Abstract: From the perspective of the air traffic network flow system, the large-scale flight delay problem is converted into the congestion problem of the air traffic network flow system, and its congestion prediction is performed. Firstly, the relationship between capacity, flow and flow demand of the air traffic network flow system is analyzed, and the four forecast indicators including flow demand, cumulative delay flight change rate, capacity change in the previous period and capacity change in period are extracted using congestion degree as a congestion evaluation index. Then based on genetic algorithm(GA) optimization back propagation(BP) neural network, a congestion prediction model is established to predict the degree of congestion. Finally, the actual data of an air traffic network flow system is used to analyze examples. The results show that the model has a good prediction and can provide a decision basis for the air traffic flow management department.

Key words: air traffic network flow system, congestion prediction, genetic algorithm(GA), BP neural network