(667h) Superstructure-Based Process Synthesis of a Pre-Combustion Membrane CO2 Capture System
This work employs a Mixed Integer Non-Linear Programming (MINLP) formulation to determine the optimal pre-combustion CO2 membrane system configuration from a pre-defined set of potential process options. In MINLP, a set of binary state variables is distinguished from continuous state variables and characterizes the flowsheet topology. Binary variables in this formulation indicate whether a piece of equipment proposed in a superstructure is present in the optimal configuration. In general, this formulation is non-linear and non-convex presenting challenges for global optimization. The total problem formulation which characterizes the membrane system superstructure consists of 2.1x106 potential process configurations characterized by 2,234 equations and 2,177 variables including 21 binary variables.
Results suggest that advanced membrane designs may be an improvement, albeit slight, over state of the art solvent (Selexol) based CO2 separation systems. If projected improvements can be achieved in full, they can be expected to reduce required membrane area by 98% and compression power by 60% over current membrane performance. The COE of an advanced membrane based solution consequently drops by ~24% and becomes competitive with solvent based technology. However, the MINLP formulation synthesizes an optimal process configuration that would not be chosen by conventional heuristic design strategies for this type of application and results in an improvement of 7-8%. For a nominal 550MW plant, this represents an annual offset in required power generation costs of nearly $31M over the performance of conventional configurations.