Support Vector Machine-Based False Data Injection Attacks Detection in Interconnected DC Microgrids

The integration of multiple DC microgrids offers a robust solution for modern energy needs, especially with the high penetration of intermittent renewable sources. However, the reliance on extensive communication networks for coordinating Multiple interlinking converters (MICs) in these microgrids introduces vulnerabilities, particularly to False Data Injection Attacks (FDIAs), which can compromise system stability and security. This paper presents an AI-based anomaly detection strategy utilizing Support Vector Machines (SVM) at the primary control level to detect FDIAs in clustered DC microgrids. The proposed SVM model, trained offline with a comprehensive dataset of operational and attack scenarios, effectively delineates boundaries between normal and attack-induced conditions, enabling real-time detection in the online phase. Upon identifying an FDIA, the system decouples the primary control and transitions to a localized power balancing control, effectively preserving microgrid stability. Extensive simulations validate the efficacy of the proposed approach, demonstrating its ability to sustain stable microgrid operation under various FDIA scenarios, including time-varying and unbounded attacks.