离港航班可变滑出时间预测方法及应用

考虑离港航班可变滑行时间的可量化影响因素，构建了基于BP神经网络的离港航班可变滑出时间预测模型，然后采用遗传算法（Genetic Algorithm, GA）优化BP神经网络的权值和阈值，并提出基于可变滑出时间预测结果的航空器推出控制策略。最后，基于我国中南某枢纽机场2周的实际运行数据对预测模型及控制策略进行了验证。结果表明：①离港航班的可变滑出时间与机场场面交通流有强相关性，与平均滑出时间中度相关，与滑行距离相关性和转弯个数较弱；②基于GA优化后的BP神经网络预测结果误差在±60s、±180s、±300s内的准确率分别提升了14%、10%和5%；预测结果的平均绝对误差百分比提升了1.87%，平均绝对误差和均方根误差分别减少了3.58s、32.45s。③基于可变滑出时间预测的离港推出策略比实际推出时间平均晚68s。研究成果为提升大型枢纽机场场面运行效率和协同决策能力提供了新的思路。

Prediction method of departure flight estimated taxi-out time and its application

Considering of the quantifiable influencing factors of taxi-out time, the prediction model of departure flight estimated taxi-out time based on BP was constructed. Then genetic algorithm (GA) is used to optimize the weights and thresholds of BP neural network, and an aircraft push-out control strategy based on taxi-out time prediction is proposed in this paper. Finally, the prediction model and control strategy are validated by two weeks'' actual operation data of a hub airport in the Central and South China. The results indicate that: (1) estimated taxi-out time has strong correlation with airport traffic flow, moderate correlation with average taxi-out time and weak correlation with taxi distance and number of turns; (2) The prediction accuracy of BP neural network optimized by GA is increased by 14% in ± 60s, 10% in ±180s and 5% in ±300s. And the mean absolute error percentage of the prediction results increased by 1.87%, the mean absolute error decreased by 3.58s, and the root mean square error decreased by 32.45s. (3) The calculated off block time based on taxi-out time prediction is 68s later than the actual off block time. It provides a new way to improve the operation efficiency and collaborative decision-making ability of large hub airports.