College of Management Science, Chengdu University of Technology, Chengdu, 610059, China.
College of Mathematics and Physics, Chengdu University of Technology, Chengdu, 610059, China.
Environ Sci Pollut Res Int. 2023 Jul;30(32):78849-78864. doi: 10.1007/s11356-023-28022-w. Epub 2023 Jun 6.
The "14th Five-Year Plan" period is a crucial phase for China to achieve the goal of carbon peaking and carbon neutrality (referred to as the "double carbon"). Thus, it is very important to analyze the main factors affecting carbon emissions and accurately predict the change of carbon emissions to achieve the goal of double carbon. For the slow data updates and the low accuracy of traditional prediction models about the carbon emissions, the key factors of carbon emissions change selected by gray correlation method and the consumption of coal, oil, and natural gas were input into four single prediction models: gray prediction model GM(1,1), ridge regression, BP neural network, and WOA-BP neural network to obtain the fitted and predicted values of carbon emissions, which serve as input to the particle swarm optimization-extreme learning machine (PSO-ELM) model together. Based on the PSO-ELM combined prediction method above and the scenario prediction indicators constructed according to relevant policy documents of Chongqing Municipality, the carbon emission values of Chongqing Municipality during the 14th Five-Year Plan period are predicted in this paper. The empirical results show that carbon emissions of Chongqing Municipality still maintain an upward trend, but the growth rate slow down compared with 1998 to 2018. In general, the carbon emission and GDP of Chongqing Municipality showed a weak decoupling state during 1998 to 2025. By calculation, the PSO-ELM combined prediction model is superior to the above four single prediction models in carbon emission prediction and has good property by the robust testing. The research results can enrich the combined prediction method about the carbon emissions and provide policy suggestions for Chongqing's low-carbon development during the 14th Five-Year Plan period.
“十四五”时期是中国实现碳达峰碳中和（以下简称“双碳”）目标的关键阶段，因此，分析影响碳排放的主要因素，准确预测碳排放变化，实现“双碳”目标至关重要。针对传统预测模型存在数据更新慢、预测精度低的问题，采用灰色关联法选取的碳排放变化关键因素和煤炭、石油、天然气消费数据，分别输入到灰色预测模型 GM(1,1)、岭回归、BP 神经网络和 WOA-BP 神经网络这 4 种单一预测模型中，得到碳排放的拟合值和预测值，将其作为粒子群优化-极限学习机（PSO-ELM）模型的输入，基于以上 PSO-ELM 组合预测方法，结合重庆市相关政策文件构建的情景预测指标，对重庆市“十四五”时期的碳排放进行预测。实证结果表明，重庆市碳排放仍呈上升趋势，但增速较 1998 年至 2018 年有所放缓。总体来看，1998 年至 2025 年重庆市碳排放与 GDP 呈弱脱钩状态。经稳健性检验，PSO-ELM 组合预测模型在碳排放预测方面优于上述 4 种单一预测模型，具有良好的性能。研究结果可以丰富碳排放的组合预测方法，为重庆市“十四五”期间的低碳发展提供政策建议。
