The electrical systems of nuclear power plants are characterized by densely interconnected nodes and short-distance network topologies. The stability of conventional model partitioning based on the ideal transformer method (ITM) is constrained by the interface impedance ratio, making it inadequate for real-time multi-core simulation of full-scale electrical systems. To address this issue, a model partitioning method combining the constant calibration method (CCM) and the linkage network (LN) is proposed. First, the instability mechanism of ITM is analyzed. By introducing coordinate transformation and a variation-threshold mechanism, data exchange is transformed from a time-driven to an event-driven paradigm, thereby eliminating the dependence of stability on the impedance ratio. Second, synchronization matrices and vector synchronization matrices are constructed to achieve dynamic control of the synchronization range for multi-core concurrent disturbances and accurate tracking of events. Simulation results on a typical nuclear power plant electrical system demonstrate that the proposed method maintains numerical stability under any impedance ratio, effectively reproduces the original system characteristics, and exhibits stronger stability and recovery capability than ITM under large fault disturbances. This method can serve as an effective solution for real-time multi-core simulation of short-distance, densely interconnected power networks in nuclear power plants.