Prediction and Analysis of Carbon Emissions under Specific Regional Scenarios in Anhui Province based on the STIRPAT Model

: In order to achieve the goal of reaching carbon peak by 2030, the STIRPAT model is used to predict carbon emissions under three simulation scenarios: baseline, optimization, and strict control of carbon emissions. Taking Anhui Province as an example, fully considering the impact of factors such as population, per capita GDP, carbon emission intensity, energy consumption intensity, energy structure, and industrial structure on carbon emissions, ridge regression and partial least squares regression were conducted respectively. Finally, the partial least squares regression method with a lower average error rate was selected to predict the model coefficients. The results show that all three model scenarios can achieve the carbon peak target by 2030, and the factors that have the greatest impact on carbon emissions are carbon emission intensity, energy consumption intensity, and per capita GDP.


Introduction
In 2022, global carbon emissions will increase slightly to a record level. The International Energy Agency points out that the energy crisis has not led to a significant increase in carbon emissions, but the continued growth of fossil fuel emissions is still hindering the achievement of global climate goals. The carbon emissions report released by the IEA shows that the global energy related carbon emissions in 2022 were 321 million tons, a year-on-year increase of 0.9%, far lower than the 3.2% growth rate of global GDP that year. Data shows that in order to achieve climate goals, global carbon dioxide emissions must be reversed and continuously reduced. By 2030, global carbon emissions need to be reduced by more than 40% in order to achieve the goal of global temperature control not exceeding 1.5 degrees Celsius.
From this, it can be seen that the task of carbon reduction is still very challenging. In order to achieve this goal, it is particularly important to study inter provincial carbon emissions. Taking Anhui Province as an example, the carbon emissions of industrial energy consumption and fossil energy consumption in Anhui Province are a microcosm of the six central provinces [1], with typical representativeness. Anhui Province is rich in coal resources and sits on two major coal mining bases. After being included in the Yangtze River Delta as a whole, its carbon emissions are increasing day by day. Therefore, it is necessary to scientifically predict carbon emissions in Anhui Province, and make specific requirements for carbon emissions in various fields in Anhui Province, in order to achieve peak carbon emissions as soon as possible.

Model Construction
The IPAT model is a method used to assess environmental stress, typically used to predict a country's carbon emissions. This model predicts carbon emissions by reflecting the comprehensive impact of population, per capita wealth, and technology. IPAT has certain limitations due to its consideration of a single influencing factor. The STIRPAT model is a regression analysis based on the IPAT model for the random effects of population, per capita wealth, and technology on the environment. Its standard form is: , , , representing environmental impact, population, per capita wealth, and technology respectively [2]; represents the model coefficient; , , the coefficients representing each variable; is the model Error term.

Data Description
The carbon emission data in the model comes from the China multi-scale emission inventory model, and the other variable data are all from the Anhui Provincial Statistical Yearbook from 2007 to 2021. After organizing the selected variables, the data shown in Figure 2 is obtained.

Empirical Analysis
The main disadvantage of STIRPAT model is the existence of multicollinearity, and the direct regression of data will have a greater impact on the results. Partial least square regression and ridge regression can solve the multicollinearity problem well. In the analysis process, the correlation between various factors is first analyzed, and then the variables are subjected to partial least squares regression and ridge regression to obtain the STIRPAT model. Finally, the optimal model is determined by comparing the error ratio. (

1) Correlation analysis between various factors
Factor analysis was conducted on the six factors mentioned above, and the correlation matrix between KMO and Bartlett's test is shown in Tables 3 and 4. By analyzing the KMO values in Table 3, the KMO value is 0.67, which is very suitable for factor analysis. Bartlett's test P<0.001 rejected the original hypothesis, indicating that there is a correlation between various variables and factor analysis can be performed [4]. Df (degree of freedom) 15 P (significance) 0.000*** Note: * * *, * *, * represent significance levels of 1%, 5%, and 10%, respectively    Table 4, it can be seen that there is a high correlation between variables. Including per capita GDP ( )Energy structure ( )Energy consumption intensity ( )There is a high negative correlation, with correlation indices of -0.984 and -0.98, respectively. And per capita GDP ( )And carbon emission intensity ( )Energy consumption intensity, ( ) Energy structure ( )There is a high positive correlation, with positive correlation coefficients of 0.974 and 0.945, respectively.
(2) Model parameter determination After conducting factor analysis on six variables, principal component analysis was used to extract three principal components, which were then subjected to partial least squares regression to obtain the STIRPAT model: By analyzing the regression equation, it can be concluded that carbon emission intensity( )Impact on carbon emissions( )The impact is the greatest, with a positive impact, that is, when other factors remain unchanged, for every 1% increase in carbon emission intensity, carbon emissions increase by 0.467%; In addition, both energy consumption intensity and energy structure have a negative impact on carbon emissions, that is, with other factors unchanged, for every 1% increase in energy consumption intensity, carbon emissions decrease by 0.143%, and for every 1% increase in energy structure, carbon emissions decrease by 0.157%. And it can be found that the impact of changes in energy structure on carbon emissions is higher than the impact of energy consumption intensity on carbon emissions. Although this result contradicts actual economic theory, it is similar to the research conclusions of Zhao Ci et al. and Sun Yi.
A comprehensive analysis of the ridge regression fitting chart and partial least squares regression fitting chart is shown in Figure 1. It can be found that the fitting effect of partial least squares regression is better than that of ridge regression. When comparing the average error rates of the two models, the average error rate of partial least squares regression is 0.19%, and the average error rate of ridge regression is 1.34%, as shown in Table 5. Therefore, it can be inferred that partial least squares regression is superior to ridge regression. The following specific scenario analysis uses partial least squares regression results to predict carbon emissions.

Setting of Specific Scenario Parameters
In order to comprehensively and clearly predict the carbon dioxide emissions in Anhui Province, this article sets three carbon emission scenarios from the perspectives of economic development and environmental protection.
(1) Base scenario: This refers to the relevant documents on climate change and energy arrangements released by Anhui Province. Based on the current development situation and speed, predict six factors: population, per capita GDP, energy consumption intensity, carbon emission intensity, energy structure, and industrial structure [4].
(2) Optimization scenario: Based on the current basic situation, in addition to vigorously promoting various climate change response and energy optimization policies, we also strengthen the transformation of carbon emissions. Starting from multiple aspects, optimize carbon emission intensity, energy structure, industrial structure, and energy consumption intensity. In order to predict the carbon dioxide emissions from 2025 to 2035, the parameter settings are shown in Table 7. (3) Strict scenario: All industries and regions strictly implement the clean energy structure route, focusing on the use of renewable energy such as solar and wind energy. By adjusting the energy structure and developing and utilizing new energy-saving energy sources, we will promote economic development.
By using the STIRPAT model, six carbon emission influencing factors, namely population, per capita GDP, energy consumption intensity, carbon emission intensity, energy structure, and industrial structure, were constructed, and three combinations of low, medium, and high were set to predict each influencing factor [4]. The specific combinations are shown in Table 6.

Carbon Emission Peak Prediction
Carbon emission peak prediction refers to the process of a stable downward trend in carbon dioxide emissions in a certain region from this year onwards. From the figure below, it can be seen that the peak time in Anhui Province varies among the three scenarios. The baseline scenario, optimized control of carbon emissions, and strict control of carbon emissions reached their peak in 2028, 2026, and 2025, with peaks of 3.21, 3.24, and 336 million tons, respectively. Among them, in the baseline scenario, carbon emissions showed a rapid decrease from 2023 to 2026, and the rate gradually slowed down after 2027. Under the optimized control of carbon emissions, the carbon emissions showed a rapid decrease from 2023 to 2028, and gradually slowed down after 2029. Under strict control of carbon emissions, carbon emissions from 2023 to 2029 showed a rapid decrease, while carbon emissions after 2030 showed a slow decrease trend.

Conclusion and Suggestions
Analyzing the above results, it can be found that Anhui Province will reach the peak under the baseline scenario in 2028, the peak under the optimized control of carbon emissions in 2026, and the peak under the strict control of carbon emissions in 2025. The carbon emission peaks corresponding to the three specific scenarios are 321 million tons, 324 million tons, and 336 million tons, respectively. Based on the above research results, our suggestions for effective measures to reduce carbon emissions are as follows: (1) Optimize the energy structure, minimize the use of coal as much as possible, and increase the use of clean energy such as natural gas and low-carbon energy. To achieve diversification and enrichment of energy use.
(2) Population has a significant impact on carbon dioxide emissions, and the government should provide appropriate guidance to improve population quality, advocate green and low-carbon travel. Minimize its impact on carbon dioxide emissions as much as possible.
(3) Accelerate the transformation of industrial structure and promote industrial transformation and upgrading through technological innovation. At the same time, corresponding preferential policies are provided to motivate residents to reduce carbon emissions and enhance the government's macroeconomic regulation and attraction.