PROBLEM
Current methods for the characterization of biomass and biocrude streams face limitations in reconciling experimental data with computational models. Conventional analytical techniques often fail to capture the full complexity of these streams, while existing optimization approaches lack flexibility in handling diverse biomass types and biochemical profiles. The complex nature of biomass and biocrude streams presents challenges in developing standardized compositional models, which hinders process design and scalability [2, 3]. This research addresses these issues by integrating experimental datasets with advanced optimization strategies, ensuring compatibility with industry-accepted standards for relevant feedstocks.
METHODS
A mixed-integer nonlinear programming (MINLP) model is herein developed to characterize biomass and biocrude streams through the reconciliation of experimental measurements (elemental analysis, biochemical composition) and first-principle property models. Implemented in GAMS using the BARON solver, the model employs integer cuts to generate solution pools that capture compositional diversity while minimizing deviations between calculated and experimental values. Datasets of 111 biomass-related and 144 biocrude-related compounds, organized into biochemical categories (proteins, carbohydrates, lipids, lignin), are created and validated using data from studies on sewage sludge, pine wood, and HTL-derived biocrudes. They comprise peer-reviewed data, experimental results, simulation data etc. to acquire specification parameters, stoichiometric composition and thermodynamic properties associated with bioconversion technologies like HTL.
RESULTS
The MINLP model demonstrates superior performance compared to previous characterization approaches, achieving reduction by an order of magnitude in objective function values for biomass and biocrude samples relative to previous work [3]. Additional case studies on pine wood [4] and sewage sludge [4] reveal accurate reconstruction of biochemical profiles, with the sum of squares of relative deviations (SSRD) ranging between 7.9*10-5 and 2.3*10-2 for variable problem setups. For biocrudes, the model again achieves excellent alignment with experimental distillation curves through systematic compound selection. Using integer cuts for we can identify key components and quantify their concentration distributions. For instance, in the sludge characterization case study, integer cuts indicate lignoceric and docosahexanoic acids as the dominant components among lipids; cellobiose and sucrose were the most prevalent carbohydrates; and phenylalanine and tyrosine were the most abundant amino acids.
CONCLUSIONS
Efficient modeling of complex streams requires analyzing first-principle models alongside experimental data, following industry standards. This work aims to match stream properties to chemical compounds and functional groups, using a new MINLP formulation for characterizing biomass and biocrude streams in HTL. Through integrating experimental data and optimization methods, this work addresses challenges in biomass stream characterization. The demonstrated improvements in prediction accuracy and compositional flexibility, position the model as a future tool for biorefinery analysis. By addressing the structural complexity of biomass and biocrude steams through optimized compound selection, the model also provides insights into depolymerization pathways and upgrading requirements. Future work includes incorporating additional properties to improve accuracy, creation of collective databases including synthetic data for machine learning applications, and consideration of digital passport technologies to simplify biomass and biocrude classification.
ACKNOWLEDGEMENTS
The project is implemented within the framework of the National Recovery and Resilience Plan "Greece 2.0" and has been financed by the European Union (NextGeneration EU).
REFERENCES
[1[ Castello D, Pedersen TH, Rosendahl LA. Continuous hydrothermal liquefaction of biomass: A critical review. Energies 11(11):3165 (2018). https://doi.org/10.3390/en11113165
[2] Taghipour A, Ramirez J, Rainey TJ. A method for HTL biocrude simulation using multi-objective optimisation and fractional distillation. Comput Chem Eng 157:107600 (2022) https://doi.org/10. 1016/ j.compchemeng.2021.107600
[3] Aslanoglou I, Anastasakis K, Michalopoulos C, Marcoulaki E, Kokossis A. A systems approach to model nonconventional streams applied to biocrude production from hydrothermal liquefaction. Comput Aided Chem Eng 53:1177-1182 (2024) https://doi.org/10.1016/B978-0-443-28824-1.50197-6
[4] Obeid R, Smith N, Lewis DM, Hall T, van Eyk P. A kinetic model for hydrothermal liquefaction of microalgae, sewage sludge, and pine wood with product characterization of renewable crude. Chem Eng J 428:131228 (2022) https://doi.org/10.1016/j.cej.2021.131228
[5] Snowden-Swan LJ, Li S, Thorson MR, Schmidt AJ, Cronin DJ, Zhu Y, Hart TR, Santosa DM, Fox SP, Lemmon TL, Swita MS. Wet waste hydrothermal liquefaction and biocrude upgrading to hydrocarbon fuels: 2022 state of technology. No. PNNL-32731. Pacific Northwest National Lab.(PNNL), Richland, WA (United States) (2022) https://doi.org/10.2172/1897670