Description
A Master of Science Thesis in Mechatronics Submitted to the School of Engineering by Jasim Al Hammadi, "Model Based Control of Amine Sweetening Unit," June 2005. Chair of Committee Dr. Nabil Abdel Jabbar & Co-Advisor Dr. Mohamed Ameen Al Jarrah. Available are Both Soft and Hard Copies of the Thesis.
Abstract
Many industrial processes often exhibit significant non-linear behavior, and amine sweetening unit comprising absorption and regeneration towers is a typical example of such chemical processing plants. Conventional linear control schemes based on rigorous mathematical models, implemented in sweetening units, show poor performance and lead to off-specification products. Therefore, the need for advanced controls based on non-linear model structure is required. The black-box approach of modeling and control an amine system using Artificial Neural Networks (ANNs) is studied in this thesis. As a first step, a rigorous model for steady state simulation of an amine unit in Abu Dhabi Gas Liquefaction Company (ADGAS) is developed. This model is used to study the effect of varying system operating and sizing parameters on H2S and CO2 sweet gas concentrations. Desired H2S specification in the sweet gas can be maintained by operating the plant with low DEA concentration and/or low DEA solution temperature. In contrast, high solution concentration and/or high DEA temperature tends to increase CO2 removal from the process gas. Then, dynamic analysis of amine sweetening plant is conducted on the developed rigorous simulation model. It aims to prioritize the system controlling variables, namely; DEA solution flow rate, DEA solution temperature, and reboiler duty as per their speed of response on sweet gas quality. Dynamic profiles of H2S sweet gas composition show that steam flow rate gives the shortest time constant among other controlling parameters. On the other side, CO2 pick up is experienced to increase faster with the increase of DEA solution flow rate. On the other side, pick up is experienced to increase faster with the increase of DEA solution flow rate. Non-linear amine process needs to be operated under tight performance specifications to meet product quality and satisfy environmental considerations. Non-linear predictive control appears to be a well suited approach for this kind of problems. Because Artificial Neural Networks (ANNs) can provide good empirical models of complex nonlinear processes, dynamic data extracted from HYSYS are used to model amine plant applying ANN technique. Two NN models are developed for H2S and CO2 and separately because they expose different dynamic behavior when reacting with DEA solution. Feed-Forward Neural Networks (FNNs) are found to give very good match with rigorous plant data. In this work, the MPC model based control strategy was applied to amine absorption plant with view to control H2S and CO2 and composition in sweet gas. The FNN model is augmented in the MPC control structure as the plant model leads to NN-MPC. The performance of the proposed MPC structure under different model uncertainties has been investigated. The closed-loop performance and stability of the proposed NN-MPC depend on setting prediction and control horizons. Accordingly, it is desirable to minimize the performance objective determined by the cost functional using long horizons. However, from computational point of view using short horizons is preferred. The shorter the horizon, the less costly the solution of the on-line optimization problem.