Multi-Agent based Distributed Optimization Architecture for Energy-Management Systems with Physically Coupled Dynamic Subsystems

Abstract

This paper introduces a special architecture for realizing multi-agent based distributed optimization for large scale dynamic systems. By combining advanced environments for multi-agent implementation and fast non-linear dynamic optimization according to a newly devised Mediator-based architecture, optimal operation of large scale power and energy management systems becomes possible. The architecture allows overcoming typical problems like convergence issues, especially high exchange of information between agents due to tightly coupled physical subsystems. Two application examples are introduced and considered schematically: first application is a Building EMS for energy optimization by controlling A/C and other energy consuming devices inside a building, and the second application is a Mediator based distributed energy matching system (for demand side). In the first application, a distributed model predictive control strategy is deployed while dividing the system (i.e. Building EMS in this context) into separate independent yet coupled subsystems. In the second application, the whole demand side is divided into several aggregators where these aggregators perform distributed energy matching within them while having their local energy optimized. In both of these applications, Mediators are introduced to occasionally intervene into the optimization. Multi-agent framework is deployed for performing the distributed optimization in both of these applications.