基于最大Lyapunov指数的铁路货物运量预测研究  

Research on Railway Freight Traffic Prediction Based on Maximum Lyapunov Exponent

作　　者：吴华稳[1] 王富章[1] 

机构地区：[1]中国铁道科学研究院电子计算技术研究所,北京100081

出　　处：《铁道学报》2014年第4期7-13,共7页Journal of the China Railway Society

基　　金：铁道部科技研究开发计划(2007X008-G;2008X015-H)

摘　　要：根据混沌理论具有分析非线性动态系统混沌特性的特点,对货物发送量相关时间序列进行了分析和研究。本文在Takens相空间重构的基础上,利用C-C方法求嵌入时延与嵌入窗、G-P方法求嵌入维数;应用小数据量法计算铁路货物发送量相关时间序列的最大Lyapunov指数,并进行混沌特性分析,结果显示:货物发送增长量和增长率符合混沌特性,货物发送量不符合混沌特性;利用基于最大Lyapunov指数方法和BP神经网络方法对1999年1月到2013年4月共172个月的铁路货物发送增长量和增长率进行预测,预测结果表明基于最大Lyapunov指数预测值能够较好地与实际值相吻合,其预测的准确度明显好于BP神经网络预测值,因而混沌理论中的最大Lyapunov指数预测在货物发送量相关时间序列预测中有广泛的实用价值。Applying the chaos theory of the feature of analyzing the chaotic characteristics of nonlinear dynamic systems ,the railway freight traffic time series were analyzed. On the basis of Takens phase space reconstruc-tion ,the C-C method was used to calculate the embedded time-delay and embedded window and the G-P method was used to calculate the embedded dimension ,and then the small-data method was used to calculate the maxi-mum Lyapunov exponent of railway freight traffic time series. The Lyapunov exponent was used to analyze the chaotic characteristics of the time series. The analytical results show as follows ：The growth amount and growth rate of railway freight volumes have chaotic characteristics whereas the dispatched freight volume does not have the same. The maximum Lyapunov exponent method and BP neural network were separately used to forecast the growth amount and growth rate of railway freight traffic from January 1999 to April 2013. The re-sult shows that the predicted data using the maximum Lyapunov exponent method is anastomotic with the real data and the predicting accuracy of the maximum Lyapunov exponent method is better than the BP neural net-work.

关 键 词：货物发送量时间序列 混沌预测 最大LYAPUNOV指数 相空间重构 混沌判定 

分 类 号：U294.13[交通运输工程—交通运输规划与管理]