The large-scale adoption of power electronics systems enables a more sustainable and flexible power grid. The offline simulation tools support the analysis of such systems. However, despite of the advancement in computer hardware and software, the accurate and efficient simulation of large-scale power converters is still a significant challenge; the simulation is usually time-consuming. In this article, we propose a state-variable-interfacd decoupling (SVID) strategy based on the discrete state event-driven (DSED) approach. The SVID strategy partitions the large-scale circuit into smaller subsystems with state variables as interfaces so that the numerical integration can be performed in a decoupled manner. Meanwhile, it ensures that no accuracy is sacrificed after decoupling by representing the interfaced variables with the high-order Taylor expansion. A 10-kV 2-MW solid-state transformer with 578 power switches is studied as a simulated case, and more than 1000-fold acceleration (from 3 h to 12 s) is achieved compared with commercial software, where the SVID strategy contributes to five- to eightfold. The simulated results are compared with both explicit and implicit solvers in commercial software and experimental results. The DSED approach with SVID strategy makes it possible to analyze and design a high-power converter more reliably and perform studies that are very difficult or impossible to do otherwise.