Prediction of energy‐related CO2 emissions in multiple scenarios using a least square support vector machine optimized by improved bat algorithm: a case study of China.

At present, China has the world's highest CO2 emissions. The reduction of China's CO2 emissions will have a direct effect on the world. Considering that CO2 emissions mainly come from the burning of fossil fuel, it is of great significance to accurately calculate and forecast China's energy‐related CO2 emissions. To improve the prediction accuracy of CO2 emissions, this paper proposed a new prediction model, which combines t‐distribution, Gaussian perturbations bat algorithm, and a least squares support vector machine, namely the TBAG‐LSSVM model. Furthermore, in order to ensure the rationality of factor selection, a stationary test, cointegration test, and Granger causality test were utilized to analyze factors affecting CO2 emissions. Through an empirical test, it was found that the proposed model has higher accuracy than LSSVM, BA‐LSSVM, WOA‐LSSVM, ELM, and BPNN. Therefore, three scenarios were put into the model to predict China's CO2 emissions up to 2030. The results indicate that CO2 emissions in 2030 will be 12 853.18, 11 378.27, and 12 008.19 million tons. According to the simulation results, industrial restructuring and the elimination of outdated production capacity should be pushed forward continuously to ensure that the Chinese government achieves emission reduction targets. © 2019 Society of Chemical Industry and John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]