【摘要】 基于为地铁站周边步行设施规划、交通组织及信号控制优化提供依据的研究目的,提出进出地铁站短时步行交通需求动态预测模型。以地铁列车到发时间间隔为单位,通过Levenberg-Marquardt算法标定的指数平滑模型对下一到发时间间隔内的进、出站客流量进行预测,基于动态行程时间分析地铁站周边人行道、无信号交叉口、信号交叉口以及立体行人过街设施处的步行交通阻抗函数,并提出侵占人行道、寒地城市冰雪路面、立体行人过街设施与水平路面高度差、缺乏方向指示4类惩罚系数对步行阻抗函数进行修正,采用多路径交通分配模型将指数平滑模型预测得到的进出站步行交通量分配至地铁站周边慢行交通网络。以哈尔滨市地铁1号线博物馆站1号站口为例进行良好天气及雪后进出站客流量预测、对比冬夏两季步行阻抗函数计算和步行流量分配情况,研究结果表明,在通过连续6个到发时间间隔客流量数据标定指数平滑模型的前提下,进出站客流量预测相对误差总体呈下降趋势,第7个到发时间间隔进、出站客流量预测相对误差分别为1.67%和-3.57%,根据地铁站周边步行阻抗函数分配得到的步行流量分布与实际情况基本吻合且能反映步行环境和季节变化引起的路线选择倾向变化,证实模型可用于地铁站周边步行设施的规划与评价,并可协同地铁站邻近交叉口进行考虑行人效益的动态交通控制。更多还原

【Abstract】 As the basis of walking facilities planning, traffic operation, and signal control optimization around metro stations, a dynamic forecasting model of short-time demand of pedestrians arrive at and leave a metro station is developed. The time interval of metro system is taken as the time unit, an exponential smoothing model calibrated by Levenberg-Marquardt method is used to forecast passenger volume at entrances and exits in the next time unit. Walking impedance of sidewalk, non-signalized intersection, signalized intersection, and tridimensional crossing road facilities are analyzed based on dynamic travel time. Four kinds of penalty coefficients are defined to revise walking impedance, then a model for multi-path traffic assignment is used to assign pedestrians forecasted by exponential smoothing model to slow-moving network around the metro system. As a case study, the passenger volume at entrances and exits on days of good weather and after snow is forecasted, calculated, then compared with walking impedance and assigned passenger volume in summer and winter at Bowuguan Station Exit Port 1 of Harbin Metro Line One. The results show that with the volume of six time intervals calibrating the exponential smoothing model, the relative error decreases as time passes by. In the seventh time interval, the relative errors of arrived and outbound passenger volume are 1.67% and-3.57%, respectively. Assignment of passenger volume based on walking impedance functions fits the actual passenger volume on the same routes. The model can reflect different route choosing tendency in different walking conditions and different seasons. The dynamic forecasting model is verified to be applied in planning and evaluation of walking facilities, as well as cooperating with intersections near metro stations for dynamic traffic control considering pedestrian benefits.更多还原