基于事件特征机器学习的高速公路收费站管控决策系统  

Expressway Toll Station Control Decision-making System based on Event Characteristics Machine Learning

作　　者：郭覃 陆启荣 刘志远[3] 杨赞 GUO Tan;LU Qi-rong;LIU Zhi-yuan;YANG Zan(Zhejiang Expressway Information Engineering and Technology Co.Ltd.,Hangzhou 310014,China;College of Civil Engineering and Architecture,Zhejiang University,Hangzhou 310058,China;School of Transportation,Southeast University,Nanjing 211189,China)

机构地区：[1]浙江高信技术股份有限公司,杭州市310014 [2]浙江大学建筑工程学院,杭州市310058 [3]东南大学交通学院,南京市211189

出　　处：《公路》2025年第2期312-318,共7页Highway

基　　金：国家自然科学基金重点项目,项目编号52131203;浙江省科技计划项目,项目编号2022C01129。

摘　　要：收费站流量管控可以平衡主路通行密度,缓解交通拥堵,降低事故风险,是高速公路运营管理的高频且有效手段。为解决高速公路收费站管控存在的人力投入大、沟通成本高、决策耗时长等问题,设计了一套基于事件特征机器学习的高速公路收费站管控决策系统。系统基于历史高速公路事件和实际管控信息,利用机器学习算法,建立高速公路收费站管控行为预测模型,实现收费站管控策略的快速生成和自动决策。在G92沪杭甬高速应用结果表明,大流量、大雨、事故3种典型事件的管控预测平均准确率分别达91.7%、85.8%及88.3%。系统上线应用后,管控预测总体准确率达80%,可在事件触发阶段超前预知并推荐科学合理的管控策略,辅助高速公路运管人员决策,具备较好的实践推广价值。Toll station flow control serves as a high-frequency and effective means for balancing main road traffic density,alleviating congestion,and reducing accident risks in expressway operation and management.However,existing approaches to toll station control suffer from challenges such as high labor input,communication costs,and lengthy decision-making processes.To address these issues,a framework for toll station control decision-making on expressways based on event characteristics and machine learning is proposed.This framework utilizes historical expressway event and actual control information,leveraging machine learning classification algorithms to predict toll station control behaviors and facilitate rapid and automated decision-making.Through application in the G92 Shanghai-Hangzhou-Ningbo Expressway,the average accuracy rates for control predictions in high traffic volume,heavy rain,and accident scenarios are 91.7%,85.8%,and 88.3%,respectively.Upon deployment,the overall accuracy rate for control predictions reaches 80%,enabling the early anticipation and recommendation of scientifically sound control strategies during event triggers,showing significant practical value for implementation and replication.

关 键 词：智慧高速 智能交通 决策系统 机器学习 收费站管控 

分 类 号：U491.4[交通运输工程—交通运输规划与管理]