Spatial-temporal Patterns Mining And Prediction Based On The Traffic Complex Network In Terminal Airspace

With the rapid development of the global air transport industry and the rapid growth of civil aviation travel demand in recent years,the airspace flow has taken on "high density and high growth" characteristics.Civil aviation companies respond to rising demand for airspace traffic by adding routes or flights,resulting in a decrease in average flight on-time rate and an increase in average passenger flight delay time.Automatic dependent surveillance broad can update aircraft flight trajectory in real time,which provides a data basis for understanding the flight status and monitoring the air traffic flow.Under the control of airspace control and ground tower,the flight path of the airport terminal airspace is relatively fixed,forming an air traffic network with a relatively fixed structure and constantly changing flow.Traditional air traffic flow monitoring ignores the impact of real-time changes in terminal airspace traffic on airport delay prediction and does not measure the state change of the air traffic network.This study will break the traditional limitations of air traffic flow measurement,and build a set of terminal airspace based on air track traffic network construction and measurement method,explore the airport terminal airspace spatial and temporal patterns of air trajectory,propose a delay rate prediction method based on the spatialtemporal characteristics of air traffic flow,which not only provides a reusable algorithm framework for the field of airport terminal airspace management,but also provides a new perspective for researchers to understand air traffic characteristic patterns in different regions.This paper aims to realise the real-time monitoring of air traffic flow by constructing the complex traffic network in terminal airspace;explore the patterns of air traffic complex networks in different countries and regions around the world;realise a spatial-temporal graph convolution model considering the air traffic network in terminal airspace.The main research contents and results of this paper are as follows:1)In terminal airspace,an adaptive traffic complex network construction method is proposed.We achieve adaptive identification of airport traffic hub points in terminal airspace,identify the trajectory skeleton line in terminal airspace,calculate the interactive flow of airspace traffic hub points,and construct a complete set of automatic traffic complex network calculation processes using aviation trajectory data.This method can adapt to the aviation trajectory data of airport terminal airspace in different regions,accurately identify important traffic hub points and trajectory skeleton lines,and build an air traffic network with spatio-temporal variation characteristics2)This study looks at the characteristics of air traffic modes at the regional and terminal levels in various countries and regions around the world.We create a regionalscale air transport network to measure the flow interactions between different airports,as well as a complex network measurement system for terminal airspace traffic and research into the air traffic flow laws of airports around the world.The structure and topological characteristics of the terminal airspace network are found to be closely related to airport throughput;under the multi-feature model,global airports can be divided into three basic types,and the type distribution is closely related to regional economic development.3)The air complex network is used to suggest a spatio-temporal graph convolution model.With the weather elements and delay rate information,the regional aviation complex network is taken as the graph structure and the spatio-temporal variation of terminal airspace traffic flow as the characteristics,a spatio-temporal graph convolution model for predicting airport delay rate is realised.The model has a decent prediction impact on various kinds of airports,according to the experimental data.Furthermore,the complex network features in terminal airspace have enhanced the model prediction outcomes to varied degrees,with increased RMSE accuracy ranging from 8% to 16%.