Introduction: The Quest for Sustainable Energy
The world’s growing energy demands and the urgent need to combat climate change have spurred a global search for sustainable energy solutions. As chemical engineers, we are committed to contributing to this effort by exploring innovative ways to harness the power of biomass, an abundant and renewable resource.
Biomass, derived from organic matter like agricultural waste, can be converted into valuable energy products through a process called pyrolysis. This process involves heating biomass in the absence of oxygen to produce bio-oil, bio-char, and gas.
Our latest research [1], published in Renewable Energy, focuses on using machine learning to predict and optimize the yields of these valuable products from biomass pyrolysis. By gaining better control over the pyrolysis process, we aim to maximize the utilization of biomass and contribute to a more sustainable energy future.
Fig.1 Illustration of fuel energy consumption
Why Biomass Pyrolysis Matters
Biomass offers several advantages as a renewable energy source. It is widely available, diverse, and often considered carbon-neutral. Furthermore, utilizing waste biomass for energy production can help reduce waste and minimize greenhouse gas emissions.
Pyrolysis is a particularly promising method for converting biomass due to its simplicity, cost-effectiveness, and versatility. The resulting products have various applications: Bio-oil can be further processed into fuels [2-4], bio-char can be used for soil amendment and carbon capture [5], and the gas can be used for energy generation [6].
However, the yield of each product from pyrolysis varies significantly depending on the type of biomass and the operating conditions, making it challenging to optimize the process.
Fig.2 Schematic illustration of biomass pyrolysis for biofuel production.
Challenges in Optimizing Pyrolysis
One of the main challenges in biomass pyrolysis is the complex interplay of factors that influence product yields. Biomass is composed of cellulose, hemicellulose, and lignin, and the proportions of these components vary between different feedstocks. Additionally, operating conditions such as temperature and heating rate play a crucial role in determining the final product distribution.
To overcome these challenges, it is essential to develop methods that can accurately predict product yields based on biomass composition and operating conditions.
Our Approach: Machine Learning to the Rescue
In our research, we explored the use of machine learning to model and optimize biomass pyrolysis. Machine learning is a powerful tool that can identify complex patterns in data and make accurate predictions. We compiled a comprehensive dataset of 273 experiments from various studies, including data on different types of biomass (e.g., agricultural residues) and pyrolysis conditions (temperature and heating rate). We then trained and compared several machine learning models to predict the yields of bio-oil, bio-char, and gas. Our analysis revealed that the ensemble models like XGBoost outperformed other machine learning models in accurately predicting product yields. This model demonstrated a high level of accuracy, capturing the complex relationships between biomass composition, operating conditions, and product distribution.
Fig.3 Machine learning-based prediction and optimization of biomass pyrolysis product yields
Furthermore, we used the XGBoost model to conduct a sensitivity analysis, which allowed us to understand how changes in temperature and heating rate affect the yields of bio-oil, bio-char, and gas for different types of biomass. This analysis provided valuable insights for optimizing pyrolysis conditions to achieve desired product outcomes.
Conclusion: Towards Sustainable Biomass Utilization
This research demonstrates the potential of machine learning, particularly the XGBoost model, to accurately predict and optimize biomass pyrolysis. Our findings provide valuable insights for developing customized pyrolysis guidelines to maximize the production of desired products from specific biomass feedstocks. We believe that this work contributes to the advancement of sustainable energy technologies and paves the way for more efficient utilization of biomass resources.
Read the full research paper for more details: K. Cheenkachorn et al. Renewable Energy 248 (2025) 123108
References
[1] K. Cheenkachorn, C. Prapainainar, T. Wijakmatee, Machine learning-driven modeling of biomass pyrolysis product distribution through thermal parameter sensitivity, Rew. Energy 248 (2025) 123108. https://doi.org/10.1016/j.renene.2025.123108.
[2] J. Jiang, J. Xu, Z. Song, Review of the direct thermochemical conversion of lignocellulosic biomass for liquid fuels, Front. Agric. Sci. Eng. 2 (2015) 13. https://doi.org/10.15302/J-FASE-2015050.
[3] S. Wang, G. Dai, H. Yang, Z. Luo, Lignocellulosic biomass pyrolysis mechanism: A state-of-the-art review, Prog. Energy. Combust. Sci. 62 (2017) 33–86. https://doi.org/10.1016/j.pecs.2017.05.004.
[4] R.E. Guedes, A.S. Luna, A.R. Torres, Operating parameters for bio-oil production in biomass pyrolysis: A review, J. Anal. Appl. Pyrolysis 129 (2018) 134–149. https://doi.org/10.1016/j.jaap.2017.11.019.
[5] S. Ren, H. Lei, L. Wang, Q. Bu, S. Chen, J. Wu, Hydrocarbon and hydrogen-rich syngas production by biomass catalytic pyrolysis and bio-oil upgrading over biochar catalysts, RSC Adv. 4 (2014) 10731–10737. https://doi.org/10.1039/C4RA00122B.
[6]W. Jung, G.B. Rhim, K.Y. Kim, M.H. Youn, D.H. Chun, J. Lee, Comprehensive analysis of syngas-derived Fischer–Tropsch synthesis using iron-based catalysts with varied acidities, Chem. Eng. J. 484 (2024) 149408. https://doi.org/10.1016/j.cej.2024.149408.
