Optimal operation of a divided-wall column with local operating condition changes

The aim of this paper is optimal operation of a divided-wall column (DWC) based on self-optimizing control (SOC). By now, the proposed SOC methods have been based on linearization of the process. The novelty of this paper is to overcome this shortcoming of the local optimality of SOC. Theoretically, changes in optimal sensitivity matrix from nominal design, due to changes in operating condition, make SOC deviate from steady state optimality. These deviations from optimal operation, in already available SOC structures, have to be counteracted by optimization layer in the control structure hierarchy which involves solving a large nonlinear optimization problem online. The proposed method in this paper solves this problem with modeling optimal sensitivity matrix with Takagi-Sugeno fuzzy inference. This fuzzy inference system is tuned offline. The proposed method is dynamically validated and compared with conventional SOC. The results showed that conventional SOC had high value of loss and deviated from optimal operation. However, in the same operating condition, the proposed method with the aid of Takagi-Sugeno fuzzy inference system, which involves online calculation of weighted average of some linear function, imposed small loss, made DWC track optimal trajectory and removed the need for online solving large nonlinear optimization problem.