摘要
为了全面了解城市轨道交通短期客流预测现有的研究进展,结合国内外相关的研究,梳理了近年来的研究状况,归纳了城市轨道交通短期客流预测的研究焦点并分类进行了讨论。重点从客流分析、预测方法、不同情况下预测方法和时间粒度选择3个方面总结归纳现有研究成果。研究结果表明:在客流分析方面,大多采用出行方式链、聚类分析等方法定量分析客流特征,缺乏定性定量综合分析;在预测方法方面,主要使用统计学、非线性以及神经网络的预测模型,并且随着预测方法研究的逐步深入,用于短期客流预测的3类解析模型更加完善,预测精度日益提高,但在完善模型缺陷方面仍有待进一步研究;在预测方法和时间粒度选择方面,主要研究正常情况和突发大客流2种情况的预测方法选择,以及工作日与非工作日、高峰与平峰时段的短期客流预测时间粒度选择,考虑的情况不够全面。未来的研究可以从客流分析、预测方法、不同情况下预测方法和时间粒度选择3个角度出发。首先,通过大数据,运用定性定量结合的方法对城市轨道交通的客流进行分析;其次,构建不同特点的客流预测综合模型,解决单一模型存在的问题,并在保证精度的基础上提高组合模型的计算速度;最后,合理选择节假日、突发事件等不同条件下线路客流和网络客流的预测方法和时间粒度。未来也可以进一步综合上述3个方面的成果,从而更加准确预测城市轨道交通的短期客流量,为合理的行车组织提供依据。
In order to comprehensively understand the current research progress of short-term passenger flow prediction of urban rail transit,the research status of recent years were summarized,the research focus of short-term passenger flow prediction of urban rail transit were summarized,and its classification at home and abroad were discussed.The existing research results from three aspects,such as passenger flow analysis,forecasting method,prediction methods under different conditions and time granularity selection were mainly summarized.The results show that in terms of passenger flow analysis,most of the travel mode chain,cluster analysis and other methods are used to quantitatively analyze the characteristics of passenger flow,but lack of comprehensive qualitative and quantitative analysis.In terms of prediction methods,statistical,nonlinear and neural network prediction models are mainly used.With the development of prediction methods,the three kinds of analytical models for short-term passenger flow prediction are more perfect and the prediction accuracy is improving day by day.However,the defects of the model still need to be further studied.In terms of prediction method and time granularity selection,the prediction method selection of normal situation and sudden large passenger flow,as well as the time granularity selection of short-term passenger flow prediction in working days and non-working days,peak and flat peak periods are mainly studied,which is not comprehensive enough.Future research can be carried out from three perspectives:passenger flow analysis,forecasting method,forecasting method under different conditions and time granularity selection.Firstly,through big data,the passenger flow of urban rail transit is analyzed by combining qualitative and quantitative methods.Secondly,the passenger flow prediction models with different characteristics are combined to solve the problems existing in the single model and improve the calculation speed of the combined model on the basis of ensuring the accuracy.Fin
作者
雷斌
张源
郝亚睿
景立竹
LEI Bin;ZHANG Yuan;HAO Ya-rui;JING Li-zhu(School of Civil Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,Shaanxi,China;Key Laboratory for Special Area Highway Engineering of the Ministry of Education,Chang'an University,Xi'an 710064,Shaanxi,China)
出处
《长安大学学报（自然科学版）》
CAS
CSCD
北大核心
2022年第1期79-96,共18页
Journal of Chang’an University(Natural Science Edition)
基金
陕西省交通科技项目(20-05R)
陕西省科学技术厅社会发展领域项目(2021SF-486)。
关键词
交通工程
轨道交通
客流预测
神经网络
短期客流
客流分析
组合模型
时间粒度
traffic engineering
rail transit
passenger flow prediction model
neural network
short-term passenger flow
passenger flow analysis
combination model
time granularity
