Online reinforcement learning in multi-agent systems for distributed energy systems

Abstract

Researchers have been taking efforts to reduce the dependency on distributed generators, due to high fuel cost and problems associated with the depletion of non-renewable energy sources. Hence, it becomes inevitable to formulate techniques, which can utilize alternate energy sources and capable of meeting power demands in a cost effective manner. The concept of Multi-Agent Systems (MAS) is novel in taking intelligent decisions in place of manual operations and thereby ensuring greater operational efficiency. MAS offer a range of benefits like flexibility, autonomy, less maintenance, reduced cost and so on. The primary objective of this paper is to develop an intelligent MAS that emulates the real-time operation of a distributed energy system. It also aims at implementing an artificially intelligent learning algorithm, which can aid the autonomous behavior of the multi agents without any human intervention. MAS are designed to accommodate decision-making modules as well as learning mechanisms based on evolutionary computation. These techniques increase the intelligence of the MAS.