The increasing integration of renewable energy sources, particularly those interfaced via inverters, is reducing the inertia of power systems, potentially threatening stability. This paper proposes a novel framework for continuous inertia estimation using convolutional neural networks (CNNs). The framework leverages state-of-the-art AI techniques and investigates power spectra analysis and input-output correlations to explain the CNN's operation. Validation on a heterogeneous power network demonstrates distinct spectral footprints for different devices, crucial for online network stability analyses by transmission system operators.

Daniele Linaro, Federico Bizzarri, Davide del Giudice, Cosimo Pisani, Giorgio M. Giannuzzi, Samuele Grillo, Angelo M. Brambilla