变时间尺度城轨客流的本征模量分解及组合深度学习预测

朱广宇; 孙歆霓; 杨荣正; 刘康琳; 魏运; 吴波

doi:10.11999/JEIT221300

变时间尺度城轨客流的本征模量分解及组合深度学习预测

doi: 10.11999/JEIT221300

1.
北京交通大学北京市城市交通信息智能感知与服务工程技术研究中心北京 100044
2.
北京市地铁运营有限公司北京 100014
3.
太原中铁轨道交通建设运营有限公司太原 030006

基金项目: 基本科研业务费(2022JBZX024)，国家自然科学基金(62272036, 62173167, 62132003)

详细信息

作者简介:
朱广宇：男，教授，博士生导师，研究方向为交通运输智能自动化

孙歆霓：女，硕士生，研究方向为城市轨道交通客流分析与预测

杨荣正：男，硕士生，研究方向为城市轨道交通知识工程与知识图谱

刘康琳：女，讲师，研究方向为应急物流与优化调度

魏运：男，教授级高级工程师，研究方向为智能交通系统和模式识别

吴波：男，高级工程师，研究方向为城市轨道交通安全管理及设施设备故障预测等

通讯作者:
朱广宇　gyzhu@bjtu.edu.cn

中图分类号: TP391.1
计量
- 文章访问数: 556
- HTML全文浏览量: 343
- PDF下载量: 104
- 被引次数: 0
出版历程
- 收稿日期: 2022-10-14
- 修回日期: 2023-02-22
- 网络出版日期: 2023-03-13
- 刊出日期: 2023-12-26

Intrinsic Mode Decomposition and Combined Deep Learning Prediction of Urban Rail Transit Passenger Flow at Variable Time Scales

1.
Beijing Research Center of Urban Traffic Information Sensing and Service Technologies, Beijing Jiaotong University, Beijing 100044, China
2.
Beijing Mass Transit Railway Operation Corporation Limited, Beijing 100014, China
3.
Taiyuan China Railway Rail Transit Construction and Operation Co., Ltd, Taiyuan 030006, China

Funds: The Fundamental Research Funds for the Central Universities(2022JBZX024), The National Natural Science Foundation of China (62272036, 62173167, 62132003)

摘要

摘要: 城市轨道交通的不同运营状态，通常对应着客流时间序列中不同的本征模态分量(IMF)及时间尺度特征。基于自适应噪声的完全总体经验模态分解(CEEMDAN)算法和双向长短期记忆(BiLSTM)网络，该文构建了地铁短时客流时间序列的组合深度学习预测模型。具体包括：基于CEEMDAN算法实现了客流时间序列的模态分解。分别使用样本熵和层次聚类对IMF分量进行复杂性和相似度分析，并在此基础上完成IMF分量的分类合并与重构；使用Optuna框架中的树形Parzen优化器(TPE)对模型的超参数进行优化，构建CEEMDAN-TPE-BiLSTM组合预测模型。采用实际数据对该文模型进行验证，结果表明，对于特定特征的客流时间序列数据，该文模型的精确性、有效性指标均达到最优。
- 城市轨道交通 /
- 短时客流时间序列 /
- 自适应噪声的完全总体经验模态分解 /
- 双向长短期记忆 /
- 组合预测
Abstract: Different operational states of urban rail transit usually correspond to different Intrinsic Mode Functions (IMFs) and time-scale characteristics in passenger flow time series. A combined deep learning prediction model for short-term passenger flow time series of subway is proposed based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Bidirectional Long Short Term Memory network (BiLSTM), including: mode decomposition of passenger flow time series based on the CEEMDAN algorithm. The sample entropy and hierarchical clustering are used respectively to analyze the complexity and similarity of IMFs. The IMFs are then classified, merged and reconstructed on this basis. The hyper-parameters of the model are optimized using the Tree-structured Parzen Estimator (TPE) in the Optuna framework, and the combined prediction model CEEMDAN-TPE-BiLSTM is established. Actual data are used to validate the model. The results show that the accuracy and validity indicators of the model all reach the optimum for passenger flow time series data with specific characteristics.

HTML全文

图 1 BiLSTM结构图

下载: 全尺寸图片幻灯片

图 2 建模及实验流程图

下载: 全尺寸图片幻灯片

图 3 进站客流序列CEEMDAN分解结果

下载: 全尺寸图片幻灯片

图 4 IMF分量聚类结果

下载: 全尺寸图片幻灯片

图 5 IMF分量重构结果

下载: 全尺寸图片幻灯片

图 6 模型预测效果对比

下载: 全尺寸图片幻灯片

图 7 高峰时段划分图像

下载: 全尺寸图片幻灯片

表 1 IMF分量样本熵计算结果

	IMF1	IMF2	IMF3	IMF4	IMF5	IMF6	IMF7	IMF8	IMF9	IMF10	IMF11
工作日	0.954	0.894	0.789	0.483	0.524	0.300	0.296	0.076	0.038	0.005	0.001
非工作日	0.812	0.983	1.178	0.544	0.514	0.282	0.240	0.102	0.002	—	—

下载: 导出CSV

表 2 待优化超参数搜索空间范围

	神经元个数	批大小	迭代次数	学习率
搜索空间范围	(16, 64)	(16, 64)	(30, 70)	(0.0001, 0.01)
步长	2	4	5	0.0001

下载: 导出CSV

表 3 优化后各模型的超参数

	模型	IMF	隐藏层神经元个数	批大小	迭代次数	学习率
工作日	BiLSTM模型	—	26	40	65	0.0072
	CEEMDAN-BiLSTM 模型	IMF1	60	24	65	0.0084
		IMF2	60	36	65	0.0034
		IMF3	42	16	60	0.0090
		IMF4	60	28	65	0.0066
		IMF5	50	20	55	0.0041
		IMF6	62	20	60	0.0054
非工作日	BiLSTM模型	—	38	52	65	0.0078
	CEEMDAN-BiLSTM 模型	IMF1	62	20	70	0.0080
		IMF2	36	16	60	0.0023
		IMF3	42	24	50	0.0062
		IMF4	54	24	70	0.0082
		IMF5	58	16	40	0.0092
		IMF6	40	16	50	0.0065

下载: 导出CSV

表 4 模型预测性能评价指标

模型	工作日				非工作日
模型	RMSE	MAE	MAPE(%)	R²	RMSE	MAE	MAPE(%)	R²
BiLSTM	137.467	92.486	21.094	0.914	147.001	94.828	22.838	0.956
TPE-BiLSTM	121.868	77.680	19.004	0.933	140.193	90.248	19.352	0.960
CEEMDAN-BiLSTM	70.945	51.258	14.627	0.977	104.186	75.919	22.573	0.978
CEEMDAN-TPE-BiLSTM	57.730	36.981	12.071	0.985	80.575	52.587	16.300	0.987

下载: 导出CSV

表 5 高峰状态下模型预测性能评价指标

模型	工作日				非工作日
模型	RMSE	MAE	MAPE(%)	R²	RMSE	MAE	MAPE(%)	R²
BiLSTM	194.380	170.480	11.823	–1.601	229.810	175.016	8.749	–2.047
TPE-BiLSTM	138.215	106.626	7.296	–0.315	222.178	167.007	8.411	–1.848
CEEMDAN-BiLSTM	115.999	107.435	7.561	0.074	171.194	154.437	7.916	–0.691
CEEMDAN-TPE-BiLSTM	73.466	57.771	4.019	0.628	117.462	108.406	5.539	0.204

下载: 导出CSV

参考文献(20)

[1]	CHEN Enhui, YE Zhirui, WANG Chao, et al. Subway passenger flow prediction for special events using smart card data[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(3): 1109–1120. doi: 10.1109/TITS.2019.2902405
[2]	RYU U, WANG Jian, KIM T, et al. Construction of traffic state vector using mutual information for short-term traffic flow prediction[J]. Transportation Research Part C:Emerging Technologies, 2018, 96: 55–71. doi: 10.1016/j.trc.2018.09.015
[3]	LIANG Shidong, MA Minghui, HE Shengxue, et al. Short-term passenger flow prediction in urban public transport: Kalman filtering combined k-Nearest Neighbor approach[J]. IEEE Access, 2019, 7: 120937–120949. doi: 10.1109/ACCESS.2019.2937114
[4]	WANG Xuemei, ZHANG Ning, ZHANG Yunlong, et al. Forecasting of short-term metro ridership with Support Vector Machine online model[J]. Journal of Advanced Transportation, 2018, 2018: 3189238. doi: 10.1155/2018/3189238
[5]	陈忠辉, 凌献尧, 冯心欣, 等. 基于模糊C均值聚类和随机森林的短时交通状态预测方法[J]. 电子与信息学报, 2018, 40(8): 1879–1886. doi: 10.11999/JEIT171090 CHEN Zhonghui, LING Xianyao, FENG Xinxin, et al. Short-term traffic state prediction approach based on FCM and random forest[J]. Journal of Electronics &Information Technology, 2018, 40(8): 1879–1886. doi: 10.11999/JEIT171090
[6]	SUN Bo, SUN Tuo, and JIAO Pengpeng. Spatio-temporal segmented traffic flow prediction with ANPRS data based on improved XGBoost[J]. Journal of Advanced Transportation, 2021, 2021: 5559562. doi: 10.1155/2021/5559562
[7]	YANG Dan, CHEN Kairun, YANG Mengning, et al. Urban rail transit passenger flow forecast based on LSTM with enhanced long-term features[J]. IET Intelligent Transport Systems, 2019, 13(10): 1475–1482. doi: 10.1049/iet-its.2018.5511
[8]	代亮, 梅洋, 钱超, 等. 基于生成对抗网络的大规模路网交通流预测算法[J]. 控制与决策, 2021, 36(12): 2937–2945. doi: 10.13195/j.kzyjc.2020.0333 DAI Liang, MEI Yang, QIAN Chao, et al. Traffic flow forecasting algorithm for large-scale road network based on GAN[J]. Control and Decision, 2021, 36(12): 2937–2945. doi: 10.13195/j.kzyjc.2020.0333
[9]	秦超, 高晓光, 万开方. 深度卷积记忆网络时空数据模型[J]. 自动化学报, 2020, 46(3): 451–462. doi: 10.16383/j.aas.c180788 QIN Chao, GAO Xiaoguang, and WAN Kaifang. Deep spatio-temporal convolutional long-short memory network[J]. Acta Automatica Sinica, 2020, 46(3): 451–462. doi: 10.16383/j.aas.c180788
[10]	曾诚, 吴佳媛, 罗霞. 城市轨道交通短时客流预测文献综述[J]. 铁道运输与经济, 2021, 43(8): 105–111,125. doi: 10.16668/j.cnki.issn.1003-1421.2021.08.17 ZENG Cheng, WU Jiayuan, and LUO Xia. Literature review of short-term passenger flow forecast for urban rail transit[J]. Railway Transport and Economy, 2021, 43(8): 105–111,125. doi: 10.16668/j.cnki.issn.1003-1421.2021.08.17
[11]	殷礼胜, 唐圣期, 李胜, 等. 基于整合移动平均自回归和遗传粒子群优化小波神经网络组合模型的交通流预测[J]. 电子与信息学报, 2019, 41(9): 2273–2279. doi: 10.11999/JEIT181073 YIN Lisheng, TANG Shengqi, LI Sheng, et al. Traffic flow prediction based on hybrid model of auto-regressive integrated moving average and genetic particle swarm optimization wavelet neural network[J]. Journal of Electronics &Information Technology, 2019, 41(9): 2273–2279. doi: 10.11999/JEIT181073
[12]	ZHANG Lizong, ALHARBE N R, LUO Guangchun, et al. A hybrid forecasting framework based on support vector regression with a modified genetic algorithm and a random forest for traffic flow prediction[J]. Tsinghua Science and Technology, 2018, 23(4): 479–492. doi: 10.26599/tst.2018.9010045
[13]	寇唯. 基于数据稳定化处理的城市轨道交通短时客流预测研究[D]. [硕士论文], 北京交通大学, 2020. KOU Wei. Research on short-term passenger flow prediction of urban rail transit based on data stability processing[D]. [Master dissertation], Beijing Jiaotong University, 2020.
[14]	ZHANG Shuaichao, ZHOU Lingxiao, CHEN Xiqun, et al. Network-wide traffic speed forecasting: 3D convolutional neural network with ensemble empirical mode decomposition[J]. Computer-Aided Civil and Infrastructure Engineering, 2020, 35(10): 1132–1147. doi: 10.1111/mice.12575
[15]	WANG Jiaxuan, WANG Rui, and ZENG Xin. Short-term passenger flow forecasting using CEEMDAN meshed CNN-LSTM-attention model under wireless sensor network[J]. IET Communications, 2022, 16(10): 1253–1263. doi: 10.1049/cmu2.12350
[16]	沈琼. 城市轨道交通进站客流短时预测方法研究[D]. [硕士论文], 北京交通大学, 2019. SHEN Qiong. Study on short-term forecasting method for boarding passenger flows in urban rail transit[D]. [Master dissertation], Beijing Jiaotong University, 2019.
[17]	AKIBA T, SANO S, YANASE T, et al. Optuna: A next-generation hyperparameter optimization framework[C]. The 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, Anchorage, USA, 2019: 2623–2631.
[18]	LACERDA P, BARROS B, ALBUQUERQUE C, et al. Hyperparameter optimization for COVID-19 pneumonia diagnosis based on chest CT[J]. Sensors, 2021, 21(6): 2174. doi: 10.3390/S21062174
[19]	黄宇, 高珊, 李其贤, 等. SCR脱硝系统的分数阶 ${\text{P}}{{\text{I}}^\lambda }{{\text{D}}^\mu }$ 参数优化控制[J]. 动力工程学报, 2022, 42(2): 122–128. doi: 10.19805/j.cnki.jcspe.2022.02.004 HUANG Yu, GAO Shan, LI Qixian, et al. Optimal control of fractional ${\text{P}}{{\text{I}}^\lambda }{{\text{D}}^\mu }$ parameters of SCR denitration system[J]. Journal of Chinese Society of Power Engineering, 2022, 42(2): 122–128. doi: 10.19805/j.cnki.jcspe.2022.02.004
[20]	石庄彬, 张宁, 邵星杰. 城市轨道交通客流高峰持续时间预测方法[J]. 城市轨道交通研究, 2016, 19(7): 35–39. doi: 10.16037/j.1007-869x.2016.07.008 SHI Zhuangbin, ZHANG Ning, and SHAO Xingjie. Prediction of passenger flow duration in urban rail transit peak hours[J]. Urban Mass Transit, 2016, 19(7): 35–39. doi: 10.16037/j.1007-869x.2016.07.008