Shipboard power system reconfiguration using reinforcement learning

Abstract

In this paper we deal with shipboard power system (SPS) reconfiguration with an integrated power system (IPS) and distributed generator. The objective of reconfiguration is to determine the status of the switches such that power is delivered to the vital loads even under fault conditions. We have used a model-free reinforcement technique called Q-learning for solving the reconfiguration problem. We don't only get the final configuration but also the sequence of switches to open and close to reach the final state. To the best of the author's knowledge this is the first time reinforcement learning is being applied to shipboard power system reconfiguration.