航空学报 > 2022, Vol. 43 Issue (9): 325585-325585   doi: 10.7527/S1000-6893.2021.25585

基于分形插值的空中交通流量短期预测

王飞, 韩翔宇   

  1. 中国民航大学 空管学院, 天津 300300
  • 收稿日期:2021-03-29 修回日期:2021-05-14 出版日期:2022-09-15 发布日期:2021-07-20
  • 通讯作者: 王飞,E-mail:feiwang@cauc.edu.cn E-mail:feiwang@cauc.edu.cn
  • 基金资助:
    国家自然科学基金(U1833103);中央高校基本科研业务费(3122019129)

Short-term prediction of air traffic flow based on fractal interpolation

WANG Fei, HAN Xiangyu   

  1. College of Air Traffic Management, Civil Aviation University of China, Tianjin 300300, China
  • Received:2021-03-29 Revised:2021-05-14 Online:2022-09-15 Published:2021-07-20
  • Supported by:
    National Natural Science Foundation of China (U1833103); the Fundamental Research Funds for the Central Universities (3122019129)

摘要: 为精准实施空中交通流量优化与管理, 对流量短期预测方法进行了研究。首先, 应用重标极差(R/S)方法计算流量时间序列的Hurst指数, 来识别分形特征。其次, 应用分形插值模型为每个相似日建立迭代函数系, 并将所有相似日的迭代函数系进行加权, 形成1个统计意义上的迭代函数系, 从任意已知点出发, 通过多次迭代获得稳定的吸引子曲线, 进而得到流量预测值。最后, 采集35天的实际运行数据进行算例分析。结果显示: 60 min尺度流量时间序列的Hurst指数为0.333 6, 具有分形特征; 预测结果的均衡系数为0.957 4、平均绝对相对误差为0.086 7;统计尺度为30 min和15 min的流量时序也具有分形特征, 预测结果的均衡系数分别为0.925 9和0.875 7;临近相似日和相同周天相似日的预测结果没有显著差异; 相较于传统模型, 本文方法对于分形时序预测具有更好适应性。结果说明, 分形插值模型用于空中交通流量短期预测是可行的和有效的, 预测准确性随着统计尺度的减小而降低。

关键词: 空中交通管理, 流量短期预测, 分型插值, 迭代函数系, 时间序列

Abstract: To implement optimization and management of air traffic flow accurately, a short-term flow prediction method is proposed based on fractal interpolation. Firstly, the Hurst index is calculated by the Rescaled Range Analysis (R/S) method to identify the fractal features of flow time series. Secondly, the fractal interpolation model is used to establish the iterated function system for each similar day, and the iterative function system of all similar days is weighted to form an iterated function system in the statistical sense. Starting from any given point, the stable attractor curve is obtained through multiple iterations, and then the flow prediction values are obtained. Finally, the actual operation data of 35 days are collected for example analysis. The results show that the Hurst index of 60 min scale flow time series is 0.333 6, so that the time series has fractal characteristics. The equilibrium coefficient and the average absolute relative error of prediction results is 0.957 4 and 0.086 7, respectively. The flow time series with the statistical scale of 30 min and 15 min also have fractal characteristics, and the equilibrium coefficients are 0.925 9 and 0.875 7, respectively. There is no significant difference between the prediction results of the similar days close to the prediction day and the similar days of the same day in different weeks. Compared with the traditional model, the method proposed has better adaptability for fractal time series prediction. The fractal interpolation model is found to be feasible and effective for the short-term prediction of air traffic flow, and the prediction accuracy decreases with the decrease of statistical scale.

Key words: air traffic management, short-term flow prediction, fractal interpolation, iterated function system, time series

中图分类号: