A protection scheme for AC microgrids based on multi-agent system combined with machine learning

Traditional protection schemes at the distribution level designed for unidirectional power flow will be compromised due to bi-directional flow of power with the increased penetration of distributed generation (DG) sources, resulting in miscoordination between protection devices. This paper proposes a new microgrid protection method based on the multi-agent system (MAS) combined with machine learning (ML) for fault detection in autonomous and grid-connected modes, protection coordination and updating relay settings to achieve adaptive protection. MAS framework with various layers and roles of each agent are described in detail. A meshed microgrid model is developed in Simulink to collect fault data for training and testing ML algorithms, while the behaviour of individual agents and interactions between them are validated in AnyLogic simulation software. The simulation results confirmed that the proposed MAS algorithm could provide primary and backup protection in both modes of microgrid.