Valorization of lignocellulosic feedstock in bioethanol production via separate hydrolysis and fermentation: An integrated machine learning approach

The increasing demand for sustainable energy sources has intensified research on lignocellulosic biomass for bioethanol production. This study investigates the valorization of pistachio shells as a feedstock for bioethanol production through separate hydrolysis and fermentation processes , incorporating ultrasonic and hot water pretreatments . The effectiveness of these pretreatments in delignification was evaluated, with ultrasonic treatment achieving a lignin removal rate of 36 %, compared to 27 % for hot water pretreatment. Enzymatic hydrolysis was performed using Cellic CTec2 at enzyme loadings of 30–60 FPU, resulting in reducing sugar concentrations of 7.34 g/L and 5.98 g/L for ultrasonic and hot water pretreated samples, respectively. Fermentation with Saccharomyces cerevisiae yielded a maximum ethanol concentration of 1.81 mg/mL and a fermentation efficiency of 52 %. Machine learning models, including XGBoost and Random Forest, were employed to predict reducing sugar concentrations with high accuracy (R 2 > 0.94), demonstrating their potential in optimizing bioethanol production. This study highlights the synergistic impact of advanced pretreatment methods and computational modeling in enhancing bioethanol yields, contributing to sustainable biofuel development. • Pistachio shells were assessed as a feedstock for bioethanol via SHF. • Ultrasonic pretreatment removed 36 % lignin, improving enzymatic hydrolysis. • Cellic CTec2 (60 FPU) yielded 7.34 g/L reducing sugars from ultrasonic pretreated samples. • Saccharomyces cerevisiae fermentation produced 1.81 mg/mL ethanol at 52 % efficiency. • XGBoost and Random Forest predicted sugar concentrations with R 2 > 0.94.