The robustness of power grid structure is crucial for ensuring load power supply under fault conditions. While existing studies have revealed macro-level correlations between grid structure and blackout risks, these correlations show low consistency and vary across systems of different scales, making them inadequate for guiding robust grid construction. By considering the impact of node heterogeneity on power supply ranges and factors like power path sharing on load outage probability, this study proposes a community theory-based power supply zone partitioning method, and an equivalent power path length metric. These form a novel grid robustness assessment framework. Validated across 62 power networks of three scales, the results demonstrate that the proposed metrics exhibit significantly improved positive correlations with blackout risks across all systems. Notably, the optimized Southern Power Grid topology designed using this method reduces average outage risk by 15.57% compared to manual designs, providing effective technical support for high-reliability grid construction.