吉林大学学报(工学版) ›› 2023, Vol. 53 ›› Issue (2): 430-438.doi: 10.13229/j.cnki.jdxbgxb20210720

• 交通运输工程·土木工程 • 上一篇    下一篇

基于循环神经网络的城市轨道交通短时客流预测

张惠臻1(),高正凯1,2,李建强2,王晨曦1,潘玉彪1,3,王成1,王靖1   

  1. 1.华侨大学 计算机科学与技术学院,福建 厦门 361021
    2.北京工业大学 信息学部软件学院,北京 100124
    3.南威软件股份有限公司,福建 泉州 362000
  • 收稿日期:2021-07-30 出版日期:2023-02-01 发布日期:2023-02-28
  • 作者简介:张惠臻(1983-),男,副教授,博士. 研究方向:交通大数据. E-mail: zhanghz@hqu.edu.cn
  • 基金资助:
    国家自然科学基金项目(61802133);福建省科技计划重点项目(2020H0016)

Short⁃term passenger flow forecasting of urban rail transit based on recurrent neural network

Hui-zhen ZHANG1(),Zheng-kai GAO1,2,Jian-qiang LI2,Chen-xi WANG1,Yu-biao PAN1,3,Cheng WANG1,Jing WANG1   

  1. 1.School of Computer Science and Technology,Huaqiao University,Xiamen 361021,China
    2.School of Software Engineering,Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China
    3.Linewell Software Co. ,Ltd. ,Quanzhou 362000,China
  • Received:2021-07-30 Online:2023-02-01 Published:2023-02-28

摘要:

为更好地预测城市轨道交通的短时客流情况,提出了基于循环神经网络模型的预测方法。首先,针对轨道交通进出站客流数据,利用Pearson相关系数确定短时客流影响因素;然后,改进K-means聚类算法划分高、中、低客流量三类轨道站点,分析客流时空分布规律及高峰时间段;最后,采用分别基于长短时记忆神经网络(LSTM)与门控循环单元(GRU)的短时客流预测方法,预测不同类型站点在不同时段的客流。实验结果表明:5 min为预测的最佳时间粒度,在此时间粒度下GRU模型整体性能优于LSTM模型。

关键词: 交通运输工程, 城市轨道交通, 短时客流预测, 循环神经网络, 长短时记忆神经网络, 门控循环单元

Abstract:

In order to better predict the short-term passenger flow of urban rail transit, a prediction method based on the recurrent neural network model is proposed. Firstly, based on the actual passenger flow data of each rail transit station, the Pearson correlation coefficient is used to determine the influencing factors of short-term passenger flow of rail transit, such as the weather conditions, historical passenger flow, whether it is a peak time period, whether it is a working day, etc. Secondly, the K-means clustering algorithm is used to classify rail transit stations into three types: high, medium, and low passenger flow stations. Then the distribution of passenger flow for each station type in time and space is analyzed, to determine the peak period of passenger flow for each station type. Finally, two urban rail transit short-term passenger flow prediction methods based on long-short term memory neural network(LSTM) and gated recurrent unit(GRU) respectively are proposed to predict the passenger flow of each type of station in different time period. The experimental results show that 5 min is the best time granularity for short-term passenger flow prediction of the two models. In this time granularity, the overall performance of the GRU model is better than the LSTM model.

Key words: engineering of communication and transportation, urban rail traffic, short-term passenger flow forecast, recurrent neural network, long-short term memory neural network, gated recurrent unit

中图分类号: 

  • U121

表1

客流量与相关影响因素的Pearson相关系数"

相关影响因素Pearson相关 系数显著性 (双尾)
[-45,-30)min客流量0.880*0.000
[-30,-15)min客流量0.930*0.000
[-15,0)min客流量0.960*0.000
温度0.304*0.000
湿度-0.176*0.000
能见度0.053*0.004
降水量-0.093*0.000
云量0.420*0.000

图1

不同时段客流影响分析"

表2

城市轨道交通短时客流影响因素"

变量说 明
x1温度/°C
x2湿度/%
x3能见度/km
x4降水量/mm
x6前一时段客流量/人次
x7是否工作日(1表示非工作日、0表示工作日)
x8

是否高峰时段(1表示高峰时段、

0表示平峰时段、-1表示停运时段)

表3

厦门市轨道交通一号线站点聚类分析结果"

类别站点数站点名日均客流量/人次
高客流量2镇海路、乌石浦≥20 000
中客流量12

中山公园、将军祠、文灶、

湖滨东路、莲坂、莲花路口、

吕厝、塘边、火炬园、集美学

村、园博苑、厦门北站

7500~20 000
低客流量10

殿前、高崎、杏林村、杏锦路、

官任、诚毅广场、集美软件

园、集美大道、天水路、岩内

<7500

图2

各类型站点工作日与非工作日客流时间分布情况"

图3

SPFF模型结构"

表4

采用LSTM/GRU的SPFF模型训练时间"

时间 粒度 /minLSTM/GRU模型训练时间/s
时间步长1时间步长3时间步长6
单轮时间总时间单轮时间总时间单轮时间总时间
50.8/0.7

629/

583

2.6/2.3

2047/

1842

6.2/5.7

4934/

4603

100.4/0.3

323/

283

1.0/0.9

834/

758

2.7/2.6

2173/

2085

150.3/0.2

259/

228

0.5/0.5

435/

393

1.6/1.6

1293/

1278

图4

各类客流量站点预测结果"

表5

模型预测结果比较分析"

站点模型时间粒度 /min时间步长
136
RMSEMAERMSEMAERMSEMAE
镇海路LSTM538.224.835.223.634.723.0
1056.338.662.443.964.444.8
1579.656.574.251.981.958.3
GRU538.525.834.223.233.822.6
1056.240.354.438.551.236.1
1578.355.375.353.576.555.1
乌石浦LSTM524.416.423.516.023.415.9
1041.628.341.028.339.527.5
1558.840.357.340.128.141.9
GRU524.416.522.915.622.315.1
1040.728.239.927.038.526.3
1560.742.454.736.855.237.9
吕厝LSTM519.113.217.512.319.713.2
1026.919.029.320.031.121.1
1540.027.538.927.938.126.3
GRU519.914.217.412.217.111.7
1028.019.728.219.831.022.0
1543.631.939.328.339.329.0
集美学村LSTM518.612.515.710.916.511.8
1023.816.123.615.929.419.7
1534.122.534.02336.224.3
GRU518.412.416.411.615.610.5
1023.415.724.015.922.114.7
1534.022.934.323.832.721.7
官任LSTM511.97.011.37.312.78.8
1019.612.619.913.418.112.1
1528.818.528.118.624.515.2
GRU512.28.111.27.211.77.9
1020.013.219.112.418.012.1
1530.120.727.218.224.415.5
集美大道LSTM57.556.64.46.64.5
1011.17.110.66.810.77
1515.810.415.410.314.19.1
GRU57.55.06.64.36.84.6
1011.67.810.66.910.57.2
1515.610.416.010.914.69.5
1 中国城市轨道交通协会. 城市轨道交通2020年度统计和分析报告[J]. 隧道建设(中英文), 2021, 41(4): 691.
China Association of Metros. Statistics and analysis Report of urban rail transit in 2020[J]. Tunnel Construction(English and Chinese), 2021, 41(4): 691.
2 Schimbinschi F, Moreira-Matias L, Nguyen V X, et al. Topology-regularized universal vector autoregression for traffic forecasting in large urban areas[J]. Expert Systems with Applications, 2017, 82: 301-316.
3 Guo J H, Huang W, Williams B M. Adaptive kalman filter approach for stochastic short-term traffic flow rate prediction and uncertainty quantification[J]. Transportation Research Part C, 2014, 43(2): 50-64.
4 张淑玉. 基于贝叶斯理论的铁路短期客流预测方法研究[D]. 北京: 北京交通大学交通运输学院, 2018.
Zhang Shu-yu. Research on forecasting method of short term passenger flow based on Bayesian theory[D]. Beijing: School of Traffic and Transportation, Beijing Jiaotong University, 2018.
5 孟品超,李学源,贾洪飞,等.基于滑动平均法的轨道交通短时客流实时预测[J]. 吉林大学学报:工学版, 2018, 48(2): 448-453.
Meng Pin-chao, Li Xue-yuan, Jia Hong-fei, et al. Short-time rail transit passenger flow real-time prediction based on moving average[J]. Journal of Jilin University(Engineering ang Technology Edition), 2018, 48(2): 448-453.
6 Zhang Y R, Zhang Y L, Haghani A. A hybrid short-term traffic flow forecasting method based on spectral analysis and statistical volatility model[J]. Transportation Research Part C: Emerging Technologies, 2014, 43: 65-78.
7 刁文君. 基于灰色模型的城市轨道交通客流预测研究[D]. 天津: 天津大学管理与经济学部, 2016.
Diao Wen-jun. Study on the forecast of urban rail transit passenger flow based on Grey Model[D]. Tianjin: College of Management and Economy, Tianjin University, 2016.
8 杨兆升,王媛,管青.基于支持向量机方法的短时交通流量预测方法[J]. 吉林大学学报:工学版, 2006, 36(6): 881-884.
Yang Zhao-sheng, Wang Yuan, Guan Qing. Short-term traffic flow prediction method based on SVM[J]. Journal of Jilin University(Engineering ang Technology Edition), 2006, 36(6): 881-884.
9 姚智胜,邵春福,熊志华,等.基于主成分分析和支持向量机的道路网短时交通流量预测[J]. 吉林大学学报:工学版, 2008, 38(1): 48-52.
Yao Zhi-sheng, Shao Chun-fu, Xiong Zhi-hua,et al. Short-term traffic volumes forecasting of road network based on principal component analysis and support vector machine[J]. Journal of Jilin University (Engineering ang Technology Edition), 2008, 38(1): 48-52.
10 陈健冲. 基于改进卡尔曼滤波算法的短时公交客流预测模型研究[D]. 大连: 大连海事大学信息科学技术学院, 2020.
Chen Jian-chong. Research on short-term bus passeng based on improved Kalman flow prediction model filter algorithm[D]. Dalian: College of Information Science and Technology, Dalian Maritime University, 2020.
11 Bai Y, Sun Z Z, Zeng B, et al. A multi-pattern deep fusion model for short-term bus passenger flow forecasting[J]. Applied Soft Computing, 2017, 58: 669-680.
12 Glisovic N, Milenkovic M, Bojovic N, et al. A hybrid model for forecasting the volume of passenger flows on Serbian railways[J]. Operational Research, 2016, 16(2): 271-285.
13 田壮壮. 城市轨道交通延伸线的进出站客流预测[D]. 南昌: 华东交通大学电气与自动化工程学院, 2019.
Tian Zhuang-zhuang. Forecasting passenger flow of inbound and outbound for urban rail transit extension line[D].Nanchang: School of Electrical and Automation Engineering, East China Jiantong University,2019.
14 王秋雯, 陈彦如, 刘媛春. 基于卷积长短时记忆神经网络的城市轨道交通短时客流预测[J/OL]. [2020-09-02]. / 10.13195/j.kzyjc.2020.0501
doi: 10.13195/j.kzyjc.2020.0501
15 赵阳阳, 夏亮, 江欣国. 基于经验模态分解与长短时记忆神经网络的短时地铁客流预测模型[J]. 交通运输工程学报, 2020, 20(4): 194-204.
Zhao Yang-yang, Xia Liang, Jiang Xin-guo. Short-term metro passenger flow prediction based on EMD-LSTM[J]. Journal of Traffic and Transportation Engineering, 2020, 20(4): 194-204.
16 Tsai T H, Lee C K, Wei C H.Neural network based temporal feature models for short-term railway passenger demand forecasting[J]. Expert Systems with Applications, 2009, 36(2): 3728-3736.
17 Pearson K. Notes on the history of correlation[J]. Biometrika, 1920, 13(1): 25-45.
18 徐威,郑长江,马庚华,等.基于k-means聚类的城市轨道交通站点分类研究[J].贵州大学学报:自然科学版, 2018, 35(6): 106-111.
Xu Wei, Zheng Chang-jiang, Ma Geng-hua, et al. Classification of urban rail transit stations based on k-means clustering[J]. Journal of Guizhou University (Natural Sciences), 2018, 35(6): 106-111.
19 Cho K, Merrienboer B V, Gulcehre C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[C] ∥EMNLP, Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, Doha, 2014: 1724-1734.
20 李梅. 基于深度学习的短时公交客流预测研究[D]. 北京: 北京交通大学经济管理学院, 2019.
Li Mei. Research on short-time bus passenger flows forecasting based on deep learning[D]. Beijing: School of Economics and Management, Beijing Jiaotong University, 2019.
21 王挺. 城市轨道交通短时客流预测时间粒度选择[D]. 北京: 北京交通大学土木建筑工程学院, 2018.
Wang Ting. Selecting time granularity of short-term passenger flow forecast in urban rail transit[D]. Beijing: School of Civil Engineering, Beijing Jiaotong University, 2018.
[1] 王菁,万峰,董春娇,邵春福. 城市轨道交通站点吸引范围及强度建模[J]. 吉林大学学报(工学版), 2023, 53(2): 439-447.
[2] 李津,孙雨彤,魏小忠,焦玉玲. 考虑柔性车道设置的公交优先信号设计[J]. 吉林大学学报(工学版), 2023, 53(2): 448-456.
[3] 张龙,徐天鹏,王朝兵,易剑昱,甄灿壮. 基于卷积门控循环网络的齿轮箱故障诊断[J]. 吉林大学学报(工学版), 2022, 52(2): 368-376.
[4] 潘晓英,魏德,赵逸喆. 基于Mask R⁃CNN和上下文卷积神经网络的肺结节检测[J]. 吉林大学学报(工学版), 2022, 52(10): 2419-2427.
[5] 徐洪峰, 高霜霜, 郑启明, 章琨. 信号控制交叉口的复合动态车道管理方法[J]. 吉林大学学报(工学版), 2018, 48(2): 430-439.
[6] 孟品超, 李学源, 贾洪飞, 李延忠. 基于滑动平均法的轨道交通短时客流实时预测[J]. 吉林大学学报(工学版), 2018, 48(2): 448-453.
[7] 赵学彧, 杨家其, 彭亚美. 城市轨道交通与地面公交竞合关系演化机制[J]. 吉林大学学报(工学版), 2017, 47(3): 756-764.
[8] 王海玮, 温惠英, 刘敏. 夜间环境驾驶员精神负荷的生理特性评估与实验[J]. 吉林大学学报(工学版), 2017, 47(2): 420-428.
[9] 姜桂艳, 刘彬, 隋晓艳, 马明芳. 基于IC卡收费系统的公交客流信息实时采集方法[J]. 吉林大学学报(工学版), 2016, 46(4): 1076-1082.
[10] 宗芳, 王占中, 贾洪飞, 焦玉玲, 吴杨. 基于支持向量机的通勤日活动-出行持续时间预测[J]. 吉林大学学报(工学版), 2016, 46(2): 406-411.
[11] 李世武, 徐艺, 孙文财, 王琳虹, 郭梦竹, 柴萌. 基于瞳孔直径的撞固定物冲突自反馈识别方法[J]. 吉林大学学报(工学版), 2016, 46(2): 418-425.
[12] 潘义勇, 马健霄, 孙璐. 基于可靠度的动态随机交通网络耗时最优路径[J]. 吉林大学学报(工学版), 2016, 46(2): 412-417.
[13] 赵淑芝, 梁士栋, 马明辉, 刘华胜, 朱永刚. 信号交叉口实时排队长度估计[J]. 吉林大学学报(工学版), 2016, 46(1): 85-91.
[14] 姚向明, 赵鹏, 禹丹丹. 基于平均策略的城市轨道交通动态O-D矩阵估计[J]. 吉林大学学报(工学版), 2016, 46(1): 92-99.
[15] 刘华胜,赵淑芝,朱永刚,李晓玉. 基于有效路径的轨道交通接运线路设计模型[J]. 吉林大学学报(工学版), 2015, 45(2): 371-378.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] 李寿涛, 李元春. 在未知环境下基于递阶模糊行为的移动机器人控制算法[J]. 吉林大学学报(工学版), 2005, 35(04): 391 -397 .
[2] 刘庆民,王龙山,陈向伟,李国发. 滚珠螺母的机器视觉检测[J]. 吉林大学学报(工学版), 2006, 36(04): 534 -538 .
[3] 李红英;施伟光;甘树才 .

稀土六方Z型铁氧体Ba3-xLaxCo2Fe24O41的合成及电磁性能与吸波特性

[J]. 吉林大学学报(工学版), 2006, 36(06): 856 -0860 .
[4] 张全发,李明哲,孙刚,葛欣 . 板材多点成形时柔性压边与刚性压边方式的比较[J]. 吉林大学学报(工学版), 2007, 37(01): 25 -30 .
[5] 杨树凯,宋传学,安晓娟,蔡章林 . 用虚拟样机方法分析悬架衬套弹性对
整车转向特性的影响
[J]. 吉林大学学报(工学版), 2007, 37(05): 994 -0999 .
[6] 冯金巧;杨兆升;张林;董升 . 一种自适应指数平滑动态预测模型[J]. 吉林大学学报(工学版), 2007, 37(06): 1284 -1287 .
[7] 刘寒冰,焦玉玲,,梁春雨,秦卫军 . 无网格法中形函数对计算精度的影响[J]. 吉林大学学报(工学版), 2007, 37(03): 715 -0720 .
[8] 杨庆芳,陈林 . 交通控制子区动态划分方法[J]. 吉林大学学报(工学版), 2006, 36(增刊2): 139 -142 .
[9] 李月英,刘勇兵,陈华 . 凸轮材料的表面强化及其摩擦学特性
[J]. 吉林大学学报(工学版), 2007, 37(05): 1064 -1068 .
[10] 曲昭伟,陈红艳,李志慧,胡宏宇,魏巍 . 基于单模板的二维场景重建方法[J]. 吉林大学学报(工学版), 2007, 37(05): 1159 -1163 .