Probabilistic and explainable data-driven modelling of SOC

Green hydrogen plays a vital role in the transition towards green energy and the decarbonization of the energy system. Its significance is particularly important in industries that significantly contribute to climate change, such as the carbon-intensive production of materials like steel, cement, fertilizers, and certain segments of the chemical industry. Hydrogen offers a means of decarbonization in these sectors either as a carbon-free, energy-dense fuel or as a necessary raw material in the production process. Solid oxide cell (SOC) technology is a highly promising advancement in the field of hydrogen technologies. It offers a unique solution by utilizing a single unit for electricity, heat, and hydrogen production through the use of solid oxide fuel cell and solid oxide electrolyser cell modes. Compared to other fuel cell technologies that rely on platinum catalysts, SOC systems utilize more affordable and readily available raw materials, such as nickel and steel, while also providing high fuel flexibility. Additionally, SOC technology boasts the highest conversion efficiency among fuel cell technologies, both in fuel cell and electrolysis regimes. As a result, SOC technology is a top contender for hydrogen production and various stationary applications. Notwithstanding its potential, achieving widespread commercialization of SOC technology remains challenging due to issues with performance and morphology degradation, as well as scale-up. The objective of this research is to develop probabilistic data-driven techniques that offer interpretable models, integrate expert knowledge to considerably reduce training duration, and enhance the model`s interpretability and its ability to be trained with limited datasets. The training of these methods with limited datasets is crucial because obtaining vast amounts of experimental data may not be feasible due to time or financial constraints. Therefore, the inclusion of expert knowledge becomes necessary to compensate for the shortage of data. The study will conduct a thorough examination of the operating conditions and fault specifications in SOEC/SOFC modes. The investigation will focus on a restricted number of irreversible fault modes, such as high fuel/steam utilization, carbon deposition, nickel reoxidation, YSZ crushing, and its deposition onto the catalyst, among others. The study will also assess the experimental requirements while considering resource and time limitations. To reduce carbon dioxide emissions, the electrochemical conversion of CO2 will be conducted to generate syngas and environmentally-friendly fuels.