This paper presents a hybrid approach that integrates data-driven and model-driven methods for system frequency response (SFR) modeling, explicitly incorporating the dynamics of inverter-based resources (IBRs) and the resulting local frequencies into the conventional SFR model based on synchronous generators (SGs). Unlike conventional models that assume a uniform system frequency, the proposed approach employs a dynamically interacting coefficients (DICs) matrix to capture slow local frequency dynamics within a unified SFR model. In this model, multiple SGs are coherently aggregated into a reduced-order equivalent, while the IBRs are represented by an equivalent model that neglects fast voltage and current control dynamics but preserves slower, frequency-relevant dynamics. The DICs are identified as frequency-dependent transfer functions using pseudo-spectral analysis via the MUSIC algorithm, with least-squares estimation (LSE) applied for parameter refinement. Simulation results on two benchmark networks confirm that the proposed model accurately reproduces system slow dynamics and systematically reflects the evolving dynamics of IBRs as penetration levels increase.
