Data-Driven Modeling of Long-Term Performance Degradation in Solid Oxide Electrolyzer Cell System

One key challenge of Solid Oxide Electrolyzer Cell (SOEC) systems is degradation over long periods of time. Degradation decreases efficiency by increasing the electrical energy required for H2 production. This paper presents the first step in managing long-term degradation in SOEC systems. In this work, the first data-driven dynamic model for the prediction of performance degradation in SOECs as a function of humidity, operating temperature, and current density was developed. The model was trained using experimental data from multiple data sets in the literature under various conditions. The model showed good agreement with validation data over 7000 h operation. One key finding is that the data show there are three distinct time regimes in which degradation behaviour is qualitatively different. This is likely due to different degradation phenomena, although the specific phenomena have not yet been isolated. This is significant for PSE applications because operators can choose to vary the operating conditions over time in order to predict, account for, or minimize the effects of long-term degradation.