Hydrogen production by a PdAg membrane reactor during glycerol steam reforming: ANN modeling study

In the present work, an artificial neural networks (ANNs) model has been developed for investigation of glycerol steam reforming (GSR) process with PdAg membrane reactor (MR) in the presence of Co/Al2O3 catalyst. Reaction pressure and sweep factor as independent variables (Inputs) and glycerol conversion, hydrogen recovery, hydrogen yield, H2 selectivity, CO selectivity and CO2 selectivity as dependent variables (outputs) are chosen for ANN modeling of GSR. The ANN model was developed by feed-forward back propagation network with trainlm algorithm and topology (2: 10: 6) and Sigmoid transfer function for hidden and output layers. A good agreement between predicted values using ANN with experimental results was observed (R2 and MSE values were 0.9998 and 3.48 × 10−6 (based on normalized data), respectively). Modeling results indicated that all selected factors (reaction pressure and sweep factor) were effective on output variables. It was found that the reaction pressure with a relative importance of 59% was the most effective parameter in the GSR process with PdAg MR in the presence of Co/Al2O3 catalyst.