Abstract:
An LNG cascade process operates under high pressure and extremely low temperature to liquefy natural gas, requiring very high work input to the compressors. This thesis performed the optimization of the single refrigerant ConocoPhillips LNG process by a global optimization technique based on the domain and image partitioning methodology (Faria and Bagajewicz, 2012) to minimize work input to the system. The procedure consists of an NLP upper bound and MILP lower bound models. The lower bound model was formulated by discretizing and linear relaxation of the upper bound non-linear equations by assigning new integer variables and a set of linear constraints. To guarantee the global optimum, the difference between objective function of the upper and lower bound model must be small. The ConocoPhillips LNG process consists of three liquefaction loops which are propane, ethylene and methane loops. Thermodynamic properties prediction model consisted of a metamodel based on quadratic polynomials regressed from Peng-Robinson EOS. Several minimum approach temperatures were performed and the results showed that the lowest approach temperature of 3 K was the most efficient case that gave the lowest refrigerant flowrates, and the lowest total work input to the compressors. A simple cost analysis was performed both CAPEX and OPEX. The problem was solved in GAMS and the results were verified by PRO/II simulation software.
