首页 | 本学科首页   官方微博 | 高级检索  
     

基于Agent模型的机场网络延误预测
引用本文:王春政,胡明华,杨磊,赵征,单晶. 基于Agent模型的机场网络延误预测[J]. 航空学报, 2021, 42(7): 324604-324604. DOI: 10.7527/S1000-6893.2020.24604
作者姓名:王春政  胡明华  杨磊  赵征  单晶
作者单位:南京航空航天大学民航学院,南京 211106;国家空管飞行流量管理技术重点实验室,南京 211106;南京航空航天大学民航学院,南京 211106
基金项目:国家自然科学基金(61903187);江苏省自然科学基金(BK20190414)
摘    要:准确可靠的机场网络航班延误预测是科学认知空中交通运行态势,动态精准实施国家空域系统容流协同调配策略的重要依据。提出了基于Agent的机场网络延误模型,表征机场网络系统中各元素及子系统间的交互作用下的延误特征涌现。针对机场节点动态容量、预计起飞时间、最小飞行与周转时间等Agent模型中的关键参数,适应性选用了贝叶斯估计、模糊k近邻等数据挖掘方法建立参数模型,并采用2015—2017年全美历史航班和气象数据进行训练学习。为综合评价模型性能及泛化能力,选取全美2018年3个不同延误程度的典型日进行测试。实验结果表明,在全美34个核心机场组成的网络中,各节点在4小时预测区间内延误最大误差不过27.9 min,其中约80%的节点误差小于5 min,验证了所提延误预测模型在时空范围内的准确性和稳健性特征。另外,通过与其他模型对比,展示了本模型优良的延误预测性能。

关 键 词:空中交通流量管理  Agent模型  机场网络  延误预测  数据挖掘
收稿时间:2020-08-05
修稿时间:2020-09-02

Airport network delay prediction based on Agent model
WANG Chunzheng,HU Minghua,YANG Lei,ZHAO Zheng,SHAN Jing. Airport network delay prediction based on Agent model[J]. Acta Aeronautica et Astronautica Sinica, 2021, 42(7): 324604-324604. DOI: 10.7527/S1000-6893.2020.24604
Authors:WANG Chunzheng  HU Minghua  YANG Lei  ZHAO Zheng  SHAN Jing
Affiliation:1. College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China;2. National Key Laboratory of Air Traffic Flow Management, Nanjing 211106, China
Abstract:Accurate and reliable airport network flight delay prediction is an important basis for scientifically understanding the air traffic situation and dynamically and accurately implementing the national airspace system capacity coordination strategy. This paper proposes an Agent airport network delay model, characterizing the emergence of delay characteristics under the interaction between elements and subsystems in the airport network system. Several data mining algorithms were selected to estimate key parameters in the Agent model such as the dynamic capacity of airport nodes, estimated departure time, minimum flight, and turnaround time. The 2015-2017 American national historical flight and weather data for training and learning were used to train these parameter models. To comprehensively evaluate the model performance and generalization ability, three typical days with different delays in the United States in 2018 were selected for testing. The experimental results show that in the network composed of 34 core airports in the United States, the maximum delay error of each node in the 4-hour prediction interval is only 27.9 minutes, and about 80% of the node errors are less than 5 minutes, verifying the accuracy and robustness of the proposed delay prediction model in the space-time range. Comparison with other models further demonstrates the excellent delay prediction performance of our model.
Keywords:air traffic flow management  Agent model  airport network  delay prediction  data mining  
本文献已被 万方数据 等数据库收录!
点击此处可从《航空学报》浏览原始摘要信息
点击此处可从《航空学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号