Data-driven optimization of proton exchange membrane fuel cell cold start strategies for safe operating boundary

Current loading strategy is crucial for the cold start of proton exchange membrane fuel cells (PEMFC). This study proposes a two-dimensional transient model for the cold start of PEMFC, incorporating electrochemical reactions, phase transitions, and mass transport processes within catalyst agglomerates. The performance of constant current, constant slope current, and combined current loading strategies during the cold start process is evaluated, revealing failure mechanisms of startup under different strategies. Results indicate that combined current loading enables faster startup, with specific current parameter ranges for successful startup. Within this range, as current parameters increase, startup time is decreased while minimum voltage drops. To further determine safe operational boundaries for successful startup and shorten startup time, a data-driven surrogate model is proposed to predict the cold start process by combining a multilayer perceptron with a recurrent sliding window approach. The safe operating boundaries at different initial temperatures have been determined through the prediction model. At −20 °C, compared to constant slope current, combined current loading can shorten the startup time by 34.5 %. Furthermore, to balance startup time and minimum voltage, the game theory diagrams are provided to facilitate the rapid selection of current parameters based on practical requirements.