基于3种可视图的进场航班流量波动特性适应性评估

在空域资源优化配置、运行效率提升、飞行安全保障等方面, 掌握空中交通流量波动规律发挥着先导性、基础性和关键性作用。为评估可视图、水平可视图、有限穿越可视图这3种图对航班流量波动特性及其演化的刻画能力, 针对同1个进场航班流的多尺度流量时间序列构建复杂网络, 分别从网络的整体结构和局部结构开展了适用性评估分析。针对网络整体结构特点, 提出了基于网络结构从属阵特点的网络细节损失率定义, 再通过k-core聚类分析考察了k阶核量化流量波动强度的适用性; 针对网络局部结构特点, 利用motif方法计算波动模式转移概率, 分析了不同长度序模体刻画波动演化的适应性水平。分析结果表明: ①当有限穿越可视图网络N值与节点数量占比在0.48%~1.442%区间时, N值的选择能够保证从属阵细节损失率在0.5范围内; ②可视图与有限穿越可视图(N=1~3)均能有效刻画航班流量时间序列的波动强度, 对时间序列波动的适应性评估值分别为2.665、4.810、6.973和9.883;③motifs序列长度过短, 将导致motifs类型数量少、不同motifs类型之间的转移概率趋于相同, 而在交通流混沌特性的影响下motifs序列过长对于预测没有意义, 因此, 可视图及N=1~3的有限穿越可视图motifs序列长度推荐使用选择4~7个节点长度。综上所述, 运用k-core聚类与motifs方法能有效分析整体网络与局部网络下波动模式的转移特征, 准确揭示空中交通时间维度的演变规律, 相关分析结果可以为航班延误预测提供依据, 能在航班实际运行管理中发挥先导性作用。 

An Evaluation method for the Suitability of Three Visibility Graphs in Analyzing the Fluctuation Characteristics of Arrival Flight Flows

Understanding the fluctuation characteristics of air traffic flows plays a leading, essential, and key role in many aspects of their control and management, such as airspace configuration optimization, efficiency promotion, and safety assurance. This paper aims to evaluate the suitability of the visibility graph(VG), horizontal visibility graph(HVG), and limited penetrable visibility graph(LPVG) in analyzing the fluctuation characteristics of air traffic flows. A complex network based on the multi-scale time series data extracted from the same arrival flow is developed and the suitability of three visibility graphs is evaluated from the global and local structure perspectives. From the global perspective, a concept of details loss rate is proposed by considering the characteristics of the network structure-dependent matrix. Then a k-core cluster is used to analyze the suitability of quantifying the strength of flight flow fluctuations. From the local perspective, a transfer probability of fluctuation patterns is calculated using the sequential motifs method, and the suitability of the sequential motif with different lengths in characterizing fluctuation characteristics of flight flows is evaluated. The results show that: ①the loss rate of detail can be limited within 0.5 when the proportion of N value of the LPVG in network nodes ranges from 0.48% to 1.442%;②VG and LPVG(N=1~3) can effectively describe the intensity of fluctuation of flight flow time series data and the suitability value is 2.665, 4.810, 6.973, and 9.883, respectively; ③a long sequential motif would reduce the number of sequential motifs and result in the similarity of transition probability among different types of the sequential motifs, while a short sequential motif is useless for prediction due the chaotic characteristics of traffic flow. Thus, it is recommended to use the sequential motif with the length of 4, 5, 6, and 7 for VG and LPVG(N=1~3). In conclusion, the k-core cluster and the motifs method provide an in-depth analysis of the transfer characteristics among the fluctuation modes and the evolution of time dimension in air traffic, which offers support for delay prediction and plays a leading role in the actual operation management of flights.