Abstract: Accurate carbon emission factors serve as the core basis for realizing the low-carbon transition of power systems. Currently, the average carbon emission factor of the power grid has coarse spatiotemporal granularity, while the carbon flow tracing methods struggle to meet computational requirements of massive nodes in actual complex power grids. Moreover, both overlook regional non-grid-connected electricity quantity. To address this, a multi-level regional carbon emission factor correction method considering non-grid-connected electricity quantity is proposed. Based on the electrical partitioning theory, this method constructs a solution framework for carbon emission factors and divides the actual power grid into three hierarchical structures (provincial-level, prefecture-level, and end-user level) from top to bottom according to voltage levels. A hierarchical and progressive strategy is adopted in the calculation of each level, ultimately attributing the environmental benefits of non-grid-connected electricity quantity to the corresponding nodes and dynamically correcting the carbon emission factors of grid-connected electricity quantity. Case studies at three levels (provincial, prefecture, and end-user) verify the feasibility of the method, achieving refined spatiotemporal granularity and fully accounting for the environmental benefits of distributed resources. It provides an engineering-practical technical pathway for carbon accounting of regional power systems.