Blended learning fitting algorithm for polarization curves of fuel cells

Fuel cell polarization curves, characterized by nonlinear models and the parameters of which are time-consuming to be identified, can represent fuel cell performance but will alter as the fuel cell degrades. For getting the information on degradation in time, a less time-consuming and an easily programmed algorithm, based on blended learning technique and linear least square estimation (LSE), is proposed to fit polarization curves obtained from the fuel cell systems. Simulations show that the proposed algorithm, compared with classical nonlinear LSE algorithms, converges much faster, features better extrapolation and less average quadratic error, and is easy to be programmed by C language. Therefore, the algorithm is a good option not only for fitting the polarization curves but also for implementation in embedded systems.