Research On Flight Delay And Recovery Forecast Methods In Low Visibility Conditions

Since the 21 st century,with the rapid development of the country’s economic construction and industrial production,air transportation has become one of the main modes of travel chosen by people.Therefore,aviation safety and punctuality have received more and more close attention.Meteorological factors such as aircraft icing,bumps,thunderstorms,wind shear,and low visibility are one of the important reasons affecting the normal operation of flights,and the southwest airports are particularly affected by low visibility.This article uses Chengdu Shuangliu Airport’s meteorological data and aerosol and atmospheric pollutant data for the fiveyear period from 2014 to 2018 to predict airport flight delays and flight recovery.The main content consists of the following three parts: multivariate nonlinear fitting research,low visibility prediction,Flight delay and recovery forecast are summarized as follows:In the multivariate nonlinear fitting research part,the change trend and correlation analysis between visibility and its influencing factors are first analyzed,and 3 major influencing factors and 3 minor influencing factors are determined.Because the 45% visibility observation value is limited to 10 km by the range of the observation instrument,in order to standardize the data,a data group with a value of 10 km is excluded,and a multivariate nonlinear fitting study is performed on the three main influencing factors.The combined equations do a reverse fit to the visibility of the culled data set,and any non-linear correlation of the resulting optimized data set is higher than the original data set,and the best non-linear correlation function between the main influence factor and the visibility is obtained It is an S-type function,so the S-type function is selected as the excitation function of the subsequent GA-BP neural network.In terms of low visibility prediction,three visibility level standards were set according to the actual low visibility operation procedures of Shuangliu Airport.Due to the limitations of the general BP neural network,a GA-BP neural network prediction model optimized by genetic algorithms is constructed to predict the visibility of the next 1h,2h,and 3h respectively,and the results show that the prediction can reach the desired accuracy rate.In terms of flight delay and recovery prediction,it is found that the hourly schedule and hourly delay of Shuangliu Airport have a strong regular pattern.Nonlinear fitting is used to obtain Y(t)and S(t)functions that represent its regularity.For the time forecast model of flight delay and recovery,the model combined with the Y(t)and S(t)functions can better predict the recovery time of delayed flights.According to the accuracy of the prediction results of the two types of prediction models in this paper,flight delays and delayed flight recovery can be predicted in advance for the airport in operation.Depending on the upcoming low-visibility weather or delayed flights that are about to resume,air traffic control,airports,and airlines can formulate corresponding emergency strategies,time,and staffing plans in advance.