Identification of room temperature models using k-step PEM for Hammerstein systems

The contribution focuses on the identification of short term room temperature prediction models using k-step ahead prediction error minimization (PEM). In the first part of the paper we describe the general identification problem and summarize the k-step PEM-algorithm for discrete time Hammerstein systems. In the second part the k-step ahead approach is applied to measured data of a conference room and compared to the standard 1-step PEM approach. The results indicate, that k-step ahead approach can indeed lead to better prediction performance in this particular practical application.