Bauchspiess, A.; Ishihara, J. Y.; Felgner, F.; Litz, L.:
First-principles structured identification for predictive HVAC control,
Proceedings of "XII. Congresso Latino-Americano de Controle Automático (CLCA)", Salvador da Bahia, Brazil, 2006.

This paper presents a first-principles structured identification approach for thermal environments equipped with Heating, Ventilation and Air Conditioning (HVAC) systems. Linear data-driven identification and first-principles modeling are combined to produce an accurate and computationally efficient model. The objective is to find out which model is best suited for energy-saving and comfort control strategies. A proportional integral (PI) controller running a conference room at the University of Kaiserslautern, using PMV (Predicted Mean Vote) comfort index, is quite satisfactory as regulator, but it is not anticipatory, wasting energy and lasts long to reach the comfort zone when room utilization changes. A model-based predictive controller is anticipatory and can cope with actuator saturations. The first-principles structured model allows the separation of the different heat flow contributions simplifying the parameter identification. The theoretical foundations and experimental results are presented considered heating, cooling, outside temperature, neighboring rooms and solar radiation.

Keywords: thermal building modeling, HVAC modeling, identification, first principles.