Integrating Human-in-the-Loop for Safe Reinforcement Learning in Optimal Operation of Community Energy Storage Systems

VAN HAI BUI, Guilherme Vieira Hollweg, Wencong Su

Track: New Research

Categories: Energy Generation and/or Transmission

Keywords: community microgrids, energy storage systems, optimization, power and energy system, safe reinforcement learning.

Abstract: Reinforcement learning (RL) is becoming a potential solution for solving optimization problems in power and energy systems. However, a major issue with conventional RL is that it does not guarantee the safe operation of critical infrastructures such as microgrids or power systems. Therefore, this paper proposes a safe RL-based optimization framework with a human-in-the-loop approach for the operation of a community energy storage system (CESS) in community microgrid (MG) systems. The proposed framework not only maximizes the CESS’s profit but also reduces the amount of load shedding in the MG during emergency situations. To demonstrate the effectiveness of the proposed framework, safe Q-learning is implemented to optimize the operation of the CESS with human input, aiming to avoid all catastrophic actions at critical states.

Submission Number: 2