(680e) Integrating Machine Learning and Process Simulation to Estimate Energy and Greenhouse Gas Emissions of Activated Carbon Production | AIChE

(680e) Integrating Machine Learning and Process Simulation to Estimate Energy and Greenhouse Gas Emissions of Activated Carbon Production


Liao, M. - Presenter, North Carolina State University
Yao, Y., Yale University
Activated carbon (AC) is a carbonaceous material that can be produced from coal and biomass. It has been used as fertilizer additives in soil amendment, adsorbents in water pollutant removal, catalysts in flue gas cleanup [1]. In 2015, the worldwide demand for AC was 12.8 million ton [2] and keeps growing. Depending on the feedstocks and technologies, the environmental footprints and resource consumption of AC production have large variations. For example, with steam activation, the cradle-to-gate global warming potential (GWP) of AC produced from hard coal is 18.28 kg CO2-eq./kg AC, while the GWP of AC from wood chip is only 8.6 kg CO2-eq./kg AC [3]. Various technologies (e.g., physical activation and chemical activation) applied to specific feedstock will also lead to different environmental impacts. It is critical to understand the environmental impacts of different combination of feedstocks, technologies and operational parameters in AC production. Such understandings could significantly enhance the feedstock selection, process optimization, and supply chain design for sustainable production of AC.

Several researches have evaluated the environmental impacts of AC production through Life Cycle Assessment (LCA) [3–7]. However, most studies have focused on AC produced from specific feedstock (e.g., coconut shell) and operational parameters, and thus their results cannot be extended to other feedstocks with different operational conditions. In this work, we integrated machine learning and process simulation to estimate the primary energy and Greenhouse Gas (GHG) emissions of AC produced from different feedstocks with varying operational conditions. Artificial neural network (ANN), a machine learning technique, was trained by a large dataset collected from literatures and used to predict the yield of steam AC production [8]. The trained ANN model was integrated with pyrolysis kinetic model [9] and Aspen Plus process simulation to estimate the overall energy and mass balance that were used to estimate overall primary energy consumption and GHG emissions [10].

The results indicated large variations in primary energy consumption and GHG emissions across 73 different woody biomass species (43.4-277 MJ/kg AC and 3.96-22.0 kg CO2-eq./kg AC). A sensitivity analysis was conducted and the results showed the large impacts of biomass composition (e.g., hydrogen and oxygen content) and operational parameters (e.g., activation temperature). Such understandings will be helpful for feedstock selection and process optimization. For example, the results showed that higher hydrogen content (and H/C molar ratio) of biomass increases the primary energy consumption, and higher oxygen content (and O/C molar ratio) increases the energy recovery ratio. This conclusion will be useful for selecting suitable biomass for energy-efficient AC production. The process-based environmental data generated from this work can be used a transparent data source for future LCA studies related to AC or AC applications (e.g., wastewater treatment). Although this study focuses on AC made from woody biomass, the integrated modeling framework can be applied to other types of biomass or bio-based products.


[1] Dias JM, Alvim-Ferraz MCM, Almeida MF, Rivera-Utrilla J, Sánchez-Polo M. Waste materials for activated carbon preparation and its use in aqueous-phase treatment: A review. J Environ Manage 2007;85:833–46. https://doi.org/10.1016/j.jenvman.2007.07.031.

[2] González-García P. Activated carbon from lignocellulosics precursors: A review of the synthesis methods, characterization techniques and applications. Renew Sustain Energy Rev 2018;82:1393–414. https://doi.org/10.1016/j.rser.2017.04.117.

[3] Gu H, Bergman R, Anderson N, Alanya-Rosenbaum S. Life Cycle Assessment of Activated Carbon From Woody Biomass. Wood Fiber Sci 2018;50:1–15.

[4] Bayer P, Heuer E, Karl U, Finkel M. Economical and ecological comparison of granular activated carbon (GAC) adsorber refill strategies. Water Res 2005;39:1719–28. https://doi.org/10.1016/j.watres.2005.02.005.

[5] Arena N, Lee J, Clift R. Life Cycle Assessment of activated carbon production from coconut shells. J Clean Prod 2016;125:68–77. https://doi.org/10.1016/j.jclepro.2016.03.073.

[6] Hjaila K, Baccar R, Sarrà M, Gasol CM, Blánquez P. Environmental impact associated with activated carbon preparation from olive-waste cake via life cycle assessment. J Environ Manage 2013;130:242–7. https://doi.org/10.1016/j.jenvman.2013.08.061.

[7] Kim MH, Jeong IT, Park SB, Kim JW. Analysis of environmental impact of activated carbon production from wood waste. Environ Eng Res 2018;24:117–26. https://doi.org/10.4491/eer.2018.104.

[8] Liao M, Kelley SS, Yao Y. Artificial neural network based modeling for the prediction of yield and surface area of activated carbon from biomass. Biofuels, Bioprod Biorefining 2019;13:1015–27. https://doi.org/10.1002/bbb.1991.

[9] Anca-Couce A, Scharler R. Modelling heat of reaction in biomass pyrolysis with detailed reaction schemes. Fuel 2017;206:572–9. https://doi.org/10.1016/j.fuel.2017.06.011.

[10] Liao M, Kelley S, Yao Y. Generating Energy and Greenhouse Gas Inventory Data of Activated Carbon Production Using Machine Learning and Kinetic Based Process Simulation. ACS Sustain Chem Eng 2020;8:1252–61. https://doi.org/10.1021/acssuschemeng.9b06522.