A stochastic optimisation approach to air traffic flow management : addressing predictable and unpredictable weather events

Abstract

With the increasing number of flights, flight delays occur more frequently due to limited airspace and airport capacity. Additionally, adverse weather conditions are becoming more severe and frequent, exacerbating the issue of flight delays. Therefore, it is essential to address this problem. In this thesis, two significant aspects are considered: predictable and unpredictable weather events. The methods required to handle these problems differ slightly due to their particular characteristics. Thus, two different models have been presented to address the corresponding issues and reduce flight delays caused by adverse weather conditions. Predictable weather events refer to those with more stable weather conditions, causing their trajectories are relatively easier to forecast. In this thesis, tropical storms are selected as an example of predictable weather event. A two-stage stochastic optimisation model, considering the effects of tropical storms, is proposed to maximise the punctuality of flights. Various scenarios are used for computation and testing to assess the performance of the proposed stochastic model. It is concluded that the performance of the model is satisfactory. Unpredictable weather events, on the other hand, refer to more dynamic weather conditions. Rainfall is selected as an example of this category. Therefore, a scenario-based two-stage stochastic optimisation model is presented. The proposed model combines the aircraft landing problem and the terminal traffic flow problem to reduce the total time of flight delays. Several computational improvement procedures have been suggested to enhance the performance of the proposed model. It is determined that the performance without using any computational improvement procedures is optimal. A comparison between the traditional scheduling method (i.e. first-come-first-serve strategy), the proposed deterministic model, and the proposed stochastic model has also been performed, and the stochastic model is concluded to be the best among the three methods.