Impact Analysis of a Bias Injection Cyber-Attack on a Power Plant

This paper addresses the concept of a bias injection cyber-attack on the load frequency control loop of a single-area power plant. The system evolves in the discrete-time domain and a convex and compact set of polyhedral state constraints represents a valid domain of safe operation under the effect of a stabilizing output-feedback controller. Whenever the safety constraints are violated, an alarm alerts the control center to a security breach. An attacker gains access to frequency sensor measurements and corrupts the data transferred to the automatic generation control unit. The objective of the adversary is to drive the electrical frequency to a safety-critical steady-state value without triggering an alarm. An analysis based on robust invariance concepts assesses the vulnerability of the system on this particular scenario and determines the maximal impact that such an attack can have without activating an alarm. Simulation studies highlight the effect of the cyber-attack on the physical plant.