Predictive neural network model and empirical equations for the physico-chemical properties and solvent characteristics of potassium carbonate solutions in carbon capture processes

Abstract Controlling and optimizing carbon capture processes is vital for improving efficiency, reducing energy consumption, and enhancing sustainability. Process analytical technology (PAT) plays a crucial role in achieving these goals. Establishing the relationship between physico-chemical properties (PCPs) and solvent characteristics, such as loading and strength, can facilitate the practical implementation of PAT. This study develops empirical models for the PCPs of potassium carbonate solutions, including density, refractive index, and electrical conductivity, as well as a mechanistic model for pH across varying temperatures, solvent concentration, and solvent loadings. The models showed strong agreement with experimental data. Density and refractive index increased with solvent strength and decreased with temperature, while conductivity correlated with solvent strength and temperature but decreased with solvent loading. A feedforward neural network was trained to predict solvent strength and loading using eight input scenarios. The highest accuracy was achieved with PCPs combined with Fourier transform infrared (FTIR) or ultraviolet-visible (UV-Vis), using only PCPs, or using PCPs with FTIR and UV-Vis while excluding pH. The findings provide essential insights into K 2 CO 3 solution behavior, contributing to advances in carbon capture technologies.