Probabilistic Predictions of Air Traffic Demand for Airspace Sectors Based on Timing Predictions of Individual Flights

摘要

The Federal Aviation Administration (FAA) uses the Traffic Flow Management System (TFMS) to track, predict and plan air traffic flow. TFMS predicts the demand for each sector, and determines alert status by comparing demand and capacity at each 15-minute interval of the time period of interest. Traffic managers use these predictions to identify possible congestion areas and to take measures to prevent it. Current TFMS provides deterministic demand predictions by aggregating the flights whose Estimated Time of Arrivals (ETAs) or Departures (ETDs) fall within a time interval of interest without considering uncertainty (random errors) in flight ETA or ETD. Those errors contribute to the uncertainty in aggregate demand count predictions. This paper presents an analytical approach and techniques for translating characteristics of uncertainty in predicting the times for individual flight events into characteristics of uncertainty in predicting one-minute sector demand counts. Characteristics of uncertainty in individual flight times predictions, such as sector entry times and time in sectors, were estimated via statistical analysis of TFMS data separately for airborne and proposed flights. The paper shows that expected one-minute sector demand predictions are determined by a probabilistically weighted average of one-minute demand predictions within a sliding time window that contains several consecutive one-minute intervals surrounding the interval of interest. The width of the window and the weight coefficients are determined depending on probability distributions of errors in flights' timing predictions. Expected one-minute sector demands along with standard deviations of demand counts allow for probabilistic predictions of sector demand, which, in turn, allow for probabilistic predictions of sector congestion.