| Popis: |
Anticipatory control methods can enable the participation of buildings in demand-side management. Model- based control techniques offer promising avenues to improve the energy efficiency of buildings. Grey-box model identification can be carried out through optimization and filtering techniques. In this work, these two techniques were compared for the identification of a second order RC grey-box model using open- loop input-output data. It was found that, generally, the filtering method identifies the model over a shorter time period as compared to the optimization technique. Moreover, using the filtering approach, the poor initial state guesses can be taken care of, while it can be problematic to deal with this issue in an optimization-based identification method. We show then that the model may not necessarily be identified accurately but may still produce accurate simulation results. Finally, we also determine the extent to which the identified parameters drift away from their true values with increasing noise intensity in the identification data. |
