Component models have become increasingly complex and variable in novel power systems. Simulation models often deviate from actual systems

resulting in reduced simulation accuracy. Additionally

the extensive areas of power grids make it challenging to pinpoint the sources of errors. This paper proposes an error-tracing method based on topological aggregation and multi-level block decoupling to address the issues above. Firstly

considering the scattered distribution of elements in the adjacency matrix

the concept of network topology aggregation is introduced. An optimization model is established and solved to obtain a new matrix. Then

employing a multi-level block decoupling algorithm based on network topology

the original power grid is decomposed into multiple subnets according to the aggregated adjacency matrix. Hybrid dynamic simulations are conducted by injecting measured PMU data at the boundary points of each subnet. Comprehensive disturbance depth and simulation credibility indices are used to identify error regions

gradually reduce the error range

and locate the error components. Simulation results confirm that the proposed algorithm significantly enhances the efficiency of error tracing and reduces the number of required hybrid dynamic simulations.