Deciphering the CO2 hydrates formation dynamics in brine-saturated oceanic sediments using experimental and machine learning modelling approach

Abstract

One of the potential strategies for limiting global warming is capturing excess amounts of industry CO2 emissions and sequestrating it inside the oceanic sediments as CO2 hydrates. The seabed consists of sediments with different porosity and sizes, which can significantly impede the kinetics of CO2 hydrates. To address these challenges, in this novel work the CO2 hydrate kinetics and morphologies have been investigated in brine-rich [3.5 wt% NaCl] coarse sediments [porosity = 0.28, Ø =0.5-1.5 mm], granule sediments [porosity = 0.43, Ø =1.5-3 mm] and dual sediments [coarse + granules] inside a high-pressure reactor [3.5 MPa, 1-2 oC] with artificial seabed. According to the key experimental results, the water-to-hydrate conversion [%] was estimated to be 67.24 (± 3.02) % for coarse sediments, 49.58 (± 4.97) % for dual sediments (coarse + granules), and 27.68 (± 5.21) % for granules. In-situ Raman spectroscopy was used to evaluate the real-time solubility of CO2 [0.023 mol/mol], and image processing thresholding was used to identify CO2 hydrate distribution patterns. A mathematical model was proposed as a major contribution to predict CO2 hydrates kinetics in sediments, using 94,203 data points. The model was trained via a supervised machine-learning [ML] algorithm and can predict CO2 hydrate kinetics with an AARD of 7.54-8.47%.