Machine Learning-Assisted Multiobjective Optimization of the Catalyst Layer in a Proton Exchange Membrane Water Electrolyzer

Catalyst layer (CL) optimization plays a crucial role in enhancing the performance of proton exchange membrane water electrolyzer (PEMWE). Herein, a multiobjective optimization framework was developed for fast PEMWE performance prediction and CL optimization toward increased current density, mass activity, and temperature uniformity. In this framework, a data-driven PEMWE performance prediction model is first constructed by incorporating the eXtreme gradient boosting algorithm into the 2D two-phase nonisothermal PEMWE agglomerate model. Using the PEMWE performance prediction model as a surrogate model, the second generation of nondominated sorted genetic algorithm was implemented into the CL parameter multiobjective optimization process. Accordingly, three sets of optimal CL parameters were selected using the technique for order preference by similarity to an ideal solution method. The performance of PEMWE with optimal CL parameters is significantly improved compared with that of the benchmark PEMWE physical model. The framework and results of this study provide important guidance for the optimization design of high-performance PEMWE.