Q-Learning based Protection Scheme for Microgrid using Multi-Agent System

Abstract

Distributed Energy Resources (DERs) such as Distributed Generators (DGs) or storage systems can be integrated with the central power distribution system. However, one of the severe challenges posed by the penetration of DGs to the utility grid system is the bi-directional power flows in the feeders. The bi-directional energy flows cause issues pertaining to the failures of protection systems because usually relays are designed to protect the network under unidirectional power flow case. Therefore, it is essential to have a robust protection scheme in support of distributed generators to protect the system from various faults. This paper proposes a novel protection scheme based on Q-learning and multi-agent system that identifies and isolates the different types of failure. The Q-learning algorithm is built to teach agents for making responsible decisions in fault identification and clearing. Furthermore, decentralized Blockchain based connections were adopted for exchanging information between agents. The system was simulated in MATLAB and JADE platforms and the results have shown that system is capable to identify different type of faults based on various states.