分车型的高速公路短时交通流量预测方法研究.docVIP

分车型的高速公路短时交通流量预测方法研究( 周 桐1,2，杨智勇1,2，孙棣华2,魏方强2 （）摘要关键词交通流中图分类号 文献标识码 A Research on Forecast Method of the Short-Term Freeway Mixed Traffic Flow ZHOU Tong1, 2, YANG Zhi-yong1, 2, SUN Di-hua2 , WEI Fang-qiang2 (1.School of Information Engineering, The Chongqing Vocational Institute of Engineering, Chongqing, 400037, China; 2.College of Automation, Chongqing University, Chongqing 400030, China) Abstract：flows behave uncertainty and random feature, which lead to traditional forecast method has low prediction precision. This paper proposed a new multi-type vehicles forecast method of the short-term freeway traffic flow. Based on the change law of traffic flow of the various type vehicles is different, the paper used improved time series method to forecast the flow value of large type car and trailer, and used second exponential smoothing method to forecast the flow value of small type car and middle type car, then obtained the prediction value of the mixed traffic flow by weight average method on the basic of vehicle conversion factor. Finally, comparing with nonparametric regression model and traditional time series model, the results of experiment shows that the new method can increase the forecasting accuracy of short-term freeway traffic flow on weekdays or holidays, and the results can supply the foundation to further improve the ability of highway management and control. Key Words：prediction; multi- type vehicle; time series method; second exponential smoothing method0引言 高速公路高精度短时交通流预测能够为交管部门的管理和控制提供决策和支持，对保证高速公路正常运行具有重要的实际意义[1]。对于短时交通流量预测，早期的预测方法有指数平滑法、时间序列法、卡尔曼滤波方法等，随着最优化理论及数理统计学等理论的发展，又出现了一些新的方法，如神经网络、支持向量机等[2-7]这些方法均能在一定程度上预测车流量，但上述方法对混合交通流中分车型的车流预测相关研究较少。由于混合交通流的车辆类型具有较强的不确定性和随机性，直接对其车流量进行预测难以获得较高的预测精度。基于上述原因，本文根据各种车型每天的车流量变化的规律性不同的特点，选用适当的预测算法分别对其流量进行预测，然后通过车辆折算系数对各车型的流量预测值进行加权求和，得到总车流量预测值，并通过实验验证新方法的预测精度。 1 分车型的车流规律性 如图1所示为某微波车检器某一天混合交通流的量时间序列图1 混合交通流的总车流量 图从图1，混合交通流受随机因素的影响较大，变化并无明显规律可预测性弱单一车型每天的流量变化却出一定的规律性其中大型车与拖挂车流量变化相对平缓，且流量较小，小客车流量变化呈现