Distributed Real-Time Power Management in Microgrids Using Multi-Agent Control with Provisions for Fault-Tolerance

This paper presents a distributed real-time control scheme based on multi-agent systems for cost optimisation of a micro-grid using real-time dynamic price estimation. The real-time prices are forecast using realistic UK energy price data via a Markov Chain Monte Carlo algorithm. A backup mechanism for main containers of the agent platform is implemented to improve fault tolerance of the control system, addressing the single point of failure problem at the hardware and software levels. The Multi-Agent system developed in JAVA and run with Raspberry Pi controls a simulated microgrid in an OPAL-RT real-time simulator to test the accuracy of the estimation method, the capacity of the control to realise power management at minimal supply cost, and uninterrupted operation in case of container faults.