Comparative Kinetic Modeling of the Fischer–Tropsch to Olefins Process Over Fe‐Cu‐K/AC

ABSTRACT

This study compares mechanistic, data‐driven, and hybrid‐surrogate modeling strategies for the kinetics of the Fischer‐Tropsch‐to‐olefins (FTO) process over a Fe‐Cu‐K/AC catalyst. Two mechanistic formulations were evaluated: the alkenyl model (ALK) and an alkenyl model including olefin re‐adsorption (ALK‐OR). The kinetic parameters were estimated using 43 experiments and validated against seven independent tests. The ALK model generally provided lower errors for light olefin predictions, particularly at low CO conversion, where secondary reactions are less pronounced. In contrast, the purely data‐driven ANN showed limited generalization because of the small experimental dataset and the absence of embedded mechanistic constraints. A supplementary hybrid ANN analysis, trained using additional mechanistic‐model‐generated data, was therefore included to contextualize the potential of ANN surrogates for computational acceleration. Overall, the results show that physically informed kinetic models remain essential for data‐scarce catalytic systems, while hybrid ANN approaches may be useful when they preserve mechanistic structure and are trained within a well‐defined operating domain.