基于边缘智能的高速公路交通流预测方法  

Highway Traffic Flow Prediction Method Based on Edge Intelligence

作　　者：王小博 张轩 高张浩 吴鑫 董云卫[2] WANG Xiao-bo;ZHANG Xuan;GAO Zhang-hao;WU Xin;DONG Yun-wei(Shaanxi Highway Electronic Engineering Co.,Ltd.,Xi’an 710010,China;School of Software,Northwestern Polytechnical University,Xi’an 710129,China;School of Computer Science,Northwestern Polytechnical University,Xi’an 710129,China)

机构地区：[1]陕西高速电子工程有限公司,陕西西安710010 [2]西北工业大学软件学院,陕西西安710129 [3]西北工业大学计算机学院,陕西西安710129

出　　处：《计算机技术与发展》2025年第4期193-201,共9页Computer Technology and Development

基　　金：陕西省交通运输厅2023年度交通科研项目(23-107K)。

摘　　要：高速公路作为国民经济发展的重要动脉之一,其网络交通流量的规模不仅反映了区域经济的发展水平,依据路网流量对车流进行引导,确保高速公路通畅和安全也体现了交通保障与管理的能力。由于高速公路运行会受到重大社会活动、极端天气、自然灾害、交通事故及其他突发事件影响,高速公路交通流量预测的准确性和及时性是高速公路管理和运维中的一个技术难题。为此,提出了一种基于边缘智能的高速公路交通流预测方法,设计并构建了结构化神经网络预测模型,能够有效地捕捉和表达复杂的不同高速公路间车辆通行的时空相关性,用于实现省域高速公路网络交通流量的预测。基于陕西省关中地区多条互联高速公路的实际运营数据,构建了模型学习数据集和实验测试集,并在多种运行场景下进行预测实验。实验结果表明,提出的“多层时空卷积”预测网络模型在高速公路交通流量预测的实时性和准确性方面具有显著优势。Highways,as vital arteries driving national economic development,reflect both the level of regional economic growth and the effectiveness of traffic management and security through the scale of network traffic flow.Due to the influence of major social events,extreme weather,natural disasters,traffic accidents,and other unexpected incidents,achieving accurate and timely highway traffic flow prediction has become a technical challenge in highway management and operations.To address this issue,a highway traffic flow prediction method based on edge intelligence is proposed.A structured neural network prediction model is designed and developed to effectively capture and represent the complex spatiotemporal correlations of vehicle movements across different highways,enabling traffic flow predictions for provincial highway networks.Based on real-world operational data from multiple interconnected highways in the Guanzhong Region of Shaanxi Province,a model training dataset and experimental testing dataset were constructed,and prediction experiments were conducted under various operating scenarios.Experimental results demonstrate that the proposed"multi-layer spatiotemporal convolution"prediction network model significantly outperforms in terms of real-time performance and accuracy for highway traffic flow prediction.

关 键 词：智能交通 “多层时空卷积”预测网络模型 边缘智能 高速公路流量预测 时空相关性 云边协同计算 交通流量大数据 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]