Model predictive control of a petlyuk distillation column for ternary separation: Aspen-Matlab cosimulation

Petlyuk type distillation columns are a promising technology in multicomponent separation that allows for efficient energy consumption. However, further research is needed on advanced control, such as Model Predictive Control (MPC). This work delves into the study of MPC in Petlyuk distillation columns of ternary mixtures. The columns are designed in Aspen Plus© V8.8 and the control in Matlab© 2013a, in this way the operation of the controlled columns is carried out by cosimulation "matlab calls aspen". The manipulable and controllable variables are chosen according to the methodology used for thermally coupled columns given in the literature. The MPC uses a linear dynamic model derived from the nonlinear dynamic model simulated in Aspen Plus Dynamics. The MPC results are contrasted with the results of proportional-integral (PI) control, under the same disturbance scenarios. Several case studies are analyzed (experimental plan with 60 tests) perturbing up to six variables at a time. During these dynamic controllability runs, responses such as maximum stability time, maximum overshoot, and sum of squares of error are measured and compared. The results show that MPC is more robust compared to PI control, since it manages to control each of the 60 study cases, while PI control stabilizes only 27 runs. The MPC is robust even when subjected to excessive perturbations in equipment of very high mathematical and operational complexity (the Petlyuk-type column), proving to maintain stability within the study region.