The Internet of Things (IoT) depends primarily on low-cost wireless sensors with limited energy capacity to allow pervasive monitoring and intelligent control. Nevertheless, unmanned aerial vehicle (UAV) can be used to connect remote terminals that are outside wireless coverage to IoT networks. This solution provides a means of extending the reach of IoT networks, offering more opportunities for monitoring and control. Despite this benefit, the UAV also suffers from low capacity onboard battery. To overcome these problems, solar energy is integrated with UAV, and wireless-powered communication (WPC) techniques are used for IoT terminals. Also, the non-orthogonal multiple access (NOMA) technique can be employed to address the massive connectivity issue of IoT terminals. By leveraging these advantages, we jointly optimize the three-dimensional UAV trajectory and time allocation for WPC powered IoT devices (IoTDs) underlaying solar-powered UAV. To achieve the target, in this paper, introduces a multiagent federated reinforcement learning (MAFAL) algorithm, which concentrates on maximizing energy efficiency (EE) while minimizing energy consumption, guaranteeing quality of service (QoS), fairness, and trajectory planning. The proposed algorithm aims to optimize the overall performance of the system by learning from the collective experience of multiple agents. Simulation result demonstrated that the proposed method achieves 56.84%, 68.45%, and 73.63% higher EE as compared to MAD2PG, DDPG, and DQN, respectively.