Monitoring and Alerting Congestion at Airports and Sectors under Uncertainty in Traffic Demand Predictions

Abstract

Important functions of the Traffic Flow Management System (TFMS) include predicting air traffic demand for National Air Space (NAS) elements (airports, fixes and en route sectors) for several hours into the future, and using these predictions to alert traffic flow management (TFM) specialists to potential congestion when predicted demand exceeds available capacity. The current TFMS Monitor/Alert functionality uses deterministic predictions, neglecting their stochastic nature. This paper focuses on improving the accuracy and stability of traffic demand predictions for airports and sectors by considering the uncertainty in aggregate demand count predictions. The emphasis is on uncertainty caused by errors inherent in TFMS during processing flight data not affected by future air traffic control. We propose a constructive approach for improving aggregate demand predictions under uncertainty based on linear regression that includes TFMS demand counts for several adjacent time intervals within a sliding time window. Numerical examples based on TFMS data showed that the regression models produce more accurate (up to 22% reduction in the standard deviation of errors in demand predictions) and more stable (fewer crossings of the Monitor/Alert threshold) predictions than current TFMS predictions. For airports, regression significantly reduced the total number of missed alerts (21%) with a small increase in the total number of false alerts (3%). For sectors, the reduction in missed alerts was 22%, with an increase in false alerts of 8%.