A building energy management system based on distributed Model Predictive Control

The emergence of information and communication technologies (ICT) in the building integrates them as important consumer-actor players in smart-grid. They then have the potential to shift and to reduce energy consumption or even to store it with various storage capacities: thermal storage, hot-water tank and also electrical battery. But to ensure their efficiency, it is necessary to develop building energy management (BEM) systems which can interact with the grid and control the building and its systems. In this paper, a BEM system based on distributed predictive control is proposed. The idea is to schedule the actions of the various controllable systems to minimize the energy cost while maintaining the occupant comfort and systems constraints. This scheduling is based on the knowledge of the future data profiles as well as the future cost of energy. The cost reduction is ensured by means of the building storage capacities and by shifting the house consumption periods if the future price is high. Each building is different from another, because of its construction, its systems and its occupants. Consequently, BEM systems have to be modular. This point is ensured by its distributed architecture: one agent is dedicated to each controllable system, and a coordinator agent ensures an optimized global behavior.