Machine Learning Model Finds Optimal Conditions for Energy-Efficient CO2 Capture with Oil Palm Ash

Researchers have combined machine learning with multi-objective optimization to identify operating conditions that maximize carbon dioxide adsorption while minimizing energy consumption in a capture process built on an agricultural waste material. The study, published in Discover Chemical Engineering, uses KOH-functionalized oil palm ash (OPA) as the adsorbent and demonstrates a framework that could support the design of more energy-efficient and sustainable CO2 adsorption processes.

The starting point is a familiar problem in carbon capture: adsorption performance depends on multiple interacting variables, so finding operating conditions that simultaneously deliver high CO2 uptake and low energy demand is difficult. Agricultural waste-derived adsorbents such as oil palm ash offer a sustainable route to carbon capture while also supporting biomass valorization, but their full potential can only be realized if the right operating window is found. To tackle this, the team assembled a dataset of 346 fixed-bed adsorption experiments and used it to train three kinds of surrogate models: Gaussian Process Regression (GPR), Support Vector Regression (SVR), and Regression Tree models. The models took five input variables: the OPA-to-KOH ratio, gas flow rate, adsorption temperature, inlet CO2 concentration, and contact time.

Among the models evaluated, GPR achieved the highest predictive performance, with an R2 of 0.9984, a root mean square error of 0.0262 mmol g−1, and a mean absolute error of 0.0119 mmol g−1. These error figures, expressed in the same units as the measured adsorption capacity, indicate that the surrogate model reproduced the experimental behavior of the fixed-bed system very closely.

Once validated, the GPR model was coupled with optimization algorithms in two complementary ways. For single-objective optimization aimed purely at maximizing adsorption capacity, the researchers integrated the model with a Genetic Algorithm (GA) and with Particle Swarm Optimization (PSO). Both algorithms converged on the same prediction: a maximum adsorption capacity of 2.954 mmol g−1. For the more demanding task of balancing adsorption performance against electrical energy consumption, the team turned to two multi-objective algorithms, Multi-Objective Particle Swarm Optimization (MOPSO) and Non-dominated Sorting Genetic Algorithm II (NSGA-II).

The authors present the work as a demonstration of the potential of integrating machine learning with multi-objective optimization for CO2 adsorption process design. Because the framework is built on experimental data from a waste-derived adsorbent, it links two sustainability goals at once: valorizing agricultural by-products such as oil palm ash and reducing the energy penalty of carbon capture. The study is published as open access, with the version shared early as peer-reviewed, accepted research carrying a permanent DOI ahead of the final Version of Record.

Subject of Research: Machine learning-assisted multi-objective optimization of CO2 adsorption on KOH-functionalized oil palm ash

Article Title: Gaussian process surrogate modelling and multi-objective optimization for energy-efficient CO2 capture using functionalized oil palm ash

Article References: Rohman, F. S., Ahmad, M. Y., Muhammad, D., Patle, D. S., Azmi, A., & Saleh, S. N. M. (2026). Gaussian process surrogate modelling and multi-objective optimization for energy-efficient CO2 capture using functionalized oil palm ash. Discover Chemical Engineering. https://doi.org/10.1007/s43938-026-00144-w

Image Credits: AI Generated

DOI: 10.1007/s43938-026-00144-w

Keywords: CO2 capture, oil palm ash, Gaussian process regression, machine learning, multi-objective optimization, genetic algorithm, particle swarm optimization, NSGA-II, adsorption, biomass valorization, energy efficiency, surrogate modelling