固体氧化物电解池动态运行下热中性目标的多目标优化：一种混合建模方法

Abstract Intermittent renewable power sources can induce thermal oscillations in solid oxide electrolysis cells (SOECs), compromising their integrity and safety. A tubular SOEC numerical model has been constructed. Under the condition of a quasi-step voltage response, it can be observed that the single-cell model can rapidly transition from one steady state to another quasi-steady state within 0.5 s while maintaining stable heat generation, indicating that the steady-state data can be used to predict the dynamic-state performance. A hybrid modelling approach that integrates a multi-physics simulation model verified by experimental data, deep neural networks, and genetic algorithms is presented to assess cell performance under stable thermal conditions, especially when interfaced with variable renewable energy inputs. Four control strategies are compared when maintaining thermal neutrality: uncontrolled, fuel ratio control, fuel flow control, and combined fuel ratio and flow control. Among these, the combined strategy demonstrates superior flexibility in adjusting operational parameters under fluctuating conditions. However, maintaining strict thermal neutrality introduces an excessive safety margin, leading to performance degradation. Therefore, multi-objective optimization is employed to explore scenarios with minor deviations from thermal neutrality. Results show that at thermal neutrality, the total co-electrolysis conversion rate peaks at 77.7 %, while slight heat absorption increases total electrolysis efficiency to 83.9 % and reduces average direct feedstock and energy cost to produce syngas (the mixture of CO and H 2 ) to $0.31/kg in a single cell.