基于多尺度空间特征融合的高速公路车道级交通流量预测方法

席宽; 张存保; 李春; 鹿宇鑫; 高思羽

doi:10.3963/j.jssn.1674-4861.2025.05.012

基于多尺度空间特征融合的高速公路车道级交通流量预测方法

doi: 10.3963/j.jssn.1674-4861.2025.05.012

席宽^{1, 2,},
张存保^{1, 2, ,},
李春^{1, 2},
鹿宇鑫^{1, 2},
高思羽^{1, 2}

1.
武汉理工大学智能交通系统研究中心武汉 430063
2.
武汉理工大学交通信息与安全教育部工程研究中心武汉 430063

基金项目:

国家重点研发计划项目 2023YFB4301800

详细信息

作者简介:
席宽（2001—），硕士研究生. 研究方向：交通管理与控制、智能交通. E-mail：482661265@qq.com

通讯作者:
张存保（1976—），博士，研究员. 研究方向：交通信息工程及控制、交通安全、智能交通.E-mail: zhangcunbao@163.com

中图分类号: U491.1
计量
- 文章访问数: 7
- HTML全文浏览量: 3
- PDF下载量: 1
- 被引次数: 0
出版历程
- 收稿日期: 2025-05-21
- 网络出版日期: 2026-03-05

Freeway Lane-Level Traffic Flow Prediction Method Based on Multi-Scale Spatial Feature Fusion

XI Kuan^{1, 2
,},
ZHANG Cunbao^{1, 2
, ,},
LI Chun^{1, 2},
LU Yuxin^{1, 2},
GAO Siyu^{1, 2}

1.
Intelligent Transportation Systems Research Center, Wuhan University of Technology, Wuhan 430063, China
2.
Engineering Research Center of Transportation Information and Safety of Ministry of Education, Wuhan University of Technology, Wuhan 430063, China

摘要

摘要: 现有高速公路交通流量预测研究多聚焦于道路断面，未充分考虑同1个断面内不同车道、上下游断面间流量的时空相关特性，且多忽略了流量与速度间的内在联系。以车道断面为对象，研究了基于多尺度空间特征融合的高速公路车道级交通流量预测方法。通过量化计算方法消除上下游断面间的空间时滞影响，降低了上下游交通流量在时序维度上的非对齐性；在空间特征提取方面，将剔除空间时滞影响的流量与速度信息进行整合，利用3个尺度的2通道三维卷积模块与注意力机制动态捕捉流量的车道间局部交互特征、上下游断面的全局传播模式，以及流量与速度间的内在关联；在时序建模上，利用长短期记忆网络同步提取多尺度空间特征变量的全局时间依赖关系，并通过全连接层输出预测结果。采用美国PeMS断面实测数据进行实例验证，结果表明：在单步预测任务中，本文方法的平均绝对误差、均方根误差和平均绝对百分比误差较其他模型平均至少降低了6.61%、5.50%、8.46%；在多步预测中，各步长平均误差最高可降低14.09%、15.25%、29.16%，验证了本文方法在挖掘交通流量多尺度细粒化时空特征方面的有效性，并在预测精度上表现出显著优势。此外，消融实验结果进一步验证了注意力机制与多尺度空间信息协同整合在提升高速公路交通流量预测性能中的关键作用。
- 高速公路 /
- 交通流量预测 /
- 深度学习 /
- 注意力机制 /
- 多尺度空间特征
Abstract: Existing studies on freeway traffic flow prediction mainly focus on single cross-sections, without fully considering spatio-temporal correlations across lanes and along upstream downstream segments. The intrinsic relationship between flow and speed is also often neglected. This study proposes a lane-level freeway traffic flow prediction method based on multi-scale spatial feature fusion. The method quantifies and compensates for spatial time-lag effects between adjacent sections, reducing temporal misalignment of upstream and downstream flow sequences. For spatial feature extraction, flow and speed data with time-lag effects removed are integrated and processed through three-scale dual-channel 3D convolutional modules with attention mechanisms. These modules dynamically capture local inter-lane interactions, global propagation patterns between sections, and intrinsic flow-speed dependencies. For temporal modeling, a long short-term memory network is employed to extract global temporal dependencies among multi-scale spatial features, and a fully connected layer generates final predictions. Empirical validation using real PeMS freeway data demonstrates that, in one-step prediction tasks, the proposed method reduces the mean absolute error, root mean square error, and mean absolute percentage error by at least 6.61%, 5.50%, and 8.46% on average compared with the other models. In multi-step prediction, average errors across horizons decrease by up to 14.09%, 15.25%, and 29.16%, confirming the method's effectiveness in capturing fine-grained multi-scale spatio-temporal features and its significant advantage in prediction accuracy. Furthermore, ablation experiments verify that the attention mechanism and the collaborative integration of multi-scale spatio information play crucial roles in improving freeway traffic flow prediction performance.
- freeway /
- traffic flow prediction /
- deep learning /
- attention mechanism /
- multi-scale spatial feature

HTML全文

图 1 空间特征提取示意图

Figure 1. Diagram of spatial feature extraction

下载: 全尺寸图片幻灯片

图 2 多尺度空间特征融合示意图

Figure 2. Schematic diagram of multi-scale spatial feature fusion

下载: 全尺寸图片幻灯片

图 3 多尺度卷积长短期记忆网络模型

Figure 3. Multi-scale convolutional long short-term memory network model

下载: 全尺寸图片幻灯片

图 4 选取的监测站点位置示意图

Figure 4. Schematic diagram of selected monitoring station locations

下载: 全尺寸图片幻灯片

图 5 Pearson相关性系数

Figure 5. Pearson correlation coefficient

下载: 全尺寸图片幻灯片

图 6 MSCLSTM单步预测结果

Figure 6. One-step prediction results of MSCLSTM

下载: 全尺寸图片幻灯片

图 7 不同预测步长下模型性能对比

Figure 7. Performance comparison of the model at different prediction horizons

下载: 全尺寸图片幻灯片

表 1 网络结构超参数的搜索范围

Table 1. Search range of hyperparameters for network structure

超参数	参数搜索范围
时间步长	[12, 24, 36, 48, 60]
训练批大小	[32, 64, 128, 256]
LSTM 隐藏单元	[16, 32, 64, 128]

下载: 导出CSV

表 2 不同预测模型单步预测结果

Table 2. One-step prediction results of different prediction models

模型	车道断面L10			车道断面L11			车道断面L12
模型	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%
SVR	10.104	13.176	24.457	7.917	10.213	13.837	6.006	8.135	27.612
KNN	8.962	12.101	21.243	7.362	9.522	14.690	4.976	6.778	21.459
GRU	8.848	11.834	19.062	6.896	8.963	12.057	5.025	6.746	22.713
LSTM	8.464	11.366	18.828	6.793	8.931	11.938	4.979	6.743	23.036
T-GCN	8.213	10.987	16.893	6.711	8.819	11.132	4.875	6.652	21.779
MSCLSTM	7.825	10.582	15.776	6.481	8.583	10.454	4.802	6.587	21.172

下载: 导出CSV

表 3 原始模型及2个变体单步预测结果

Table 3. One-step prediction results of the original model and two variants

模型	MAE	RMSE	MAPE/%
消融模型1	6.775	8.805	10.652
消融模型2	6.803	8.842	11.179
MSCLSTM	6.581	8.642	10.016

下载: 导出CSV

参考文献(22)

[1]	姚俊峰, 何瑞, 史童童, 等. 基于机器学习的交通流预测方法综述[J]. 交通运输工程学报, 2023, 23(3): 44-67. YAO J F, HE R, SHI T T, et al. Review on machine learning-based traffic flow prediction methods[J]. Journal of Traffic and Transportation Engineering, 2023, 23(3): 44-67. (in Chinese)
[2]	KASHYAP A A, RAVIRAJ S, DEVARAKONDA A, et al. Traffic flow prediction models-a review of deep learning techniques[J]. Cogent Engineering, 2022, 9(1): 2010510. doi: 10.1080/23311916.2021.2010510
[3]	XIE P, LI T R, LIU J, et al. Urban flow prediction from spatiotemporal data using machine learning: a survey[J]. Information Fusion, 2020, 59: 1-12.
[4]	唐继强, 钟鑫伟, 刘健, 等. 基于时间序列季节分类模型的轨道交通客流短期预测[J]. 重庆交通大学学报(自然科学版), 2021, 40(7): 31-38, 60. TANG J Q, ZHONG X W, LIU J, et al. Short term forecast of rail transit passenger flow based on time series seasonal classification model[J]. Journal of Chongqing Jiaotong University (Natural Science), 2021, 40(7): 31-38, 60. (in Chinese)
[5]	王代君, 李明, 鹿守山. 基于Bayes-ARIMA的景区公路短时交通流量预测[J]. 公路, 2024, 69(4): 225-234. WANG D J, LI M, LU S S. Prediction of short-term traffic flow of scenic roads based on Bayes-ARIMA[J]. Highway, 2024, 69(4): 225-234. (in Chinese)
[6]	白伟华, 张传斌, 张塽旖, 等. 基于异常值识别卡尔曼滤波器的短期交通流预测[J]. 计算机应用研究, 2021, 38(3): 817-821. BAI W H, ZHANG C B, ZHANG S Y, et al. Outlier-identified Kalman filter for short-term traffic flow forecasting[J]. Application Research of Computers, 2021, 38(3): 817-821. (in Chinese)
[7]	LYU Y S, DUAN Y J, KANG W W, et al. Traffic flow prediction with big data: a deep learning approach[J]. IEEE Transactions on Intelligent Transportation Systems, 2015, 16(2): 865-873.
[8]	乔建刚, 李硕, 刘怡美, 等. 改进PSO-LSTM算法预测高速公路交通量[J]. 科学技术与工程, 2024, 24(15): 6466-6472. QIAO J G, LI S, LIU Y M, et al. Improved PSO-LSTM algorithm for forecasting expressway traffic volume[J]. Science Technology and Engineering, 2024, 24(15): 6466-6472. (in Chinese)
[9]	崔文岳, 谷远利, 赵胜利, 等. 基于有向图卷积与门控循环单元的短时交通流预测方法[J]. 交通信息与安全, 2023, 41 (2): 121-128. doi: 10.3963/j.jssn.1674-4861.2023.02.013 CUI W Y, GU Y L, ZHAO S L, et al. A method of predicting short-term traffic flows based on a DGC-GRU model[J]. Journal of Transport Information and Safety, 2023, 41(2): 121-128. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2023.02.013
[10]	LIAN Q Y, SUN W, DONG W. Hierarchical spatial-temporal neural network with attention mechanism for traffic flow forecasting[J]. Applied Sciences (Switzerland), 2023, 13 (17): 9729. doi: 10.3390/app13179729
[11]	李劲业, 李永强. 融合知识图谱的时空多图卷积交通流量预测[J]. 浙江大学学报(工学版), 2024, 58(7): 1366-1376. LI J Y, LI Y Q. Spatial-temporal multi-graph convolution for traffic flow prediction by integrating knowledge graphs[J]. Journal of Zhejiang University(Engineering Science), 2024, 58(7): 1366-1376. (in Chinese)
[12]	ZHENG Y, LI W Q, ZHENG W, et al. Lane-level heterogeneous traffic flow prediction: a spatiotemporal attention-based encoder-decoder model[J]. IEEE Intelligent Transportation Systems Magazine, 2023, 15(3): 51-67.
[13]	MA Y X, ZHANG Z J, IHLER A. Multi-lane short-term traffic forecasting with convolutional LSTM network[J]. IEEE Access, 2020, 8: 34629-34643.
[14]	LI B, YANG Q, CHEN J J, et al. A dynamic spatio-temporal deep learning model for lane-level traffic prediction[J]. Journal of Advanced Transportation, 2023, 2023(1): 1-14.
[15]	侯越, 崔菡珂, 邓志远. 横向相关性及参数影响下的车道级交通预测[J]. 公路交通科技, 2022, 39(5): 122-130. HOU Y, CUI H K, DENG Z Y. Lane level traffic prediction under influence of lateral correlation and parameters[J]. Journal of Highway and Transportation Research and Development, 2022, 39(5): 122-130. (in Chinese)
[16]	江辉, 张阳, 杨书敏, 等. 基于动态时空卷积网络的车道级交通流预测[J]. 武汉理工大学学报(交通科学与工程版), 2024, 48(2): 242-247. JIANG H, ZHANG Y, YANG S M, et al. Lane-level traffic flow prediction based on dynamic spatio-temporal convolutional networks[J]. Journal of Wuhan University of Technology(Transportation Science & Engineering), 2024, 48(2): 242-247. (in Chinese)
[17]	雷旭, 李立, 李光泽, 等. 多车道交通流理论与应用研究综述[J]. 长安大学学报(自然科学版), 2020, 40(4): 78-90. LEI X, LI l, LI G Z, et al. Review of multilane traffic flow theory and application[J]. Journal of Chang'an University (Natural Science Edition), 2020, 40(4): 78-90. (in Chinese)
[18]	郭嘉宸, 杨宇燊, 王研, 等. 精细化短时交通流预测模型及迁移部署方案[J]. 计算机应用, 2022, 42(6): 1748-1755. GUO J C, YANG Y S, WANG Y, et al. Refined short-term traffic flow prediction model and migration deployment scheme[J]. Journal of Computer Applications, 2022, 42(6): 1748-1755. (in Chinese)
[19]	LANA I, DEL SER J, VELEZ M, et al. Road traffic forecasting: recent advances and new challenges[J]. IEEE Intelligent Transportation Systems Magazine, 2018, 10(2): 93-109.
[20]	李春, 张存保, 陈峰, 等. 考虑车道间差异和上下游断面关联的快速路交通流量预测方法[J]. 交通信息与安全, 2024, 42(4): 102-109. doi: 10.3963/j.jssn.1674-4861.2024.04.011 LI C, ZHANG C B, CHEN F, et al. An expressway traffic flow prediction method considering inter-lane differences and upstream and downstream cross-section correlations[J]. Journal of Transport Information and Safety, 2024, 42(4): 102-109. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2024.04.011
[21]	LU W Q, RUI Y K, RAN B. Lane-level traffic speed forecasting: a novel mixed deep learning model[J]. IEEE transactions on intelligent transportation systems, 2020, 23(4): 3601-3612.
[22]	ZHAO L, SONG Y J, ZHANG C, et al. T-GCN: a temporal graph convolutional network for traffic prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 21 (9): 3848-3858.