Shapley值在中国城市天然气需求量组合预测中的应用  

Application of Shapley value in the combination forecasting of city gas demand in China

作　　者：秦步文 张吉军[1] 李岚 曾镰 文静 韩咪 QIN Buwen;ZHANG Jijun;LI Lan;ZENG Lian;WEN Jing;HAN Mi(School of Economics and Management,Southwest Petroleum University,Chengdu,Sichuan 610500,China;Sichuan Changning Gas Development Co.,Ltd.,Chengdu,Sichuan 610056,China;Development Division,PetroChina Southwest Oil&Gasfield Compa-ny,Chengdu,Sichuan 610000,China;Discipline Inspection and Audit Center,PetroChina Southwest Oil&Gasfield Company,Cheng-du,Sichuan 610051,China)

机构地区：[1]西南石油大学经济管理学院,四川成都610500 [2]四川长宁天然气开发有限责任公司,四川成都610056 [3]中国石油西南油气田公司开发事业部,四川成都610000 [4]中国石油西南油气田公司纪检审计中心,四川成都610051

出　　处：《天然气技术与经济》2022年第2期50-55,共6页Natural Gas Technology and Economy

基　　金：四川省科技计划项目“四川天然气供需预测预警机制研究”(编号:2021JDR0241)。

摘　　要：天然气作为清洁、高效的优质能源,对减少碳排放量,促进城市生态环境和经济高质量发展,实现中国“碳达峰、碳中和”战略目标具有重要意义,因此精准预测城市天然气需求量是城市健康发展的重要保障。天然气需求量的预测主要有单一预测模型和构建组合预测模型两种研究方法。组合预测模型可以有效的对单一预测模型提供的信息进行整合,从而进一步减小单一预测模型带来的误差,提高预测的精确度。为了对城市天然气需求量进行准确预测,通过借鉴合作博弈论Shapley值利益分配理论,考虑均值GM(1,1)模型、指数平滑模型、灰色Verhulst模型之间的差异以及各自的特点,建立天然气需求组合预测模型,可以有效减少预测的误差。研究结果表明:(1)Shapley值能减少各单一预测模型差异的影响,更好的对三种单一预测模型的平均绝对百分比误差进行分配,确定的单一预测模型权重更合理,有效提高了城市天然气需求量预测的精度;(2)“十四五”期间中国城市天然气需求量将呈持续增长趋势,预计2025年将达到1995×10;m;。Natural gas,as a clean and efficiently quality energy,is of great significance to reduce carbon emission,promote quality development of city ecological environment and economy,and realize the China’s goal of"carbon peak and carbon neutrality".Therefore,forecasting the demand of city gas accurately is an important guarantee to city’s healthy development.There are mainly two forecasting methods,namely single model and combination model.The combination one can effectively integrate the information in single model,so as to further reduce the errors of single model and improve forecasting accuracy.In order to accurately forecast the demand of city gas,a combination forecasting model on natural gas demand has been established by referring to the Shapley benefit distribution theory of cooperative game theory,considering the difference among gray GM(1,1)model,exponential smoothing model,and gray Verhulst model and their characteristics,and modifying the Shapley value,which may reduce some forecasting errors effectively.Results show that(1)the Shapley value can reduce certain influence caused by the difference among single models,better distribute the average absolute and relative errors among three single models,determine their weight more reasonably,and effectively improve the forecasting accuracy;and(2)the demand of domestic city gas will maintain an increasing trend during the 14th Five-Year Plan and is expected to reach 1995×10;m;in 2025.

关 键 词：天然气 城市燃气 需求量预测 SHAPLEY值 组合预测 

分 类 号：F224[经济管理—国民经济] F426.22