中国省际碳排放预测及达峰情景模拟研究

Research on Prediction and Peak Scenario Simulation of Provincial Carbon Emissions in China

中国不同省级的碳排放量呈现明显差异。文章采用STIRPAT模型分析得到碳排放影响因素,运用遗传算法优化的极限学习机模型和中国30个省(自治区、直辖市)1997—2020年的面板数据,对不同发展情景下中国30个省(自治区、直辖市)未来20年的碳排放量进行了预测分析,并将预测结果和误差指标与ELM、BP、GWO-SVM模型进行对比。同时,文章以行政区域为单位划分东北、华北等7个区域进行碳达峰、碳减排能力分析。研究结果显示:使用遗传算法改进的极限学习机模型可以克服ELM模型容易陷入局部最优解的缺点,获得更高的预测精度;在绿色发展情景下中国7个区域均能在2030前实现碳达峰。

China's carbon emissions show obvious inter provincial differences.This paper uses STIRPAT model to analyze the influencing factors of carbon emissions,and then uses the limit learning machine model optimized by genetic algorithm and the panel data of various provinces,cities and autonomous regions in China from 1997 to 2020 to predict the carbon emissions of 30 provinces,cities and autonomous regions in China under different development scenarios over the next 20 years.The prediction results and error indexes are compared with ELM,BP and GWO-SVM models.Finally,the administrative region is divided into seven regions,including Northeast and North China,to analyze the capacity of carbon peak and carbon emission reduction,and give policy suggestions.The results show that:The improved Extreme Learning Machine(ELM)model using genetic algorithm can overcome the disadvantage that(ELM)model is easy to fall into local optimal solution and obtain higher prediction accuracy.Under the green development scenario,all seven regions in China have the ability to reach the carbon peak by 2030.