Novel data-driven methodologies for parameter estimation and interpretation of fuel cells performance

Fuel cell based power generation systems are expected to become more widespread in the near future. Stationary fuel cells may be used as an uninterruptible or back-up power supply, or to supply micro-grids. In particular, proton exchange membrane fuel cells (PEMFC) are an attractive technology due to its high energy density, rigid and simple structure, low operating temperature and fast start-up characteristics. The power quality assessment of fuel cells as a viable power sources requires a good understanding of the fuel cell performance characteristics. This paper presents two novel data-driven methodologies for the identification of the main steady state (polarization curve) and the dynamic (impedance response) characteristics for fuel-cells allowing the development of rapid, accurate and empirical models based on the experimental data. M-NMF is a modified non-negative matrix factorization technique developed for the analysis of polarization curve data that allows to identify the three main contributions for the fuel-cell power degradation, while for impedance spectroscopy data, this paper proposes the use of fractional order transfer functions (FC-FOTC) to describe the main dynamic modes present in the fuel-cell. A brief description of these two approaches is presented, together with the analysis of a real experimental dataset obtained from a 12W open cathode PEMFC stack to illustrate their potential and scope. While the former is instrumental for the deconvolution of the fuel cell polarization curves into its major components, the latter enables the estimation of the parameters related to the inherent transport and kinetic phenomena, thus opening the way, in both cases, for the interpretation of the effect of the operating conditions on the relative dominance and magnitude of these components and phenomena