Predicting bioconversion from physical and chemical characteristics of multiple lignocellulosic biomass types

Blending biomass is a concept that allows incorporation of additional biomass into the bioenergy supply chain rather than relying on a single, abundant, low-cost feedstock. Quality and variability become increasingly important as the range of feedstocks used for bioenergy broadens. Understanding the relationships between feedstock quality attributes and conversion is important for supplying feedstock that meets conversion needs. The project objectives were to develop models to predict convertibility of dilute-acid pretreated biomass from physical and chemical characteristics of biomass, and determine characteristics most important for prediction of convertibility. Models will further be applied to predict conversion of blended feedstocks. Thirteen samples were included in the models from seven biomass types, including corn stover, switchgrass, Miscanthus, lawn clippings, sorghum, wheat straw, and non-recyclable paper. Physical and chemical properties measured were composition, fuel properties, particle size, and elemental ash. The response variables used to develop the conversion models were glucose and xylose released from pretreatment and enzymatic hydrolysis. Predictor variables determined to have the greatest impact on conversion based on this preliminary dataset were comprised mostly of composition and fuels properties variables. Partial least-squares regression models could be formed to predict sugar yields from dilute-acid pretreatment and enzymatic hydrolysis based on physical and chemical properties; however, results are preliminary due to a limited number of calibration samples. Future work will focus on (1) adding more calibration samples and predictor variables, including pore area, water retention value, and lignin functional groups, and (2) reselecting important predictor variables based on a larger calibration set.