基于多时间尺度的城市轨道交通短时OD估计

基于状态空间模型构建了城市轨道交通短时OD估计的多模型组合方法,估计早晚高峰期间15 min内进站客流的去向目的站.组合模型以不同时间尺度下的进站客流分流率为状态变量,并利用历史数据预估其中通勤客流的分流率,然后通过交互多模型算法加权融合不同时间尺度下的分流率估计结果.以北京地铁为案例,研究表明:早高峰期间的15 min分流率估计误差的平均值和最大值分别为16.4%和21.8%,晚高峰期间分别为22.7%和24.6%,比既有文献的估计误差减小了约一半.本文的研究成果可为实时的线网客流分布预测提供更准确的输入数据,以辅助运营管理部门实现客流预警和应急响应.

Short-term Origin-destination Estimation for Urban Rail Transit Based on Multiple Temporal Scales

To estimate the destinations of the entrance passenger flows at urban rail transit stations in each 15 minutes at rush hours, based on state-space models, a multi-model composition approach of short-term origin- destination estimation is proposed. Firstly, the split rates of the entrance passenger flows under different temporal scales are used as the state variables for the composition model. Then the split rates of the commuter entrance passenger flows are estimated based on the historical data. Secondly, the results of the estimated split rates under different temporal scales are weightily fused by the interacting multiple model algorithm. The research of the Beijing subway demonstrates that the average and maximum estimation errors of the split rates at morning rush hours are 16.4% and 21.8% respectively. And those of evening rush hours are 22.7% and 24.6% respectively. The result reduces the estimation errors by nearly one half in comparison with those in present literatures. The results of this paper can provide more accurate input data for the realtime prediction of network traffic distribution to assist the management agency in implementing the early warning and emergency response system on mass passenger flows.