(84f) Development of Machine Learning [ML] Based Model for Predicting CO2 Hydrate Formation Kinetics in Porous Media

Carbon capture, utilization, and storage (CCUS) is the process of capturing CO₂ emissions and either using them to create products like construction materials (utilization) or permanently storing them below a suitable medium, such as depleted oil and gas reservoirs or deep-saline aquifers. Capturing CO₂ emissions from industry and injecting them into deep oceanic sediments to be stored as gas hydrates is a viable alternative CCUS technique. Gas hydrates are crystalline ice-liked compounds formed under high pressure and low-temperature conditions, already prevalent at depths of 300 m and above in the ocean sediments. In this work, experimentally, the formation kinetics of CO₂ hydrates in different sediments have been investigated. An ML-based kinetic model has been trained to predict CO₂ hydrate formation kinetics in sediments using experimental data. Four parameters based on a new ML-based algorithm were proposed to predict the stochastic kinetics nature of CO₂ hydrate formed in the presence of sediments. Out of four parameters, two were taken from the published literature studies.

The experiments were carried out inside a specialized high-pressure reactor in which CO₂ was injected via injection tube directly into the sediments such as silica sand (0.5-1.5 mm) and granules sand (1.5-3mm) at 3.5 MPa and 1-2 ^oC. Then using the experimental kinetic data, a new ML-based CO₂ hydrate kinetic model was developed to predict the CO₂ hydrate formation kinetics in the sediments. The water-to-hydrate conversion was estimated to be more in the presence of mixed sand [>85 %] followed by silica sand [>70%] and granules sand [>65%]. The proposed ML-based model was trained using a total of 32843 experimental kinetics data and could predict the water-to-hydrate conversion with accuracy in terms of %AARD (<10%) for all kinds of experiments done in this study. The experimental results and proposed ML-based kinetic model will help bridge the critical knowledge gap and serve as a step forward for further developing a sustainable hydrate-based CCUS technology for the energy sector.