Enhanced CO2 capture using KOH-functionalized oil palm ash adsorbent: experimental and applied machine learning approach

Abstract This work aimed to develop a cost-effective, environmentally friendly, and sustainable adsorbent for CO 2 capture using biomass waste, specifically oil palm ash (OPA), which is abundantly available as agricultural waste in Malaysia. Pristine OPA was first subjected to acid washing, followed by carbonization and chemical activation with KOH to enhance its physicochemical properties. Since machine learning (ML) can provide new insights into the design of adsorption processes and the understanding of CO 2 adsorption mechanism, this paper presents a detailed comparative analysis of experimental and ML approaches. In this study, the surface and pore characteristics of OPA-KOH(1:2) were significantly enhanced through carbonization and KOH-functionalization, achieving an optimized mesoporous structure (average pore size: 72.71 Å), with a surface area of 30.95 m 2 /g. Despite having a moderate surface area, the tailored pore structure promoted efficient CO 2 diffusion, thus enabling high CO 2 adsorption capacity of 2.9 mmol/g, which was comparable or exceeds that some Acs with higher surface areas. Systematic analyses of adsorption isotherms, kinetics, and thermodynamics have confirmed that the CO 2 adsorption onto OPA-KOH(1:2) occurred exothermically and was predominantly driven by a physisorption mechanism, supported by a weak chemisorption mechanism. ML-based models employed in this study have demonstrated that the bilayered neural network (NN) model could accurately predict CO 2 adsorption onto OPA-KOH(1:2), with exceptionally high accuracy of R 2 > 0.99. These findings have provided valuable insights into the capture, conversion, utilization, and storage (CCUS) technology, while highlighting OPA-KOH(1:2) as an affordable and environmentally friendly adsorbent for effective CO 2 capture. Graphical Abstract