A Parameter Estimation Method for Fuel Cell Diagnostics

Researchers and developers typically use I-V curves (current vs. potential) as a means to demonstrate good data fitting of a model under experimental validation. Despite the popularity of this method, assessment of the results can be intuitive rather than mathematical, and usually, when such a demonstration occurs, information on the quality of the parameter estimation is not given. On the contrary, it has been shown that simple model fitting on I–V curves cannot give statistically reliable results on the values of the model parameters since the confidence intervals are quite high. In previous work we used a model-based Design of Experiments method to establish optimal measurements on SOFC’s, in order to improve the parameter estimation’s reliability. Based on this theoretical analysis we proceeded to experimental validation with measurements on SOFC button cells. This work led to the development of a different method for sampling and treatment of data taken from I-V curves. This method facilitates production of histograms for each parameter –instead of simple values– allowing the researcher to observe their stochastic behaviour and assess the discrepancies of the calculated parameters. Furthermore the calculation of their variances and covariances are based on experimental data instead of the theoretical linear approach of the output model’s partial derivatives with respect to the parameters. In this work we present part of the produced results from this method using an advanced electrochemical and mass transport model for the cell. The proposed method can be used as a tool to improve model validation procedures, for fuel cell diagnostic applications, research on degradation etc.

Presented at:
10th European Fuel Cell Forum, Lucerne, Switzerland, June 28 - July 1st, 2011

Note: The status of this file is: Involved Laboratories Only

 Record created 2011-05-23, last modified 2018-01-28

External link:
Download fulltext
Rate this document:

Rate this document:
(Not yet reviewed)