In order to reveal the traffic decay patterns in terminal areas under convective weather conditions, and further enhance existing models’ ability to capture complex dependencies in long sequences, a model integrating a multi-head self-attention mechanism was developed for predicting the degree of terminal area traffic impact. This approach first established an airspace block partitioning framework, incorporating both weather hazard index and airspace blockage rate as dual-dimensional features. Subsequently, a long short-term memory (LSTM) network improved with a multi-head self-attention mechanism (MHSA) was constructed. The model was validated using actual operational data from the Guangzhou terminal area. The results show that the model predicts best when the input sequence length is 12. Compared with classical models such as support vector regression (SVR), random forest (RF), recurrent neural network (RNN), long short-term memory network (LSTM), and gated recurrent unit (GRU), the mean absolute error (MAE) is reduced by 61.97%, 53.45%, 43.75%, 10.00%, and 20.59%, respectively. The root mean square error (RMSE) is reduced by 58.89%, 53.16%, 41.27%, 13.95%, and 15.91%, respectively. Additionally, the coefficient of determination (R2) is improved by 48.10%, 30.18%, 15.13%, 2.30%, and 2.97%, respectively. It is concluded that the proposed model can effectively capture the complex relationship between meteorological factors and traffic flow decay. The traffic decay effect is most significant when convective weather is located in the airspace adjacent to the runways and in the key arrival/departure areas in the northern part of the terminal area. Therefore, it is suggested that air traffic controllers enhance real-time monitoring and early warnings for these critical airspace sectors.
PENG Ying, WANG Ting-ting, YE Wen-jie, RAN Long-hui.
Prediction of Airport Terminal Area Flow Impact under Convective Weather[J].
Science Technology and Engineering, 2026, 26(16): 7092-7101 DOI:10.12404/j.issn.1671-1815.2504676
为验证MHSA-LSTM模型的预测效果,将其与SVR、RF、RNN、LSTM、门控循环单元(gated recurrent unit, GRU)进行对比,其中RNN及其各种变体的实验设置保持不变,输入序列长度均取12。测试集上模型表现如表5所示。其中,MHSA-LSTM模型预测效果最好,与SVR、RF、RNN、GRU相比,MAE分别降低了61.97%、53.45%、43.75%、20.59%;RMSE分别降低了58.89%、53.16%、41.27 %、15.91%;R2分别提高了48.10%、30.18%、15.13%、2.97%。就拟合效果而言,MHSA-LSTM模型与真实值拟合得最好,R2达93.6%,LSTM次之,SVR最差。此外,与未加多头注意力机制的LSTM相比,MHSA-LSTM模型MAE降低了10.00%,RMSE降低了13.95%,R2提高了2.30%,表明多头注意力机制可以进一步提高模型预测性能。
YangShangwen, HuMinghua, ZhaoZheng, et al. Terminal area dynamic capacity prediction model[J]. Journal of Nanjing University of Aeronautics and Astronautics, 2012, 44(1): 113-117.
TianYong, YangShuangshuang, WanLili, et al. Research on the method of sector dynamic capacity evaluation[J]. Systems Engineering-Theory & Practice, 2014, 34(8): 2163-2169.
[5]
赵征. 空域容量评估与预测技术研究[D]. 南京: 南京航空航天大学, 2015.
[6]
ZhaoZheng. Research on airspace capacity assessment and forecast[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2015.
[7]
MatthewsM P, DelauraR, VenutiJ. Strategic forecasts of TRACON airspace capacity during convective weather impacts[C]// 17th AIAA Aviation Technology, Integration, and Operations Conference. Denver: AIAA, 2017: 3430-3441.
WenRuiying, HeJiaxing, WangHongyong. Multifaceted complexity calculation method for air traffic supporting stable trajectory optimization[J]. Journal of Transportation Systems Engineering and Information Technology, 2025, 25(1): 258-269.
[10]
毛利民. 对流天气下空域容量预测方法研究[D]. 南京: 南京航空航天大学, 2020.
[11]
MaoLimin. Research on forecasting method of terminal area capacity under the influence of convective weather[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2020.
MaoLimin, PengYing, LiJie, et al. Random-forest based terminal capacity prediction under convective weather[J]. Systems Engineering-Theory & Practice, 2021, 41(8): 2125-2136.
[14]
郭聪聪. 对流天气影响下的终端区容量预测方法研究[D]. 南京: 南京航空航天大学, 2021.
[15]
GuoCongcong. Research on forecasting method of terminal area capacity under the influence of convective weather[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2021.
[16]
WangS J, YangB T, DuanR R, et al. Predicting the airspace capacity of terminal area under convective weather using machine learning[J]. Aerospace, 2023, 10(3): 288-309.
WangWeili, WangYiwen. Explainable prediction of inbound flight delays based on meteorological factors and feature selection[J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(5): 162-171.
ZhouRui, QiuShuang, MengShuangjie, et al. Short-term prediction of air traffic flow based on EMD-LSTM of differential processing[J]. Science Technology and Engineering, 2025, 25(2): 842-849.
[21]
PengY, WangH, MaoL M, et al. Terminal traffic flow prediction method under convective weather using deep learning approaches[J]. Transactions of Nanjing University of Aeronautics and Astronautics, 2021, 38(4): 634-645.
SongWeikai, ZhangHonghai, WanJunqiang, et al. Short-term prediction of air traffic flow based on dynamic time warping-long short-term memory[J]. Science Technology and Engineering, 2022, 22(25): 11256-11261.
WangLili, ZhaoYunfei. A method for predicting air traffic flow based on a combined GA, RBF, and improved Cao method[J]. Journal of Transport Information and Safety, 2023, 41(1): 115-123.
YangGuozhou, WangYunyi. Simulation of air traffic flow prediction in terminal area considering fluctuation factor[J]. Computer Simulation, 2024, 41(11): 6-10, 177.
WangXin, MengTianyu, ZhouJunxi. Remaining useful life prediction for aircraft engine using attention based LSTM[J]. Science Technology and Engineering, 2022, 22(7): 2784-2792.