Artificial neural network modeling of forced cycling operation between propane steam reforming and CO2 carbon gasifier

This investigation has employed artificial neural network (ANN) modeling to describe the complex relationship between the forced cycling parameters and the reactor performance during periodic operation between propane steam reforming and CO2—carbon gasifying agent. Experimental data from our laboratory were assessed against different ANNs and based on a 2-way ANOVA treatment of various error indices, a two-hidden layer network with 5 neurons emerged as the best model for both descriptive and predictive purposes. Cycle split has the most significant (85%) positive effect on the improvement in H2 and CO production and the appearance of resonant peaks while cycle period appeared to have detrimental effect on product yield.