Fractional order models for system identification of thermal dynamics of buildings

Abstract

This study presents a class of fractional order models for system identification of thermal dynamics of buildings. Fractional order dynamics has been found to be inherent in the nature of heat transfer problems. It is thus instinctive to use fractional order models to describe the overall thermal dynamics of a building. Besides, fractional time series modeling is known by its long memory effect and capability of representing high-order complicated models in lower-order and compact forms. The reduction of model parameters can then relieve the computational overhead in the system identification procedure. This is of particular significance in model-based predictive control for building energy efficiency. In particular, a fractional order autoregressive model with exogenous input (FARX) is formulated and a corresponding parameter estimation using least squares technique is also provided. Furthermore, the FARX model is validated using simulation data from a detailed model built via IES<VE> software and compared with the prediction using traditional ARX model. It is found that the FARX model can reduce the computational time largely while retaining the prediction accuracy.