Model Selection for the Control of Temperature in a Centralized Air Conditioning System

Abstract

In the building sector, air-conditioning systems contribute a very high percentage of the total energy consumption. The increased focus on energy efficiency and savings has resulted in the need for the use of intelligent and efficient controllers. These controller designs would require accurate models that describe the dynamics of the system. However, studies of modelling a centralized multi-circuit air-conditioning system are scarce. This work presents a study on the selection of a dynamic model using system identification approach for an operating centralized multi-circuit water-cooled package unit air-conditioning system. Black box linear parametric models such as the Autoregressive with exogenous terms (ARX), Autoregressive-moving average with exogenous terms (ARMAX), Box Jenkins (BJ) and Output-Error (OE) were used to process, estimate and validate the collected data. The analysis showed that the Autoregressive-moving average with exogenous terms (ARMAX) of the third order model, demonstrated the best model structure with the highest Best Fit and Lowest Mean Square Error.