Abstract
Automatic Generation Control (AGC) is essential for power grid stability butremains vulnerable to stealthy cyberattacks, such as False Data InjectionAttacks (FDIAs), which can disturb the system's stability while evadingtraditional detection methods. Unlike previous works that relied on blackboxapproaches, this work proposes Kolmogorov-Arnold Networks (KAN) as aninterpretable and accurate method for FDIA detection in AGC systems,considering the system nonlinearities. KAN models include a method forextracting symbolic equations, and are thus able to provide moreinterpretability than the majority of machine learning models. The proposed KANis trained offline to learn the complex nonlinear relationships between the AGCmeasurements under different operating scenarios. After training, symbolicformulas that describe the trained model's behavior can be extracted andleveraged, greatly enhancing interpretability. Our findings confirm that theproposed KAN model achieves FDIA detection rates of up to 95.97% and 95.9% forthe initial model and the symbolic formula, respectively, with a low falsealarm rate, offering a reliable approach to enhancing AGC cybersecurity.