基于组合模型的中国能源需求预测
|
摘要
能源是人类生存、经济发展、社会进步和现代文明不可缺少的重要物质基础。随着社会经济的不断发展,能源需求也在不断增长。因此对能源需求的研究具有重要的理论意义和现实意义。能源需求预测是从研究一个国家、地区或特定范围内能源消费的历史与现状开始,根据其消费行为,归纳影响能源消费的各种因素,寻求消费与这些因素的关系,根据这些关系对未来能源需求发展趋势作出估计和评价。因此对能源需求进行建模与预测是制订能源发展战略、规划部署的基础之一。
近年来,很多学者对能源需求的预测进行了研究,有很多的预测方法。本文将经济增长、产业结构、能源消费结构、人口和城市化、居民消费水平、技术进步和环境政策等作为对能源需求的影响因素。利用1978~2008年的能源需求总量时间序列数据,通过时间序列、灰色理论以及BP神经网络分别对2010-2015中国能源需求总量进行了预测。
组合预测模型比单个预测模型具有更高的预测精度,能增强预测的稳定性,具有较高的适应未来预测环境变化的能力,本文以最小预测误差平方和为目标函数的线性组合预测模型,计算出时间序列、灰色理论以及BP神经网络组合模型的权重系数,利用1978~2008年的数据进行建模和检验,结果表明组合模型预测结果的平均相对误差为2.19%,比时间序列、灰色理论以及BP神经网络的平均相对误差分别小1.93%、2.14%以及1.12%,利用此组合模型预测到2015年我国能源需求总量将达到385781万吨标准煤。
Energy is an indispensable material basis of human survival, economic development, social progress and modern civilization,With the development of social economy, energy demand is also growing.Therefore, the demand for energy research has important theoretical and practical significance. Energy demand forecast is from a country, region or specific energy consumption‘s past and present . According to their consumption behavior, summarized the various factors affecting energy consumption, for consumption and the relationship between these factors. Therefore, the demand for energy modeling and forecasting is one of the foundations to develop energy development strategy, planning the deployment.
In recent years, many scholars has been studied the energy demand forecast, there are many prediction methods. The article take the economic growth, industrial structure, energy consumption, population and urbanization, the level of consumption, technological progress and environmental policies as energy demand factors. Use the total energy demand time-series data from 1978 to 2008, through the time series, gray theory, and BP neural network respectively, forecasted the total energy demand of China from 2010 to 2015.
Combination forecasting model has higher prediction accuracy than a single prediction model ,it can enhance the stability of prediction, and the higher forecast to adapt to future environmental change, In this paper, the minimum sum of squares prediction error as the objective function of the linear combination forecast model Calculated the weight factor of the time series, gray theory, and BP Neural Network Model, Using data from 1978 to 2008 to build and test model. The results show that the the average relative error of combined model forecasted is 2.19%,which is 1.93% smaller than time series, 2.14% smaller than gray theory and 1.12% smaller than BP neural network. The combined model forecasted that by 2015 China's total energy demand will reach 3,857,810,000 tons of standard coal.
近年来,很多学者对能源需求的预测进行了研究,有很多的预测方法。本文将经济增长、产业结构、能源消费结构、人口和城市化、居民消费水平、技术进步和环境政策等作为对能源需求的影响因素。利用1978~2008年的能源需求总量时间序列数据,通过时间序列、灰色理论以及BP神经网络分别对2010-2015中国能源需求总量进行了预测。
组合预测模型比单个预测模型具有更高的预测精度,能增强预测的稳定性,具有较高的适应未来预测环境变化的能力,本文以最小预测误差平方和为目标函数的线性组合预测模型,计算出时间序列、灰色理论以及BP神经网络组合模型的权重系数,利用1978~2008年的数据进行建模和检验,结果表明组合模型预测结果的平均相对误差为2.19%,比时间序列、灰色理论以及BP神经网络的平均相对误差分别小1.93%、2.14%以及1.12%,利用此组合模型预测到2015年我国能源需求总量将达到385781万吨标准煤。
Energy is an indispensable material basis of human survival, economic development, social progress and modern civilization,With the development of social economy, energy demand is also growing.Therefore, the demand for energy research has important theoretical and practical significance. Energy demand forecast is from a country, region or specific energy consumption‘s past and present . According to their consumption behavior, summarized the various factors affecting energy consumption, for consumption and the relationship between these factors. Therefore, the demand for energy modeling and forecasting is one of the foundations to develop energy development strategy, planning the deployment.
In recent years, many scholars has been studied the energy demand forecast, there are many prediction methods. The article take the economic growth, industrial structure, energy consumption, population and urbanization, the level of consumption, technological progress and environmental policies as energy demand factors. Use the total energy demand time-series data from 1978 to 2008, through the time series, gray theory, and BP neural network respectively, forecasted the total energy demand of China from 2010 to 2015.
Combination forecasting model has higher prediction accuracy than a single prediction model ,it can enhance the stability of prediction, and the higher forecast to adapt to future environmental change, In this paper, the minimum sum of squares prediction error as the objective function of the linear combination forecast model Calculated the weight factor of the time series, gray theory, and BP Neural Network Model, Using data from 1978 to 2008 to build and test model. The results show that the the average relative error of combined model forecasted is 2.19%,which is 1.93% smaller than time series, 2.14% smaller than gray theory and 1.12% smaller than BP neural network. The combined model forecasted that by 2015 China's total energy demand will reach 3,857,810,000 tons of standard coal.
引文
[1]吴德春,董继斌.能源经济学.北京:中国工人出版社,1991
[2]周德群,万红飞,高亚平.可持续发展的能源、环境、经济(3E)关联模型.连云港化工高等专科学校学报,2000,13(1):50-53
[3]王海建.经济结构变动与能源需求的投入产出分析.统计研究,1999,第6期:30-34
[5]于立.能源价格理论研究.辽宁:东北财经大学出版社,1994.12
[4] D.梅多斯等.增长的极限.北京:商务印书馆,1984
[6]Jorgenson, Dale W. and Barbara Fraumeni.“Relative Prices and Technology change”in Modeling and Measuring Natural Resource Substitution. Cambridge, MA:MIT Press, 1981
[7]Griffin and Gregory. An Intercountry Translog Model of Energy Substitution Responses. American Economics Review, Vol.66, 1976, 845-857
[8]Prosser, R.D. Demand Elasticity in OECD: Dynamical Aspects. Energy Economics January, 1985, 9-12
[9]McFadden, D. and Jeffrey Dubin. Econometric Analysis of Residential Electric Appliance Holdings and Consumption. Econometrics, Vol.52, 1984
[10]Pindyck, Robert. Gains to Producers from the Certelization of Exhaustible Resources, Review of Economics and Statistics, Vol.60, 1978: 238-251
[11]李文彦. 21世纪前期我国能源战略的若干问题.经济地理,2000,20(1):7-12
[12] Energy forecasting. http://www.enerdata.fr/enerdatauk/services/ model_MEDEE. html[WB].2007
[13] Joost Siteur.The Long-range Energy Alternatives Planning model(LEAP)andWood Energy Planning. http://www.rwedp.org/index.html[EB].2007
[14]刘彦民.我国十五期间油品供需预测分析[J].石油规划设计.2001.12(5):16~18
[15]陈军才.广东省能源需求预测及供需分析[J].统计与预测.2002.116(2): 35~37
[16]曹建海,肖兴志中国市场前景报告[M].北京:中国时代经济出版社,2005.
[17]魏一鸣.等中国能源报告战略与政策研究[M].北京:科学出版社,2006.
[18]邱大雄.能源规划与系统分析.北京:清华大学出版社,1995
[19]朱达能.源一环境的经济分析与政策研究.中国环境出版社.2003年3月
[20]吴宗鑫,陈文颖.以煤为主多元化的清洁能源战略.北京:清华大学出版社,2001
[21]周总瑛,张抗,周庆凡.中国能源消费特征与发展战略分析.资源战略,2001,第5期: 25-27
[22]卢二坡.我国能源需求预测模型研究[J].统计与决策,2005,(10):12~15
[23]国家统计局.中国统计年鉴[M].2006.北京中国统计出版社
[24]刘思峰,郭天榜,党耀国.灰色系统理论及应用(第三版)[M].北京:科学出版社,2003
[25]Wankeun Oh,and Kihoon Lee. Energy consumption and economic growth inKorea: testing the causality relation. Journal of Policy Modeling[J]. 2004.26(8-9):973~981
[26]王维,贺京同,张建勋,等.人工神经网络在非线性经济预测中的应用.系统工程学报,2000,15(6):202-207
[27]司昕.预测方法中的神经网络模型.预测,1998,17(2):32-35
[28]程相君等著.神经网络原理及其应用.北京:国防工业出版社,1995
[29]Y.F.Zhang J.Y. Hfuh. A Neural Network Approach for Early Cost Estimation of PackagingProducts. Computer and Engineering. Vol.34, No.2, 1998, 433-450
[30]张婷.基于灰色神经网络组合模型的能源需求预测[D],天津大学管理学院,2007
[31]王燕.应用时间序列分析.北京中国人民大学出版社,2005
[31]李俊.中国区域能源供求及其因素分析.资源科学,1994,第4期:34-40
[32]周凤起. 2020年:中国的能源消费和能源发展战略.中外管理导报,1997,第4期:12-14
[33]龚灏,周仲礼,罗强,等.中国能耗的GM(1,N)模型及模型评价.成都理工学院院报,2000,27(4):413-417
[34]Gunter S.I.,Aksu C .,N-step Combination of Forecasts[J],Journal of Forecasting,1989,8(3): 253-267
[35]郭涤,周军.基于Matlab的神经网络预测模型研究.北京:物流科技,2006:125-128
[36]荆全忠,张健.GM(1,1)模型在煤炭需求预测中的应用[J].中国煤炭,2004,30(1):17-19.
[37]平凡,张长水.人工神经网络与模拟进化计算[M].北京:清华大学出版社,2005.
[38]陈华友.组合预测方法及其应用研究[D].中国科学技术大学,2002:76-79.
[39]李宝慧,项静恬.浅谈权重综合方法及组合预测[J].统计教育,2000,(6):28—29.
[40]王应明,广义加权几何平均组合预测方法研究[J],厦门大学学报自然科学版,1998 17(1)6-10
[41]张志涌,精通MATLAB 6.5版,北京:北京航空航天大学出版社,2003.1~233
[42]郝海,顾培亮,卢奇,中国和世界能源消费因素对比分析.中国能源,2002,(5):36-39
[43]王振龙.时间序列分析[M].北京:中国统计出版社,2000
[44]Martin T.Hagan,Howard B .Demuth,Mark Beale.Neural network design.北京:机械工业出版社.2002.
[45]刘豹,胡代平.神经网络在预测中的一些研究,系统工程学报。1999.14(4)338-344
[46]emen R.T.,Combining Forecasts:A Review and Annotated Bibliography[J],International Journal of Forecasting,1989,5(4):559-583
[47]唐纪,王景,组合预测方法评述[J],预测,1999,18(2):42-43
[48]王明涛,确定组合预测权系数最优近似解的方法研究,系统工程理论与实践,2000(3):104-109
[49]马永开,唐小我,线性组合预测模型优化问题研究[J] ,系统工程理论与实践,1998,18(2)110-114
[50]王应明,傅国伟,基于不同误差准则和范数的组合预测方法研究[J],控制与决策,1994,9(1)20-28
[51]杨桂元,唐小我,非负权重组合预测模型优化方法研究[J],数量经济技术经济研究,1998,(3)56-60
[52]王应明,罗英,广义加权算术平均组合预测技术研究[J],预测,1998,17(1),51-53
[53]唐小我,组合预测计算方法研究,预测,1991
[54]吕明,中国煤炭消费预测模型研究与应用[D],北京交通大学,2008
[2]周德群,万红飞,高亚平.可持续发展的能源、环境、经济(3E)关联模型.连云港化工高等专科学校学报,2000,13(1):50-53
[3]王海建.经济结构变动与能源需求的投入产出分析.统计研究,1999,第6期:30-34
[5]于立.能源价格理论研究.辽宁:东北财经大学出版社,1994.12
[4] D.梅多斯等.增长的极限.北京:商务印书馆,1984
[6]Jorgenson, Dale W. and Barbara Fraumeni.“Relative Prices and Technology change”in Modeling and Measuring Natural Resource Substitution. Cambridge, MA:MIT Press, 1981
[7]Griffin and Gregory. An Intercountry Translog Model of Energy Substitution Responses. American Economics Review, Vol.66, 1976, 845-857
[8]Prosser, R.D. Demand Elasticity in OECD: Dynamical Aspects. Energy Economics January, 1985, 9-12
[9]McFadden, D. and Jeffrey Dubin. Econometric Analysis of Residential Electric Appliance Holdings and Consumption. Econometrics, Vol.52, 1984
[10]Pindyck, Robert. Gains to Producers from the Certelization of Exhaustible Resources, Review of Economics and Statistics, Vol.60, 1978: 238-251
[11]李文彦. 21世纪前期我国能源战略的若干问题.经济地理,2000,20(1):7-12
[12] Energy forecasting. http://www.enerdata.fr/enerdatauk/services/ model_MEDEE. html[WB].2007
[13] Joost Siteur.The Long-range Energy Alternatives Planning model(LEAP)andWood Energy Planning. http://www.rwedp.org/index.html[EB].2007
[14]刘彦民.我国十五期间油品供需预测分析[J].石油规划设计.2001.12(5):16~18
[15]陈军才.广东省能源需求预测及供需分析[J].统计与预测.2002.116(2): 35~37
[16]曹建海,肖兴志中国市场前景报告[M].北京:中国时代经济出版社,2005.
[17]魏一鸣.等中国能源报告战略与政策研究[M].北京:科学出版社,2006.
[18]邱大雄.能源规划与系统分析.北京:清华大学出版社,1995
[19]朱达能.源一环境的经济分析与政策研究.中国环境出版社.2003年3月
[20]吴宗鑫,陈文颖.以煤为主多元化的清洁能源战略.北京:清华大学出版社,2001
[21]周总瑛,张抗,周庆凡.中国能源消费特征与发展战略分析.资源战略,2001,第5期: 25-27
[22]卢二坡.我国能源需求预测模型研究[J].统计与决策,2005,(10):12~15
[23]国家统计局.中国统计年鉴[M].2006.北京中国统计出版社
[24]刘思峰,郭天榜,党耀国.灰色系统理论及应用(第三版)[M].北京:科学出版社,2003
[25]Wankeun Oh,and Kihoon Lee. Energy consumption and economic growth inKorea: testing the causality relation. Journal of Policy Modeling[J]. 2004.26(8-9):973~981
[26]王维,贺京同,张建勋,等.人工神经网络在非线性经济预测中的应用.系统工程学报,2000,15(6):202-207
[27]司昕.预测方法中的神经网络模型.预测,1998,17(2):32-35
[28]程相君等著.神经网络原理及其应用.北京:国防工业出版社,1995
[29]Y.F.Zhang J.Y. Hfuh. A Neural Network Approach for Early Cost Estimation of PackagingProducts. Computer and Engineering. Vol.34, No.2, 1998, 433-450
[30]张婷.基于灰色神经网络组合模型的能源需求预测[D],天津大学管理学院,2007
[31]王燕.应用时间序列分析.北京中国人民大学出版社,2005
[31]李俊.中国区域能源供求及其因素分析.资源科学,1994,第4期:34-40
[32]周凤起. 2020年:中国的能源消费和能源发展战略.中外管理导报,1997,第4期:12-14
[33]龚灏,周仲礼,罗强,等.中国能耗的GM(1,N)模型及模型评价.成都理工学院院报,2000,27(4):413-417
[34]Gunter S.I.,Aksu C .,N-step Combination of Forecasts[J],Journal of Forecasting,1989,8(3): 253-267
[35]郭涤,周军.基于Matlab的神经网络预测模型研究.北京:物流科技,2006:125-128
[36]荆全忠,张健.GM(1,1)模型在煤炭需求预测中的应用[J].中国煤炭,2004,30(1):17-19.
[37]平凡,张长水.人工神经网络与模拟进化计算[M].北京:清华大学出版社,2005.
[38]陈华友.组合预测方法及其应用研究[D].中国科学技术大学,2002:76-79.
[39]李宝慧,项静恬.浅谈权重综合方法及组合预测[J].统计教育,2000,(6):28—29.
[40]王应明,广义加权几何平均组合预测方法研究[J],厦门大学学报自然科学版,1998 17(1)6-10
[41]张志涌,精通MATLAB 6.5版,北京:北京航空航天大学出版社,2003.1~233
[42]郝海,顾培亮,卢奇,中国和世界能源消费因素对比分析.中国能源,2002,(5):36-39
[43]王振龙.时间序列分析[M].北京:中国统计出版社,2000
[44]Martin T.Hagan,Howard B .Demuth,Mark Beale.Neural network design.北京:机械工业出版社.2002.
[45]刘豹,胡代平.神经网络在预测中的一些研究,系统工程学报。1999.14(4)338-344
[46]emen R.T.,Combining Forecasts:A Review and Annotated Bibliography[J],International Journal of Forecasting,1989,5(4):559-583
[47]唐纪,王景,组合预测方法评述[J],预测,1999,18(2):42-43
[48]王明涛,确定组合预测权系数最优近似解的方法研究,系统工程理论与实践,2000(3):104-109
[49]马永开,唐小我,线性组合预测模型优化问题研究[J] ,系统工程理论与实践,1998,18(2)110-114
[50]王应明,傅国伟,基于不同误差准则和范数的组合预测方法研究[J],控制与决策,1994,9(1)20-28
[51]杨桂元,唐小我,非负权重组合预测模型优化方法研究[J],数量经济技术经济研究,1998,(3)56-60
[52]王应明,罗英,广义加权算术平均组合预测技术研究[J],预测,1998,17(1),51-53
[53]唐小我,组合预测计算方法研究,预测,1991
[54]吕明,中国煤炭消费预测模型研究与应用[D],北京交通大学,2008