In power systems with a high penetration of renewable energy

the inertia support capability from the generation side continues to decline. This inertia reduction significantly impacts system frequency stability. The load side dynamic inertia response exhibits substantial spatiotemporal variations that conventional approaches fail to adequately capture. Traditional methods lack both dynamic partitioning capability and effective quantification of these spatial differences in inertia response. To address these limitations

a novel load side equivalent inertia estimation method based on spectral clustering and dynamic feature fusion was proposed. First

a systematic classification framework for load side inertia resources was established according to their dynamic characteristics. This framework categorizes loads into three distinct types of inertial equivalent components

each representing a specific dynamic behavior pattern in frequency response. Subsequently

a dynamic spectral clustering algorithm was developed to achieve adaptive inertia partitioning based on the similarity of load spectral energy distribution. Furthermore

a computational framework for loadside inertia that integrates spatial and temporal features was constructed. By introducing a node inertia contribution index and combining it with dynamic weight adjustment for load types and a zonal inertia aggregation strategy

a hierarchical inertia calculation framework was formed. Simulation results on the IEEE 29-node system show that the proposed method reduces the root mean square error of inertia estimation by 52.17% compared to traditional methods

with a maximum relative error not exceeding 5.3% under multiple operating conditions

verifying its effectiveness and adaptability.