Evaluating CO2 hydrate kinetics in multi-layered sediments using experimental and machine learning approach: Applicable to CO2 sequestration.

The transition to a low-carbon economy requires the implementation of effective carbon capture and sequestration (CCS) strategies. One of the potential CCS strategies is to capture industrial CO 2 emissions and inject them into the oceanic sediments to be stored as CO 2 hydrates. However, the success of this technique depends on a few key factors such as the type of sediments where CO 2 is injected, the kinetics of CO 2 hydrate formation and dissociation, the accuracy of the models for prediction of formation kinetics, and the CO 2 hydrates morphology. So, in this first-gen work, a highly complex set of experiments was carried out to examine the CO 2 hydrate formation and dissociation processes by injecting CO 2 via injection tube into different size wet sediments, i.e., coarse (diameter: 0.5–1.5 mm), granules (diameter:1.5–3.0 mm) and dual-layered sand (coarse + granules), embedded inside high-pressure reactor as the artificial seabed. The experiments were carried out at 3.5 MPa at T = 1.5–2.0 °C with 500 ppm of the eco-friendly hydrate promotor (l -tryptophan). The images of morphological changes during hydrate formation/dissociation, the Scanning Electron Microscope analysis of the sediments, and the estimated water-to-hydrate conversations have been reported in this work. A novel mathematical four-parameter-based CO 2 hydrate kinetics model was also developed. A set of 32,843 experimental data points was used to train a supervised machine learning algorithm using two parameters with the other two taken from published literature. The water-to-hydrate conversion was estimated and follows the order of dual-layered sand [88.26 (±4.62) %] > coarse [77.77 (±5.72) %] > granules [65.36 (±2.3) %]. The proposed ML-based model predicted the water-to-hydrate conversion with an Average Absolute Relative Deviation [%AARD] of 4.23–13.29 %. This work serves as a step forward in developing a sustainable hydrate-based oceanic carbon storage technology. [Display omitted] • CO 2 hydrate kinetics and morphology in dual-layered sediments. • Effect of injecting CO 2 inside the sediments directly via vertical injection tube. • Developed and trained CO 2 hydrate formation kinetics model in sediments via a supervised machine learning (ML) algorithm. • ML model can predict CO 2 hydrate formation kinetics in sediments with accuracy in terms of AARD of 4.23–13.29 %. [ABSTRACT FROM AUTHOR]