Building thermal models reduction: improving exixting methods by taking spectral imputs characteristics into account

Modal-based model reduction techniques have been modified in order to improve their performance when applied to building thermal models. The basic idea is to take the spectral characteristics of meteorological inputs into account in such reduction techniques, which are very sensitive to the hypothesis concerning inputs nature. The resulting modified modal-based reduction methods have been implemented in Matlab and tested. From the performed analyses, it may be concluded that the introduced modifications significantly improve results.