基于路网在途车辆数的拥堵指数短期预测  

Short-term prediction of the congestion index based on the number of vehicles on the road network

作　　者：孙士宏 宋国华[1] 孙建平 雷雪 朱瑞仪 范鹏飞[1] SUN Shi-hong;SONG Guo-hua;SUN Jian-ping;LEI Xue;ZHU Rui-yi;FAN Peng-fei(Key Laboratory of Big Data Application Technologies for Comprehensive Transport of Transport Industry,Beijing Jiaotong University,Beijing 100044,China;Beijing Transport Institute,Beijing 100073,China;School of Mathematical Sciences,Tongji University,Shanghai 200092,China)

机构地区：[1]北京交通大学,综合交通运输大数据应用技术交通运输行业重点实验室,北京100044 [2]北京交通发展研究院,北京100073 [3]同济大学,数学科学学院,上海200092

出　　处：《交通运输工程与信息学报》2023年第3期98-107,共10页Journal of Transportation Engineering and Information

基　　金：国家自然科学基金项目(71871015,51678045);国家重点研发计划项目(2018YFB1600701)。

摘　　要：交通拥堵指数是定量评价路网车辆运行状况的指标,准确预测交通拥堵指数的发展趋势是交通工程师关注的重点。为克服天气、节假日等复杂因素对交通高峰时段拥堵指数预测精度的干扰,本文以预测未来2 h范围内每5 min的拥堵指数为目标,首先,利用交通流基本图和交通波理论,从拥堵蔓延角度分析路网在途车辆数与拥堵指数的峰值滞后现象机理;其次,分析路网在途车辆数与拥堵指数的相关性,结合非线性回归方法,构建基于路网在途车辆数的拥堵指数预测模型(NV-NR模型);最后,利用北京市2018年6~9月工作日的数据集进行验证分析。结果表明,NV-NR模型在验证集中预测精度达91.9%;在拥堵指数达到中度拥堵阈值及以上的范围时效果更优,预测精度达93.8%,平均绝对误差为0.412。该方法能够有效预测交通高峰时段的拥堵指数短期变化趋势,为交通拥堵指数预测提供一种新的思路。The traffic congestion index quantitatively evaluates the operation statuses of vehicles on the road network,and the focus of traffic engineers is to accurately predict the changes in the trends of the traffic congestion index.To overcome the interference of complex factors,such as the weather and holidays,on the prediction accuracy of the congestion index during peak traffic hours,this study aims to predict the congestion index every 5 min within the next 2 h.First,we used the basic traffic flow diagram and the traffic wave theory,and analyzed the mechanism of the peak lag phenomenon between the number of vehicles on the road network and the congestion index from the perspective of congestion spread.Second,we analyzed the correlation between the number of vehicles on the road network and the congestion index,and then used a nonlinear regression method to construct a congestion index prediction model based on the number of vehicles on the road network(NV-NR model).Finally,the dataset of the working days in Beijing from June to September 2018 was used for validation analysis.The results show that the prediction accuracy of the NV-NR model in the vali-dation set is 91.9%;the effect is improved when the congestion index either reaches or exceeds the moderate congestion threshold,and the prediction accuracy and average absolute error are 93.8%and 0.412,respectively.This method can effectively predict the short-term changes in trends of the congestion index during the peak traffic hours,and provide a new approach for the prediction of the traffic congestion index.

关 键 词：城市交通 拥堵指数 短期预测 回归模型 路网在途车辆数 

分 类 号：U491[交通运输工程—交通运输规划与管理]