Fusing data in adaptive agent control systems for electrical grids

Abstract

This paper reports the results of Monte Carlo study of a microgrid agent scenario in which n interconnected microgrids cooperate in detecting the stability condition of a shared feeder line from the main electrical grid. Agents sample the stability behavior of the feeder line, rate the observed patterns, and use votes to express and communicate their individual inferences about system stability. Agents sent their votes through a centralized or decentralized network to be processed. In the centralized network scenario, one agent receives votes from all other agents and aggregates them into a collective inference about feeder line stability. In the decentralized network or peer-to-peer (P2P) scenario, each agent sends its votes to every other agent and each agent aggregates received votes to make a collective inference about feeder line stability.