In order to predict the flight parameters of UAV engine, a vector auto regression integrated moving average (VARIMA) model with multi-parameter correlation was established. The model takes the UAV flight data as the input, uses the simulated annealing algorithm to optimize the VARIMA model parameters, constructs the data pattern of multi-source flight parameters association, and uses the constructed data pattern to realize the prediction of state parameters. The flight data of multiple flights of UAV are selected for the experiment. The experimental results show that the optimized VARIMA prediction model has good prediction performance, and the prediction time is saved by 0. 23 s compared with that before optimization.