Identification of acceleration parameters for SOC degradation: Data-driven and Experimental Approaches

Solid oxide cells (SOCs) have the potential to enable the transition to a greener energy system. These electrochemical devices convert the chemical energy of fuels to electrical energy without being limited by Carnot cycle efficiency. Although SOCs have reached a high level of technology readiness between 6-9, limited lifetimes continue to challenge the goal of large scale commercialization. Thus, novel approaches for faster, cheaper and reliable degradation studies are required. Accelerated Stress Testing (AST) is a proven way to ensure reliable testing of shorter durations across various manufacturing fields. To develop AST protocols for SOCs, relationships between the degradation mechanisms and the relevant stresses need to be mapped and tested. This was the primary motivation for this work.

A data-driven approach using Machine learning (ML) algorithms was proposed to investigate any underlying patterns between different operating parameters and cell performance across cell generations. To draw broad generalizations, a large number of data points are needed. A unique dataset of 2135 in-house SOC tests was created for this purpose. The test conditions recorded in the form of time series, characterization tests and other metadata were saved into a single file. Exploratory data analysis and visualizations were done on the dataset to gain further insight into the different test conditions. The correlation between chosen target parameters such as the change in resistance, voltage drop after a specific time etc. were modelled using a selection of regression algorithms. Such an approach would aid in identification of relevant stress parameters based on the extent of degradation observed across different tests without human bias.

Microstructural changes have been studied widely in the past to understand the effect of operating parameters on the long-term cell behaviour. The fuel electrode was the primary focus of this study because previous studies have shown that fuel electrodes contribute to the majority of the degradation contributions in SOC operating conditions. In this work, cells were characterized after more than 20,000 hours of testing in the field using microscopy and spectroscopy techniques to identify the changes observed in the fuel electrodes. Anode Supported Cells (ASCs) with Ni-YSZ and Electrolyte Supported Cells (ESCs) with Ni-CGO fuel electrodes were investigated in this work. Observations were made in terms of Ni migration, agglomeration, pore formation, etc. depending on the operation mode and electrode composition.

A detailed experimental campaign was designed to evaluate three chosen parameters - steam, temperature and current density - based on literature, field test analysis and ML models. For this purpose, firstly a deconvolution of the polarization contributions and evolution over long-term operation was carried out. Thereafter, the ASCs and ESCs were aged under various conditions using in-situ (under current load) and ex-situ (at OCV conditions) treatments for 1000 hours. The cells were characterized using electrochemical tools during operation. In addition, pre- and post-test characterization were performed
on the cells using impedance and microscopy techniques. Steam was shown to stimulate the majority of the degradation seen on the fuel electrode in Ni-YSZ cells under nominal fuel cell and electrolysis operation, but it caused additional degradation in case of ESCs. This was investigated further with the help of Raman spectroscopy to probe the surface changes observed primarily on the 3YSZ electrolyte. Further, the presence of impurities such as Si were observed to have a significant effect on the Ni-CGO based electrodes due to the formation of cerium silicates.

Finally, the lab tests were assessed by comparison with the microstructural changes investigated in the field-tested samples to ensure that the changes observed under the chosen ageing approach using specific parameters induced the same degradation mechanisms as observed in on-field long-term tests.