(404d) Multi-Objective Bayesian Optimization for Guiding Biomass Pretreatment Experiments

Ash is an important component of biomass, accounting for up to 53% of its weight. Ash consists of alkali and alkaline oxides such as Ca₂CO₃, K₂O, CaO, MgO, Al₂O₃, and SiO₂ [1]. These minerals are essential for biomass growth but pose significant challenges to thermochemical biomass processing. Specifically, ash may reduce efficiency and cause catalyst poisoning, corrosion, slagging, and fouling in processing equipment [2]. Similarly, the selective removal of nitrogen from biomass, which is associated with proteins and amino acids, is essential. Nitrogen compounds, when present during thermochemical processing, can lead to the formation of HNO₃, which can be catastrophic for biofuel production processes [3].

Biomass demineralization is crucial to maximize the utilization of biomass as a feedstock. This study investigates the demineralization of wheat middlings in an HCl solution under different conditions of temperature, concentration, and contact time. Demineralization experiments are resource-intensive and time-consuming, requiring numerous trials to effectively explore the experimental design space. To overcome these limitations, we employ a Multi-Objective Bayesian Optimization (MOBO); this approach builds probabilistic models that predict ash removal, protein content, and costs as a function of operating conditions. The probabilistic models also enable the identification of experiments that maximize performance and information content. We show that this approach streamlines experimentation and identifies optimal demineralization conditions with few experiments.

References:

[1] W.T. Chen, W. Qian, Y. Zhang, Z. Mazur, C.T. Kuo, K. Scheppe, L.C. Schideman, B.K. Sharma, Effect of ash on hydrothermal liquefaction of high-ash content algal biomass, Algal Res 25 (2017) 297–306. https://doi.org/10.1016/j.algal.2017.05.010.

[2] A.M. Smith, S. Singh, A.B. Ross, Fate of inorganic material during hydrothermal carbonisation of biomass: Influence of feedstock on combustion behaviour of hydrochar, Fuel 169 (2016) 135–145. https://doi.org/10.1016/j.fuel.2015.12.006.

[3] A. Howard, Why the Lessons of the Fulcrum Fiasco must not be Wasted, The Chemical Engineer (2024). https://www.thechemicalengineer.com/features/why-the-lessons-of-the-ful….