Publication: Steady state based adaptive model predictive control for optimized ethylene glycol reactor
Loading...
Date
2021-11-01
Authors
Sulaiman, Muhammad Syafiq
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Ethylene glycol (EG) production is important for automotive, chemical, pharmaceutical and cosmetics industries to be used as coolant, antifreeze, and reactants for final products. To this end, process control is significant to EG production process. Conventional controllers such as Proportional-Integral (PI) and Proportional-Integral-Derivative are less effective to provide good performance and robustness. This leads to a superior alternative control approach for EG reactor control. In this work, model predictive control (MPC) and adaptive MPC (aMPC) were designed to control the hydrogenation process of dimethyl oxalate to EG in a plug flow reactor (PFR). Prior to that, a PFR model was developed using Aspen Plus
software. The reactor model was compared to literature and the results showed that the error obtained for EG was 6.63% while the error obtained for methanol and hydrogen were only 0.10% and 0.13%, respectively. The simulated model had the accuracy of 93% which shows the acceptability of the reactor model. Using the developed model, sensitivity analysis was performed. From the analysis, EG production and reactor temperature were selected as the controlled variables with hydrogen flowrate and coolant flowrate as manipulating variables. Subsequently, two optimization approaches which are single objective optimization (SOO) and multi objective optimization (MOO) were implemented and the results obtained are compared. The EG flowrate and energy consumption results obtained from SOO and MOO were 173.61 kmol/hr, -7258262 kJ/hr and 111.75 kmol/hr, -2100010 kJ/hr, respectively. The results of EG hydrogenation reactor optimization show that MOO produce better results compared to SOO with an overall improvement of 30.92%. To ensure the optimum results are followed, MPC and aMPC were designed and implemented. Their results are compared with PI and decoupled PI (dPI) controls. The hydrogen flowrate and coolant flowrate were selected as manipulated variables while the EG flowrate and reactor temperature were chosen as controlled variables (CV). State-space model was developed as the prediction model with the best fit for CV1 and CV2 are 91.51% and 83.34%, respectively. The set point tracking results showed that aMPC has the best set point tracking for EG production while dPI has the best set point tracking for reactor temperature. However, aMPC had the best overall performance. As for disturbance rejection, the analyses showed that dPI is better in controlling disturbance rejection. Meanwhile, the robustness test showed that aMPC control has the highest degree of robustness on tracking CV1 changes while dPI for CV2. Despite that, the overall analysis results in dPI having the most robust control. Based on the overall integral time squared absolute error (ITAE)
results for set point tracking, disturbance rejection and robustness test, dPI has the best response with the value of 74541.6 followed by aMPC, PI and MPC with the value of 84605.9, 85657.3 and 97946.0. The results showed that the MPC developed had poor performance compared to dPI and PI. On the contrary, aMPC showed exceptional control performance for CV1 but with a degraded control performance for CV2 which is caused by the boundary constraint and limitations of the state-space model. Based on the primary focus and objective of the reactor, which is on the production of EG, aMPC was found to be the best control scheme in controlling EG hydrogenation reactor.