Prediction of Solvent Component and Composition for Absorption-Based Acid Gas Removal Unit using Optuna-LightGBM and K-means
Rafi Jusar WISHNUWARDANA, Madiah OMAR, Haslinda ZABIRI, Kishore BINGI, Rosdiazli IBRAHIM
Abstract. Acid gas removal unit (AGRU) is a pivotal component of a natural gas processing plant. The primary purpose of acid gas removal is to reach the industrial pipeline standard of H2S below four ppm and CO2 below 2% per volume for pipeline quality. The most widely used technique is an absorption-based AGRU using amine as a solvent. MDEA is the most utilized solvent but has the drawback of absorbing CO2. The mixture of other amine and physical solvents is necessary to assist the absorption of CO2. However, the main problem of mixing solvents is parameter complexity. The machine learning method is utilized to find the most optimal solvent based on its operational parameters. LightGBM tuned with Optuna are used to classify the solvent component, followed by K-means to identify solvent composition. The algorithm is applied to six different solvent blends and two feed gas compositions, resulting in 37,786 data points. The LightGBM model tuned with Optuna performed excellently with accuracy of 0.98 and training time under 0.2 seconds. K-means showed the silhouette score averaging 0.5, showing that the data is not well clustered. This model demonstrates reliable capability in analyzing and distinguishing the solvent component and its respective composition.
Keywords
Acid Gas Removal Unit, Solvent Component, Solvent Composition, LightGBM, Optuna, K-Means
Published online 1/15/2026, 8 pages
Copyright © 2026 by the author(s)
Published under license by Materials Research Forum LLC., Millersville PA, USA
Citation: Rafi Jusar WISHNUWARDANA, Madiah OMAR, Haslinda ZABIRI, Kishore BINGI, Rosdiazli IBRAHIM, Prediction of Solvent Component and Composition for Absorption-Based Acid Gas Removal Unit using Optuna-LightGBM and K-means, Materials Research Proceedings, Vol. 59, pp 162-169, 2026
DOI: https://doi.org/10.21741/9781644903957-21
The article was published as article 21 of the book Separation Technology
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 license. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
References
[1] R. Faiz, M. Al-Marzouqi, Insights on natural gas purification: Simultaneous absorption of CO₂ and H₂S using membrane contactors, Sep. Purif. Technol. 76 (2011) 351–361. https://doi.org/10.1016/j.seppur.2010.11.005
[2] S. Sheikh, Enhancements in modeling gas sweetening, in: Proc. Abu Dhabi Int. Petroleum Exhibition and Conference (ADIPEC), Society of Petroleum Engineers, 2015. https://doi.org/10.2118/177656-MS
[4] A.M. Tamidi, N.H. Yasin, CO₂ and H₂S co-removal from natural gas using membrane contactor technology, in: Proc. SPE Asia Pacific Oil and Gas Conference and Exhibition (APOG), Society of Petroleum Engineers, 2020.
[5] L.A. Pellegrini, S. Moioli, F.M. Munari, P. Vergani, B. Picutti, A. Uccelletti, The acid gas removal unit at Gasco’s Habshan 5: Simulation and comparison with field data, in: Proc. Offshore Mediterranean Conference and Exhibition (OMC), 2015.
[6] M. Faqih, M.B. Omar, R.J. Wishnuwardana, N.I. Ismail, M.H.M. Zaid, K. Bingi, Prediction of solvent composition for absorption-based acid gas removal unit on gas sweetening process, Molecules 29 (19) (2024) 4591. https://doi.org/10.3390/molecules29194591
[7] W. Zhu, H. Ye, X. Zou, Y. Yang, H. Dong, Analysis and optimization for chemical absorption of H₂S/CO₂ system: Applied in a multiple gas feeds sweetening process, Sep. Purif. Technol. 276 (2021) 119301. https://doi.org/10.1016/j.seppur.2021.119301
[8] A. Esmaeili, T. Yoon, T.A. Atsbha, C.J. Lee, Rate-based modeling and energy optimization of acid gas removal from natural gas stream using various amine solutions, Process Saf. Environ. Prot. 177 (2023) 643–663. https://doi.org/10.1016/j.psep.2023.07.030
[9] M. Hakimi, M.B. Omar, R. Ibrahim, Application of a neural network for predicting H₂S from an acid gas removal unit (AGRU) with different compositions of solvents, Sensors 23 (2023) 1020. https://doi.org/10.3390/s23021020
[10] I. Izonin, R. Muzyka, R. Tkachenko, I. Dronyuk, K. Yemets, S.-A. Mitoulis, A method for reducing training time of ML-based cascade scheme for large-volume data analysis, Sensors 24 (15) (2024) 4762. https://doi.org/10.3390/s24154762
[11] P. Anandan, A. Manju, M.R. Reddy, Classification of massive data sets using a revolutionary grey wolf optimization algorithm and a deep learning model in a cloud-based setting, in: Proc. 2023 Int. Conf. Data Sci., Agents & Artif. Intell. (ICDSAAI), 2023, pp. 1–6.
[12] L.C. Law, N.Y. Azudin, S.R. Syamsul, Optimization and economic analysis of amine-based acid gas capture unit using monoethanolamine/methyl diethanolamine, Clean Technol. Environ. Policy 20 (2018) 451–461. https://doi.org/10.1007/s10098-017-1430-1
[13] B.K. Choi, S.M. Kim, K.M. Kim, U. Lee, J.H. Choi, J.S. Lee, I.H. Baek, S.C. Nam, J.H. Moon, Amine blending optimization for maximizing CO₂ absorption capacity in a diisopropanolamine–methyldiethanolamine–H₂O system using the electrolyte UNIQUAC model, Chem. Eng. J. 419 (2021) 129517. https://doi.org/10.1016/j.cej.2021.129517
[14] A. Sharifi, E.O. Amiri, Effect of the tower type on the gas sweetening process, Oil Gas Sci. Technol. 72 (2017). https://doi.org/10.2516/ogst/2017018
[15] S. Zhang, The study of model generalization ability for spam classification based on machine learning models, in: Proc. 2024 9th Int. Symp. Comput. Inf. Process. Technol. (ISCIPT), 2024, pp. 43–46.
[16] H.A. Behrooz, Robust synthesis of amine-based natural gas sweetening plants, Gas Sci. Eng. 113 (2023) 204970.
[17] Y. Li, Y. Cao, J. Yang, M. Wu, A. Yang, J. Li, Optuna-DFNN: An Optuna framework driven deep fuzzy neural network for predicting sintering performance in big data, Alexandria Eng. J. 97 (2024) 100–113. https://doi.org/10.1016/j.aej.2024.04.026
[18] L. Zhou, D. Garg, Y. Qiu, S.M. Kim, I. Mudawar, C.R. Kharangate, Machine learning algorithms to predict flow condensation heat transfer coefficient in mini/micro-channel utilizing universal data, Int. J. Heat Mass Transf. 162 (2020) 120351. https://doi.org/10.1016/j.ijheatmasstransfer.2020.120351
[19] S. Banerjee, K. Bhavna, T. Raychoudhury, Prediction of transport behavior of nanoparticles using machine learning algorithm: Physical significance of important features, J. Contam. Hydrol. 258 (2023) 104237. https://doi.org/10.1016/j.jconhyd.2023.104237
[20] O. Alshboul, G. Almasabha, A. Shehadeh, K. Al-Shboul, A comparative study of LightGBM, XGBoost, and GEP models in shear strength management of SFRC-SBWS, Structures 61 (2024) 106009. https://doi.org/10.1016/j.istruc.2024.106009
[21] R. Bakır, C. Orak, A. Yüksel, Optimizing hydrogen evolution prediction: A unified approach using random forests, LightGBM, and Bagging Regressor ensemble model, Int. J. Hydrogen Energy 67 (2024) 101–110. https://doi.org/10.1016/j.ijhydene.2024.04.173
[22] K. Yuan, G. Chi, Y. Zhou, H. Yin, A novel two-stage hybrid default prediction model with k-means clustering and support vector domain description, Res. Int. Bus. Finance 59 (2022) 101536. https://doi.org/10.1016/j.ribaf.2021.101536
[23] G. Niu, Y. Ji, Z. Zhang, W. Wang, J. Chen, P. Yu, Clustering analysis of typical scenarios of island power supply system by using cohesive hierarchical clustering based K-means clustering method, Energy Rep. 7 (2021) 250–256. https://doi.org/10.1016/j.egyr.2021.08.049
