考虑天气因素影响的离港航班滑出时间预测

Prediction of Departure Flights'Taxi-out Time Considering Weather Factors

针对现有滑出时间预测研究成果未考虑天气因素影响的问题,基于航空例行天气报告(meteorological terminal aviation routine weather report,METAR),构建了考虑天气因素的离港航班滑出时间预测模型。首先,通过分析航空器场面运行态势,厘清了进离港航班滑行过程的时空交叠关系,重新定义了滑出时间的影响因素,并分别阐述了航班运行数据和气象数据的分析流程。基于相关性分析结果构建了滑出时间的反向传播(back propagation,BP)神经网络预测模型,并采用蝗虫优化算法(grasshopper optimization algorithm,GOA)对模型进行优化。以深圳宝安机场2周的实际运行数据对模型进行了验证,结果表明:(1)天气因素是滑出时间的主要影响因素之一,引入量化后的天气因素可显著提升滑出时间预测结果;(2)重新定义的同时段推出及滑行的进离港航班数量、进离港队列的概念和数据样本更加精准,相关性分析结果更加客观;(3)基于GOA-BP的滑出时间预测结果精度有明显提升,平均绝对误差(mean absolute error,MAE)和均方根误差(root mean square error,RMSE)分别减少了11.40 s、12.62 s,平均绝对百分比误差(mean absolute percentage error,MAPE)提升了0.37%;±3 min和±5 min的准确率分别高达81%和94%。

In view of the problem that the existing research results of taxi-out time prediction do not consider the influence of weather factors,based on the data of meteorological terminal aviation routine weather report(METAR),a prediction model of taxi-out time of departure flights considering weather factors was constructed.Firstly,by analyzing the operation situation of the aircraft on the surface,the space-temporal overlapping relationship of the taxi process of inbound and outbound flights was clarified,the influencing factors of the taxi-out time were redefined,and the analysis process of flight operation data and meteorological data was described respectively.Based on the results of correlation analysis,the back propagation(BP)neural network prediction model of taxi-out time was constructed,and the grasshopper optimization algorithm(GOA)was used to optimize the model.The model was validated with the actual operation data of Shenzhen Baoan airport for two weeks.The results show these as follows.Weather factor is one of the main influencing factors of taxi-out time,and the introduction of quantified weather factor can significantly improve the prediction results of taxi-out time.The concept and sample data of the number of inbound and outbound flights pushed and taxied at the same time,inbound and outbound queues redefined are more accurate,and the correlation analysis results are more objective.The prediction accuracy of taxi out time based on GOA-BP is significantly improved,MAE and RMSE are reduced by 11.40 s and 12.62 s respectively,and MAPE is increased by 0.37%;The accuracy of±3 min and±5 min was up to 81%and 94%respectively.