Optimal design and flow-field pattern selection of proton exchange membrane electrolyzers using artificial intelligence

The design of a proton exchange membrane (PEM) electrolyzer and various hardware options remain key research areas for green hydrogen production technology. We have proposed an approach to select optimum design factors, including a flow-path design for the PEM electrolyzer using a machine learning (ML) technique for the first time. Efficient multiple ML models employing k-nearest neighbors and decision tree regression approaches predict the optimal hydrogen-generating system design for the PEM electrolyzer cell. The proposed ML model was trained and validated using 1062 design data points. The model predicts 17 parameters of the electrolyzer assembly for five input parameters: the hydrogen production rate, electrode area, anode flow area, cathode flow area, and type of cell design (e.g., single or stack). The model shows an absolute mean square error of 0.31 when compared to the experimental results (e.g., potential), which indicates that the model has excellent reliability. Finally, this study presents that the ML model can predict the optimal design of a PEM electrolyzer for commercial-scale hydrogen production rates at 50–3000 mL/min. This research will contribute to reducing the cost and time required to develop future water electrolyzers for hydrogen production.