Investigating Reliability on Fuel Cell Model Identification. Part II: An Estimation Method for Stochastic Parameters
An alternative way to process data from polarization measurements for fuel cell model validation is proposed. The method is based on re- and subsampling of IV data, with which repetitive estimations are obtained for the model parameters. This way statistics such as standard deviations and correlations between the parameters may be experimentally derived. Histograms may also be produced, approximating the probability distributions that they follow. Two experimental case studies are discussed. In the first case, observations are made on the behavior of the parameter values for two mathematical models. As the number of data points (measurement points) employed in the estimation of the parameters increases, parameters with high variances converge to specific values. On the contrary, parameters with small variances diverge linearly. The parameters' histograms do not usually follow normal distributions rather they show a connection between the number of peaks in the graphs and correlations of the parameters. The second case study is an application on a fast degraded SOFC button cell, where the values and the histograms of the parameters are compared before and after degradation.
Keywords: Data Fitting ; Design of Experiments ; Diagnostics ; Fast Degradation ; Identification ; Parameter Estimation ; Polarization Curves ; Robust Regression ; Solid Oxide Fuel cells ; Stein's Paradox
Record created on 2013-02-27, modified on 2016-08-09