【作者】 艾德春；

【导师】 韩可琦；

【作者基本信息】 中国矿业大学 ， 资源开发与规划， 2008， 博士

【摘要】 “过剩—缓和—再过剩”一直是我国煤炭工业未走出的“怪圈”,究其根本原因,是由煤炭供需总量的预测偏差造成的,且缺乏对煤炭供需的监测预警体系。目前国内对煤炭供需预测及预警的研究处于碎片式状态,发达国家的研究,由于涉及经济安全、国家安全和对外战略等方面,对外基本不公开,因此对煤炭行业的预测及预警可借鉴的模式极少。本论文在煤炭供需预测及预警方面做了一些开拓性的研究。本论文在理论分析和实证分析相结合的基础上,利用Eviews和SPSS统计软件,详细地分析了影响我国煤炭供给和需求的主要因素,从而初选了影响煤炭供需的指标,然后,根据时差分析理论和因果关系理论,从影响煤炭供需初选指标中,构建了煤炭供需预测预警指标体系。受指标样本长度的限制,为了提高预测精度,本论文建立了不同的预测模型。对煤炭需求进行预测时,建立了BP神经网络模型、ARMA模型和逐步多元回归模型的组合模型,各模型在1986-2006年的预测误差平方和分别为308387316、487386757、411932952和166631387,组合模型预测结果优于单一模型。利用组合模型对2008-2010年的煤炭需求量进行预测,分别为:26.6、27.5和28.5亿t。经单位根检验及约汉森协整检验知,煤炭供给量、国内生产总值、铁路运输量和煤炭消费量四个变量(对数)均为非平稳变量,但它们存在协整关系且协整向量个数为2,采用差分的方法构造VAR模型将丢失重要的非均衡误差信息,因此,论文建立了向量误差修正模型(VEC)进行煤炭供给短期预测。从VEC对历史的拟和评价结果看,最大误差仅为4.95%,最小误差仅为0.08%,精度很高,运用VEC模型对2008-2010年的煤炭供给量进行预测,分别为:26.9、27.5和28.6亿t。为了煤炭供需中长期预测,论文建立了系统动力学(SD)模型,煤炭供需动力学模型由煤炭供应、煤炭消费、铁路运输和GDP和煤矿投资五个子系统组成。从SD模型对历史的仿真结果看,煤炭消费量的最大误差为4.92%,最小误差仅为0.67%;煤炭供给量的最大误差为9.54%,最小误差为1.43%。利用SD模型预测了2015年、2020年的煤炭供给和消费量,分别为:30.1亿t、29.1亿t和35.4亿t、33.7亿t。设置了不同的情景对煤炭供需进行了仿真计算。论文用周期波动理论研究了煤炭供需波动规律,通过趋势分离法测算了煤炭供需波动周期并进行了检验,结果表明,我国煤炭供需存在周期波动。为把煤炭供给和需求看作一个系统而非独立的两个变量,该文提出了系统供需速率概念,采用经济力学与突变理论相结合的方法,确定了煤炭系统供需速率警限,最后确定了煤炭供需综合警限值及警度,编写了煤炭供需预警分析软件,并对2008-2020年的煤炭供需进行了预警。更多还原

【Abstract】 “Overplus—abating—again overplus”is the odd circle that our coal industry has been in. The basic reason is the incorrect prediction of total supply and demand,lack of a monitoring and early warning system. For coal forecast and early-warning system, there is little study about it in our country, in developed countries, the studies are not in open due to involving economy safety, national security and foreign strategy, therefore a useful model that can be used for reference for the study of coal early-alarming and prediction is not existed. This paper done some pioneering work about study on coal prediction and early-alarming.Based on statistic software SPSS and Eviews, by the method of combining theoretical analysis and substantial evidence analysis,the paper analysed the factors that influencing coal demand and supply in detail, then selected the indicators initially, set up a foundation for the system of coal early-alarming and prediction.Based on the theory of analysis on the time difference and the causation theory, the paper set up a target system of coal demand and supply respectively from the factors that influencing coal demand and supply by using software Eviews.To improve forecast precision, the paper established different forewarning models according to index sample size. Using the BP neural networks theory, ARMA theory and stepwise regression theory, the paper set up a combination forecasting model for short-term prediction of coal demand, in which including BP neural networks model, ARMA model and stepwise regression model, their forecasting error square sum were 171162159, 496420810, 125467320 and 85886631 from the year 1986 to 2006, from which we get the result that the forecasting result of combination model is much better than the local estimation precision for each single model. During the years 2008-2010, the forecasting of coal demands were 26.9, 27.5 and 28.6 million tons by using combination forecasting model.Unit root test shows that coal supply, GDP, consumption, railway transportation and coal consumption are all nonstationary variables, but the result of the Johansen Cointegration Test demonstrates cointegrated relationship among them, and there are two cointegration vectors, unbalanced information errors will lost if designing a VAR model with difference method, so a VEC model was established to predict the coal supply in this paper. Maximum errors of actual and forecasting value of VEC is 4.95%, minimum error is 0.08%. During the years 2008-2010, the forecasting results are 26.9, 27.5 and 28.6 million tons by using VEC model.The paper established SD model for coal demand and supply medium-term and long-term forecast, in which including five interaction subsystems of coal supply, coal consumption, railway transportation GDP and coal investment. Maximum errors of actual and forecasting value of coal consumption is 4.92%, minimum error is 0.67%, maximum errors of actual and forecasting value of coal demand is 9.54%, minimum error is 1.43%.In 2015 and 2020 years, the forecasting results of coal consumption and coal demand are 30.1、29.1和35.4、33.7 million tons. Set up different situations for simulating calculation of coal demand and supply system.Based on cyclic fluctuation theory, the paper studied fluctuation law of coal demand and supply, measured and tested its fluctuation cycle by using tendency separating method; the examination result showed that the coal demand and supply are periodic wave. In order to regard demand and supply as a system, not two independent variables, the paper put forward a conception, that is the system demand-supply velocity. By using the method of the combination of economic mechanics and catastrophe theory, the paper calculated critical values of growth rate of coal demand and supply and coal demand-supply velocity alert limits, eventually designed the comprehensive warning limit value and warning degree interval, designed early-warning programme with VB.NET, which is used to provides early warning of coal demand and supply from 2008 to 2020.更多还原