Resilient Control Strategy for Smart Grid Under Delayed Sporadic Measurements: Attack Estimation via Hybrid Observer

This paper investigates the safe control problem for smart grid and proposes a resilient control scheme via attack estimation, in which both the sampled output signals and command signals may be corrupted by false data injection attacks. Meanwhile, the communication channels between power plant and controller suffer from time-varying delay. Firstly, the i-th power area is modeled as a hybrid system model composed of continuous system signals and discrete attack signals. And then, a hybrid observer is constructed for each area to estimate the attacks, by which a modified load frequency control scheme is designed to effectively attenuate the effect of attacks on the actuator signals. In particular, the observer gains for each area are obtained via only local information, which can avoid privacy leakage and reduce computational complexity. Moreover, an optimization algorithm is proposed for each area to search a sub-optimal observer gain. Finally, a three-area power system is utilized to verify the proposed resilient control scheme. Note to Practitioners—It is challenging to realize frequency regulation of power systems via communication networks that are vulnerable to cyber-attacks. Conventional robust control methods from a passive perspective are inefficient against prolonged and intense attacks, thus, this paper proposes an active defense scheme based on attack estimation, in which the attacks are estimated via the designed hybrid observer. Using only delayed sporadic measurements, the developed resilient control scheme can solve the practical problems in existing works from following aspects: 1) Continuous output signals are no longer required for attack estimation, which is of significant importance because frequency control in practical cyber-physical systems typically employs sampled-data control; 2) Model discretization is no longer needed, a discretized model obtained with a relatively large sampling period may lack sufficient accuracy; 3) The attack signals in both measurement and controller channel can be estimated, and the inevitable transmission delay is considered; 4) Centralized estimation is avoided, the designed hybrid observer works in a distributed way, making it more scalable. The above merits make the proposed resilient control strategy more practical to be adopted for improving the resilience in cyber-physical power systems against false data injection attacks.