基于AFC数据的城市轨道交通OD客流动态预测  

Dynamic Prediction of Urban Rail Transit OD Passenger Flow Based on AFC Data

作　　者：田玙珩 罗霞[1,2] TIAN Yuheng;LUO Xia(School of Traffic and Logistics,Southwest Jiaotong University,Chengdu 611756,China;National and Local Joint Engineering Laboratory for Intelligent Integrated Transportation,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学交通运输与物流学院,四川成都611756 [2]西南交通大学综合交通运输智能化国家地方联合工程实验室,四川成都611756

出　　处：《综合运输》2022年第5期80-86,共7页China Transportation Review

基　　金：省科技厅科技计划项目(2020YJ0255)。

摘　　要：路网客流实时状态是城市轨道交通系统进行日常运营及关键决策的重要基础,针对目前城市轨道交通客流预测中站点层次预测方法较成熟,而客流分布预测较少的情况,提出基于时序神经网络的量测方程OD客流动态预测方法。利用地铁AFC数据,确定时序神经网络预测的最优数据粒度为15mins和最优时间序列阶数为4,以此构建时序神经网络框架,对站点进站量进行预测;对于站点进站客流与OD客流间的时空关联性,主要体现在进站客流的不同去向以及相同去向下不同的到达时间,建立量测方程反应这一联系,将进站客流转化为OD客流,并以成都地铁为例,对路网条件下不同分布特征OD客流进行预测验证,加权相对误差为14.08%,验证了模型的有效性。The real-time state of the network passenger flow is an important basis for the daily operation and key decision of the urban rail transit system.In view of the situation that the prediction method of station passenger flow is more mature in the current passenger flow prediction of urban rail transit,while the passenger flow distribution prediction is less,therefore,the measuring equation OD passenger flow dynamic prediction method based on the time-series neural network was proposed.Based on the AFC data of Chengdu Metro,this paper determined the optimal data granularity of 15mins and the optimal time series order of 4 for the prediction of the time-series neural network,and built the framework of the time-series neural network to predict the inbound passenger flow.In addition,aiming at the spatiotemporal correlation between the inbound passenger flow and OD passenger flow,which is mainly reflected in the different destinations of the inbound passenger flow and the different arrival times of the same destination passenger flow,the measuring equation was established to convert the inbound passenger flow into OD passenger flow.Moreover,taking the data of Chengdu Metro as an example,the prediction and verification of the OD passenger flow with different distribution characteristics under the road network condition were carried out.The weighted relative error is 14.08%,which proved the validity of the model.

关 键 词：城市轨道交通 客流动态预测 量测方程 客流分布 时序神经网络 

分 类 号：U293.13[交通运输工程—交通运输规划与管理]