Cyber-Secured Modular Multilevel Converters Against False-Data Injection Attacks Through Concurrent Estimators

Over the past two decades, modular multilevel converters (MMCs) and similar cascaded circuits have become a key technology in the electricity grid, primarily for medium- and high-voltage applications. They perform important functions for grid stability as they convert, transport, and inject large power levels, connect wind farms, or compensate for power quality issues, and may in the future also contribute to grid forming. However, their complex control and monitoring subsystems as well as the sensitive nature of electronics turns them into a target for sabotage. This study investigates the operation of MMCs under various false data injection attacks (FDIA), e.g., in the power supply of data centers. The proposed solution uses two concurrent estimators, a module-level Kalman filter and an arm-level neural network estimator, to detect and mitigate complicated cyber-attacks. The proposed technique would significantly improve the resilience of MMCs against complex FDIAs and thus the resilience of critical power supplies, e.g., of DC lines and data centers. The simulations and experimental results demonstrate that the proposed detection technique outperforms state-of-the-art methods when facing sophisticated attacks and still avoids their often significant computational demands or complexity.