A new plant-wide control structure for the Tennessee Eastman process systematically obtained

GONZALO MOLINA; DAVID ZUMOFFEN; MARTA BASUALDO

Rosario, Argentina

Congreso; RPIC 2009; 2009

Universidad Nacional de Rosario

In this work an approach for addressing the problem of designing a plant-wide control structurein a systematical way is presented . The goals of this control structure are: keep the process at the optimum energy consumption and reject the disturbances. The main novelty of this methodology consists on the metric used for searching the controlled variables and the way in which it is implemented in genetic algorithm for systematically find the most suitable set without accountingany heuristic concept. Therefore, in the last step of this approach, a combinatorial optimization problem is successfully solved, even though the high dimension involved in large scale chemical plants. It is tested in a well known process, the Tennessee Eastman (TE) plant proposed by Downs and Vogel [6]. It It must be remarked that only steady-state plant information is usedfor selecting the control structure. The methodology begins choosing only variables with no steady-state effects and active constraints to be controlled. It follows with the decision of the most suitable inventory control structure in order to achieve the plant stabilization. Finally, the remaining controlled variables (CV’s) are selected using the metric cited before, based on  inimizing the sum of the square errors (SSE) defined for specific uncontrolled variables. In this last  tage, a good tradeoff between servo and regulator problem is achieved together with high  erformance on self-optimizing properties. Finally, a set of dynamic simulation results are given in  rder to demonstrate the potentiality of this approach.