Fast internal distribution prediction of key parameters in a tubular SIS-SOFC based on the POD-ANN reduced-order model

Predicting the internal distributions of key parameters is critical for the development of solid oxide fuel cells but is complicated and time-consuming. In this study, a framework for fast parameter distribution prediction based on proper orthogonal decomposition (POD) and an artificial neural network (ANN) is proposed and applied to a tubular 20-cell segmented-in-series solid oxide fuel cell. The characteristics of the POD mode for temperature, hydrogen and potential distributions are analyzed, and the variation in the coefficient with respect to the operating parameters is investigated. The predicted distributions under 20 random operating conditions are compared with those simulated from the multi-physics model. The calculation time is reduced from 14 h to 24 min to 160 ms. The distributions predicted by the POD-ANN model have good agreement with the results simulated by the multi-physics model, both globally and locally.