High-Fidelity Cluster-Equivalent Modeling and Fast Deadband-Aware Active Power Control for Primary Frequency Support in Grid-Following Wind Farms

High penetration of grid-following wind power reduces the natural inertial contribution of synchronous generation and increases the importance of primary frequency control (PFC). This paper presents a unified framework for deadband-aware modeling and fast active-power-based PFC of grid-following wind farms. Detailed aerodynamic, mechanical, generator-converter, PLL, measurement-filtering, and control-delay effects are represented, and heterogeneous turbines are reduced to a cluster-equivalent model. For the archived validation case, the cluster model reproduces the detailed-model frequency response with an RMSE of approximately 0.003 Hz and a peak total-power difference of about 0.6%. Building on this model, a two-layer control strategy combines turbine-level multi-stage fast follow-up and recovery with farm-level optimization-based allocation according to regulation margin and a normalized regulation-effort index. Comparative studies against fixed droop, multi-segment droop, pure virtual inertia, hierarchical virtual inertia, and MPC-based PFR show that the proposed strategy provides the highest frequency nadir in the investigated cases and reduces the normalized regulation-effort index by about 20% relative to conventional droop. This study explicitly distinguishes the physical initial RoCoF, fixed by the post-contingency power imbalance and inertia before delayed control action develops, from subsequent controlled frequency recovery. A surrogate measurement-chain perturbation test is used only to examine sensitivity to PMU/SCADA-like noise and is not presented as independent field validation.