A Hybrid Framework for Dynamic Inertia Estimation of Power Grid Considering Virtual Inertia Support From Wind Farms

Rapid growth of converter-interfaced renewable energy sources has made the estimation of system inertia increasingly complex, as virtual inertia control (VIC) introduces nonlinear and time-delayed dynamics that traditional methods often overlook. This article proposes a data-driven framework for online estimation of effective inertia under varying wind power penetration (WPP) and VIC gains. The approach integrates wavelet-based denoising, Isolation Forest-based onset detection with an optimally tuned decision threshold, robust rate of change of frequency estimation, and an Attention-Augmented Encoder–Decoder Architecture model for accurate estimation of net power imbalance. Unlike existing data-driven methods, the proposed framework explicitly accounts for the activation delay and nonlinear response of VIC, yielding superior robustness across diverse operating conditions. Extensive simulations are conducted on a modified 11-bus and the truncated model of Indian Eastern Regional Grid system, considering a wide range of disturbance magnitudes, WPP levels, and VIC gains. Further, the framework is shown to be scalable for area-level inertia estimation in large systems by leveraging boundary bus frequency and tie-line power flow.