Abstract: A nonlinear predictive PID control algorithm is designed for the situation that the drone control performance is susceptible to wind disturbance. Firstly, the dynamic model of drone height and attitude angle directly controlled by PID parameters is established. Secondly, the NNI and NNC composite neural network is used to predict the load of the UAV at the next moment and the NMPC-PID control algorithm is designed. The comparison simulation shows that compared with the traditional PID control algorithm, the NMPC-PID control algorithm has obvious advantages in robustness and anti-interference ability. The analysis of the experimental data of the UAV hovering under actual wind disturbance shows that the UAV adopting the NMPC-PID control algorithm has the characteristics of being able to adapt to the wind speed change.