(51e) A Data-Driven Framework for Biomass Selection and Process Optimization of Activated Carbon Production

Liao, M. - Presenter, North Carolina State University
Kelley, S., North Carolina State University
Yao, Y., Yale University
Activated carbon (AC) is a carbon material with high surface area and wide applications in the industry.1,2 AC can be produced from different carbonaceous sources such as coal and biomass through pyrolysis and activation (i.e., chemical and physical activation).3 Previous studies indicated that AC production performance (e.g., AC yields and BET surface area) and environmental footprints (e.g., energy consumption and Greenhouse Gas (GHG) emissions) are highly driven by feedstock properties and operational conditions of AC.4,5 However, few studies have developed quantitative understanding of complex relationships among feedstock property, AC quality, and environmental footprints of AC production. Thus it is challenging to screen different biomass and optimize AC production using traditional process modeling approaches.

In this study, a predictive modeling framework was developed for AC production by integrating a data-driven approach Artificial Neural Network (ANN) with traditional chemical process simulation. Specifically, we focused on AC produced from woody biomass using pyrolysis and steam activation. ANN model was trained by a large dataset containing 168 data samples of biomass composition (i.e., ultimate and proximate analysis), operational conditions (i.e., pyrolysis time, pyrolysis temperature, activation time, activation temperature, steam to biochar ratio) and AC quality (i.e., yields and BET surface area).6 By providing the characterization data of target biomass and operational conditions, the well-trained ANN is capable of predicting the key process parameters such as yields of overall AC production.6 The composition of materials flows within the AC production was estimated by a pyrolysis kinetic model adapted from the previous studies.7 Aspen Plus process simulation was developed using the data generated by two models mentioned previously.8 The integrated modeling framework is able to generate information of primary energy consumption and GHG emissions of AC production using inputs of biomass characterization and operational conditions.

The modeling framework was tested for 251 data samples of woody biomass collected from literature to provide quantitative understandings of biomass species on primary energy and GHG emissions of AC production. Biogenic and fossil-based GHG emissions are tracked separately. Different scenarios regarding energy recovery were developed to identify potential opportunities of energy savings. The preliminary results indicated that AC from different biomass species have large variations in the primary energy consumption (43.4 – 276.7 MJ/kg AC product without energy recovery) and GHG emissions (3.7 – 20.6 CO2 eq./kg AC product without energy recovery). Impacts of specific biomass compositions (e.g., atomic H/C ratios) were further explored to understand the relationships between biomass characterization and energy/carbon footprints. By varying operational conditions, the integrated modeling framework can also provide insightful information of process optimization for specific biomass species.


  1. Sirichote O, Innajitara W, Chuenchom L, Chunchit D, Naweekan K. Adsorption of iron (III) ion on activated carbons obtained from bagasse, pericarp of rubber fruit and coconut shell. Songklanakarin J Sci Technol. 2002;24(2):235–42.
  2. Bansal RC, Goyal M. Activated carbon adsorption. 1st ed. Boca Raton: Taylor & Francis; 2005. 1-520 p.
  3. Danish M, Ahmad T. A review on utilization of wood biomass as a sustainable precursor for activated carbon production and application. Renew Sustain Energy Rev [Internet]. 2018;87(February):1–21. Available from: https://doi.org/10.1016/j.rser.2018.02.003
  4. Cagnon B, Py X, Guillot A, Stoeckli F, Chambat G. Contributions of hemicellulose, cellulose and lignin to the mass and the porous properties of chars and steam activated carbons from various lignocellulosic precursors. Bioresour Technol. 2009;100(1):292–8.
  5. Chen D, Chen X, Sun J, Zheng Z, Fu K. Pyrolysis polygeneration of pine nut shell: Quality of pyrolysis products and study on the preparation of activated carbon from biochar. Bioresour Technol. 2016;216:629–36.
  6. 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 [Internet]. 2019;1–13. Available from: https://onlinelibrary.wiley.com/doi/full/10.1002/bbb.1991
  7. Anca-Couce A, Sommersacher P, Scharler R. Online experiments and modelling with a detailed reaction scheme of single particle biomass pyrolysis. J Anal Appl Pyrolysis. 2017;127:411–25.
  8. AspenTech. Aspen Plus user guide. Cambridge; 2000.