An Assessment of the Dilute Acid Pretreatment of Coastal Bermudagrass for Bioethanol Production

Abstract

There is a clear interest domestically to examine alternative liquid fuels which are more sustainable and environmentally friendly than gasoline. Bioethanol is a leading candidate for this replacement, but limitations exist on current starch based production. As a result, lignocellulosics are being examined. Lignocellulosics require a pretreatment step to degrade the biomass enough to allow enzymes to access to the carbohydrates. Dilute acid pretreatment has been demonstrated across many lignocellulosic feedstocks as a leading method compared to other pretreatment options. Coastal bermudagrass was identified as a promising lignocellulosic feedstock for bioethanol production. It is well suited for the Southeastern United States where it is currently grown for nutrient management in concentrated animal farming operations and as a source of hay. In a full factorial experimental design, dilute sulfuric acid pretreatment was used to pretreat coastal bermudagrass at 120, 140, 160, and 180ºC using 0.3, 0.6, 0.9, and 1.2% w/w sulfuric acid over residence times of 5, 15, 30, and 60 minutes. After enzymatic hydrolysis, the highest yield of total sugars (combined xylose and glucose) was 97%, which was shared by the pretreatment condition combinations of 0.9% acid at 160ºC for 15 minutes and 1.2% acid at 160ºC for 5 minutes. The prehydrolyzate liquor was analyzed for inhibitory compounds (furfural, hydroxymethylfurfural (HMF)) in order to assess potential risk for inhibition during fermentation. Accounting for the inhibitory compounds, a pretreatment with 1.2% acid at 140ºC for 30 minutes with a total sugar yield of 94% o r0.9% acid at 160 ºC for five minutes with a total sugar yield of 91% may be more favorable for fermentation because furfural levels remain under the inhibitory threshold concentration of 1 g/L. Additionally, due to significant interactions between factors, there are likely optimal pretreatment condition combinations possible other than those found experimentally. Both kinetic and multiple linear regression (MLR) models have been developed in other studies to describe dilute acid pretreatment, however no study has yet applied an artificial neural network (ANN). In this study, the utility of an ANN was assessed for modeling the dilute acid pretreatment of coastal bermudagrass using statistics that quantified the error between the predicted data and actual data and through a comparison with an MLR model. The statistics used were the coefficient of determination (R2), the root mean squared error (RMSE), and the root percent deviation (RPD). A standard 2nd-order polynomial multiple linear regression (MLR) model was developed for comparison with the ANN model. Time (minutes), acid concentration (% w/w), and temperature (ºC) were input into the models to generate xylose in the prehydrolyzate (PreH), glucose in the PreH, furfural in the PreH, HMF in the PreH, xylose in the enzymatic hydrolyzate (EH), and glucose in the EH. It was found that the two types of models predicted most of the outputs closely with the exception of the xylose in the PreH, which the ANN predicted more accurately. An ANN model with six hidden layer neurons was found to be the best overall model and confirmed the utility of utilizing ANN modeling in the area of biomass pretreatment.