(558b) Predictive Model and Bioconversion of Mixed Feedstocks

Authors: 
Narani, A., Lawrence Berkeley National Laboratory
Coffman, P., Lawrence Berkeley National Laboratory
Miller, M., Lawrence Berkeley National Laboratory
Tachea, F., Lawrence Berkeley National Laboratory
Chen, C. S., Lawrence Berkeley National Laboratory
Li, C., Idaho National Laboratory
Ray, A. E., Idaho National Laboratory
Pray, T., Lawrence Berkeley National Laboratory
Tanjore, D., Lawrence Berkeley National Laboratory
Currently, biorefineries are utilizing single feedstock inputs but in order to reduce feedstock dependency and the related risks, researchers are exploring the option of mixing multiple feedstocks. To understand the implication of mixing feedstocks on sugar yield, we relied on statistical design of experiments to build a predictive model for sugar yield based on varying feedstocks fractions, pretreatment types, and pretreatment reaction parameters. We performed pretreatments at low biomass loading, but for a biorefinery to be economically viable biomass deconstruction has to be performed at higher solid loading. Also, lignocellulosic sugars should be fermentable to produce high energy density molecules such as hydrocarbons.

In this study, three biomass feedstocks: corn stover, switchgrass, and energy cane were chosen based on the least cost formulation model developed by Idaho National Laboratory. These single and mixed feedstocks were pretreated with dilute alkali (DL), dilute acid (DA), and ionic liquid (IL) at severities predetermined by the statistical experimental design and were subsequently, enzymatically hydrolyzed with CTec2® at a loading of 10 mg protein/g glucan. The resulting sugar yields from all treatments, which varied from 25 to 100% w/w untreated biomass, were then fed into SAS JMP® to develop a predictive model that generated a continuous envelope of optimal feedstock mixtures. In order to achieve 90% sugar yield, we have to perform IL pretreatment at 140°C for 120 mins with 12%, 13%, and 75% (w/w) corn stover, switchgrass, and energy cane. More predictions of such kind were generated from this model that can potentially overcome the dependence on a single feedstock while maintaining the overall product yields.

These feedstock/treatment combinations, selected from the model, were tested at higher biomass loading to study the impact on downstream fermentation process to produce ethanol and hydrocarbon, such as bisabolene. At higher biomass loading, mixed feedstock conversion to ethanol performed very similar to single feedstock conversion to ethanol. Rheological characterization was performed on high solid pretreated biomass mixtures to better understand mixing, material handling, and mass transfer issues. This information will be applied in techno-economic analysis of this end-to-end study of deconstruction of optimal biomass mixtures at high solid loading for ethanol and hydrocarbon production.