Data-driven degradation analysis of proton exchange membrane electrolyzer stacks under dynamic load operation

Dynamic load operation of proton exchange membrane (PEM) electrolyzers offers a significant opportunity to reduce the cost of green hydrogen, by quickly adjusting their hydrogen production rates in response to low electricity prices. However, this operation mode affects the degradation rate of PEM electrolyzer stacks differently compared to steady-state operation. This study aims to analyze the effects of dynamic load operation parameters on the degradation rate of PEM electrolyzer stacks and to provide suggestions to mitigate some of their negative implications. A feed-forward deep neural network model was developed using experimental data from two 50 kW PEM stacks (each operated for 1134 h). This model was able to evaluate the PEM stack degradation at different operating conditions. It was observed that partial load operation decreased the degradation rate of PEM electrolyzer stacks by up to 69%. Moreover, the degradation rate at 0.3 A/(cm 2 ⋅ s) rate of change in current density and low load cycling time was 266% higher than the degradation rate during steady-state operation. Although dynamic load operation increases the degradation rate compared to steady-state operation, optimizing the operation parameters can limit this increase in the degradation rate.