Prediction and Evaluation Methods for Large-Scale Flight Delay Propagation Based on Spatiotemporal Network
-
摘要: 针对航班运行在时空维度上的高度耦合性所引发的大面积延误在多机场间传播问题，采用动态网络分析方法，深入探究空中交通延误的传播规律。为精准捕捉延误传播动态，构建了1个以机场为节点、以离港航班为边、5 min为时间分辨率的时空网络。在航班延误时空图的构建过程中，将边的权重从直接采用延误时间的统计平均值或通过简单的经验规则估算，改进为基于深度学习预测得到的权重。针对航班延误预测任务，研究利用多任务学习的NR-DenseNet模型，同时预测航班延误时间(回归)与是否延误(分类)，提升了权重的准确性和实时性。通过对比不同网络深度的性能，实验表明：16层NR-DenseNet在双任务中表现最优，其回归预测的均方误差(mean squared error，MSE)和平均绝对误差(mean absolute error，MAE)分别达到58.30和3.28，分类准确率提升至94.8%。在指标度量方面，研究发现单一指标难以全面评估空中交通延误传播的复杂性，因此构建了3个评估指标：强度、传播率和速度，以定量分析延误传播的多维度特征。以华东空管局提供的国内数据为研究对象，结果表明：本文方法能有效阐明空中交通在计划航班时刻表上大面积延误传播的时空细节。Abstract: The highly coupled nature of flight operations in spatial and temporal dimensions leads to the widespread propagation of large-scale delays across multiple airports. This issue is addressed through using dynamic network analysis to explore the patterns of air traffic delay propagation. To accurately capture the dynamics of delay spread, a spatiotemporal network is constructed where airports are nodes, departing flights are edges, and a temporal resolu-tion is 5 min. In constructing the flight delay spatiotemporal graph, edge weights are improved. Initially estimated using statistical averages of delay times or simple empirical rules, the weights are predicted via deep learning. For the flight delay prediction task, a multi-task NR-DenseNet model is employed to simultaneously predict flight delay duration (regression) and delay occurrence (classification), enhancing the accuracy and timeliness of the weights. By comparing performances across different network depths, experiments demonstrated that a 16-layer NR-DenseNet achieved optimal performance in both tasks. The regression prediction yielding a mean squared error (MSE) of 58.30 and a mean absolute error (MAE) of 3.28, while classification accuracy reached 94.8%. Regarding metric evaluation, single metrics are found to be insufficient for fully assessing the complexity of air traffic delay propagation. Therefore, three evaluation metrics, intensity, propagation rate, and velocity, are established to quantita-tively analyze the multidimensional characteristics of delay spread. Using domestic data from the East China Air Traffic Management Bureau as the study, the results indicated that the proposed method effectively reveals the spa-tiotemporal details of large-scale delay propagation within the scheduled flight timetable.
-
图 7 24 h内延误航班的分类准确率
Figure 7. Classification accuracy of delayed flights within 24 hours
表 1 42个机场基本信息
Table 1. Basic information for 42 airports
机场 城市 航班量 ZBAA 北京市 212 553 ZGGG 广州市 167 676 ZGSZ 深圳市 144 139 ZUUU 成都市 133 813 ZLXY 咸阳市 129 282 ZSSS 上海市 124 940 ZUCK 重庆市 123 735 ZSPD 上海市 120 399 ZSHC 杭州市 120 004 ZPPP 昆明市 116 685 ZSNJ 南京市 97 186 ZHCC 郑州市 88 120 ZHHH 武汉市 79 714 ZSAM 厦门市 79 060 ZGHA 长沙市 78 343 ZSQD 青岛市 77 458 ZBTJ 天津市 68 747 ZJHK 海口市 68 354 ZWWW 乌鲁木齐 66 971 ZYHB 哈尔滨 58 068 ZSJN 济南市 57 917 ZYTL 大连市 56 714 ZYTX 沈阳市 56 203 ZJSY 三亚市 54 555 ZGNN 南宁市 51 595 ZBYN 太原市 50 070 ZSCN 南昌市 48 435 ZLLL 兰州市 46 182 ZSFZ 福州市 45 531 ZSOF 合肥市 41 315 ZYCC 长春市 40 677 ZGSD 珠海市 40 322 ZSWZ 温州市 39 408 ZSNB 宁波市 36 984 ZBSJ 石家庄 35 157 ZLIC 银川市 34 747 ZGKL 桂林市 34 184 ZLXN 海东市 27 392 ZGOW 揭阳市 23 679 ZGZJ 湛江市 11 663 ZWAK 阿克苏 5 692 ZBLA 呼伦贝尔 5 317 
下载: 导出CSV
表 2 特征属性
Table 2. Feature attributes
特征属性 特征含义 CRSDepTime 航班计划起飞的时间 CRSDepTime 航班计划降落的时间 CRSArrTime 航班实际起飞的时间 REALDepTime 航班实际降落的时间 REALArrTime 航班计划起飞的机场 PlanOriginAirport 航班实际起飞的机场 OriginAirport 航班计划降落的机场 PlanDestAirport 航班实际降落的机场 DestAirport 航班号 FlightNum 计划机型 PlanAircraft 巡航速度 CruSpeed 巡航高度 CruAltitude 任务类型 TaskType 保障等级 GuaranteeLevel 预计航路 EstimatedRoute 气象数据记录时间 Record_time 温度 TP 露点温度 DT 道面温度 PRE 风速2 min最小值 WS_2_MIN 风速2 min平均值 WS_2_AVG 风速2 min最大值 WS_2_MAX 风速10 min最小值 WS_10_MIN 风速10 min平均值 WS_10_AVG 风速10 min最大值 WS_10_MAX 相对湿度 RD 场压 FP 修正海压 CSP 降雨量 RAIN 能见度1 min平均值 VIS_1_AVG 能见度10 min平均值 VIS_10_AVG 跑道视程1 min最小值 RVR_1_MIN 跑道视程1 min平均值 RVR_1_AVG 跑道视程1 min最大值 RVR_1_MAX 跑道视程10 min最小值 RVR_10_MIN 跑道视程10 min平均值 RVR_10_AVG RVR_10_MAX 跑道视程10 min最大值 MOR_1_AVG 气象光学视程1 min平均值 MOR_10_MIN 气象光学视程10 min最小值 MOR_10_AVG 气象光学视程10 min平均值 MOR_10_MAX 气象光学视程10 min最大值 BGB_1_AVG 背景亮度1 min平均值
下载: 导出CSV
表 3 超参数配置
Table 3. Hyperparameter configuration
主要参数 参数值 损失函数 均方误差，交叉熵 学习率 0.001 优化器 SGD 最大迭代次数 400 损失权重分配 $\left[\alpha_{M S E}, \alpha_{A C C}\right]$ 批处理量 1 024 隐藏层层数 10/13/16/19
下载: 导出CSV
表 4 不同网络层数的预测结果
Table 4. Prediction results for different numbers of network layers
任务 评价指标 NR-DenseNet预测模型 L=10 L=13 L=16 L=19 MSE 348.90 128.49 58.30 81.20 回归 MAE 11.16 5.99 3.28 3.79 R2 0.902 3 0.964 0 0.978 7 0.967 3 ACC 0.818 1 0.903 6 0.948 0 0.933 7 分类 P 0.819 8 0.904 2 0.947 9 0.933 9 F1 0.818 1 0.903 6 0.948 0 0.933 8
下载: 导出CSV
表 6 不同模型预测结果对比
Table 6. Comparison of prediction results of different models
预测模型 评价指标 MSE MAE R2 ACC P F1 LightGBM 2 062.30 25.61 0.422 4 0.754 6 0.754 4 0.766 2 XGBoost 1 939.45 25.44 0.456 8 0.760 0 0.765 7 0.768 6 随机森林 2 353.60 27.45 0.340 8 0.743 8 0.741 4 0.757 5 NR-DenseNet(16) 58.3 3.28 0.978 7 0.948 0 0.947 9 0.948 0
下载: 导出CSV
-
[1] 中国民用航空局. 民航发展统计公报[R]. 北京: 中国民用航空局, 2023.Civil Aviation Administration of China. Bulletin on civil avia-tion development statistics[R]. Beijing: Civil Aviation Admin-istration of China, 2023. (in Chinese) [2] 李鹏. 航班延误分析[D]. 南京: 南京航空航天大学, 2016.LI P. Research of delay propagation mechanism and flight de-lays analysis[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2016. (in Chinese) [3] 中国民用航空局. 航班正常管理规例[S]. 北京: 中国民用航空局, 2016.Civil Aviation Administration of China. Flight regularity man-agement regulations[S]. Beijing: Civil Aviation Administra-tion of China, 2016. (in Chinese) [4] GUI G, LIU F, SUN J, et al. Flight delay prediction based on aviation big data and machine learning[J]. IEEE Transactions on Vehicular Technology, 2019, 69(1): 140-150. [5] LU J, SHI Y, REN Z, et al. Research on flight training predic-tion based on incremental online learning[J]. Applied Intelli-gence, 2023, 53(21): 25662-25677. doi: 10.1007/s10489-023-04930-9 [6] QU J, WU S, ZHANG J. Flight delay propagation prediction based on deep learning[J]. Mathematics, 2020, 11(3): 494. [7] 贾萌. 基于蔓延动力学的航空网络中航班延误链式波及及机制研究[D]. 南京: 南京航空航天大学, 2020.JIA M. A research on chain propagation mechanism of flight delay in air transport network based on propagation dynam-ics[D]. Nanjing: Nanjing University of Aeronautics and Astro-nautics, 2020. (in Chinese) [8] BASPINAR B, KOYUNCU E, et al. A data-driven air trans-portation delay propagation model using epidemic process models[J]. International Journal of Aerospace Engineer-ing, 2016(1): 4836260. [9] KAFLE N, ZOU B. Modeling flight delay propagation: a new analytical econometric approach[J]. Transportation Research Part B: Methodological, 2016, 93: 520-542. doi: 10.1016/j.trb.2016.08.012 [10] WU W, WU C. Enhanced delay propagation tree model with bayesian network for modelling flight delay propagation[J]. Transportation Planning and Technology, 2018, 41(3): 319-335. doi: 10.1080/03081060.2018.1435453 [11] STERNBERG A, CARVALHO D, MURTA L, et al. An anal-ysis of brazilian flight delays based on frequent patterns[J]. Transportation Research Part E: Logistics and Transportation Review, 2016, 95: 282-298. doi: 10.1016/j.tre.2016.09.013 [12] EVLER J, LINDNER M, FRICKE H, et al. Integration of turnaround and aircraft recovery to mitigate delay propaga-tion in airline networks[J]. Computers & Operations Re-search, 2022, 138: 105602. [13] ZHENG Z, WEI W, HU M. A comparative analysis of delay propagation on departure and arrival flights for a chinese case study[J]. Aerospace, 2021, 8(8): 212. doi: 10.3390/aerospace8080212 [14] KILIC K, SALLAN J M. Study of delay prediction in the us airport network[J]. Aerospace, 2023, 10(4): 342. doi: 10.3390/aerospace10040342 [15] BEATTY R, HSU R, BERRY L, et al. Preliminary evalua-tion of flight delay propagation through an airline sched-ule[J]. Air Traffic Control Quarterly, 1999, 7(4): 259-270. doi: 10.2514/atcq.7.4.259 [16] 丁建立, 陈坦坦, 徐涛. 基于时间Petri网的航班延误链式反应模型构建[J]. 系统仿真学报, 2008, 20(14): 2334-2340.DING J L, CHEN T T, XU T. Modeling of flight delays chain reaction based on timed petri net[J]. Journal of System Simulation, 2008, 20(14): 2334-2340. (in Chinese) [17] AHMADBEYGI S, COHN A, LAPP M. Decreasing airline delay propagation by re-allocating scheduled slack[J]. ⅡE Transactions, 2010, 42(7): 478-489. [18] PYRGIOTIS N, MALONE K M, ODONI A. Modelling de-lay propagation within an airport network[J]. Transportation Research Part C: Emerging Technologies, 2013, 27: 60-75. doi: 10.1016/j.trc.2011.05.017 [19] SUN X, WANDELT S, LINKE F. Temporal evolution analy-sis of the european air transportation system: air navigation route network and airport network[J]. Transportmetrica B: Transport Dynamics, 2015, 3(2): 153-168. doi: 10.1080/21680566.2014.960504 [20] CAI Q, ALAM S, DUONG V N. A spatial-temporal network perspective for the propagation dynamics of air traffic de-lays[J]. Engineering, 2021, 7(4): 452-464. doi: 10.1016/j.eng.2020.05.027 [21] JIA Z, CAI X, HU Y, JI J. Delay propagation network in air transport systems based on refined nonlinear granger causali-ty[J]. Transportmetrica B: Transport Dynamics, 2022, 10(1): 586-598. doi: 10.1080/21680566.2021.2024102 [22] BARTHÉLEMY M. Spatial networks[J]. Physics re- ports, 2011, 499(1-3): 1-101. [23] HOLME P, SARAMÄKI J. Temporal networks[J]. Physics Reports, 2012, 519(3): 97-125. doi: 10.1016/j.physrep.2012.03.001 [24] HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]. The IEEE Conference on Computer Vision and Pattern Recognition, Honolulu: IEEE, 2017. [25] OBAID H S, DHEYAB S A, SABRY S S. The impact of da-ta pre-processing techniques and dimensionality reduction on the accuracy of machine learning[C]. Electromechanical En-gineering and Microelectronics Conference (IEME-CON), Jaipur, India: IEEE, 2019. [26] PROKHORENKOVA L, GUSEV G, VOROBEV A, et al. CatBoost: unbiased boosting with categorical features[C]. The 32nd International Conference on Neural Information Pro-cessing Systems. Red Hook, USA: NIPS, 2018. [27] 屈景怡, 肖敏, 李佳怡, 等. 基于多任务NR-DenseNet网络的航班延误预测模型[J]. 信号处理, 2023, 39(3): 550-560.QU J Y, XIAO M, LI J Y, et al. Flight delay prediction mod-el based on NR-DenseNet[J]. Journal of Signal Process-ing, 2023, 39(3): 550-560. (in Chinese) [28] JIANG C, JIANG C, CHEN D, et al. Densely connected neu-ral networks for nonlinear regression[J]. Entropy, 2022, 24 (7): 876. doi: 10.3390/e24070876 [29] KHAN W A, MA H L, CHUNG S H, et al. Hierarchical inte-grated machine learning model for predicting flight depar-ture delays and duration in series[J]. Transportation Research Part C: Emerging Technologies, 2021, 129: 103225. doi: 10.1016/j.trc.2021.103225 [30] LAMBELHO M, MITICI M, PICKUP S, et al. Assessing strategic flight schedules at an airport using machine learn-ing-based flight delay and cancellation predictions[J]. Jour-nal of Air Transport Management, 2020, 82: 101737. [31] ANTUNES R N, NG W, TAY J, et al. Delay predictive ana-lytics for airport capacity management[J]. Transportation Re-search Part C: Emerging Technologies, 2025, 171: 104947. doi: 10.1016/j.trc.2024.104947 -
点击查看大图
图(10) / 表(6)
计量
- 文章访问数: 2
- HTML全文浏览量: 1
- PDF下载量: 0
- 被引次数: 0