Title : Machine learning-enabled techno-economic and environmental analysis of succinic acid production from biodiesel byproduct glycerol
Abstract:
This study establishes a machine learning (ML)-enabled framework for a comprehensive techno-economic analysis of bio-based succinic acid (SA) production using glycerol derived from biodiesel production. The primary objective is to optimize the minimum product selling price (MPSP) across varied operational and economic conditions. Our baseline production model assumes a daily glycerol input of 9,792 kg, obtained from a large-scale biodiesel mill producing 80,000 kg/day of biodiesel, with an SA production target of 5,739 kg/day. Using Yarrowia lipolytica as a biocatalyst, the process integrates SA production within a biodiesel facility to explore cost-efficient resource utilization. Through automated simulations in SuperPro Designer via the COM interface, the ML framework evaluates the impacts of varying glycerol concentration in the crude glycerol feedstock, solvent prices (trioctylamine, 1-octanol, and trimethylamine), labor costs, and SA production throughput. Machine learning models are trained on simulation data to predict MPSP, unit product cost, IRR, net profit, and revenue, providing rapid sensitivity analyses across economic and operational scenarios. This approach reduces reliance on time-intensive, traditional simulations by enabling scalable, data-driven economic analysis. Findings indicate that variations in crude glycerol concentration and solvent pricing are significant factors in optimizing production costs, with labor costs and throughput also impacting MPSP. The integration of SA production with biodiesel facilities provides economic synergy, making this ML-enabled framework a valuable tool for developing cost-effective bio-based SA production pathways, particularly in biodiesel-producing regions like Brazil.
