THE EFFECTS OF OPERATIONAL AND DESIGN CHARACTERISTICS OF CATALYTIC CRACKING REACTORS ON THE CLOSED-LOOP PERFORMANCE OF LINEAR MODEL PREDICTIVE CONTROLLERS

Abstract

In closed-loop simulations of the catalytic cracker, it was observed that the model identified at a low throughput operating point, when used in MPC did a better job on the high throughput operating point than a model identified at the latter operating point. At the same time MPC utilizing a model identified at a high throughput operating point did not perform well when used at the low throughput operating points. In this paper we examine this behavior through an analysis of the closed-loop catalytic cracker under unconstrained MPC. The persistent appearance of an upward spike in stack carbon monoxide concentrations immediately following a disturbance, despite extensive tuning efforts suggests that it is due to process rather than control system design characteristics. At high throughput operating points this phenomenon is of concern because it seriously impairs the ability of a control system to keep CO below its upper constraint. In the paper we trace the source of this dynamic behavior to the design of the catalytic cracker and suggest modifications that may favorably affect control performance.