Available air channel capacity prediction by weather-capacity Graphical Neural Network(wcGNN)

The available air channel capacity ratio is a major concern of air traffic control. Predicting channel capacity from weather condition is an urgent requirement by aviation management and airport.

To analysis and predict capacity variation with meteorological factor, this research builds a capacity analysis and prediction system, which uses radar and satellite data to analysis real-time capacity status, and numerical model forecasting data to predict capacity change.

To model multiple meteorological variables to flight flow, as well as the complex interactions between air channels, we propose a novel weather-capacity Graphical Neural Network (wcGNN) to analysis data relativity. The prediction from multiple model are also assembled to further increase prediction accuracy. Eventually, our work outperforms previous researches as well as human expert prediction results.