Jump to Main Content
Composition–Temperature Cascade Control of Dividing-Wall Distillation Columns by Combining Model Predictive and Proportional–Integral Controllers
- Qian, Xing, Huang, Kejin, Jia, Shengkun, Chen, Haisheng, Yuan, Yang, Zhang, Liang, Wang, Shaofeng
- Industrial & engineering chemistry process design and development 2019 v.58 no.11 pp. 4546-4559
- controllers, distillation, models, process design, temperature
- Although dividing wall technology is very effective for process intensification of distillation, the resultant dividing-wall distillation columns (DWDCs) exhibit strongly interactive and highly nonlinear behaviors, which may pose great challenges to process operations. Proportional–integral (PI) controllers can stabilize the operation of the DWDC, but their performances are frequently far from satisfactory. Model predictive control (MPC) can potentially improve performance sharply, but the stringent requirements on the qualities of the embedded model considerably lower its robustness. To suppress these two deficiencies, in this work a cascaded control structure by combining the MPC and PI controllers is proposed for the DWDC. The PI controllers regulate the stage temperature in the inner layer, serving to quickly stabilize the DWDC, while the MPC directly controls the product compositions in the outer layer, addressing the strongly interactive and highly nonlinear behaviors of the DWDC. Two typical cases, that is, a three-product Petlyuk DWDC and a four-product Kaibel DWDC, and five control schemes, that is, two impurity composition controls (PI-ICC and MPC-ICC), two purity composition controls (PI-PCC and MPC-PCC), and one pure temperature control PI-TC (which represents the case wherein MPC in the outer layer fails to work), are studied to evaluate the feasibility and effectiveness of the proposed control structure. It is found that the cascaded MPC/PI controllers (i.e., MPC-ICC and MPC-PCC) substantially outperform the pure PI controllers (i.e., PI-ICC and PI-PCC), with lower maximum and steady-state deviations, shortened settling times, and suppressed oscillations. In the case of failure of the MPC, the cascaded MPC/PI controller can still maintain stable operation of the DWDC. These outcomes highlight that the combination of MPC and PI controllers can be a potential method for the operation of highly interactive and complicated DWDCs.