Gas composition modeling in a reformed Methanol Fuel Cell system using adaptive Neuro-Fuzzy Inference Systems

This work presents a method for modeling the gas composition in a Reformed Methanol Fuel Cell system. The method is based on Adaptive Neuro-Fuzzy-Inference-Systems which are trained on experimental data. The developed models are of the H2, CO2, CO and CH3OH mass flows of the reformed gas. The ANFIS models are able to predict the mass flows with mean absolute errors for the H2 and CO2 models of less than 1% and 6.37% for the CO model and 4.56% for the CH3OH model. dels have a wide range of applications such as dynamic modeling, stoichiometry observation and control, advanced control algorithms, or fuel cell diagnostics systems.