Bayesian Optimization of Electrode Conditioning of Ni(‐Fe) Electrodes for the Alkaline Oxygen Evolution Reaction

Efficient alkaline water electrolysis requires highly active electrodes for the sluggish oxygen evolution reaction (OER). NiFeO x H y materials are among the most active OER electrocatalysts, and their activity can be tailored, among other methods, by electrochemical conditioning. However, there is a lack of systematic approaches to optimize the conditioning process to achieve the best electrode activation. A promising way to develop such approaches is to use mathematical models. While mechanistic models are not readily available and hard to develop, data‐driven models might offer a straightforward alternative. The use of Bayesian optimization (BO) with Gaussian processes to improve the electrode conditioning process of a Ni‐Fe bulk electrode is proposed. With this approach, an electrode conditioning process yielding a stronger activity enhancement compared to previous manual optimization is identified; at the same time, fewer experiments are also required. It is further shown that this also allows to transfer knowledge to new materials: transfer learning starting from the experimental data for the Ni‐Fe electrode allows optimization of the conditioning of a Ni electrode with fewer experiments than applying BO to the Ni electrode from scratch. Overall, the potential of using data‐driven numerical optimization in a hardware‐in‐the‐loop approach for electrode conditioning is highlighted.