Research On Flight Departure Time Prediction For A-CDM

Airport Collaborative Decision-Making(A-CDM)is one of the important methods for the efficient operation of large-scale hub airports and the milestone node time prediction is the core content of A-CDM.The departure of the flight as the last node of the entire ground operation,its accurate estimation is of vital importance for flight launch control and scene operation scheduling.It is the starting point of this paper to realize the accurate dynamic estimation of flight departure time and the construction of the outbound flight milestone node dynamic prediction model.Therefore,a model for estimating flight departure time based on dynamic Bayesian networks and an improved dynamic Bayesian network model based on Gibbs sampling reasoning algorithm are established to predict the departure time of the flight,which is expected to provide support for the prediction of the airport collaborative decision-making and promotes the implementation of the airport collaborative decision-making concept.As for the prediction of flight departure time,a model for estimating flight departure time based on dynamic Bayesian networks is constructed in this paper,which is applied to the flight departure time estimation for the first time.Firstly,the actual outbound flight departure process and the factors affecting the flight departure process were analysed based on different flight attribute.According to the influencing factors,the data was classified and processed,and the Monte Carlo simulation method was combined with the historical data classification to obtain the joint and prior distribution of each link.The Kolmogorov test was used to determine the joint distribution model of each link and the parameters of the dynamic Bayesian network model were obtained;Secondly,the Bayesian network architecture and conditional probability were used to infer the dynamic estimation of the departure time and the completion time of each link;Finally,the destination operation data of an airport in the middle of the country was selected for simulation verification.The simulation results show that the model could update the forecast value of departure time in real time based on the flight departure process,and realize the dynamic perception and prediction of the airport operation situation,and the stability of the dynamic estimation result would be better.As for the problem of link propagation error in the flight departure time estimation model,The Gibbs sampling method is also introduced in this paper and an improved dynamic Bayesian network model based on Gibbs sampling reasoning algorithm is constructed.The Gibbs sampling reasoning algorithm was used to extend the data,and the Bayesian theorem was used to modify the extended data to calculate the Bayesian posterior probability density.Finally,the experimental results were analyzed and the evaluation and analysis of the model proved the feasibility of this method,which could effectively reduce the propagation error of the link and improve the prediction accuracy.