(278h) Model Predictive Control of Post-Combustion CO2 Capture Process Integrated with Supercritical Pulverized Coal Plant

Qiang Zhang, Debangsu Bhattacharyya, and Richard Turton

Department of Chemical Engineering, West Virginia University,

Morgantown WV, 26506, USA

The process of CO₂ capture and storage is a promising and effective way to control the emissions of CO₂ from large-scale, stationary sources such as conventional, pulverized-coal, power plants. In this work, a liquid MEA-based, post-combustion CO₂ capture and compression system was simulated in Aspen Plus Dynamics, under the constraint of 90% overall CO₂ capture efficiency. The process was also integrated with the steam system of a 550MWe supercritical pulverized coal power plant. Using this basis, the simulation shows that a total of 564 m³/s of flue gas at around 1 atm pressure and containing approximately 13 vol% CO₂ needs to be treated. The sizing calculations show that 6 parallel trains of absorber and strippers with diameters of 9.2 m and 7 m, respectively, would be required to treat all the flue gas. The pressure drop across the absorber using selected packing was found to be small. Most of the published works on the control of CO₂ capture processes focus on finding the optimal pairs of controlled variables for the control structure design for one train. However, less attention has been paid on the control of large-scale parallel trains for CO₂ capture. Due to the pressure-flow dynamics and taking into account the unavoidable variations in the column and plant hardware during the course of operation, a precise distribution of the flue gas between the trains is a challenge for this nonlinear and highly interactive multivariable system. Variations in distribution of the flue gas in the absorbers can lead to variations in CO₂ capture achieved in any train. In addition, due to structural changes in the packing, the stage efficiencies are expected to be different between the columns.  Therefore, variability in performance among the columns would be expected. On the other hand, the pressure drop that can be afforded for manipulating the distribution of the flue gas among the absorbers is very small. However, the penalty due to CO₂ capture should be minimized without violating the CO₂ capture target. This leads to a challenging control problem. To effectively handle this control problem, both linear and nonlinear model predictive controllers (LMPC and NMPC) are developed and their performances are compared. Various linear and nonlinear models such as autoregressive with exogenous inputs, autoregressive moving average with exogenous inputs, nonlinear autoregressive moving average with exogenous inputs, Volterra, and Hammerstein models are identified and their performances are compared. Our study shows that the disturbance model can greatly improve the performance of the MPCs. It also shows that for optimal operation, the extent of CO₂ capture can greatly vary among the trains. The study provides insights that can be very valuable for optimal design and operation of the multiple trains for CO₂ capture.