NONLINEAR MODEL-PREDICTIVE CONTROL OF A FIXED-BED WATER-GAS SHIFT REACTOR - AN EXPERIMENTAL-STUDY

This paper describes new results on the experimental application of no nlinear model-predictive control (NMPC) to a fixed-bed water-gas shift (WGS) reactor. The development and experimental validation of an appr opriate first-principles WGS reactor model, and how it impacts control ler performance is discussed. The implementation of NMPC is computatio nally intense, requiring that a large nonlinear program (NLP) be solve d at each sampling period. The significant computational burden dictat es that a relatively slow sampling rate be used. Infrequent sampling, however, diminishes disturbance rejection capabilities. To combat this problem, NMPC was implemented in a master-slave cascade configuration where a low-level liner controller, having a significantly faster sam pling rate, was employed. The control study was performed using a PC-b ased distributed control system (DCS) . One of the three processors wa s dedicated to NMPC calculations. A complete and rigorous implementati on strategy is described in the paper, and the performance of NMPC for set-point tracking of this nonlinear process is shown to be superior to adaptive or linear control. We also illustrate the ease with which NMPC accommodated feedforward control.