Aircraft recovery with precancellation

Airlines often adopt a wait-and-see strategy for disruptions, resulting in canceling flights at the last moment. This not only incurs extra compensation costs but also significantly affects passengers’ travel experiences. To mitigate these losses, we introduce the concept of flight precancellation, which is defined as canceling flights one to several days before departure. To make precancellation decisions with respect to stochastic future weather conditions, we develop a two-stage stochastic model aimed at minimizing the overall recovery cost. To solve this model, we design a Lagrangian dual decomposition (LDD) approach, which efficiently decomposes the model into scenario-independent submodels. These submodels are then solved by a column generation framework. Additionally, we propose a dual-based variable evaluation strategy (DVS) to accelerate the solving process of LDD. We evaluate the effectiveness and efficiency of our model and algorithms using real operational data from three airlines, which are tested via real typhoon data. The computational results show that LDD can obtain optimal linear programming (LP) solutions and near-optimal integer programming (IP) solutions. Compared with the baseline column generation algorithm, the solution times for LDD and LDD-DVS are reduced by 41% and 46%, respectively. Additionally, tests conducted on real typhoon data demonstrate that, by incorporating precancellation decisions, it achieves an average cost savings of 17% compared with solutions that consider only real-time cancellation decisions.