An adaptive purge control strategy with spatial awareness of purge difficulty for hydrogen leakage in proton exchange membrane fuel cell stacks

Proton exchange membrane fuel cells are promising for vehicle power systems; however, hydrogen accumulation due to leakage within enclosed cavities remains a major barrier to inherent safety. To prevent the in-cavity hydrogen concentration from entering the flammable range, active purging with an inert gas is typically employed. Existing purging strategies are predominantly conservative and commonly rely on constant high flow rates, which makes it difficult to achieve rapid hazard mitigation while avoiding redundant high-flow purge operation. Accordingly, a dimensionless Purgeability Index is first proposed based on standardized purging tests, and a three-dimensional Purgeability Index Map is constructed to translate the cavity’s complex flow-topology structure into a quantitative field of purging difficulty. Building on this foundation, a spatial purge-difficulty–aware model-predictive feedforward–feedback control strategy is developed, in which spatial purging difficulty is mapped to the controller’s feedforward gain and the safety and purge-effort weighting is adaptively adjusted across leak locations. In addition, a probability density function of Purgeability Index over the entire cavity is established to quantify the volumetric fraction associated with different Purgeability Index ranges, thereby providing a mechanistic explanation for the inefficiency of constant-flow purging in dead zones. Finally, experiments and simulations demonstrate that, while safety-threshold constraints are continuously satisfied, cumulative purge-gas use is reduced by 51.89% and 20.90% in two representative high-Purgeability-Index regions relative to a constant high-flow baseline. In two low-Purgeability-Index dead-zone regions, the hazard-suppression time is shortened by an average of 0.83 s, while nitrogen consumption is simultaneously decreased by 33.08% and 20.58%.