Prediction of Carbon Emissions in China Based on Arima-BP Neural Network

The objective of this research is to leverage a hybrid model integrating ARIMA and BP Neural Network (BPNN) techniques to forecast China's long-term carbon emissions. Initially, a comprehensive time series dataset is compiled, incorporating provincial energy consumption figures, historical carbon emission data, and economic indicators from across China. Subsequently, the ARIMA methodology is applied to analyze these datasets, extracting linear trends and cyclical patterns. The residuals from the ARIMA predictions are then processed through a BPNN to account for any non-linear elements, thereby enhancing the overall predictive accuracy. Empirical findings indicate that the hybrid ARIMA-BPNN approach outperforms standalone ARIMA and BPNN models in terms of predictive precision. With an average absolute error (MAE) of 6.67 and a root mean square error (RMSE) of 8.16, the model demonstrates notable reliability and accuracy. Residual analysis further validates the efficacy of BPNN in mitigating ARIMA's prediction errors while identifying non-linear features within the dataset. Conclusively, the ARIMA-BPNN integrated model offers a robust analytical framework supporting China's efforts towards reaching its “carbon peak” and “carbon neutrality” objectives. This framework also serves as a valuable tool for policymakers in developing strategies aimed at promoting energy efficiency and reducing emissions.