To address the strong time-varying characteristics and pronounced regional differences of power system inertia under high renewable energy penetration, this paper proposes an online regional inertia assessment method based on frequency dynamic partitioning. Real-time correlations among frequency trajectories are used to perform online frequency partitioning within a sliding time window via a Pearson-based spectral clustering approach. Dynamic Time Warping (DTW) and K-Medoids clustering are employed to select representative measurement points within each region, constructing equivalent regional frequency signals. Furthermore, the method integrates N4SID state-space identification with multiple inertia estimation techniques through statistical fusion to enhance robustness and accuracy. Simulation results on the modified IEEE 10-generator 39-bus system demonstrate that the proposed approach achieves high estimation accuracy and strong engineering applicability across different operating scenarios.
