Securing AGC Systems Against False Data Injection Attacks Using Federated Learning and Model Predictive Control
Mohammad Mahdi Soleymani, Masoud Babaei Vavdareh, Mohsen Ghafouri, Chadi M. Assi · 2024
The automatic generation control (AGC) system is essential for maintaining power system stability by regulating even minor frequency fluctuations to prevent disruptions and equipment damage. However, cyberattacks targeting communication channels can compromise the AGC system's functionality. False data injection attacks (FDIAs) are a common type of cy-berattack that aims to manipulate the AGC system's behaviour by adding false data to the measurement signals. To tackle the impact of FDIAs on multi-area interconnected power systems, this study proposes a coordinated strategy using federated learning (FL) combined with model predictive control (MPC). The proposed strategy first estimates the true measurement values using the FL approach. Subsequently, MPC utilizes these estimated values to generate mitigation signals, effectively counteracting the FDIAs' impact. The proposed strategy is assessed across various stealthy FDIA scenarios within a three-area interconnected power system. The findings demonstrate the efficacy of the proposed strategy in mitigating the impact of FDIAs on the AGC system.