Commercial-scale bio-refineries are designed to process 2000tons/day of single lignocellulosic biomass. Several geographical areas in the United States generate diverse feedstocks that, when combined, can be substantial for bio-based manufacturing. Blending multiple feedstocks is a strategy being investigated to expand bio-based manufacturing outside Corn Belt. In this study, we developed a model to predict continuous envelopes of biomass blends that are optimal for a given pretreatment condition to achieve a predetermined sugar yield or vice versa. For example, our model predicted more than 60% glucose yield can be achieved by treating an equal part blend of energy cane, corn stover, and switchgrass with alkali pretreatment at 120°C for 14.8h. By using ionic liquid to pretreat an equal part blend of the biomass feedstocks at 160°C for 2.2h, we achieved 87.6% glucose yield. Such a predictive model can potentially overcome dependence on a single feedstock.

Previously, a predictive model was developed to identify optimal blends of expensive high-quality and cheaper low-quality feedstocks for a given geographical location that can deliver high sugar yields. In this study, the optimal process conditions were tested for application at commercially-relevant higher biomass loadings. We observed lower sugar yields but 100% conversion to ethanol from a blend that contained only 20% high-quality feedstock. The impact of applying this predictive model simultaneously with least cost formulation model for a biorefinery location outside of the US Corn Belt in Lee County, Florida was investigated. A blend ratio of 0.30 EC, 0.45 SG, and 0.25 CS in Lee County was necessary to produce sugars at high yields and ethanol at a capacity of 50 MMGY. This work demonstrates utility in applying predictive model and LCF to reduce feedstock costs and supply chain risks while optimizing for product yields.