Model-Free Predictive Frequency Control Under Sensor and Actuator FDI Attacks

This brief investigates the load frequency control (LFC) issue of an interconnected power system under sensor and actuator false data injection (FDI) attacks. To this end, a novel model-free control method, ultra-local model-free predictive control (ULMFPC), is proposed as the secondary frequency controller to deal with the coordinated FDI attacks and system physical limitations. In contrast to the approaches available in the literature, the presented ULMFPC approach does not need the state estimation of observers and the mathematical model of the system. This method can independently cope with disruptions injected into the system. The ULMFPC considers the employed plant as a dependent system and utilizes predictive control in a constantly updated linear model to track the reference and control limitation. Unlike other predictive control methods, there is no training data set in the presented control strategy, and only small-scale convex optimization is essential for control computation. The proposed control method performance is evaluated by the Speedgoat real-time emulator and compared with the MPC and distributed economic MPC (DEMPC) methods under sensor and actuator FDI attacks. The experimental results indicate that the proposed ULMFPC technique provides much better dynamic responses compared to other methods.