Forecast data of provincial carbon emissions in China from 2025 to 2035: based on ARIMA-BP model
Description
China is an important contributor to global carbon emissions. Accurately estimating carbon emissions is crucial for reducing carbon emissions in accordance with the United Nations Sustainable Development Goals and China's carbon neutrality strategy. This study selected energy consumption data from Chinese provinces from 2000 to 2021, calculated carbon emissions using the carbon emission factor method, and then predicted carbon emissions data for 2022-2035 using the ARIMA-BP model. This dataset can be used to describe the spatiotemporal evolution trend of carbon emissions in China. This study can provide policy basis and wisdom for China's carbon reduction efforts.
Files
Steps to reproduce
Considering that the total carbon emission is a typical time series data, this paper uses ARIMA-BP neural network combination model to forecast. 1.ARIMA model In this study, ARIMA model is constructed by three steps: stationarity test of time series, determination of order P of ARIMA model and parameter estimation and diagnostic test, and the prediction results and error sequence are obtained (see Formula 2 for details). Firstly, the ADF method is used to test the stationarity of time series, and the DCI series is obtained by taking the first-order difference of each series. After the ADF test, the adjoint probability T statistic is -2.97~-2.99, and the unit root hypothesis is rejected at the statistical level of 1%. After the first-order difference, the carbon emission intensity series is stable. Then, the autocorrelation (AC) coefficient and partial autocorrelation (PAC) coefficient are adopted, and according to the order determination principles such as AIC, as can be seen from Figures 3 and 4, the (1,1,0) model is adopted according to the comprehensive analysis of AC and PAC images and data and the inspection table. Finally, the significance of model parameters is tested and predicted. （2） 2. ARIMA-BP neural network combination model Establish a two-layer hidden layer (Formula 4)BP neural network with one neuron (Formula 3) and three neurons. Input the error data into the neural network in time sequence, with the input node set to 4 and the output node set to 1. By rolling the window, the error of the previous period will continue to be transmitted into the neural network, and as part of the input, the model will be continuously revised and predicted. Finally, the combination model is used to predict the change of carbon emission intensity. The specific formula is as follows: （3） （4） (3) Display of data prediction effect From Table 6 and Figure 5, it can be seen that the MSE fluctuation range of carbon emission intensity of the three major industries is 0 to 0.633, and the regression coefficient is high and close to 1, which shows that the simulation prediction effect is good and the prediction data is reliable. As can be seen from Figure 12, from 2000 to 2035, China's industrial energy consumption intensity was "secondary industry > tertiary industry > primary industry", in which the energy intensity of primary industry decreased from 0.0039 to 0.0001, that of secondary industry decreased from 0.0525 to 0.0041, that of tertiary industry decreased from 0.0024 to 0.0003, and that of secondary industry decreased. It shows that the reduction of China's energy consumption intensity is mainly concentrated in the secondary industry. Although there is still a certain gap between China's energy intensity and that of European and American countries, with the adjustment of China's industrial structure, diversification of energy structure and continuous improvement of industrialization level, there is still room for a certain decline in energy intensity