Attention-Based Transformer Model for Inertia Estimation of Power Systems

Abstract

Inertia represents the system's resistance to changes in frequency and plays a crucial role in maintaining grid stability. With diminishing reliance on synchronous generators (SGs) due to the integration of renewable energy sources (RES), accurate inertia estimation becomes increasingly complex, and traditional methods have proved to be insufficient. This paper proposes a transformer-based framework to estimate system inertia that maps power imbalance to resulting frequency deviations. It leverages the transformer self-attention mechanism to effectively capture complex temporal dependencies in high-resolution, timesynchronized data streams from phasor measurement units (PMUs) and exploits the richness of temporally dense measurement data. The proposed approach can be applied beyond traditional methodologies that rely solely on the swing equation model, which typically considers only SGs' contribution to inertia. This enables a more comprehensive data-driven approach for estimating system inertia in modern power grids. © 2026 IEEE.