离场航空器滑行时间预测研究

Taxi Time Prediction of Departure Aircraft

为准确预测离港航班滑行时间,结合北京首都国际机场实际运行情况,分析航空器滑行距离、场面滑行航空器数量(进,离港)、跑道运行模式对航班滑行时间的影响;并运用DBSCAN算法按每小时航班流量对机场运行时间段进行分类;根据分类结果建立多元回归模型,分别采用传统统计学和机器学习(Lasso回归)预测航空器离场滑行时间。结果表明:与传统统计学多元线性回归模型相比,机器学习交叉训练集下模型的预测准确度较高,预测与实际误差值在5 min内的占87%。研究结果可用于大型机场实际运行航班滑行时间预测。

In order to accurately predict the taxi time of departing flights,combined with the actual operation condition of Beijing Capital International Airport,a multiple regression model was established according to the classification results obtained by using DBSCAN algorithm to classify the airport operation time according to the flight flow per hour.The impact of the taxiing distance of aircrafts,the number of taxiing aircrafts on the ground(departure and arrive),the runway operation mode on the taxiing time of flights was analyzed.And two methods,traditional statistics and machine learning(Lasso regression),were used to predict the taxiing time of departure aircraft.The results show that:compared with the traditional statistical multiple regression model,the prediction accuracy of the proposed model under machine learning cross-training set is higher and 87%of errors between the predicted and actual values are within 5 minutes,which can be used to predict the actual flight taxiing time of large airports.