Dueling DQN优化下的航班延误自适应图卷积循环网络预测方法

doi:10.12305/j.issn.1001-506X.2025.02.23

摘要/Abstract

摘要：

为充分挖掘机场网络航班间的时空动态相关性以减小预测误差, 提出一种基于对偶深度Q网络(dueling deep Q network, Dueling DQN)优化的多组件自适应图卷积循环网络航班延误预测模型。首先, 结合自适应图卷积网络与多头空间注意力机制, 并行捕获并融合多个子空间的延误信息, 充分挖掘非线性空间动态特征。其次, 采用门控循环单元为时间特征提取模块的基础结构, 并引入时间注意力机制以学习历史延误数据间的关注权重。然后, 设置多个时间维输入组件, 增加对不同时间模式构建的多样性。最后，采用Dueling DQN优化多组件自适应图卷积门控循环单元(multi-component adaptive graph convolution-gated recurrent unit，MAGC-GRU)模型的超参数。实验结果表明，所提模型的平均绝对误差相对于历史平均法、随机森林法、梯度增强回归树法、门控循环单元法、时空图卷积网络法, 分别降低了10.6%、6.07%、9.18%、3.79%和3.12%。

关键词: 航班延误预测, 深度学习, 强化学习, 多组件融合, 图卷积

Abstract:

To fully explore the spatio-temporal dynamic correlation between airport network flights to reduce prediction errors, a multi-component adaptive graph convolutional recurrent network flight delay prediction model based on dueling deep Q network (Dueling DQN) optimization is proposed. Firstly, by combining adaptive graph convolutional network (GCN) with multi-head spatial attention mechanisms, parallel capture and fusion of delay information from multiple subspaces is achieved, fully exploiting nonlinear spatial dynamic features. Secondly, grated recurrent unit (GRU) is used as the basis for the time feature extraction module, and time attention mechanism is introduced to learn the attention weights between historical delay data. Then, multiple time dimension input components are set up to increase the diversity of constructing different time patterns. Finally, Dueling DQN is used to optimize the hyperparameters of the multi-componet adaptive graph convolution-GRU (MAGC-GRU) model. The experimental results show that the mean absolute error (MAE) of the proposed model decreased by 10.6%, 6.07%, 9.18%, 3.79%, and 3.12% compared to historical average, random forest, gradient boosting regression tree, GRU, and spatial-temporal GCN, respectively.

Key words: flight delay prediction, deep learning, reinforcement learning, multi-component fusion, graph convolution network (GCN)

中图分类号:

V351

刘晓琳, 郭梦娇, 李卓. Dueling DQN优化下的航班延误自适应图卷积循环网络预测方法[J]. 系统工程与电子技术, 2025, 47(2): 568-579.

Xiaolin LIU, Mengjiao GUO, Zhuo LI. Adaptive graph convolutional recurrent network prediction method for flight delay based on Dueling DQN optimization[J]. Systems Engineering and Electronics, 2025, 47(2): 568-579.

图/表 15

图1

图2

图3

图4

图5

图6

图7

图8

表1

图9

图10

图11

表2

图12

表3

参考文献 29

1	HENRIQUES R , FEITEIRA I . Predictive modelling: flight delays and associated factors, Hartsfield-Jackson Atlanta International Airport[J]. Procedia Computer Science, 2018, 138, 638- 645. doi: 10.1016/j.procs.2018.10.085
2	BASPINAR B , KOYUNCU E . A data-driven air transportation delay propagation model using epidemic process models[J]. International Journal of Aerospace Engineering, 2016, 2016 (1): 4836260.
3	王春政, 胡明华, 杨磊, 等. 基于Agent模型的机场网络延误预测[J]. 航空学报, 2021, 42 (7): 452- 465.
	WANG C Z , HU M H , YANG L , et al. Airport network delay prediction based on Agent model[J]. Acta Aeronautica et Astronautica Sinica, 2021, 42 (7): 452- 465.
4	罗赟骞, 陈志杰, 汤锦辉, 等. 采用支持向量机回归的航班延误预测研究[J]. 交通运输系统工程与信息, 2015, 15 (1): 143- 149.
	LUO Y Q , CHEN Z J , TANG J H , et al. Flight delay prediction using support vector machine regression[J]. Journal of Transportation Systems Engineering and Information Technology, 2015, 15 (1): 143- 149.
5	LI Q , JING R Z . Characterization of delay pro-pagation in the air traffic network[J]. Journal of Air Transport Management, 2021, 94, 102075. doi: 10.1016/j.jairtraman.2021.102075
6	程华, 李艳梅, 罗谦, 等. 基于C4.5决策树方法的到港航班延误预测问题研究[J]. 系统工程理论与实践, 2014, 34 (S1): 239- 247.
	CHENG H , LI Y M , LUO Q , et al. Study on flight delay with C4.5 decision tree based prediction method[J]. Systems Engineering Theory and Practice, 2014, 34 (S1): 239- 247.
7	KIM Y J, CHOI S, BRICENO S, et al. A deep learning approach to flight delay prediction[C]//Proc. of the 35th Digital Avionics Systems Conference, 2016.
8	YU B , GUO Z , ASIAN S , et al. Flight delay prediction for commercial air transport: a deep learning approach[J]. Transportation Research Part E: Logistics and Transportation Review, 2019, 125 (5): 203- 221.
9	CAI K Q , LI Y , FANG Y P , et al. A deep learning approach for flight delay prediction through time-evolving graphs[J]. IEEE Trans.on Intelligent Transportation Systems, 2021, 23 (8): 11397- 11407.
10	姜雨, 陈名扬, 袁琪, 等. 基于时空图卷积神经网络的离港航班延误预测[J]. 北京航空航天大学学报, 2023, 49 (5): 1044- 1052.
	JIANG Y , CHEN M Y , YUAN Q , et al. Departure flight delay prediction based on spatio-temporal graph convolutional networks[J]. Journal of Beijing University of Aeronautics and Astronautics, 2023, 49 (5): 1044- 1052.
11	ZENG W L , LI J , QUAN Z B , et al. A deep graph-embedded LSTM neural network approach for airport delay prediction[J]. Journal of Advanced Transportation, 2021, 2021 (1): 6638130.
12	CAI K Q , LI Y , ZHU Y W , et al. A geographical and operational deep graph convolutional approach for flight delay prediction[J]. Chinese Journal of Aeronautics, 2023, 36 (3): 357- 367.
13	TAN R K, BORA S. Parameter tuning in modeling and simulations by using swarm intelligence optimization algorithms[C]//Proc. of the International Conference on Computational Intelligence and Communication Network, 2018: 148-152.
14	龙远, 邓小龙, 杨希祥, 等. 基于PSO-BP神经网络的平流层风场短期快速预测[J]. 北京航空航天大学学报, 2022, 48 (10): 1970- 1978.
	LONG Y , DENG X L , YANG X X , et al. Short-term rapid prediction of stratospheric wind field based on PSO-BP neural network[J]. Journal of Beijing University of Aeronautics and Astronautics, 2022, 48 (10): 1970- 1978.
15	杨帆, 申亚, 李东东, 等. 基于GA-GNNM的极地光伏发电功率预测方法[J]. 太阳能学报, 2022, 43 (4): 167- 174.
	YANG F , SHEN Y , LI D D , et al. Polar photovoltaic power forecasting method based on GA-GNNM[J]. Acta Energiae Solaris Sinica, 2022, 43 (4): 167- 174.
16	陈超逸, 鲁娟, 陈楷, 等. 车削表面粗糙度解析模型与DDQN-SVR预测模型研究[J]. 机械工程学报, 2021, 57 (13): 262- 272.
	CHEN C Y , LU J , CHEN K , et al. Research on analytical model and DDQN-SVR prediction model of turning surface roughness[J]. Journal of Mechanical Engineering, 2021, 57 (13): 262- 272.
17	陈海鹏, 李赫, 阚天洋, 等. 考虑风电时序特性的深度小波-时序卷积网络的超短期风功率预测[J]. 电网技术, 2023, 47 (4): 1653- 1665.
	CHEN H P , LI H , KAN T Y , et al. Ultra-short-term wind power prediction of deep wavelet-timing convolutional network considering wind power timing characteristics[J]. Power System Technology, 2023, 47 (4): 1653- 1665.
18	SHI Y , HUI L . Wind power prediction based on outlier correction, ensemble reinforcement learning, and residual correction[J]. Energy, 2022, 250, 123857.
19	YU C Q , YAN G X , YU C M , et al. A multi-factor driven spatio temporal wind power prediction model based on ensemble deep graph attention reinforcement learning networks[J]. Energy, 2023, 263, 126034.
20	BRUNA J, ZAREMBA W, SZLAM A, et al. Spectral networks and locally connected networks on graphs[EB/OL]. [2023-11-05]. https://doi.org/10.48550/arXiv.1312.6203.
21	DEFFERRARD M, BRESSON X, VANDERGHEYNST P. Convolutional neural networks on graphs with fast localized spectral filtering[C]//Proc. of the Advances in Neural Information Processing Systems, 2016.
22	KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[EB/OL]. [2023-11-05]. https://doi.org/10.48550/arXiv.1609.02907.
23	ASHISH V, NOAM S, NIKI P, et al. Attention is all you need[C]//Proc. of the Conference on Neural Information Processing Systems, 2017: 5998-6008.
24	CHUNG J, GULCEHRE C, CHO K H, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[EB/OL]. [2023-11-05]. https://doi.org/10.48550/arXiv.1609.
25	BAHDANAU D, CHO K, BENGIO Y. Neural machine translation by jointly learning to align and translate[J]. https://arXivpreprintarxiv:1409.0473, 2014.
26	SUN G L , AYEPAHMENSAH D , XU R , et al. End-to-end CNN-based dueling deep Q-Network for autonomous cell activation in Cloud-RANs[J]. Journal of Network and Computer Applications, 2020, 169, 102757.
27	MNIH V, KAVUKCUOGLU K, SILVER D, et al. Playing atari with deep reinforcement learning[EB/OL]. [2023-11-05]. https://people.engr.tamu.edu/guni/ccsce642/files/clqn.pdf.
28	REBOLLO J J , BALAKRISHNAN H . Characterization and prediction of air traffic delays[J]. Transportation Research Part C: Emerging Technologies, 2014, 44, 231- 241.
29	XIA Y, CHEN J G. Traffic flow forecasting method based on gradient boosting decision tree[C]//Proc. of the 5th International Conference on Frontiers of Manufacturing Science and Measuring Technology, 2017: 436-439.

时间/天	评价指标	模型
时间/天	评价指标	HA	RF	GBRT	GRU	STGCN	MAGC-GRU
1	RMSE	8.907	8.404	8.471	8.219	8.186	8.003
	MAE	6.589	6.271	6.486	6.122	6.080	5.890
	MAPE/%	8.685	8.280	8.315	8.118	8.065	7.864
2	RMSE	8.907	8.652	8.471	8.621	8.688	8.320
	MAE	6.589	6.490	6.486	6.313	6.493	6.180
	MAPE/%	8.685	8.602	8.613	8.494	8.647	8.236
3	RMSE	8.907	8.742	8.784	8.679	8.786	8.476
	MAE	6.589	6.565	6.565	6.541	6.565	6.239
	MAPE/%	8.685	8.716	8.729	8.661	8.762	8.364
4	RMSE	8.907	8.772	8.802	8.753	8.821	8.546
	MAE	6.589	6.597	6.593	6.686	6.662	6.334
	MAPE/%	8.685	8.762	8.768	8.794	8.846	8.472
5	RMSE	8.907	8.793	8.823	8.797	8.861	8.645
	MAE	6.589	6.623	6.627	6.703	6.697	6.463
	MAPE/%	8.685	8.793	8.806	8.838	8.895	8.611

优化方法	评价指标
优化方法	MAE	RMSE	MAPE/%
PSO	5.965	8.073	8.067
DDQN	5.893	8.042	8.023
Dueling DQN	5.890	8.003	7.864

预测模型	评价指标
预测模型	MAE	RMSE	MAPE/%
HA	6.418	8.788	8.682
GRU	6.080	8.260	8.235
RF	6.248	8.448	8.416
GBRT	6.243	8.506	8.435
STGCN	5.967	8.140	8.087
MAGC-GRU	5.874	8.061	8.010

[1]	付可, 陈浩, 王宇, 刘权, 黄健. 基于不确定性的贝叶斯策略重用方法[J]. 系统工程与电子技术, 2025, 47(2): 535-543.
[2]	闫循良, 王宽, 张子剑, 王培臣. 基于LSTM-DDPG的再入制导方法[J]. 系统工程与电子技术, 2025, 47(1): 268-279.
[3]	蔡伟, 王鑫, 蒋昕昊, 杨志勇, 陈栋. 基于解耦的小样本目标检测方法研究[J]. 系统工程与电子技术, 2024, 46(9): 2941-2950.
[4]	张庭瑜, 曾颖, 李楠, 黄洪钟. 基于深度强化学习的航天器功率-信号复合网络优化算法[J]. 系统工程与电子技术, 2024, 46(9): 3060-3069.
[5]	夏雨奇, 黄炎焱, 陈恰. 基于深度Q网络的无人车侦察路径规划[J]. 系统工程与电子技术, 2024, 46(9): 3070-3081.
[6]	杨志鹏, 陈子浩, 曾长, 林松, 毛金娣, 张凯. 复杂环境下的飞行器在线航路规划决策方法[J]. 系统工程与电子技术, 2024, 46(9): 3166-3175.
[7]	陈晓萱, 徐书文, 胡绍海, 马晓乐. 基于卷积与自注意力的红外与可见光图像融合[J]. 系统工程与电子技术, 2024, 46(8): 2641-2649.
[8]	彭莉莎, 孙宇祥, 薛宇凡, 周献中. 融合三支多属性决策与SAC的兵棋推演智能决策技术[J]. 系统工程与电子技术, 2024, 46(7): 2310-2322.
[9]	汪强龙, 高晓光, 吴必聪, 胡子剑, 万开方. 受限玻尔兹曼机及其变体研究综述[J]. 系统工程与电子技术, 2024, 46(7): 2323-2345.
[10]	郭宏达, 娄静涛, 徐友春, 叶鹏, 李永乐, 陈晋生. 基于MADDPG的多无人车协同事件触发通信[J]. 系统工程与电子技术, 2024, 46(7): 2525-2533.
[11]	孙先涛, 江汪洋, 陈文杰, 陈伟海, 智亚丽. 基于感兴趣区域的物体抓取位姿检测[J]. 系统工程与电子技术, 2024, 46(6): 1867-1877.
[12]	陈雪梅, 刘志恒, 周绥平, 余航, 刘彦明. 基于HRNet的高分辨率遥感影像道路提取方法[J]. 系统工程与电子技术, 2024, 46(4): 1167-1173.
[13]	张梦钰, 豆亚杰, 陈子夷, 姜江, 杨克巍, 葛冰峰. 深度强化学习及其在军事领域中的应用综述[J]. 系统工程与电子技术, 2024, 46(4): 1297-1308.
[14]	张天文, 张晓玲, 邵子康, 曾天娇. 基于掩模注意型交互的SAR舰船实例分割[J]. 系统工程与电子技术, 2024, 46(3): 831-838.
[15]	李彦铃, 罗飞舟, 葛致磊. 基于鲁棒观测器的深度强化学习垂直起降运载器姿态稳定研究[J]. 系统工程与电子技术, 2024, 46(3): 1038-1047.