考虑省份特征变量的火电行业碳达峰预测模型  

Prediction Model of Carbon Peak in Thermal Power Industry Considering Provincial Characteristic Variables

作　　者：崔和瑞[1] 徐昭 CUI Herui;XU Zhao(Department of Economics Management,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学经济管理系,河北保定071003

出　　处：《电力科学与工程》2024年第5期28-37,共10页Electric Power Science and Engineering

基　　金：河北省科技厅创新能力提升计划项目(18456214D)。

摘　　要：不同省份碳排放量存在的差异会对碳达峰预测模型的预测精准度产生一定影响。结合Kaya恒等式的内涵并进行扩展,分解出影响火电行业碳排放特征变量并进行相关性分析;同时,考虑到省份差异,利用独热编码,将省份这一非数值型特征变量转化为数值型特征变量加入到模型中。与未考虑省份特征变量的原模型进行对比,结果显示,考虑省份特征变量的火电行业碳排放KRR(Kernel ridge regression)预测模型的均方根误差明显降低,决定系数明显提高。这说明模型的精准度有所提高,且泛化能力更强。用所提模型对2023—2045年火电行业碳排放进行多情景预测,数据结果显示,低碳发展情景下于2030年可以实现碳达峰目标,且峰值最低。Due to the differences in carbon emissions in different provinces,the accuracy of carbon peak prediction model is influenced.This paper combined the connotation of Kaya identity and extended it to decompose the characteristic variables affecting the carbon emission of thermal power industry,and carried out correlation analysis.Meanwhile,considering the differences of provinces,the non-numerical characteristic variable of provinces was transformed into numerical characteristic variable by using one-hot coding and added into the model.Compared with the original model without considering the characteristic variable of provinces,the results show that the RMSE(root mean square error)of the KRR prediction model considering provincial characteristic variables is significantly lower,and the R^(2)(coefficient of determination)is significantly improved,indicating that the model has improved accuracy and stronger generalization ability.Finally,the carbon emissions of thermal power industry from 2023 to 2045 are predicted in multiple scenarios.The results show that the carbon peak target of 2030 can be achieved under the low-carbon development scenario,and the peak value is the lowest.

关 键 词：火力发电 碳达峰 非数值型特征变量 独热编码 KRR预测模型 

分 类 号：TK-9[动力工程及工程热物理] F224[经济管理—国民经济]