Statistical modelling and forecasting schemes for air-conditioning system

Abstract

System identification is a procedure to characterize the dynamic behaviour of a system, a sub-system or individual components based on measured data. This thesis presents a study on the modelling and parameter identification of air-conditioning processes by using the mathematical black-box modelling technique -AutoRegressive Moving Average eXogeneous (ARMAX) structure. A generic Multiple Input Multiple Output (MIMO) ARMAX structure has been developed for typical air-conditioning systems and the Recursive Least Square (RLS) method is used for model parameters identification. The performance of the model developed is compared with that of the Single Input Single Output (SISO) ARMAX model. Different combinations of ARMAX orders and forgetting factors are examined for the SISO and MIMO models in order to evaluate the optimum settings that maximize their accuracy. Measured data from an air-conditioning system in an existing building were used to test and validate the SISO and MIMO models. Using the MIMO ARMAX structure, a forecasting scheme for the air-conditioning system has been developed. The technique of conditional expectations at one step ahead is explained in the scheme development. The forecasting scheme is further applied to examine the system performance and predict the cooling load by a day-ahead forecast. The ARMAX model proves remarkably to examine the dynamic performance of an air-conditioning system. In addition, with the use of field data, the model shows great potential for system performance forecasting. An advanced control strategy for the air-conditioning system based on ARMAX model is also proposed.