Robust Decentralised Agent Based Approach for Microgrid Energy Management

This paper introduces a decentralized agent based implementation of an off-line optimization technique applied to the configuration of electrical microgrids with distributed generation. This implementation conforms the ''Agent Based Energy Management System'' (ABEMS). Software agent principles are used to decentralize the microgrid control without having an implicit or explicit control hierarchy. Unlike most approaches, they work as peers. Orders to be sent to the energy micro-generation units are produced off-line by an evolutionary computation method. The orders are encoded into the agents once the micro-grid is started and agents decide when to apply them. The paper includes experiments that show the resulting behavior in several scenarios executed in real time. Results indicate the technique is stable enough and outperforms more simple approaches, a necessary initial step prior to a more comprehensive comparison with other solutions.