STIRPAT variable coefficient model to obtain the final clean energy output prediction model.

[0025] The dynamic meta-optimization includes:

[0026] According to the STIRPAT variable coefficient model and data attributes, the size of the time window is set, and the data set is divided into multiple subsets according to the time window, and each subset contains the data within one time window; within each time window, the STIRPAT variable coefficient model is trained using the observed data;

[0027] According to the performance metrics, a dynamic feedback rule is designed for adjusting the learning rate and model coefficients;

[0028] Based on the error between the predicted value of the initial STIRPAT variable coefficient model and the actual observed value within each time window, the prediction error is obtained;

[0029] Based on the prediction error, the gradient information of the STIRPAT variable coefficient model coefficients is obtained using backpropagation;

[0030] According to the gradient information and the preset dynamic feedback rule, the adjustment strategy of the learning rate is confirmed, and the adjustment strategy includes learning rate decay and learning rate restart;

[0031] In the inner layer optimization, grid search is selected as the adaptive optimizer, and the grid search updates the STIRPAT variable coefficient model coefficients according to the gradient information; the formula is expressed as:

[0032]

[0033] Among them, represents the parameter value of task t in the kth iteration; η represents the learning rate of meta-optimization, and the initial value is 10 -3 ;

[0034] In the outer layer optimization, the hyperparameters of the inner layer optimization are adjusted through the meta-optimization method, and the formula is expressed as:

[0035]

[0036] Among them, θ s represents the value of the hyperparameter.

[0037] For the inner layer optimization, the grid search is selected as the most adaptive optimizer. The grid search updates the coefficients of the STIRPAT variable coefficient model according to the gradient information, including the following steps:

[0038] Define the set of all possible values of the parameters in the STIRPAT variable coefficient model as the parameter space; divide the grid in the parameter space, and each grid point represents a set of parameter values;

[0039] For each grid point in the parameter space, use the STIRPAT variable coefficient model for training and obtain the corresponding loss function value;

[0040] Compare the loss function values at different grid points, and select the parameter value with the minimum loss as the current optimal solution;

[0041] According to the gradient information of the loss function with respect to the parameters, guide the direction and step size of the grid search to obtain the optimal parameters; the formula is expressed as:

[0042] θ t+1 = θ t - α▽L(θ t )

[0043] where θ t is the parameter in the t-th iteration; α is the learning rate, and α▽L(θ t ) is the gradient of the loss function L at θ t ;

[0044] Update the parameters using the gradient information, find the new parameter values by searching on the grid until the loss function value is lower than the set threshold, and obtain a set of optimized parameter values as the parameter values determined by the STIRPAT variable coefficient model.

[0045] A clean energy production prediction system, comprising:

[0046] A data acquisition module, configured to acquire multiple indicators related to clean energy and determine the factors affecting the indicators. The indicators include historical clean energy production, environmental protection expenditure, and per capita GDP, and the factors affecting the indicators include technological level, urbanization level, and export volume;

[0047] A prediction model construction module, configured to construct a linearized STIRPAT model by combining multiple indicators, obtain the logarithmic function of the STIRPAT model, and construct a STIRPAT variable coefficient model based on the factors affecting the indicators; construct a loss function using the logarithmic function, and perform dynamic meta-optimization based on an adaptive optimizer until the loss function reaches the upper limit of the number of iterations, and obtain the coefficients of the determined STIRPAT variable coefficient model;

[0048] The production prediction module is used to obtain the environmental protection expenditure, per capita GDP, scientific and technological level, urbanization level, export volume of the area to be predicted in the prediction year, and the historical clean energy production of the area to be predicted before the prediction year, input the STIRPAT variable coefficient model determined by the coefficients, and obtain the prediction result of the clean energy production in the area to be predicted.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows: Through the adaptive optimizer of dynamic meta-optimization, combined with grid search and gradient information, the coefficients of the STIRPAT variable coefficient model can be determined more precisely, thereby improving the prediction accuracy. The STIRPAT variable coefficient model can adaptively adjust the model coefficients according to the data characteristics of different regions and time periods, enhancing the adaptability of the model. Using the dynamic meta-optimization method to adjust the hyperparameters can more efficiently find the optimal or approximately optimal model parameters, reducing the consumption of computing resources and time. By training and optimizing the model within multiple time windows, the contingency and volatility that may be brought by a single time window are reduced, improving the stability of the prediction. By constructing and optimizing the loss function, the difference between the predicted value and the actual observed value can be minimized, thereby reducing the prediction error. By accurately predicting the clean energy production, it helps to optimize the resource allocation and improve the energy utilization efficiency. It can better evaluate the effect of environmental protection measures and promote environmental protection and sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application.

[0051] Figure 1 It is a schematic diagram of the framework structure of the present invention;

[0052] Figure 2 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The present invention will be further described below with reference to the drawings and embodiments.

[0054] Example:

[0055] A clean energy production prediction model and a corresponding prediction method, refer to Figure 1 - Figure 2 Specifically include:

[0056] Step 1: Construct a theoretical framework for analyzing the relationship between environmental protection expenditure, per capita GDP and historical clean energy production.

[0057] Step 2: Environmental protection expenditure and historical clean energy production are the core measurement indicators of the present invention, and a suitable measurement scheme needs to be designed.

[0058] Step 3: Use the STIRPAT variable coefficient model to model the obtained historical measurement data.

[0059] Step 4: For the STIRPAT variable coefficient model, use the dynamic meta-optimization adaptive optimizer method to obtain efficient estimates of the parametric and non-parametric parts of the model.

[0060] Step 5: According to the hypotheses obtained in the theoretical framework section, use the STIRPAT variable coefficient model to construct an empirical analysis framework and predict future clean energy production.

[0061] In some embodiments, in Step 1: Starting from economic theory, explore the influence mechanism of environmental protection expenditure, per capita GDP, and historical clean energy production, and form the following three theoretical hypotheses: environmental protection expenditure has a non-linear impact on historical clean energy production, per capita GDP has a non-linear impact on historical clean energy production, and the synergy effect of environmental protection expenditure and per capita GDP promotes the increase of historical clean energy production.

[0062] In some embodiments, in Step 2: Select multiple indicators related to clean energy and their influencing factors. First, determine multiple preselected factor variables related to the clean energy, including historical clean energy production, environmental protection expenditure, per capita GDP, technological level, urbanization level, and export volume; second, obtain the data of optional factor variables; finally, impute the missing data of provinces. When constructing the historical clean energy production variable, from the perspective of the energy supply side, measure the historical clean energy production based on the total production of clean energy in nine provincial-level regions along the Yellow River Basin, that is, the data of primary electricity (electricity generated by nuclear power, hydropower, wind power, and solar power), other energy sources (clean energy such as biomass fuel, geothermal energy, hydrogen energy, and tidal energy), etc. Since the measurement methods of each type of clean energy are different, in order to accurately calculate each indicator, the entropy method is used to evaluate the historical clean energy production level, and the selected indicators are standardized. The calculation steps are as follows:

[0063] Standardization of data:

[0064] where R ij is the original value of the j-th evaluation indicator in the i-th evaluation unit, is the standardized value, (R jt ) max , (R jt ) min are the maximum and minimum values of the selected indicators respectively. j = 1,..., N, and a represents the number of indicators.

[0065] Calculate the information entropy E jt :

[0066]

[0067] Calculate the weights of evaluation indicators:

[0068]

[0069] Calculate the comprehensive score ENP of historical clean energy production it :

[0070]

[0071] Environmental protection expenditure, local finance is the general budget expenditure.

[0072] The per capita GDP is measured by the per capita GDP. The control variables selected are the urbanization level, technological progress, and export volume. Among them, the calculation method of the urbanization level is the urban population / total population, and the technological progress is measured by the number of patents obtained by each province.

[0073] In some embodiments, in step three: Consider the STIRPAT model as follows:

[0074]

[0075] Where, I t is the constructed environmental index, here referring to historical clean energy production (ENP); P t represents the population size, here referring to the urbanization level (UR); T t represents the technological level, here referring to technological progress (RD), e t represents the error term.

[0076] In order to eliminate the influence of heteroscedasticity, take the logarithm to obtain the following form of the STIRPAT model:

[0077] LI it = La + bLP it + cLA it + dLT it + ε it

[0078] According to the above constructed environmental variable (ENGP), economic variable (GDP) and control variables, the logarithmic STIRPAT model can be evolved into:

[0079] LENP it = La + β 1 LGDP it + β 2 LENGP it + β 3 (LENGP × GDP)it +β 4 LRD it

[0080] +β 5 LUR it +β 6 LEX it +ε it .

[0081] where ENP it represents the observed value of historical clean energy production in the i-th individual at the t-th period and the t-th period; GDP it represents the observed value of economic development in the i-th individual at the t-th period and the t-th period; ENGP it represents the observed value of environmental protection expenditure in the i-th individual at the t-th period; (ENGP×GDP) it represents the observed value of the cross-term of environmental protection expenditure and per capita GDP in the i-th individual at the t-th period; RD it represents the observed value of technological progress in the i-th individual at the t-th period; UR it represents the observed value of the urbanization level in the i-th individual at the t-th period; EX it represents the observed value of the export volume in the i-th individual at the t-th period.

[0082] The above empirical model can only study the linear impact relationship of existing variables on historical clean energy production. In fact, there is a wide range of non-linearity between economic and environmental variables. To more accurately predict the future production of clean energy, the present invention constructs a new STIRPAT variable coefficient model based on the linear STIRPAT model, and the form is as follows:

[0083] LENP it = La + β 1 LGDP it +β 2 LENGP it +β 3 (LENGP×GDP) it +β 4 LRD it +β 5 LUR it +β 6 LEX it +Zα(u it ) + ε it .

[0084] where α(·) represents the non-parametric impact mechanism of each variable on ENP, x' it = (GDP it ', LENGP it ', (LENGP×GDP) it',RD it ',UR it ',EX it )', β = (β 1 , β 2 , β 3 , β 4 , β 5 , β 6 )'.

[0085] In some embodiments, in step four: an adaptive optimizer using dynamic meta-optimization is used to obtain estimates of the above unknowns. This method can automatically adjust the optimization strategy according to the dynamic changes of the data, helping to improve the accuracy and prediction performance of the model. First, the following loss function is established:

[0086]

[0087] where θ represents the hyperparameter of the dynamic meta-optimization strategy.

[0088] Secondly, the optimization process is divided into multiple stages, a time window and dynamic feedback are set, and the learning rate is updated according to the error and gradient information of the current model. Then, based on the gradient-based meta-optimization method, the values of the inner-layer optimization and the outer-layer optimization are obtained:

[0089]

[0090] where represents the parameters of task t in the k-th iteration, θ k represents the value of the hyperparameter, η represents the learning rate, and the initial value is 10 -3 .

[0091] Finally, monitor the convergence of the model, calculate the loss function, and use the cross-validation method to evaluate the prediction error.

[0092] In some embodiments, in step five: the STIRPAT variable coefficient model is used to conduct an empirical study on this part of the content, analyze the influencing factors of clean energy, and can accurately predict the future output of clean energy, providing a theoretical basis for effectively improving the efficiency of green and low-carbon development in a certain section of the Henan River Basin. Provide scientific and accurate prediction data support for the investment, planning, and policy formulation of the government and enterprises in the field of clean energy. By accurately predicting the output of clean energy, it helps to optimize the allocation of resources and improve the energy utilization efficiency.

[0093] It should be noted that in the present invention, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0094] The above embodiments are merely illustrative examples of the present invention and do not constitute a limitation on the protection scope of the present invention. Any design identical or similar to the present invention falls within the protection scope of the present invention.

Claims

1. A clean energy production prediction method, characterized in that: The following steps are involved: Obtain multiple indicators related to clean energy and determine factors affecting the indicators, including historical clean energy production, environmental protection expenditures and per capita GDP, and factors affecting the indicators include technology level, urbanization level and export volume; Combine multiple indicators to build a linearized STIRPAT model, obtain the logarithmic function of the STIRPAT model, and build a STIRPAT variable coefficient model based on the influencing indicator factors; The loss function is constructed using a logarithmic function, and dynamic meta-optimization is used based on an adaptive optimizer until the loss function reaches the upper limit of the number of iterations, and the coefficients of the determined STIRPAT variable coefficient model are obtained; The environmental protection expenditure, per capita GDP, technological level, urbanization level, export volume of the region to be predicted in the prediction year and the historical clean energy production of the region to be predicted before the prediction year are obtained, and the STIRPAT variable coefficient model with determined input coefficients is used to obtain the prediction results of clean energy production in the region to be predicted.

2. The clean energy production prediction method according to claim 1, characterized in that: The linearized STIRPAT model is expressed as follows: Among them, I t represents the environmental index representing the historical clean energy production ENP; a represents the constant term, A t represents GDP per capita, P t represents the urbanization level UR representing the population size; T t Represents the technological progress RD, e t represents the error term; The STIRPAT variable coefficient model is constructed, and the formula is expressed as follows: LENP it =La+β1LGDP it +β2LENGP it +β3(LENGP×GDP) it +β4LRD it +β5LUR it +β6LEX it +Zα(u it )+e it Where, L() represents a logarithmic function; ENP it represents the observed value of historical clean energy production in the tth period of the ith individual; GDP it represents the observed value of economic development in the tth period and the tth period of the i-th individual; ENGP it represents the observed value of environmental protection expenditure in the ith individual in period t; (ENGP×GDP) it represents the observed value of the cross term between environmental protection expenditure and per capita GDP in the ith individual in period t; RD it represents the observed value of scientific and technological progress in the tth period of the i-th individual; UR it represents the observed value of urbanization level in the ith individual in period t; EX it represents the observed value of export volume in the tth period of the i-th individual; u it represents the observed value of energy consumption structure in the ith individual in period t, β i (i=1,2,3,4,5,6) represents the coefficient of the independent variable; Z=(GDP it ',LENGP it ',RD it ',UR it ',EX it '), α(·) represents the non-parametric influence mechanism of each variable on ENP, ε it represents the error term, i=1,…,N, t=1,…,T represent the number of individuals and time respectively.

3. The clean energy production prediction method according to claim 1, characterized in that: The method of obtaining the STIRPAT variable coefficient model with coefficient determination comprises the following steps: The initial coefficients of the linearized STIRPAT model are the initial weights of multiple indicators, and the hyperparameters and coefficients of the linearized STIRPAT model are set, including the initial learning rate, the time window size, and the dynamic feedback mechanism; Based on the linearized STIRPAT model structure, the initialization coefficients are substituted into the STIRPAT variable coefficient model to construct the initial STIRPAT variable coefficient model; According to the clean energy output as the prediction target and the actual observation data, the loss function is constructed using a logarithmic function; the formula is as follows: Dynamic element optimization is used until the loss function reaches the upper limit of the number of iterations, and the coefficients of the determined STIRPAT variable coefficient model are obtained; Substitute the determined coefficients into the STIRPAT variable coefficient model to obtain the final clean energy output prediction model.

4. The clean energy production prediction method according to claim 3, characterized in that: The dynamic meta-optimization includes: According to the STIRPAT variable coefficient model and data attributes, the time window size is set, and the data set is divided into multiple subsets according to the time window, each subset contains data within a time window; in each time window, the STIRPAT variable coefficient model is trained using the observed data; Design dynamic feedback rules based on performance indicators to adjust learning rate and model coefficients; The prediction error is obtained based on the error between the initial STIRPAT variable coefficient model prediction value and the actual observation value in each time window; Based on the prediction error, back propagation is used to obtain the gradient information of the coefficients of the STIRPAT variable coefficient model; Determine a learning rate adjustment strategy based on the gradient information and a preset dynamic feedback rule, wherein the adjustment strategy includes learning rate decay and learning rate restart; In the inner layer optimization, grid search is selected as the adaptive optimizer, and the grid search updates the STIRPAT variable coefficient model coefficients according to the gradient information; the formula is expressed as: in, represents the parameter value of task t in the kth iteration; η represents the learning rate of meta-optimization, with an initial value of 10 -3 ; During the outer layer optimization, the hyperparameters of the inner layer optimization are adjusted by the meta-optimization method, and the formula is expressed as: Among them, θ s Represents the value of a hyperparameter.

5. The clean energy production prediction method according to claim 4, characterized in that: In the inner layer optimization, a grid search is selected as an adaptive optimizer, and the grid search updates the STIRPAT variable coefficient model coefficients according to the gradient information, including the following steps: The set of all possible values ​​of the parameters in the STIRPAT variable coefficient model is defined as a parameter space; a grid is divided in the parameter space, and each grid point represents a set of parameter values; For each grid point in the parameter space, the STIRPAT variable coefficient model is used for training and the corresponding loss function value is obtained; Compare the loss function values ​​at different grid points and select the parameter value with the minimum loss as the current optimal solution; According to the gradient information of the loss function to the parameters, the direction and step size of the grid search are guided to obtain the optimal parameters; the formula is expressed as: Among them, θ t is the parameter in the tth iteration; α is the learning rate, is the loss function L at θ t The gradient of The parameters are updated using the gradient information, and new parameter values ​​are found by searching on the grid until the loss function value is lower than the set threshold, and a set of optimized parameter values ​​are obtained as the parameter values ​​determined by the STIRPAT variable coefficient model.

6. A clean energy production prediction system, characterized in that: include: A data acquisition module, used to acquire multiple indicators related to clean energy and determine factors affecting the indicators, wherein the indicators include historical clean energy output, environmental protection expenditures and per capita GDP, and the factors affecting the indicators include technological level, urbanization level and export volume; The prediction model building module is used to build a linearized STIRPAT model by combining multiple indicators, obtain the logarithmic function of the STIRPAT model, and build a STIRPAT variable coefficient model based on the influencing indicator factors; the loss function is built by using the logarithmic function, and dynamic meta-optimization is used based on the adaptive optimizer until the loss function reaches the upper limit of the number of iterations, and the coefficients of the determined STIRPAT variable coefficient model are obtained; The production forecasting module is used to obtain the environmental protection expenditure, per capita GDP, scientific and technological level, urbanization level, export volume of the region to be forecasted in the forecast year and the historical clean energy production of the region to be forecasted before the forecast year, and input the STIRPAT variable coefficient model determined by the coefficient to obtain the forecast result of the clean energy production of the region to be forecasted.

7. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

8. A computer device, characterized in that: The method comprises 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 method described in any one of claims 1 to 5 is implemented.

Images