In view of the problem that current loading control processes of a trailing suction hopper dredger is high personnel dependence and low efficiency,prioritized experience replay (PER) and random network distillation (RND) are combined,and a reinforcement learning control algorithm is proposed on the basis of priority replay and random network distillation soft actor-critic (PRND-SAC).By designing appropriate state space,action space,and reward function,the PRND-SAC controller is compared with the traditional SAC controller.Furthermore,comparative experiments are conducted between the PRND-SAC controller and actual operational data on the basis of a full-process loading phase simulation environment.The results demonstrate that the proposed controller converges quickly and stably.Furthermore,compared to the traditional SAC controller,the PRND-SAC controller not only enhances the stability of the control process but also significantly increases loading efficiency.