Forecast Method for Medium-Long Term Air Traffic Flow Considering Periodic Fluctuation Factors

To accurately characterize the trend and periodic fluctuation of the future traffic demand in a specific airspace unit, an improved dynamic linear model that is based on the linear growth model was developed to forecast the medium-long term air traffic flow, by taking into full account periodic fluctuation factors such as the climate influence, seasonal fluctuation, actual traffic demand, and so on. Then, the Bayesian state estimation and forecasting method was used to solve the proposed model, and the medium-long term air traffic flow and its variation trend was predicted using the historical data of air traffic flow in a specific airspace unit. In addition, a case study on a real data set of a typical domestic airspace unit was carried out to compare the forecasting performance of the models. The results show that, compared with the linear growth model not considering periodic fluctuation factors, the air traffic flow obtained by the improved model has a periodic fluctuation characteristic, and is more in line with the real situation of air transportation in China; simultaneously, the mean absolute error of the yearly traffic flow decreases from 3.14% to 1.71%, and the standard deviation of forecast error decreases from 2.01% to 0.02%.