地球信息科学学报 ›› 2023, Vol. 25 ›› Issue (4): 698-713.doi: 10.12082/dqxxkx.2023.220817

• 轨迹与交通 • 上一篇    下一篇

基于注意力机制的城市轨道交通网络级多步短时客流时空综合预测模型

张金雷1(), 陈奕洁1, Panchamy Krishnakumari2, 金广垠3,*(), 王骋程4, 杨立兴1   

  1. 1.北京交通大学 轨道交通控制与安全国家重点实验室,北京 100044
    2.代尔夫特理工大学 交通与规划学院,代尔夫特 2623 CN
    3.国防科技大学 系统工程学院,长沙 410005
    4.山东省交通规划设计院集团有限公司 科技研发中心, 济南 250000
  • 收稿日期:2022-10-24 修回日期:2023-02-03 出版日期:2023-04-25 发布日期:2023-04-19
  • 通讯作者: *金广垠(1996—),男,安徽芜湖人,博士,主要从事时空数据挖掘研究。E-mail: jinguangyin96@foxmail.com
  • 作者简介:张金雷(1993—),男,河北保定人,博士,讲师,主要从事人工智能与交通大数据挖掘,城市轨道交通运营管理等相关研究。E-mail:zhangjinlei@bjtu.edu.cn
  • 基金资助:
    国家自然科学基金项目(72201029);国家自然科学基金项目(71825004);国家自然科学基金项目(72288101);中国博士后科学基金资助项目(2022M720392)

Attention-based Multi-step Short-term Passenger Flow Spatial-temporal Integrated Prediction Model in URT Systems

ZHANG Jinlei1(), CHEN Yijie1, Panchamy Krishnakumari2, JIN Guangyin3,*(), WANG Chengcheng4, YANG Lixing1   

  1. 1. State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing 100044, China
    2. Department of Transport and Planning, Delft University of Technology, Delft, 2623 CN, Netherlands
    3. College of System Engineering, National University of Defense Technology, Changsha 410005, China
    4. Research and Development Center, Shandong Provincial Communications Planning and Design Institute Group Co., Ltd., Jinan 250000, China
  • Received:2022-10-24 Revised:2023-02-03 Online:2023-04-25 Published:2023-04-19
  • Contact: JIN Guangyin
  • Supported by:
    National Natural Science Foundation of China(72201029);National Natural Science Foundation of China(71825004);National Natural Science Foundation of China(72288101);China Postdoctoral Science Foundation(2022M720392)

摘要:

准确、可靠的短时客流预测可为城市轨道交通提供运营决策支持。本研究以基于Transformer机制的LSTM网络、深度注意力模块和CNN网络为基础,提出了城市轨道交通网络级多步短时客流预测模型(STIPM)。该预测模型由3个分支组成,分支一以时间序列进站客流为输入,提出了基于Transformer机制的LSTM网络提取该数据中的时间相关性;分支二以基于时间步的OD数据为输入,提出了深度注意力模块挖掘数据中大量的时间、空间相关性,利用基于时间步的OD数据能够更好地展现站间联系紧密程度和全局信息,从而完成了拓扑网络信息提取;分支三的输入为POI数据,使用CNN网络获取其时空相关性,并作为时间与空间特征之间的纽带。为了保证在预测精度足够高的条件下,获得更长的预测时间和更详细的预测信息,本文采用“神经网络多输出”策略,完成了多步预测任务。本文在2个大规模城市轨道交通真实数据集中对该模型进行测试,并将预测结果与10个基准模型和4个消融实验模型进行对比,在RMSE、MAE与WMAPE评估指标中,STIPM模型均得到最高的预测精度,结果表明该模型具有一定的优越性与鲁棒性。

关键词: 城市轨道交通, 短时客流预测, 多步预测, 深度学习, 交通大数据, 时空特征挖掘, 特征融合

Abstract:

Accurate and reliable short-term passenger flow prediction can support operations and decision-making of the URT system from multiple perspectives. In this paper, we propose a URT multi-step short-term passenger flow prediction model at the network level based on a Transformer-based LSTM network, Depth-wise Attention Block, and CNN network, named as Spatial-Temporal Integrated Prediction Model (STIPM). The STIPM comprises three branches. The first branch takes time-series inflow data as input, and a Transformer-based LSTM network is selected to extract the temporal correlations. The second one takes timestep-based OD data as input, and many spatial and temporal features are captured using Depth-wise Attention Blocks. Meanwhile, timestep-based OD data can better include inter-station relations and global information. The third branch takes Point of Interest data (POI) as input and CNN network is utilized for spatiotemporal features extraction, which can also become the bridge between spatial and temporal features. Moreover, the “Multi-input-multi-output Strategy” for multi-step prediction is used to obtain a longer prediction period and more detailed information under a relatively high forecasting accuracy. The STIPM is applied to two large-scale real-world datasets from the URT system, and the obtained prediction results are compared with ten baselines and four variants from itself, in which STIPM model achieves highest prediction accuracy indicated by RMSE, MAE, and WMAPE evaluations, which demonstrates the superiority and robustness of the STIPM.

Key words: urban Rail Transit, short-term passenger forecasting, multi-step forecasting, deep learning, traffic big-data, spatiotemporal features mining, features fusion