Intelligent Wireless Power Scheduling for Lunar Multienergy Systems: Deep Reinforcement Learning for Real-Time Adaptive Beam Steering and Vehicle-to-Grid Energy Optimization

Abstract

The integration of wireless power transfer (WPT) and vehicle-to-grid (V2G) technologies is essential for the sustainable operation of lunar multienergy virtual power plants (MEVPPs), where rovers, habitats, and in situ resource utilization (ISRU) facilities rely on adaptive energy management. Unlike terrestrial systems, lunar environments present extreme challenges, including long-duration night cycles, regolith dust accumulation, severe temperature fluctuations, and dynamic rover mobility, all of which disrupt efficient power delivery. This paper proposes a reinforcement learning–based adaptive beam steering framework to optimize WPT scheduling, ensuring continuous and efficient energy transmission for both mobile and stationary lunar assets. Unlike traditional fixed-beam or heuristic-based WPT methods, the proposed system utilizes deep reinforcement learning (DRL) with proximal policy optimization (PPO) to autonomously adjust beam direction, power intensity, and charging priority in response to real-time rover movements, V2G interactions, and fluctuating energy demands. The proposed framework models WPT optimization as a Markov decision process (MDP), where the agent learns to dynamically adapt beam steering based on rover speed, response delay, solar power availability, and charging station congestion. The reward function penalizes energy misallocation and misalignment losses while maximizing charging efficiency and systemwide energy resilience. A case study simulating a 30-day mission near Shackleton Crater evaluates the effectiveness of the AI–driven WPT system, demonstrating a 54.6% reduction in energy downtime and a 41.3% improvement in beam alignment efficiency compared to static power scheduling methods. In addition, the system reduces latency-induced power deficits by 39.8%, ensuring reliable power distribution for ISRU oxygen extraction, habitat life support, and rover recharging stations. This study represents a novel advancement in lunar power infrastructure, integrating AI–driven adaptive WPT with intelligent energy scheduling to enhance V2G interactions in extraterrestrial environments. The results validate the feasibility of DRL–based WPT control, paving the way for scalable, resilient, and self-optimizing wireless power grids on the Moon. Future work will explore the integration of hybrid energy storage models, quantum-inspired optimization for real-time decision-making, and predictive beamforming algorithms to further enhance the reliability and efficiency of lunar energy networks.