Improving catalysts and operating conditions using machine learning in Fischer-Tropsch synthesis of jet fuels (C8-C16)

Fischer-Tropsch synthesis (FTS) offers a promising route for producing sustainable jet fuels from syngas. However, optimizing catalyst design and operating conditions for the ideal C8-C16 jet fuel range is challenging. Thus, this work introduces a machine learning (ML) framework to enhance Co/Fe-supported FTS catalysts and optimize their operating conditions for a better jet fuel selectivity. For this purpose, a dataset was implemented with 21 features, including catalyst structure, preparation method, activation procedure, and FTS operating parameters. Moreover, various machine-learning models (Random Forest (RF), Gradient Boosted, CatBoost, and artificial neural networks (ANN)) were evaluated to predict CO conversion and C8-C16 selectivity. Among these, the CatBoost model achieved the highest accuracy (R² = 0.99). Feature analysis revealed that FTS operational conditions mainly affect CO conversion (37.9 %), while catalyst properties were primarily crucial for C8-C16 selectivity (40.6 %). The proposed ML framework provides a first powerful tool for the rational design of FTS catalysts and operating conditions to maximize jet fuel productivity.