Artificial intelligence-driven modeling of biodiesel production from fats, oils, and grease (FOG) with process optimization via particle swarm optimization

Abstract

This study presents the design and optimization of a biodiesel production process, emphasizing the integration of machine learning (ML) models and process optimization techniques. The biodiesel production process involves multiple stages, including feedstock preparation, esterification, and transesterification, with catalysts Amberlyst-15 and KOH used in continuous stirred-tank reactors (CSTRs). Sensitivity analysis reveals that high conversions of free fatty acids (94 %) and triglycerides (97 %) are achievable under optimized operating conditions. To enhance process efficiency, adjustments to reaction temperature, time, and methanol-to-oil ratios are proposed, resulting in lower energy consumption and material costs. A ML model evaluation, using various algorithms, identify XGBoost, Extra Trees, Gradient Boosting, LGBM, and Random Forest demonstrate the best performer for predicting process parameters, achieving an R² value of nearly to 1. Particle Swarm Optimization (PSO) is then employed to optimize the selected ML model (XGBoost), leading to the identification of optimal input parameters for biodiesel production. The optimized process, combined with heat integration, reduces pre-heating energy requirements by 80.9 % and total heat duties by 19.9 %. The findings demonstrate the effectiveness of combining ML and optimization techniques to enhance biodiesel production efficiency while reducing costs and energy consumption.