During the actual operation of flight schedules, the change of wind direction affects the arrival time of flights at the shared waypoints in the terminal area, which in turn causes capacity overload or capacity waste. Therefore, flight schedule is adjusted based on statistical probabilities of wind direction, with the aim of developing a flight schedule that can reduce to some extent the capacity overload or waste of shared waypoints. The concept of a benchmark wind direction is proposed based on the impact of wind direction on the allocation of runways for departing flights. Using the probability of the benchmark wind direction at the airport for each month in the past five years during the flight season, the probability of the benchmark wind direction for each month in the next year is predicted. Based on the similarity of the probabilities of the benchmark wind direction, the months of the next year are clustered using the sum of squared errors and silhouette coefficients as clustering indicators. On the basis of the clustering results, a flight schedule optimization model considering wind direction uncertainty is established, and the -constraint method is combined with an improved particle swarm algorithm to solve the multi-objective model,which is called -constraint-PSO combination algorithm. The departure flights from the Beijing terminal area are used as the research object for verification. The results show that compared with the initial flight schedule, the maximum value of the hourly flow of shared waypoints decrease by 12%, and the variance of the shared waypoint flow at different benchmark wind directions decrease by 49% and 56%, respectively. Compared with the linear weighting method, this method can reduce the total number of overflow flights at shared waypoints by 70%. Research results indicate that under uncertain wind conditions the model can to some extent achieve a more balanced flow of traffic at shared waypoints, reducing occurrences of capacity overload or waste at these waypoints.