摘要
基于历史数据挖掘实现精准的高速铁路车站短期客流预测能有效支撑客运站工作组织的动态调整,提升铁路运输服务水平。考虑列车开行方案、车站关系对客流的影响,提出基于时空图神经网络的铁路车站短期到发客流预测方法,在空间卷积模块中,用关系图卷积融合铁路物理网络、基于列车开行方案的服务网络和车站关系网络以挖掘空间关联性,在时间注意力模块中用注意力机制获取时间关联特征,并用多层长短期记忆人工神经网络实现路网上多站的多步客流预测。选取京沪高速铁路沿线车站到发客流为研究对象,并对比不同步长下的短期到发客流预测效果,结果表明STGNN明显优于对比预测方法。
Accurate prediction of short-term passenger flow at high-speed railway stations based on the analysis of the historical data can effectively support the dynamic adjustment of the railway passenger operation plan in passenger stations to enhance the service level of railway transportation. Considering the influence of train service and station relationship on passenger flow, a prediction method of short-term arrival and departure passenger flow of railway stations based on Spatial-Temporal Graph Neural Network(STGNN) was proposed. In the spatial convolution module, the relationship graph convolution was used to fuse the physical railway network, the service network based on the train plan and station correlation network to dig the spatial correlation. In the time attention module, the attention mechanism was used to obtain the temporal correlation with the multi-layer Long Short-Term Memory(LSTM) to predict multi-step results of each station in the railway network. The experiment test was conducted on the daily data of the stations along the Beijing—Shanghai high-speed railway. The results on various prediction steps at the station-level and network-level show the STGNN outperforms the baselines.
作者
何必胜
朱永俊
陈路锋
闻克宇
HE Bisheng;ZHU Yongjun;CHEN Lufeng;WEN Keyu(School of Transportation and Logistics,Southwest Jiaotong University,Chengdu 611756,China;Comprehensive Transportation Key Laboratory of Sichuan Province,Chengdu 611756,China;National United Engineering Laboratory of Integrated and Intelligent Transportation,Chengdu 611756,China;School of Automation Engineering,University of Electronic Science and Technology of China,Chengdu 611756,China;China Railway Economic and Planning Research Institute,Beijing 100038,China;School of Economics and Business Administration,Southwest Jiaotong University,Chengdu 611756,China)
出处
《铁道学报》
EI
CAS
CSCD
北大核心
2022年第9期1-8,共8页
Journal of the China Railway Society
基金
国家重点研发计划(2017YFB1200701)
国家自然科学基金(61603317,52005082)
中国铁路总公司科技研究开发计划(J2018Z403,2017X010-K)
四川省软科学研究计划(2020JDR0129)。
关键词
高速铁路
客流预测
关系图卷积
时间注意力
时空图神经网络
high-speed railway
passenger flow prediction
relationship graph convolution
temporal attention
Spatial-Temporal Graph Neural Network
