Sorption capacity evaluation of industrial flue gas mixture using South African coal seams: Conventional and ANN modelling

This study investigates the sorption efficiency of synthetic flue gas compared to pure CO 2 in two South African coal seams. The samples, designated AN and TD, represent distinct coal ranks, with AN being a high-rank coal (2.91% vitrinite reflectance) and TD a medium-rank coal (1.26% vitrinite reflectance).The experiments involved a high-pressure volumetric adsorption method to evaluate the adsorption capacities of pure CO 2 and a synthetic flue gas mixture (21% CO 2 , 4% O 2 , 0.03%SO 2 and 74.97% N 2 ,) at temperatures of 35°C and 65°C, and pressures up to 87 bar. The results indicate a significant effect of temperature on CO 2 adsorption in flue gas, with a substantial drop of 49% and 37% observed for AN and TD coals samples, respectively, at 65°C compared to 35°C. Both coal samples exhibit a high adsorption preference for CO 2 in flue gas, with AN showing greater affinity across all operating conditions. The presence of flue gas components significantly impacted CO 2 adsorption, causing reductions of 94% for AN and 91% for TD at 35°C. AN coal (high rank) showed superior adsorption capacity for all flue gas components, attributed to its favourable properties including high inertinite content, low moisture content and low ash content offering minimal adsorption hindrance. This study evaluated multi-component adsorption using the Extended Langmuir and Modified Competitive Langmuir isotherm models. Both models effectively captured the experimental data, demonstrating preferential CO 2 adsorption as reflected by higher maximum adsorption capacity across all scenarios. Additionally, artificial neural network modelling of the adsorption data demonstrated a strong fit with the experimental data, yielding low MSE values and R 2 values above 0.99 for training, validation and testing. This study aims to evaluate CO 2 selectivity in a multi-component flue gas adsorption system in South African coals and explore the feasibility of direct flue gas injection into coal seams. • AN and TD sample achieved maximum CO 2 adsorption of 0.21 mmol/g, and 0.17 mmol/g, respectively, under multi-component flue gas conditions. • Both coal samples exhibited a clear adsorption preference for CO 2 , followed by N 2 , O 2 , and finally SO 2 . • The Modified Competitive Langmuir (MCL) model yielded better approximation of the experimental data over the Extended Langmuir (EL) model. • ANN modelling achieved higher predictive performance relative to conventional multicomponent adsorption models.