Data-Driven Modeling of a CO2 Refrigeration System

This paper describes a data-driven method for system identification of a CO
₂
refrigeration system. Traditionally, the interaction between the measured variables is not utilized as they are highly dependent on the refrigeration system. In this work a data-driven method, namely subspace identification, is investigated for deriving a control-oriented model such that the dynamic interaction in the refrigeration systems can be utilized for e.g. fault detection and diagnosis. The subspace identification is applied on laboratory data obtained from a test setup located at NTNU in Trondheim, Norway. The obtained results offer promising perspectives for performance improvement in fault detection and diagnosis methods as well as control strategies.