Maritime confrontation mission planning is a difficult task planning problem. The difficulty of planning stems from the dynamic characteristics and high complexity of the maritime environment. These factors make it extremely difficult to describe the state and refine the decision level. Traditional task planning methods based on rules or optimization algorithms suffer from limitations in dealing with these problems. They are not adaptable to dynamic environments, and the decision level is too rough. This paper proposes an Advanced-PER based Deep Reinforcement Learning Method based on double deep Q-Network (DDQN) for complex maritime confrontation task planning problems, and a new Prioritized Experience Replay method is introduced to improve the deep reinforcement learning method. The model significantly improves the learning efficiency and decision-making performance of the model by partitioning the storage experience and dynamically balancing the adoption ratio of different action experiences. Experimental results indicate the effectiveness of the proposed method.