Enabling thermal-neutral electrolysis for CO₂-to-fuel conversions with a hybrid deep learning strategy

High-temperature co-electrolysis of CO₂/H₂O through the solid oxide electrolysis cells (SOECs) is a promising method to generate renewable fuels and chemical feedstocks. Applying this technology in flexible scenario, especially when combined with variable renewable powers, requires an efficient optimisation strategy to ensure its safety and cost-effective in the long-term operation. To this purpose, we present a hybrid simulation method for the accurate and fast optimisation of the co-electrolysis process in the SOECs. This method builds multi-physics models based on experimental data and extends the database to develop the deep neural network and genetic algorithm. In the case study, thermal-neutral condition (TNC) is set as the optimisation target in various operating conditions, where the SOEC generates no waste heat and needs no auxiliary heating equipment. Small peak-temperature-gradient (PTG) inside the SOEC is found at the TNC, which is vital to prevent thermal failure in the operation. For the cell operating with 1023 K and 1123 K of inlet gas temperatures, the smallest PTGs reach 0.09 and 0.31 K mm⁻¹ at 1.13 and 1.19 V, respectively. Finally, a 4-D map is presented to show the interactions among the applied voltage, required power density, inlet gas composition, and temperature under the TNC. The proposed method can be flexibly modified based on different optimisation targets for various applications in the energy sector.