TEA of the CO2 capture process in pre-combustion applications using thirty-five physical solvents: Predictions with ANN

A Techno-Economic Analysis (TEA) of the CO2 capture process in a pre-combustion 543 MWe power plant was performed for thirty-five physical solvents employed to capture 90 % of CO2 from the sulfur-free fuel gas of the power plant. The process was built in Aspen-Plus v.11 and included one countercurrent fixed-bed CO2 absorber, containing structured (Mellapak 250Y) or random (IMTP50) packing, three flash-drums to regenerate the solvent, and multi-stage compressors to condition the captured CO2 for subsequent sequestration. The process operating expenditure (OPEX), capital expenditure (CAPEX), and levelized cost of CO2 captured (LCOC) were obtained.Aspen-Plus calculations revealed that Mellapak 250Y, having larger specific surface area than IMTP50, improved gas-liquid mass transfer and decreased LCOC; low-temperature operation enhanced CO2 solubility in the solvents and decreased solvent loss, lowering LCOC; high-temperature operation increased volatile solvent loss, increasing LCOC; and among the thirty-five solvents used, diethyl sebacate (DES) had the lowest LCOC ($7.14 per ton of CO2 captured). All Aspen-Plus TEA results were used to develop an Artificial-Neural-Network (ANN), consisting of one-input layer with 11 nodes, two-hidden layers with 10 nodes, and one-output layer with 3 nodes. The ANN was trained, then tested with 40 % and 60 % of the data, respectively; and it predicted the process CAPEX, OPEX, and LCOC with high accuracy. Thus, this ANN can perform TEA of numerous potential new solvents for CO2 capture in seconds compared to hours with Aspen-Plus, which is a significant timesaving, provided that plant capacity, solvent properties, and operating conditions are within its boundaries.