Cyberattack Aware Load Frequency Control in Power Systems

The load frequency control (LFC) is crucial for stabilizing the frequency of the power grid in the intermittency of renewable energy sources. Modern LFC systems utilize open communication and automation networks that expose power systems to potential cyberattacks, which could lead to frequency instability throughout the network. Current research often treats frequency control and cybersecurity independently, which could lead to ineffective solutions and increased system instability. This paper introduces a cyber-resilient LFC strategy that improves both control performance and security. We employ reinforcement learning for adaptive frequency control and integrate it with a graph convolutional neural network to enhance control responses during a cyberattack event. Our strategy also includes a graph auto encoder-based attack detector that is trained in various scenarios and has shown more than 92% in detection rate while tested on the Iberian power system topology, which includes 486 buses. This highlights the robustness and reliability of our approach in ensuring resilience against cyberattacks within the realm of frequency control in power systems.