Adaptive sliding mode observer for automotive PEMFC membrane water content estimation under hybrid dynamic tests

Proton exchange membrane fuel cell is an ideal solution to the energy shortage and environmental pollution. The membrane water content (MWC), as a crucial internal state, indicates two major failures: drying and flooding, but cannot be measured directly. In order to estimate the MWC in real time, a model-based adaptive sliding mode observer is proposed as a virtual sensor. Compared to the classical sliding mode observer, the errors between reference and estimated values are considered by the proposed adaptive sliding mode algorithm, which demonstrates better performance. Firstly, the convergence of algorithm is validated. The MWC converges to the same value in 5 s starting with different values, where the error raised by initial value was promptly eliminated. The performance of the proposed observer and sliding mode observer is compared: mean average errors are reduced by 42.5%, 48.5% and 60.6%; mean square errors by 30.8%, 13.3% and 38.1%; root of mean square errors by 30.9%, 12.9% and 35.4%, respectively. Finally, under the degradation test, relationship between the MWC over 0–1000 h and degradation is explored, concluding that enhancing the membrane hydration has a positive effect on the performance of the PEMFC in the early stage, however, as the extent of hydration increases, the liquid water accumulates, inducing a degradation of the PEMFC performance.