Abstract

The integration of multiple direct current (DC) microgrids offers a resilient and efficient solution for modern energy demands, particularly with the increasing adoption of intermittent renewable energy sources. However, the reliance on communication networks for coordinating multiple interlinking converters (MICs) introduces vulnerabilities, particularly to False Data Injection Attacks (FDIAs), which can significantly disrupt system stability and operation.
This thesis presents AI-driven cyber-defense strategies to protect clustered DC microgrids interconnected via MICs against FDIAs.
At the primary control level, a Support Vector Machine (SVM)-based anomaly detection framework is developed to identify FDIAs in real time. Once an attack is detected, the system autonomously transitions to a localized power-balancing control to maintain operational stability.
At the secondary control level, an Adaptive Neuro-Fuzzy Inference System (ANFIS)-based signal estimation strategy is proposed to detect injected FDIAs and subsequently reconstruct compromised control signals, thereby maintaining MIC coordination.
Extensive simulation studies validate the effectiveness of the proposed methods, demonstrating their ability to enhance microgrid resilience against various FDIA scenarios, including time-varying and unbounded attacks. The results confirm the efficacy of both the SVM and ANFIS frameworks in safeguarding clustered DC microgrids interconnected via Multiple Interlinking Converters against cyber threats, ensuring stable and secure operation.