摘要
现有轨道交通车站高峰客流预测方法简化了车站高峰形成过程,基于默认假设,即车站高峰小时与所属线路高峰小时一致进行预测,忽略了车站与线路间存在的高峰偏差现象,造成部分车站高峰客流量被低估,导致车站能力设计不足,站台拥挤风险增加。从车站高峰形成机理出发,基于用地发生率模型,考虑不同目的出行行为的差异化,对客流属性进行划分,引入不同目的的出行时间概率分布函数,建立站点尺度的高峰小时与高峰客流预测模型框架。该模型真实反映了车站高峰与高峰偏差现象形成的这一复杂过程,可解释性强、符合实际,且能适用于建成环境、车站特征和轨道交通网络服务等变化情形下的车站高峰客流预测。验证结果显示:1)提出模型较传统模型提升了43%~47%的车站预测精度(高峰客流相应的MAPE值下降了5.7%~6.38%,高峰时间相应的APE值下降了23~50 min),具有更广泛的适用性和更稳定、更准确的预测结果,能为车站设计和运营方案制定提供更可靠的决策依据;2)各类出行目的的峰值和峰尺度存在差异,按不同比例叠加后,会产生不同的叠加曲线。揭示了车站高峰客流形成机理为不同用地产生的不同出行目的客流时间分布叠加曲线的高峰。
Existing subway station-level peak hour ridership(PHR) prediction methods often simplif the formation process of the station peak.The prediction process is conducted on the defaulted assumption that the passenger flow peak hour of each station always overlaps with its attributed line,which ignores the peak deviation phenomena between stations and lines and may produce underestimated peak hour ridership values.This may result in the design of stations with smaller capacity and increase the potential of congestion on the platforms.This paper presented a framework of a station-level peak period and peak ridership prediction method calibrated.It started from the formation mechanism of station peaks,and introduced the probability distribution function of travel time for different purposes based on the land use rate model by considering the differentiation of travel behaviors for different purposes and dividing passenger flow attributes.The proposed framework reflected the complex formation process of the station peak and station peak deviation phenomena in reality that had highly interpretability and can still be applied under changing conditions such as the built environment,station characteristics,and rail transit network services.The model validation results are drawn.(1) Compared with the traditional model,the proposed model has improved the prediction accuracy of 43%~47% stations(the corresponding MAPE value of peak ridership prediction has dropped by 5.7%~6.38%,and the corresponding APE value of peak periods prediction has dropped by 23~50 min),exhibiting wider applicability and more stable and accurate prediction results.It can provide a more reliable decision-making basis for station design and operation plan formulation.(2) The peaks and peak scales vary with travel purposes.Superposition of different proportions will form different superimposed curves.It reveals that the formation mechanism of the station peak ridership is the peak of the ridership time distribution overlay curve for different travel purposes generated by different land use.
作者
魏杰
余丽洁
任璐
陈宽民
WEI Jie;YU Lijie;REN Lu;CHEN Kuanmin(College of Transportation Engineering,Chang’an University,Xi’an 710064,China)
出处
《铁道科学与工程学报》
EI
CAS
CSCD
北大核心
2023年第3期867-877,共11页
Journal of Railway Science and Engineering
基金
国家自然科学基金资助项目(71871027)
陕西省自然科学基金资助项目(2022JQ-455)。
