Abstract: This paper proposes a new mixed-flows model to capture the main statistical features of traffic volume flows over Beijing transportation networks. The model consists of three parts: the major part decided by residents’ everyday travel, which remains relatively stable and follows a certain M-shape curve; random fluctuations whose magnitude follow time variant super-gaussian distribution, and “peaks” whose arriving rates obey Negative exponential distribution. This model can nicely describe the arriving behaviors of the "peaks" in the traffic volume flow. Further analysis of "peaks" arrives in negative exponential distribution and traffic flow volume is in a multi-fractal time series. The obtained results sheds lights on our understandings of the origins and characteristics of the urban traffic flow, which are thus useful for traffic prediction and control.

Key words: traffic flow theory, statistical analysis, mixed-flows model

摘要： 针对北京城市网络交通流的统计特性，提出了一种新颖的混合流量模型，通过三类交通流量的组合得出：M型日均流量、振荡幅值符合时变Super-Gaussian分布的分形随机波动和到达率符合Poisson分布的流量“尖峰”，以此来描述实际的交通流。通过该模型得出北京市交通流数据的去日平均概要曲线，对去日概要曲线进行离差统计分析可以较好的反映实际观察的交通流“尖峰”现象，这种现象和Internet通讯流的研究结果相符。进一步对交通流“尖峰”的到达规律进行研究发现，“尖峰”到达时间间隔呈负指数分布，并且这种随机波动具有与高速公路交通流相似的多分形特性。对交通流“尖峰”的认识和其到达率的结论对于进一步理解城市交通流复杂现象的起因和特性，提高交通流预测和仿真效果有着重要作用。

关键词: 交通流理论, 统计分析, 混合流量模型