<p>Ship-mounted crane is a complex underactuated system and works in a complex marine environment, which makes the control of ship-mounted crane a great challenge. The control method based on accurate model often has poor effect in the face of system uncertainty. Reinforcement learning is an effective learning method to obtain the optimal control strategy by interacting with the environment, so as to solve the target control problem. Based on the model-based reinforcement learning, this paper proposes an optimal control method. Firstly, each state space of the ship-mounted crane is quantified according to the prior knowledge and combined with fuzzy control, these quantized state quantities are fuzzy divided, and finally the optimal control strategy is obtained by approximate value iteration. In order to speed up the learning process, the reward function and evaluation system are specially designed. Simulation results show the effectiveness of the proposed controller in dealing with unmodeled periodic disturbances.</p>