Investigation of Choice of State Space for Reinforcement Learning Based Battery Scheduling in Microgrid

This paper models the battery scheduling problem in a microgrid as multi-stage decision process. A Reinforcement Learning(RL) algorithm named Q-learning is used to solve this problem. Even though RL has been applied for scheduling of Microgrids, there has not been any study investigating the choice of state space. Appropriate choice of state space can reduce the computation time and help in finding the adequate schedule. In this paper, we investigate the choice of state space and action space.