1. Introduction
Energy is an important force to promote social development. A large amount of greenhouse gases are generated in the process of energy consumption, which leads to global warming and poses a huge threat to human production and life. As the world’s second largest economy, China consumes a lot of energy and releases a lot of carbon dioxide every year. Without timely and effective management measures, CO
2 emissions from the consumption of high-carbon energy sources will more than double by 2050. It is imperative to achieve low-carbon emission reduction. China made a public commitment at the United Nations Biodiversity Summit [
1] that China will achieve carbon peaking in 2030 and carbon neutrality in 2060. In particular, China’s power production has been dominated by coal power for a long time. The renewable energy has a small proportion of power generation. The power industry consumes a large amount of high-carbon energy and has become an important source of greenhouse gas emissions. Facing the approaching trend of electrification, the demand for electricity has grown rapidly under the stimulation of the whole society. The energy consumption required for electricity production will also increase rapidly and its carbon emissions will also continue to grow rapidly. As electricity is the main source of carbon emissions in the power industry, realizing the clean, efficient, and low-carbon development of the power industry in the future has become an urgent problem.
In the process of accelerating the construction of a new power system, achieving the dual-carbon goal not only requires new carbon reduction technologies but also needs effective carbon reduction strategies. The important premise of formulating a scientific carbon emission reduction policy is to achieve a reasonable carbon emission forecast. The reasonable power carbon emission forecast [
2] has an important reference value for the capacity and proportion of renewable energy connected to the new power system. At the same time, it can help China to achieve dual-carbon targets.
At present, many scholars have carried out various studies on how to achieve reasonable carbon prediction. In terms of research methods, the initial support vector machine model [
3] has gradually transitioned to the STIRPAT model [
4], which has suitable flexibility and a certain expansion space. The STIRPAT model has become a highly recognized model in the study of carbon emission peaks. The research on the attributes of factors affecting carbon emissions has gradually developed from single-attribute models, such as the logistic regression prediction model [
5] and the grey GM model [
6], to multi-attribute factors affecting carbon emissions, such as the ideal-point multi-attribute decision-making algorithm [
7]. In the application of prediction models, some scholars have used system dynamics with a causal mechanism [
8] and analyzed the relationship between the dynamic behavior of carbon emissions and production structure, which has predicted the development trend of carbon emissions for the whole society in the future. Combining the bat algorithm and simulated annealing algorithm for carbon emission prediction, the carbon emission of China’s power industry has been analyzed [
9]. Scholars have used the machine learning method [
10] to achieve carbon prediction. Some scholars have proposed a novel approach to predict CO
2 emissions in the agriculture sector of Iran based on the inclusive multiple model [
11]. In the regional carbon prediction model, some scholars have conducted wide-area modeling research on the carbon emission trajectories of China, the United States, and India under the dual-carbon goal [
12]. Additionally, scholars have explored the optimal low-carbon behavior in multiple scenarios in China [
13]. Taking Guangdong [
14] and Baoding [
15] as examples, scholars have explored ways for urban carbon dioxide emissions to peak under different paths. Some scholars have used the improved Gaussian algorithm to establish a local carbon emission prediction research model [
16]. Scholars have used the improved STIRPAT model [
17,
18] to predict the peak carbon emissions in typical energy-intensive regions, such as Shanxi and Xinjiang.
China’s power structure is dominated by thermal power, which has declined in recent years, but still occupies a dominant position. Thermal power generation fell from 80.8% of China’s total in 2010 to 67.9% in 2020. The power generation structure where most of the electricity demand is dominated by coal is the main reason that China’s power industry has become the largest carbon-emitting sector [
19]. Meanwhile, the power sector has high potential and cost advantages in decarbonization compared to other carbon-emitting sectors, such as industry [
20]. There are two main types of domestic and international studies on the characteristics of regional electric energy carbon emissions in China: one is top-down research on the characteristics of electric energy carbon emissions [
21,
22,
23]. Through the macroeconomic and energy data, the relationship between macroeconomic data, material consumption, and carbon emissions are established, thus reducing the dependence on micro-level material consumption data. The second is bottom-up explorations of the emission reduction path of the power industry and analyzing the impact of factors, such as power supply structure, energy saving, and consumption reduction, on the power industry [
24,
25,
26,
27]. However, the above studies are usually based on the traditional economic and energy growth model. Although it can adequately quantify the emission reduction potential of certain factors for the power industry, it does not fully consider the actual demand for the construction of a new power system with the dual-carbon target, especially after the large-scale access of clean energy to the power grid. In summary, existing studies have explored the potential and contributing factors of carbon emission reduction in China’s power industry in the future. However, few scholars have studied the overall consideration of macro-influencing factors and the construction demand of new power systems, such as economic development and gradually tightened regional carbon emission reduction policies. We analyzed the influencing factors of electric energy emissions and decomposed the influencing factors to identify the main factors. At the same time, the impact of economic, population, clean energy, and other factors on carbon emissions was also quantitatively studied. In this study, we comprehensively considered the needs of economic development and new power system construction and propose specific paths and key measures for carbon emission reduction in the power industry.
China aims to reach the carbon peak by 2030 and carbon neutrality by 2060. Shanghai is the frontier area of China’s economic development. It has a large population and a large economic volume. At the same time, the traditional electric power energy in Shanghai is mainly thermal power, which is facing the pressure of transformation to the new power system. On 25 July 2022, the Shanghai municipal government issued the Action Plan for Reaching Peak Carbon Emission in Shanghai, which states that the city will achieve the goal of reaching peak carbon emission by 2025. Based on the diversified energy data of various energy users and social production, it is necessary to conduct electricity carbon emission prediction research and formulate a scientific carbon reduction plan to ensure the green, low-carbon, clean, and efficient development of Shanghai. Through the research model and results of this paper, we provide scientific guidance to the Shanghai municipal government and put forward specific paths to achieve carbon peak.
In this study, the original STIRPAT model was improved and analyzed. According to the development characteristics and energy structure of Shanghai, the factors of regional carbon emissions, such as population, per capita GDP, urbanization rate, industrial structure, energy consumption, energy structure, regional electrification rate, and degree of opening to the outside world, were considered. The improved STIRPAT model was used to conduct modeling analysis and the ridge regression algorithm was used to analyze the biased estimation of historical data. Finally, according to the “14th Five-Year” development plan of Shanghai [
28], the parameters affecting the carbon emissions of the power industry were predicted and analyzed. The prediction model of carbon emissions from power energy in Shanghai based on elastic relationship was established, which realizes the effective prediction of carbon emissions from the power industry in Shanghai in the future and provides an important reference value for the regional realization of dual-carbon goals.
The rest of the paper is organized as follows. In
Section 2, the original STIRPAT model is introduced and the analysis of the influencing factors of energy carbon emissions in Shanghai and the extension of the STIRPAT model are presented. In
Section 3, we demonstrate how the ridge regression method was used to fit the carbon emission data of electric power energy in Shanghai, and an improved STIRPAT model is introduced. We also present the evaluation and analysis of the collinearity of the data in this section. Additionally, taking the Shanghai area as an example and according to the regional development plan, we discuss the model predictions of when Shanghai’s future electric energy carbon will peak, based on different scenarios, and put forward corresponding countermeasures. In
Section 4, we summarize the content of the article and illustrate its practical implications for achieving the dual-carbon goal.
2. Improved STIRPAT Model
In order to achieve effective prediction of regional power energy carbon emissions, it is necessary to analyze various factors affecting regional power energy carbon emissions. In this work, the improved STIRPAT model was used to model the influencing factors with significant influence. Since regional power energy carbon prediction is a long-term process, it is necessary to estimate the forecast data, such as terminal power consumption, total energy consumption, and clean energy proportion. Finally, the peak prediction of regional power energy carbon emissions was realized. The structural framework of this paper is shown in
Figure 1. The rest of this section shows the analysis of the composition of the model, and the solution of the model is explained by the ridge regression analysis in
Section 3.
2.1. Stochastic Impacts by Regression on Population, Affluence, and Technology Model
The STIRPAT model was improved from the original IPAT model (Environmental impact = Population × Affluence × Technology) by nonlinear extension. The IPAT model [
29] has been widely used to assess the impact of population, affluence, and technological development on environmental changes. Scholars have continuously improved the basic IPAT model and proposed different forms of analysis models, adding resource consumption to form the IPACT equation. Due to some limitations of the IPAT model, when a certain influencing factor changes, other influencing factors cannot remain fixed, resulting in unnecessary fluctuations in the analysis results.
Scholars have established the STIRPAT model by considering the retention of the multiplicative structure of the IPAT model and introducing a change in the index to overcome the defect of the proportional change in various influencing factors. The model is as follows:
In this formula,
I represents the degree of impact on the environment, which is specifically referred to as the predicted carbon emissions from electricity in this paper;
a represents the model coefficient;
P represents population;
A represents affluence, which is specifically referred to as the GDP per capita in this paper;
T represents technology, which is specifically referred to as energy consumption intensity per unit of GDP in this paper; and
ε represents the random interference term. Taking the logarithm of both sides of the model, the model can be written as:
The STIRPAT model allows each coefficient to be used as a parameter to evaluate the degree of influence on the results. The STIRPAT model not only properly decomposes various influencing factors, but also has adequate space for expansion. At the same time, it makes up for the lack of quantitative analysis of the impact of various factors on the environment to a certain extent. It modifies each influencing factor and establishes different demand models to meet different research needs and results analysis.
2.2. Extended Impact Factor
We studied the carbon emissions of Shanghai’s electric power by improving the environmental impact factor variables of the original STIRPAT model. Through the research on the influencing factors of regional power energy carbon emissions, the important factors affecting Shanghai’s energy consumption were selected and a model was established for analysis.
Population (P) is one of the important factors that measures regional energy carbon emissions. Human production activities are the source of carbon emissions. The population in the region is directly related to the energy carbon emissions generated by human life and production. Therefore, the regional population changes in quantity inevitably cause fluctuations in energy consumption, thereby changing the intensity of regional carbon emissions. There is a positive relationship between the expected population and energy demand.
The urbanization rate (C) is used to measure the level of regional population and economic development. Generally, it is believed that the urban population is denser than that of the rural areas and the development of the city needs the promotion of industrial development. Therefore, the increase in the urbanization rate represents the regional population. The improvement of the economic development level reflects that the urbanization rate has a positive effect on regional energy carbon emissions.
Affluence (A) represents an important indicator of the degree of human energy use and the activity of production activities in the region, which is measured by 10 thousand CNY in this paper. According to the usual situation, the richer the area, the more active the corresponding production activities, and the more energy consumption is generated. The development of China is associated with many energy-consuming items, such as traffic emissions and factory production emissions. Here, we used per capita GDP to measure the wealth of the region; the total GDP is converted into real GDP, with 1985 as the base period, and then is divided by the population in that year.
The development of technology (T) has brought rapid economic growth to the region, which is measured by ton of standard coal/10 thousand CNY in this paper. At the same time, technological progress has promoted the continuous innovation of industrial structure and the improvement of energy efficiency. Due to the existence of the energy rebound effect, not only will it not reduce energy consumption but it will also consequently promote a further increase in energy consumption. The progress of technology is measured by energy efficiency. The improvement of technology means that less raw materials produce a higher value. Therefore, the energy consumption intensity per unit of GDP is used to represent the degree of technological development.
The industrial structure (IS) selects the output value of the secondary industry, which is obtained by multiplying the proportion of the secondary industry and the real GDP measured by 100 million CNY in this paper. According to past experience, the carbon emissions from production activities are mainly generated by straw burning in the primary industry, pollutant treatment and industrial production in the secondary industry, and transportation in the tertiary industry. Among them, the carbon emission of the secondary industry occupies a relatively high proportion in the overall carbon emissions. The growth of the secondary industry is a very important factor leading to an increase in energy demand and it also accounts for the vast majority of energy consumption in Shanghai. Therefore, the growth of the secondary industry is expected to have a positive relationship with the increase in energy consumption in Shanghai.
The electrification rate (ER) is an important indicator that measures the efficiency of regional energy utilization. It converts fossil energy into electric energy for production and life. Compared to the direct use of fossil energy, not only does electric energy release less carbon emissions, but it also allows for rapid cross-domain flow through the grid structure. The increase in the electrification level of final energy consumption in the region means that the large-scale centralized treatment of fossil energy can be realized and the regional carbon emission level can be effectively reduced. Therefore, the regional electrification level has a positive relationship with energy carbon emissions.
The green electricity ratio (GE) refers to the proportion of renewable energy power generation in the total power generation in grid power generation. Electric power carbon emissions are an important source of regional energy carbon emissions. Building a new low-carbon and high-efficiency power system is an important part of achieving the dual-carbon goal. One of the important means is to increase the proportion of clean energy generation in the power grid. Selecting the green power ratio as an indicator to measure the utilization of regional renewable energy has important reference significance for regional energy carbon emissions.
Openness (OP) is the ratio of total foreign trade to GDP. Since the reform and opening up, Shanghai has absorbed a large number of industries, companies, and equipment. At the same time, with the deepening of opening up to the outside world, a large amount of trade has promoted the economic development of Shanghai, which requires a large amount of energy as support. Therefore, it can be expected that the degree of openness and energy clearance is a positive relationship.
To sum up, on the basis of the STIRPAT model combined with the actual situation of this work, the influencing factors of electricity energy consumption in Shanghai were studied and the following model was established:
In this formula, I represents the predicted carbon emissions from electricity; a represents the model coefficient; P represents population; C represents the urbanization rate; A represents GDP per capita; T represents energy consumption intensity per unit of GDP; IS represents the industrial structure occupied by the secondary industry; ER represents the electrification rate level; GE represents the green electricity ratio; OP represents the degree of opening up to the outside world; and ε represents the random interference term. The letters b, c, d, e, f, g, h, j represent model coefficients, which represent the elasticity of each influencing factor to energy consumption.
3. Results and Discussion
3.1. Data Sources
The data used for the above variables are from the Shanghai Statistical Yearbook [
30], China Energy Statistical Yearbook [
31], and China Technology Statistical Yearbook [
32]. The method published in the IPCC National Greenhouse Gas Guidelines [
33] is used to estimate the energy carbon emissions generated by the power industry and the model is as follows:
In this formula, Carbon represents the actual historical carbon emissions of electricity energy, which is measured by 10 kilotons; represents the consumption of carbon-containing energy; represents the carbon emission coefficient of energy; and represents the carbon oxidation factor.
By querying the Shanghai Statistical Yearbook, the types and quantities of primary electricity, and energy consumption in Shanghai over the years, the population over the years and the GDP over the years were obtained. According to formula (4), the total carbon emissions of electricity and energy in Shanghai over the years can be calculated and the statistics are shown in
Table 1.
3.2. Ridge Regression Analysis
Since the variables in the STIRPAT model usually have strong multicollinearity, the ordinary least squares method requires that there is no correlation between the respective variables. If there is a serious correlation between the independent variables, the sampling variability of the estimated value of the regression coefficient would be greatly increased. At this time, if the ordinary least squares method is still used to fit the regression coefficient, the regression results would be abnormal. The regression coefficient does not match the actual situation and the accuracy and reliability of the model cannot be guaranteed. Here, we have used the Pearson coefficient for correlation verification because the Pearson coefficient has suitable applicability to test variable correlation. The analysis of the correlation between the variables is to follow and the specific results are shown in
Table 2.
It can be seen in
Table 2 that there is a high correlation between variables and the Pearson correlation coefficient between variables is mostly above 0.7, which is statistically significant. From this, it can be judged that the correlation between variables is very high and there may be serious multicollinearity. Therefore, multicollinearity of the independent variables must be eliminated to obtain robust results.
In correlation analysis, principal component analysis can be used to propose principal components. Although it can summarize the information in the independent variable system well, it often lacks explanatory power for the dependent variable. In order to quantitatively and qualitatively explain each variable, we used ridge regression to solve the multicollinearity problem.
Ridge regression analysis is a biased estimation regression method for collinear data analysis, which is essentially an improved least squares estimation method. At the expense of accuracy, we sought a regression equation that is less effective but more realistic and more reliable. Therefore, the residual standard deviation obtained by ridge regression is larger than that of least squares regression but the tolerance to ill-conditioned data is much stronger than that of least squares. When there is collinearity among the independent variables, the determinant of the correlation matrix of the independent variables is approximately 0 or is called singular. The correlation matrix of the independent variables is added to the matrix of normal numbers and the singularity is improved compared to the original matrix. When the normal matrix is 0, it degenerates to the least square estimation. Since the least squares estimator has the smallest variance among all linear unbiased estimators, the variance is not necessarily small, so a biased estimator can be found. Although this estimator has a slight deviation, its accuracy can be much higher than an unbiased estimator. According to this principle, ridge regression analysis obtains the regression estimator by introducing a biased constant into the normal equation. The R-squared value of the ridge regression equation is slightly lower than that of the ordinary regression analysis but the significance of the regression coefficient is often significantly higher than that of the ordinary regression.
Therefore, the estimation method of ridge regression was selected to make the regression coefficients close to the truth, which can solve the problem of multicollinearity and reflect the actual situation through very significant regression coefficients. The syntax of the Ridge Regression program was written using IBM SPSS Statistics (The version: 26.0) to perform fitting regression. The graph relationship between the K values and the goodness of fit R
2 values for different ridge parameters were obtained. The corresponding relationship between the K value and R
2 of ridge regression is shown in
Figure 2.
It can be seen in the figure that there is a negative correlation between the R
2 of the goodness of fit and the ridge parameter K. When the K is smaller, the discarded information and accuracy are smaller and the accuracy of the prediction model is higher. The value of K should be as small as possible. As shown in
Figure 3, the value range of K is judged by the model stability in the ridge regression graph.
It can be seen in
Figure 3 that when K is between 0.1 and 0.8, the ridge regression coefficient tends to be stable and the fitting degree is good. The selection of K should be within this range. Combined with
Figure 2, when K = 0.1, the corresponding model fits the best. The ridge parameter K was selected to be 0.1. The specific results of ridge regression are shown in
Table 3.
From the regression results, population, affluence, technological progress, secondary industry, and openness have all contributed to the increase in carbon emissions from electric power in Shanghai. The level of the urbanization rate, electrification rate, and green power ratio are all negatively correlated with the carbon emissions of electric energy. Judging from the coefficients of the model, the population and industrial structure are important factors affecting the carbon emissions of electric power in Shanghai. The preferential policies and positive employment environment in Shanghai have attracted many outstanding talents. The impact of population on energy carbon emissions has been stable for a long time. From the data over the years, the proportion of Shanghai’s secondary industry has continued to decline and the impact of industrial structure on energy carbon emissions will be reduced year by year in the future. Affluence, technological progress, and openness have relatively little impact on Shanghai’s electric power carbon emissions, indicating that Shanghai’s energy efficiency is high, energy consumption is dominated by clean and efficient energy, and energy utilization technology is relatively advanced. At the same time, Shanghai continues to expand its opening to the outside world. The presence of many foreign companies and investments are mostly low-energy and carbon-intensive power companies, which have not had a significant impact on Shanghai’s electric power carbon emissions. With the increase in the urbanization level, the carbon emissions of electric energy have decreased, which shows that the utilization rate of electric energy in cities is relatively high. The increase in the regional electrification rate has increased the utilization rate of terminal energy, which is also an important factor affecting regional energy conservation and emission reduction. It can be seen from the coefficient that the increase in the proportion of regional renewable energy utilization is an important means to achieve carbon reduction. With the construction of new power systems under the dual-carbon goal, increasing the consumption of renewable energy and reducing the proportion of fossil fuels is an important way to achieve carbon reduction.
3.3. Carbon Prediction
Based on the above analysis, in order to more accurately predict the carbon emissions of electric power in Shanghai, it is necessary to understand the changes in the above-mentioned explanatory variables in Shanghai in the future. Moreover, due to the different degrees of interference of various factors, it is easy to produce different development trends. Therefore, according to the development of the Shanghai region and the “14th Five-Year Plan”, the adopted scenario analysis method can effectively avoid large deviations; the prediction part of the model is separated by systematic and scientific methods and the uncertainty of prediction is greatly reduced. This is to better predict the development trend of carbon emissions from electricity and energy in the future.
According to historical data, the population of Shanghai will show a steady growth in the future, the urbanization rate in Shanghai will remain at a high level, and the degrees of affluence, technology, and openness will affect the carbon emissions of Shanghai’s electric power. Here, we set scenario analysis forecasts for the industrial structure, electrification rate level, and green power ratio with a greater impact, as shown in
Table 4. According to Shanghai’s development planning goals and the current regional development level, we assumed four scenarios for this analysis. Here, we set three situations: the industrial structure, the electrification rate, and the growth rate of the green power ratio. The specific growth rate is shown in
Figure 4,
Figure 5 and
Figure 6.
Scenario I: The proportion of renewable energy rapidly develops and the industrial structure upgrade and electrification rate steadily develop.
Scenario II: The proportion of renewable energy and the industrial structure rapidly increase and the regional electrification rate steadily develops.
Scenario III: The proportion of renewable energy and the electrification rate rapidly increase and the industrial structure upgrades and steadily develops.
Scenario IV: The industrial structure upgrades, the electrification rate rapidly develops, and the proportion of renewable energy steadily develops.
The prediction results of the above relevant data were brought into the obtained model and the prediction results of the carbon emissions of Shanghai’s electric power under different scenarios were obtained, as shown in
Figure 7.
Under the predictions of the above-mentioned different scenarios, the overall trend of carbon emissions from electricity and energy in Shanghai presents different changes, and during this process, the peak times of different scenarios are also different. As can be seen in
Figure 7, under the scenario of a large number of renewable energies being added to the power grid and the rapid development of clean energy, the carbon emissions of Shanghai’s electric power can reach a peak in 2030. The carbon emissions can reach a peak in 2028 under the accelerated electrification level. Under the scenario wherein renewable energy steadily adds to the power grid development, the industrial structure can rapidly upgrade and the regional electrification rate level can significantly improve. Additionally, the carbon emissions of Shanghai’s electric power would not reach the carbon peak in 2030 and the curve tends to be stable.
However, due to the lack of direct historical data references for electric energy carbon emissions, we cannot directly obtain electric energy carbon emissions data. In this research, we used an indirect method to calculate the historical data of electric energy carbon emissions. According to the general development trend, we considered the factors affecting the carbon emissions of electric energy in the study area. We did not consider the abnormal situation of some factors, such as the massive inflow and outflow of people and the failure of the large-scale popularization of clean energy. Therefore, these conditions may cause partial bias in the results of our model.
Through the analysis of factors affecting carbon emissions of electric power energy in Shanghai and the peak forecast, it is shown that the scale of carbon emissions of power energy in Shanghai is greatly affected by the industrial structure, regional electrification level, and green power ratio. The carbon emissions of Shanghai’s electricity and energy are likely to reach the peak of carbon emissions before 2030. Therefore, in order to achieve the goal of reducing consumption and contributing to society’s dual-carbon goal, it is necessary to optimize the industrial structure, improve the electrification level, and increase the utilization rate of renewable energy to achieve the goal of energy saving and emission reduction as much as possible.
At present, Shanghai’s electricity supply is still dominated by thermal power and supplemented by clean energy. From the above fitting results, it can be seen that renewable energy has a non-negligible impact on carbon emission reduction. Therefore, in response to the call for energy conservation and emission reduction, it is imperative to optimize the current situation of Shanghai’s power energy structure and reduce the use of coal. The government should further strengthen the construction of the power grid structure. The power grid can increase the proportion of renewable energy connected to the grid, encourage the development of distributed energy transactions, and absorb more clean energy to increase the proportion of clean energy in the region. Shanghai is not a resource-rich city and is not suitable for the development of high-carbon heavy industries. The government should upgrade and optimize the industrial structure to reduce the proportion of secondary industry, introduce enterprises with low energy and power carbon emission intensity, and gradually promote the green and low-carbon transformation of regional industries.