Abstract: This paper addresses the issue of energy interaction between offshore fish farms and islands with surplus wind and solar resources by developing an energy transportation strategy involving fully electric ships for the island-fish farm-coast system. The proposed strategy utilizes deep reinforcement learning, which is well-suited to managing the uncertainties of offshore wind and solar resources and can accommodate large-scale energy transfer models. First, the mobile energy storage battery group is detailed into fully charged, unloaded, and partially charged batteries. Then, the energy transportation problem is modeled as a Markov Decision Process with a hybrid action space. To solve the hybrid action space issue, a parameterized dual deep Q-network based on multi-batch forward propagation is proposed. This method decouples the unrelated discrete and continuous actions using a multi-step forward pass strategy, reducing volatility during the agent's training process and converging to a more optimal solution. Finally, simulation results verify that the proposed strategy effectively facilitates energy transfer between locations. Compared to traditional deep reinforcement learning methods suited for discrete action spaces, the proposed algorithm demonstrates greater flexibility and achieves superior performance in the target scenario. Additionally, comparative analysis in expanding model scales further validates the advantages of the proposed method in addressing large-scale energy transportation challenges.