Adaptive cyber-physical system attack detection and reconstruction with application to power systems

Timely and automatic detection and reconstruction of cyber attack is crucial in maintaining safe operation of cyber-physical systems. This study investigates the problem of unexpected cyber attack and its automatic recovery. First, detectability of attacks based on linear system theory is explored and some sufficient conditions of detecting state attacks and sensor attacks, respectively, are established. Subsequently two adaptive sliding mode observers with on line parameter estimation are designed to estimate state attacks and sensor attacks separately. Second, appropriate residual signals are constructed and it is shown that they approach to the attacks with ultimately uniformly bounded errors. Finally, as an illustration of attack detection and reconstruction, the results are applied to an IEEE 39 bus power system with 10 generators and 39 buses, which shows the effectiveness of the proposed adaptive schemes.