改进时间序列模型在高速公路短时交通流量预测中的应用  

Application of improved time series model in forecasting of short-term traffic flow for freeway

作　　者：唐毅[1,2,3] 刘卫宁[1,2] 孙棣华[4,2] 魏方强 余楚中[4] 

机构地区：[1]重庆大学计算机学院,重庆400044 [2]重庆大学信息物理社会可信服务计算教育部重点实验室,重庆400044 [3]重庆高速公路集团有限公司机电分公司,重庆401121 [4]重庆大学自动化学院,重庆400044

出　　处：《计算机应用研究》2015年第1期146-149,共4页Application Research of Computers

基　　金：国家交通运输部科技资助项目(2011318221230);中国工程院重点咨询项目(2012-xz-22);国家教育部博士点基金资助项目(20120191110047)

摘　　要：为了提高短时交通流预测精度,针对传统时间序列模型在进行交通流量预测时存在无法动态调整模型参数、样本量过大导致序列的平稳性减弱、建模过程复杂等不足,从样本序列的动态选取及模型识别两方面进行优化,提出了一种改进的时间序列模型。利用渝武高速公路微波车检器的实测流量数据对改进前后的时间序列模型进行了实验验证和对比分析,结果表明改进后的时间序列模型有效克服了传统时间序列模型的不足,并对不同的交通流状况具有较好的适应性,无论在工作日还是节假日均具有更高的预测精度。In order to improve the prediction accuracy of short-term traffic flow, aiming at the shortcomings of the traditional time series model when used to forecast traffic flow, such as the parameter of the model could not adjust dynamically, smooth- ness of the time series abated due to sample size was too large and the complicated process of modeling, this paper improved time series model from two aspects including the dynamic selection of sample time series and model identification. Experiments and comparison analysis between the improved time series model and the previous one were carried out by using the actual traf- fic flow data of microwave vehicle detector on YuWu freeway. The results show that the improved time series model overcomes the shortcomings of the traditional time series model effectively, and has preferable adaptability and higher forecast accuracy no matter on weekdays or holidays under different traffic flow conditions.

关 键 词：交通工程 交通流量预测 时间序列 样本序列 动态建模 参数调整 

分 类 号：TL91[核科学技术—核技术及应用] U495[交通运输工程—交通运输规划与管理]