Focusing on the problem of encountering the moving obstacle for the water quality monitoring platform in the autonomous navigation,a new dynamic obstacle avoidance method based on the obstacle motion prediction model and the velocity obstacle avoidance model was proposed. Firstly, the ultrasonic ranging module and image acquisition module were used to measure the distance and the azimuth angle between the obstacle and the platform, and then the obstacle's speed and direction were calculated using the coordinate transformation method. Secondly, obstacle motion prediction model was built based on maximum likelihood estimation method, and obstacle speed and direction of the next sampling instant were obtained by this model. Finally, the platform's course angle of the next sampling instant was calculated using the velocity obstacle avoidance model. The experimental results prove that the proposed obstacle avoidance method can plan out a more realistic and optimal path. Compared with the obstacle avoidance method without the obstacle motion prediction model, the proposed obstacle avoidance method could improve the success rate of obstacle avoidance.