Dynamic inconsistent analysis and diagnosis of abnormal cells within a high-power fuel cell stack.

• Dynamic inconsistent behavior of the fuel cell stack with severe aging cells is investigated. • Theoretical changing trend of voltage difference between normal and abnormal cells is derived. • Slower dynamic response inside the abnormal fuel cell compared to normal cells is revealed. • Deterioration of the abnormal fuel cell to the dynamic consistency of the stack is quantified. • Accurate diagnosis of in-stack abnormal cells based on Bollinger band and entropy is realized. The abnormal cells within a fuel cell stack can reduce the stack's lifetime and reliability due to the "barrel" effect. In this paper, the dynamic inconsistency of a 20-kW fuel cell stack with two abnormal cells is studied. These abnormal fuel cells with severe aging have a lower voltage but without faults. First, the voltage difference change trend between abnormal and normal fuel cells under different current densities is analyzed by theoretical derivation and experiments. Further, the inconsistent dynamic response of the fuel cell stack under different current steps is investigated. The results show that the voltage difference's variation rate between normal and abnormal fuel cells is greater than 0.005 V·cm2/A at any current density. The particular is that this variation rate increases significantly at high current densities due to the increasing concentration polarization loss inside the abnormal cell. Besides, the slower dynamic response of abnormal cells during the current step deteriorates the stack's dynamic consistency. The voltage standard deviation of the stack is reduced by a maximum of 37.5% when abnormal cells are not taken into account. Therefore, a diagnosis method utilizing voltage information with small computation is proposed. The method can accurately identify in-stack abnormal fuel cells by combining qualitative analysis of voltage values based on the Bollinger band and quantitative measurement of voltage fluctuation degree using Shannon entropy. These efforts provide guidance for fuel cell system control and fault diagnosis. [Display omitted] [ABSTRACT FROM AUTHOR]