Agentenbasierte Gebäudeautomation raumlufttechnischer Anlagen

This thesis describes an agent based automation system for building supply systems. The described system is able to create an automation system for combinations of building supply components by Plug & Play. The thesis describes the system using an HVAC system. The system consists of decentralized control units, so-called agents. Each component or supply unit is supplied with at least one of these agents. There are two different types of agents: supply agents and consumption agents. Consumption agents evaluate the thermal demand (e.g. of an office room). Supply agents control heating or cooling devices of the building. The agents communicate with each other. They negotiate in order to evaluate the system’s most efficient operation point regarding the actual thermal demand. Supply agents can be implemented in different levels using cascaded cost functions. Thus, also complex building supply systems can be controlled using MAS. The system works independent of any centralized control system. Each operating decision is made by agents. The agents are using cost functions in order to negotiate with each other. These cost functions represent the operating costs of the components or assess monetarily the users’ comfort inside a room. Consumer agents are able to find the most cost efficient and thereby energy efficient supply unit comparing the supply agents’ cost functions. The multi agent system (MAS) is tested using a test bench for building supply systems. The test bench consists of an HVAC system and four rooms equipped with chilled beams. Furthermore, there is a simulated heating supply system which is connected to the test bench using Hardware-in-The-Loop method. The agents are modelled within JADE according to FIPA standard. The agents are able to communicate with the test bench as well as with the simulation program. The test results proof the functionality of the system. Therefore, four scenarios are shown. The agents are responsible for the control of building devices in each of these scenarios. Thereby, the components with the lowest cost functions are activated. The comfort level inside the test rooms is good during all four test runs. The results of the scenarios show, that the agents are able to work with any kind of building supply system without special configuration. The MAS allows us to create building management systems using Plug & Play components equipped with agents