• • Frequency nadir estimation error remains within 0.016–0.027% (49.7359–49.7624 Hz predicted vs. 49.7280–49.7525 Hz simulated) and RoCoF error within 0.21–0.41% (−1.454 to −1.667 Hz/s predicted vs. −1.451 to −1.674 Hz/s simulated) across three parameter cases, confirming that fault-location-resolved LVRT cluster mapping eliminates the systematic bias inherent in single-aggregate wind farm representations.
• • Single-fault-scenario computation time is 0.0873 s for the proposed SFR model versus 0.3574 s for the unified structure model and 14.7240 s for detailed time-domain simulation, a 168-fold acceleration that makes exhaustive fault-location enumeration computationally tractable for control-room deployment.
• • Batch scanning of 10 fault locations requires 0.892 s total (0.0892 s average per scenario) for the proposed model, compared with 3.729 s for the unified structure model and 154.326 s for detailed time-domain simulation, a 173-fold reduction that directly enables the offline enumeration/online matching architecture required for emergency frequency control.
• • The fault-location-to-LVRT-cluster mapping quantifies differential active power recovery rates at a minimum of 20% PN/s per wind farm, avoiding the error introduced by collapsing all renewable stations into a single equivalent model and thereby capturing the time-varying recovery of the wind power deficit after fault clearing.