Identificação e Controle Preditivo NEPSAC utilizando Hammerstein-Wiener e Rede Neural Recorrenteaplicado a um Sistema de Tanques de 2 Colunas

Abstract

Non-linear multiple input and multiple output (MIMO) systems represent a challenge for control- ler design. In this work, the use of a predictive controller type NEPSAC-MIMO is presented in a

problem of interest in the chemical industry: reagent level control in a two-column tank system. In this type of predictive controller a model of the system to be controlled is necessary. In order to obtain the model of the plant, a Recurrent Neural Network was used to adequately represent the dynamics of the tank system used by the controller as an application in this work. In addition, identification based on the Hammerstein-Wiener structure was also used because of its simplicity to identify the plant and compare it to the RNR. Results in simulation are presented to show the performance of the system identification using the recurrent neural network as well as the predictive controller, comparing the phenomenological model of the system and the Recurrent Neural Network model.