Identifiability study and receding-horizon experiment design for solid oxide fuel cell

In this work the identifiability and estimability of a seventeen-state nonlinear solid oxide fuel cell model are studied and the results are used to design experiments for parameter estimation. A model is identifiable if the value of each model parameter can be uniquely determined under ideal conditions. A model is estimable if each parameter can be determined from a specific experimental design or data set. The identifiability of the fuel cell model was tested by partitioning the main model into four submodels whose identifiability could be verified. The estimability of the model was studied using sensitivity analysis. The model of interest was found to be identifiable in all parameters. The estimability of parameters was tested by varying the external load as a step and also as a square pulse. The charge-transfer capacitance was found to be inestimable from a step change in the external load for a specific cut off value, whereas all five parameters are estimable from a square pulse design of the external load input. This study forms a basis for experimental studies.;Based on the estimability analysis, a new method to design experiments for parameter estimation is proposed. A receding horizon approach by selecting an optimum input is utilised. The proposed method was tested on a numerical example to examine its effectiveness before it was applied to the fuel cell model. Some of the issues related to the application of the proposed method were examined and guidelines for selecting appropriate experimental settings are provided.