An application of hybrid automata for the MIMO model identification of a gasification plant

Abstract

In this paper hybrid automata have been used in order to identify the more suitable model of a gasification process, at the occurrence of variations of plant's operative conditions and specifications. From the identification process performed on historical series at disposal and from the model validation phase it was concluded that one single model was not sufficient to assure good predictions, especially in the case of large prediction horizons. This called, from a methodological point of view for the application of a hybrid model, in particular of hybrid automata in order to guarantee best performances in the choice of the best model. Performances of the proposed hybrid model are discussed and compared with the ones obtained with a fuzzy supervisor based on the same linear models.