Behavior Modeling and Auction Architecture of Networked Microgrids for Frequency Support

As intermittent generation profiles introduce more stress due to the sudden imbalance in supply and demand, power systems are becoming more dependent on online aggregated support of networked utility-independent private microgrids. During the aggregation process, microgrid operators' preferences and bidding behaviors are priority. New aggregation approaches are required considering the individual behaviors of the networked and geographically dispersed independent microgrids. In this paper, we present a novel bidding behavior modeling and an auction architecture consisting of a central aggregator and networked microgrid agents. The bidding behavioral states of the microgrid agents are formalized as partially observable Markov processes for the belief updates and short-term policy determination in order to maximize the individual profit. A reverse auction model is adopted to enable competitive negotiations between the central aggregator and networked microgrid agents. The auction and aggregation processes were implemented in a power system control area along with an automatic generation control (AGC) scheme to contribute frequency control. The proposed AGC and auction mechanism were verified with an industrial multi-agent framework in a laboratory-based real-time application.