Development of an Agent-Based System for Decentralized Control of District Energy Systems

Abstract

This contribution introduces a new system for distributed model predictive control of energy systems. This system uses multiple agents where each agents optimizes a subsystem. A central instance coordinates the individual agents and takes care of feasibility of the combination of the single solutions. The advantages of this approach are increased maintainability and privacy for the individual agents, thus increasing applicability to real-world systems where often multiple parties are involved in a single energy system. Where adequate, the agents perform MPC to control their subsystems. The system and method can be chosen for each agent individually. In order to build this system a framework is developed as existing frameworks lack one or more required features for the overall system. A small example is presented together with first results.