PREDICTIVE MODELLING AND AI-BASED OPTIMIZATION OF INDUSTRIAL BIOETHANOL PRODUCTION: EFFECTS OF FEEDSTOCK YIELD, PROCESS EFFICIENCY AND ENERGY CONSUMPTION

Abstract

Industrial bioethanol production requires efficient use of feedstock, processing technologies, and energy resources to improve production performance and sustainability. The study investigated the impact of feedstock yield, processing technology performance, and energy usage on industrial bioethanol production and the potential for applying predictive modelling and artificial intelligence for optimization. A secondary data of 2,000 observations and 15 variables related to production, economic, environmental and market were analysed. Production patterns and relationships among major variables were described using descriptive statistics and correlation analysis. Production capacity and bioethanol growth were predicted by Multiple Linear Regression, Random Forest, Gradient Boosting and XGBoost models. The coefficient of determination, mean absolute error and root mean square error were used to evaluate the models, and the feature importance analysis was used to determine the important predictor variables. Results indicated a weak direct link between feedstock yield, processing efficiency, energy use and production. The importance of the models was comparatively higher for processing technology efficiency, but no single predictor was much more important than the others for production performance. Predictive ability of all predictive models evaluated was poor, with negative test set R² values. These results indicate that AI optimization can be a potential exploratory analytical technique, but more process-level information is needed to apply it successfully in industry. Future studies should combine detailed fermentation, pretreatment, feedstock composition, conversion-yield and real-time operational variables together to enhance predictability and facilitate more effective bioethanol production optimization.