Abstract: Numerical simulation is crucial for the development of solid oxide fuel cell (SOFC), facilitating the exploration of internal physical parameter distribution and electrochemical characteristics. It ensures the reliable operation of SOFC and accelerates structural optimization. However, the high computational cost of SOFC stacks hinders large-scale simulations and optimizations. To address this challenge, a combined simulation approach with built-in cell surrogate models in the three-dimensional stack model is proposed. Data-driven surrogate models are employed to replace certain physical processes at the cell scale, allowing for the substitution of partial differential equations with algebraic equations in the computational domain. This approach significantly improves the efficiency of numerical stack simulations. Simulation results demonstrate that the simplified model maintains consistency with the original model, capable of predicting the distribution of inconsistencies within the stack. By reducing degrees of freedom and memory requirements, the computational efficiency is improved by 35%, enabling numerical calculations for larger-scale stacks. This simplified modeling approach demonstrates practical value for large-scale SOFC simulations and provides meaningful guidance for scaled-up stack design.