In this article, we propose a novel solution: a nonorthogonal multiple access (NOMA) maritime network employing power allocation facilitated by federated multiagent deep reinforcement learning (DRL).