CONICET | Buscador de Institutos y Recursos Humanos

In this work a methodology for optimal decentralized plant-wide control (PWC) design is proposed. The overall procedure can be represented by a combinatorial problem where the multivariable control alternatives were adequately parameterized. The main PWC tasks addressed by the suggested approach are the following: 1- optimal selection of controlled and manipulated variables, 2- input-output pairing, and 3- controller sizing. The functional cost quantifies the deviations on manipulated or uncontrolled variables in a sum of square sense (Frobenius norm) by using internal model and perfect control theories. Thus, the methodology suggested here providesa systematic and generalized tool for PWC design, minimizing the heuristic considerations. In thiscase, the multi-objective optimization was efficiently solved by genetic algorithms based on binary decision variables. The well-known Tennessee Eastman (TE) process is selected here for testing the methodology.