AI-Based Intrusion Detection Enabling Cyber-Resilience in Grid Forming Inverters

This work proposes an event-triggered strategy to detect false data injection (FDI) cyber-attacks that impact the dynamic frequency behavior of a grid-forming inverters (GFMIs) network. The proposed approach employs an artificial intelligence (AI) technique to enable a hidden detection strategy (HDS) that triggers based on variations in system frequency and optimal power sharing commands (OPSC). The developed approach is tested on multiple attack vectors, including a data-driven coordinated FDI attack to manipulate the OPSC and voltage measurements in the communication lines between primary and supervisory control layers, and a malware attack to manipulate inertia and damping coefficient (IDC) values in the primary controller. The proposed method enables real-time estimation of system trajectories, specifically frequency and rate of change of frequency (ROCOF), under system variations and disturbances. These trajectories are used to assess system dynamics and detect the presence of attacks. The proposed AI-IDC-HDS is validated on a modified 9-bus IEEE system. Furthermore, a small-scale hardware setup is used for experimental validation.