Semi-Hybrid Models for Determining Gas Solubility in Brines with Salt Mixtures: Application to CCS and Gas Processing

Abstract

Abstract Gas solubility in brine plays crucial role in designing various industrial applications such as oil recovery, CCS, corrosion, and gas processing. However, most studies include only standard salts and may not capture the full spectrum of formation brines. The objective of this work is to develop a semi-hybrid framework that can determine the gas solubility in brine solution at extended pressure/temperature ranges, which is applicable to any gas and salt mixture of choice. The work includes the coupling of semi-empirical model and machine learning (ML) approach. In particular, it is an extension to Setschenow's correlation where coefficients are evaluated using ML tool based on decision tree (DT). The features in the ML models include the ionic properties of cations and anions, and thermodynamic properties of gases. This work captures combinations of various salts such as chlorides, carbonates/bicarbonates, and sulphates (as they are seen in real formation brines and water utilities applications), and various standard gases (including hydrocarbon, non-hydrocarbon/polar and acidic gases). A semi-hybrid (physics augmented) framework is developed to estimate gas solubility in brines for a generic gas-brine systems. It is applicable for a wide range of pressures, temperatures, and brine compositions. The prediction from semi-hybrid models were validated against the available experimental data. The main results are as follows: The Setschenow's coefficients for any cations, anions and gases can be generated within 1 – 3% accuracies. The semi-hybrid models predict the experimental trends of gas solubility in brine solution accurately, within the relative error of 1 – 6% for complex gas-brine systems. Most importantly, the framework is general, fast, convenient and can easily be extended for a novel species including greenhouse or hydrocarbon gases, as well as for variety of salts. Additionally, it can fill the gaps in experimental data for the gas-brine systems, and can extrapolate to elevated pressure and temperature conditions. In this work, the applicability is demonstrated for many salts that are seen in formation brine, and many gases that are used in gas injection/storage and gas processing applications. The most ML, correlation and EOS-based studies in the literature on estimating gas solubility in brine are restrictive and valid only for specific gases such as CO2 as well as few salts (NaCl/KCl/CaCl2). Here, we develop a semi-hybrid framework that can estimate the solubility of any gas in a given brine composition that could consists of wide range of salts and salt mixtures, which is the main novelty of the work.