The utilization of unmanned aerial vehicles (UAVs) in internet of things (IoT) communications has gained significant attention in recent years due to their ability to adapt to different positions and access areas with limited infrastructure. To support a large number of IoT devices, UAV networks have extensively integrated non-orthogonal multiple access (NOMA) technology. In this study, we propose an innovative Deep Reinforcement Learning (DRL) agent based on Proximal Policy Optimization (PPO) to improve the Energy Efficiency (EE) while considering far-near fairness (FNF) in NOMA-UAV networks. This agent is designed to simultaneously control dynamic factors such as UAV 3D trajectory, downlink transmit power, IoT nodes association, and power allocation (PA). We compare our solution with the Hybrid Decision Framework (HDF) approach through comprehensive simulations. The results clearly demonstrate the superiority of our proposed scheme over HDF in terms of effectiveness.