Multi-objective optimization of two-phase flow in the proton exchange membrane fuel cells based on a data driven surrogate model

Abstract

Flooding is detrimental to Proton Exchange Membrane Fuel Cells (PEMFCs). Reducing two-phase pressure drop and flooding relief time alleviates the devastating effect of flooding. To this end, an innovative surrogate model based on Artificial Intelligence is developed for flooding conditions. Volume of Fluid is employed to capture two-phase flow during the flooding in the simulations. A dataset is generated by 3D two-phase CFD simulation. An Artificial Neural Network is developed based on the dataset to predict the dynamic behavior of two-phase during flooding in time series. Flooding criteria are formulated based on the time series. The results show that the model predicts two-phase flow behavior accurately. Low gas velocity compels small parasitic power, but flooding conditions are exacerbated by the low flow rate, especially in the geometries with corners and virtual bottlenecks. Two distinct methods are used to develop multi-objective optimization, fed by the presented formula. Features are cathode channel characteristics, and objectives are relief time and stabled two-phase pressure. The optimum case brings a time of 9.11 ms and a pressure drop of 27.44 Pa. Also, it reduces the stable pressure drop and the time by 83% and 90%, compared with the worst conditions.