Statistical method-based calibration and validation of a solid oxide fuel cell model

A 2-stage model validation strategy for the previously developed solid oxide fuel cell model using the data from custom-designed experiments is presented. The strategy is based on the identification of model structural and parametric errors. In the preliminary validation, the causes that can result in the voltage error during changing temperature and fuel flow rate conditions are analysed. It is identified that the convection heat transfer process contributes significantly towards the cell performance in a temperature controlled test environment. Rectification of this error results in the reduction of the maximum voltage error from 14% to 0.5%. Graphical methods for data visualisation are utilised to examine goodness of fit of the model. Input sensitivity analysis reveals that the air flow rate has negligible influence on the output quantities of current density, fuel utilisation, and cell temperature owing to temperature-controlled conditions of the test. Parameter sensitivity analysis through the elementary effects method reveals that most of the electrochemical parameters in general and the activation energies in particular have dominant effects on the considered system outputs. Model validation is carried out through a classical statistical method of hypothesis testing by using the parameter uncertainty information obtained through nonlinear least squares fitting. The efficacy of the model validation strategy is demonstrated through the model acceptance with 8% maximum error in cell current.