中国民航大学学报 ›› 2024, Vol. 42 ›› Issue (6): 52-60.

• 航空运输管理 • 上一篇    下一篇

基于二次分解集成的机场流量短期预测

王
飞 1 ,韩翔宇 2
  

  1. 1. 中国民航大学空中交通管理学院,天津 300300;2. 内蒙古自治区民航机场集团有限责任公司,呼和浩特 010050
  • 收稿日期:2023-09-14 修回日期:2023-11-22 出版日期:2025-04-08 发布日期:2025-04-08
  • 作者简介:王飞(1982— ),男,安徽凤阳人,副教授,博士,研究方向为空中交通管理
  • 基金资助:
    天津市应用基础研究多元投入基金重点项目(21JCZDJC00840);中央高校基本科研业务费专项(3122019129)

Short-term forecast of airport flow based on twice decomposition integration

WANG Fei 1 , HAN Xiangyu2
  

  1. 1. College of Air Traffic Management, CAUC, Tianjin 300300, China; 2. Inner Mongolia Autonomous Region Civil
    Airport Group Co., Ltd., Hohhot 010050, China
  • Received:2023-09-14 Revised:2023-11-22 Online:2025-04-08 Published:2025-04-08

摘要:

为实现准确的机场流量短期预测,本文建立了基于二次分解方法的分解集成预测模型。 首先,应用局部加
权回归周期趋势分解(STL, seasonal and trend decomposition procedure based on Loess)算法将原始时间序
列分解为趋势项、季节项和余项 3 个分量,并计算其样本熵。其次,应用遗传算法(GA,genetic algorithm)优化
变分模态分解(VMD,variational mode decomposition)参数,对熵值较大的分量进行二次分解。 再次,使用极
端梯度提升(XGBoost, extreme gradient boosting)对二次分解后的所有分量进行预测,采用加和集成得到
最终的预测值。 最后,采集国内典型机场实际运行数据进行实例分析。 针对北京首都国际机场 60 min 进
场、离场流量时序,本文模型预测的均等系数(EC,equal coefficient)值分别为 0.970 3、0.995 9,相比其他常
用模型均有所提高。 此外,对于上海浦东、上海虹桥、广州白云 3 个大型国际机场,本文模型在 60 min、
30 min 统计尺度下进场和离场流量预测的 EC 值均在 0.970 0 以上,15 min 统计尺度下预测的 EC 值均在
0.950 0 以上。 结果表明,本文建立的二次分解集成预测模型具有良好的准确性和普适性,用于机场流量短
期预测是可行和有效的。

关键词:

Abstract:

In order to achieve accurate short-term forecast of airport flow, a decomposition integration forecast model based on
twice decomposition method is established in this paper. Firstly, the seasonal and trend decomposition procedure
based on Loess (STL) algorithm is applied to decompose the original time series into three components, including
trend term, seasonal term and residual term, and their sample entropy are calculated. Secondly, genetic algorithm
(GA) is applied to optimize the parameters of variational mode decomposition (VMD), and the components with
larger entropy values are subjected to twice decomposition. Thirdly, extreme gradient boosting (XGBoost) is applied
to predict all components after twice decomposition, and the final predicted value is obtained by adding and integrating. Finally, the actual operation data of domestic typical airports are collected for case analysis. For the 60 min
arrival and departure flow time series of Beijing Capital International Airport, the equal coefficient (EC) values predicted in this paper are 0.970 3 and 0.995 9 respectively, which has an improvement compared to other common
models. In addition, for the three large international airports of Shanghai Pudong, Shanghai Hongqiao, and
Guangzhou Baiyun, the EC values predicted by the proposed model are all above 0.970 0 for arrival and departure
flow at 60 min and 30 min scales, and the predicted EC values at the 15 min scale are all above 0.950 0. The results
indicate that the twice decomposition integration forecast model established in this paper has good accuracy and universality, and is feasible and effective for short-term forecast of airport flow.

Key words:

twice decomposition

中图分类号: