Development of control-oriented models for a building under regular heating, ventilation and air-conditioning operation - a comparative simulation study and an experimental validation

Abstract

The development of models is a major barrier to the fast and widespread adoption of model predictive control for building HVAC systems. This paper proposes the subspace identification technique, refined through the prediction error method, to quickly obtain a model for the accurate indoor temperature prediction, even with little identification data, even in the presence of large unmeasured disturbances and noisy identification data, and even using data which was collected during the regular HVAC operation of a building. The identification issues associated with grey-box models were thoroughly investigated. In particular, the development of a grey-box model was found to be a complex, lengthy and computationally intensive process, even for a single-zone building, and the models were not physically meaningful. The proposed method was found to be much easier and faster, with a potential for direct practical application. Analysis on experimental data from an existing building provided promising results.