Simulation and control of complex distillation processes
MetadataShow full item record
The proper choice and implementation of control method improve reliability and performance of distillation column control, which can translate into a reduction of energy usage while maintaining product quality and rates, hence economic benefit. However, clear guidelines to determine which and when advanced control strategies should be used instead of traditional control strategies are still not available. Previous work has been focused on two-product single columns. In this study, two complex distillation processes, a fluid catalytic cracker unit (FCCU) main fractionator and a gas recovery unit, are simulated with rigorous models. Traditional decentralized and model predictive control (MPC) are applied to both processes, and their performances are compared in terms of their capability to handle constrained multivariable processes. A detailed tray-to-tray rigorous model for the FCCU main fractionator is developed, in which the Soave-Redlich-Kwong (SRK) equations are used to model vapor-liquid phase equilibrium. The feed is characterized as a mixture of 36 pseudo-components and 9 defined components including water, hydrogen and light hydrocarbons from CI to C4. An efficient algorithm is developed to solve the dynamic model equations. Two decentralized control systems, one without decoupler, one with a simple decoupler are implemented, and compared with a DMCPlus™ controller. The DMCPlus™ controller performs better than both decentralized controls due to its superior decoupling power. The gas recovery unit consists of three distillation columns operated in series with feed-bottoms heat integration for the first column. Rigorous models are developed for the columns and the heat exchanger, including pressure and heat transfer dynamics. The process is a highly coupled system and has interactive constraints that exist in different units. A decentralized control system with override controls for constraints is designed, implemented on the GRU simulator, and is compared with a DMCPlus™ controller with 10 independent variables and 12 dependent variables. The DMCPlus™ controller outperforms the decentralized control system in terms of constraint handling due to its flexibility. The effects of including level control into MPC are also investigated. Three DMCPlus™ controllers with different strategies for controlling the bottom level of the first column are implemented for the GRU process. The first DMCPlus™ controller does not control the level, while the second one moves setpoint to the PI level controller, and the third one controls the level directly by manipulating the deethanizer bottoms flow. The results show that including level into MPC controller improves composition control in cases that the manipulated variable for the level control has significant impact on compositions.