A Learning Estimation Approach for Arrival and Departure Capacity considering Weather Impact

During severe weather, traffic managers in all facilities of terminal area are busy managing the air traffic within their airspace. In this case, shared capacity information is expected to assist decision making thus better coordinating the arrival and departure operation. Therefore, an integrated Estimation of the Arrival and Departure Capacity under Weather Impact is a crucial task to improve the operational safety and efficiency level in terminal airspace. However, currently, capacity estimation relies heavily on controller's empirical judgment and existing learning-based methods have limited application scenarios as they ignore the temporal and spatial effects of weather. This paper presents a Spatio-Temporal Graph Convolutional Network with Self-Enhancement mechanism and Initial residual (STGCNSEI) estimation model that extracts temporal characteristics and topological influences from coarse estimates to finer ones. Firstly, the XGBoost regression and the max-flow min-cut approach are adopted to preliminarily obtain a rough estimation of airport and fixes capacities respectively. Then, the STGCNSEI model is applied to further integrate the detailed temporal and spatial features in order to elevate estimation precision. Finally, the feasibility and accuracy of the proposed method is validated via a case study of Chengdu Terminal airspace, China. Results demonstrate that the proposed STGCNSEI model has a promising application prospect.