Fischer – Tropsch synthesis over Co-Ni/Al2O3 catalyst: Comparison between comprehensive kinetic modeling, Artificial Neural Network, and a novel hybrid GA-Fuzzy models

Abstract

BackgroundFischer – Tropsch synthesis (FTS) is a direct route for producing liquid fuels. MethodsTo predict the behavior of FTS over Co-Ni/Al2O3 catalyst, three different models including a novel hybrid GA-Fuzzy model, an Artificial Neural Network (ANN) model, and a comprehensive kinetic model were developed. The models’ outputs were CO conversion and the selectivity of CH4 and C5+, whereas the models’ inputs were pressure, H2/CO ratio, temperature, and GHSV. For the GA-Fuzzy model, a novel technique was developed to optimize the qualitative definition of input/output variables along with the fuzzy rules to improve the model performance. For the ANN model, an optimum two-layer neural network was developed. For the kinetic model (derived based on Langmuir-Freundlich technique), the elementary reactions were suggested based on alkyl mechanism and included re-adsorption and hydrogenation on the secondary active sites of the catalyst. Significant FindingsThe models’ prediction errors were 8.27%, 4.59%, and 12.56% for GA-Fuzzy, ANN, and kinetic models, respectively. The GA-Fuzzy model had higher interpretability, transparency, and generality compared to the ANN model. Besides, the kinetic model provided a more realistic representation of the reactor behavior. The optimized operating conditions of the reactor were also found, which were 240°C, 24 bar, 1233.78 h−1 for GHSV, and 2.5 for H2/CO ratio.