Detecting Cyber-Physical-Attacks in AC microgrids using artificial neural networks

In this paper, we are using a Nonlinear AutoRegressive eXogenous Neural Network NARX to diagnose the existence of cyber intrusion in a fully simulated microgrid An online power estimator is placed at the point of common coupling to predict the normal active power signals. Whereas, Detected Faults or abnormalities in the estimated signal could indicate the presence of manipulated data and hence, cyber intrusion. The proposed method is able to capture different types of attacks including False Data Injection FDI and replay attacks.