CO2 MEA Capture: A Gaussian Process-Based Global Sensitivity Analysis

Abstract

In this work, we present a variance-based global sensitivity analysis (GSA) using Gaussian process (GP) surrogate models to calculate the semi-analytic Sobol’ indices for a carbon dioxide (CO2), post-combustion capture (PCC) process. Using the open-source ROM-COMMA software library, the generated GP surrogate model enables accurate process output prediction, and calculation of the semi-analytic Sobol’ indices for GSA. We apply this methodology to a case study of a CO2 PCC process with 30 wt.% Monoethanolamine (MEA). Results showed an excellent prediction quality, further identifying key process parameters for outputs variables such as the cost of CO2 capture (US$/tCO2) and the CO2 capture rate (% mol.).
These findings have applications in dimensionality reduction, process optimisation, operation and control of PCC processes, as PCC via MEA absorption is at the forefront of amine absorption technologies owing to its performance and maturity. It is also a key part of carbon capture, utilisation and storage (CCUS) which has been identified as a growing and important decarbonisation route across a wide range of industry application and regions. This GP surrogate model presented meets the growing need for computationally efficient models to aid quick and accurate prediction of process conditions, and the identification of key process parameters over a wide range of industrial applications, as opposed to rigorous, application-specific, and computationally expensive simulation models.

Keywords: CO2 capture, CCS, Global Sensitivity Analysis (GSA), Surrogate Model, Gaussian Process (GP)