摘要
交通拥塞在形成和消散过程中车辆运行风险均处于较高的水平,而拥塞环境下影响车辆运行风险的因素较多,分析复杂.如何准确识别城市交通拥塞环境下车辆运行风险的关键因素并对其进行评估,在缓解城市交通拥塞以及降低行车风险方面具有重要的意义.首先,将城市道路交通拥塞环境下车辆运行风险解析为拥塞形成过程和拥塞消散过程中的车辆运行风险.根据交通系统四要素"人、车、路、环境",初步选取换道频次、车型比例、拥塞时长等11个风险因子.其次,通过专家打分法将车辆运行风险等级划分为低风险、中等风险、高风险、极高风险.结合主成分分析法对标准的BP神经网络进行改进,并对模型进行训练.将改进前后的模型进行对比分析,改进后的模型拟合优度判定系数达97.13%,较改进前高出5.67%.最后,进行实例应用.采用改进的BP神经网络,建立了5+8+1模式的拥塞环境下车辆运行风险识别模型.研究表明换道频次、车型比例、平均密度、拥塞时长、拥塞等级、天气情况等6个因子对车辆运行风险影响较大,其中换道频次权重最高,其次为拥塞时长.11个影响因子中车头时距主成分系数平均值最小为0.109,其影响最小.建立的风险识别模型能够为规避城市道路拥塞环境下的车辆运行风险以及拥塞治理提供参考.
Vehicle operation risks are at a high level during the formation and dispersion of traffic congestion. However, there are many factors influencing the risk of vehicle operation in the congestion environment, and the analysis is complex. It is great significant that how to accurately identify and evaluate the critical factors of the risk of vehicle operation under urban traffic congestion in alleviating urban traffic congestion and reducing traffic risk. First, the vehicle operating risk in urban road traffic congested environment is analyzed as the vehicle operating risk in the process of congestion formation and dissipation. According to the 4 elements of traffic system, uhuman , vehicle, road and environment", 11 risk factors such as frequency of lane change, vehicle type ratio, and duration of congestion are initially selected. Then , the risk grades of vehicles are classified as low risk, medium risk, high risk and extremely high risk through expert scoring. The standard B P neural network is improved by principal component analysis and the model is trained. According to the comparative analysis of the models before and after improvement, the goodness of fit judgment coefficient of the improved model reached 97. 13%, which is 5. 67% higher than that before improvement. Finally, an example application is m a d e . Using the improved B P neural network, the vehicle operating risk identification model in 5 + 8 + 1 mode is established. The result shows that (1) Six factors, including frequency of lane change, vehicle type ratio, average density, duration of congestion, congestion level, and weather situation have a greater impact on vehicle operating risk. Among them, the weight of lane change frequency is the highest, followed by the duration of congestion.(2) Among the 11 influencing factors, the average value of the principal component coefficient of the headway is the smallest and the value is 0.109, which has the least influence. The established risk identification model can provide a reference for evading vehicle operation risk in urban road congestion environment and congestion control.
作者
胡立伟
杨锦青
何越人
孟玲
罗振武
HU Li-wei;YANG Jin-qing;HE Yue -ren;MENG Ling;LUO Zhen-wu(School of Traffic Engineering,Kunming University of Science and Technology,Kunming Yunnan 650500,China)
出处
《公路交通科技》
CAS
CSCD
北大核心
2019年第10期105-113,共9页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金项目(61863019)
关键词
城市交通
风险识别
BP神经网络
交通拥塞
主成分分析
urban traffic
risk identification
BP neural network
traffic congestion
principal component analysis
