A Non-Iterative Model-Based Intrusion Detection System for AGC System

The amalgamation of interactive technologies and information has significantly improved the efficiency of the power systems; however, it also presents cyber threats at the vulnerable points of the cyber-physical networks. Adversaries can manipulate the sensor readings by inj ecting false data causing dire infrastructural damage. This paper studies the performance of the automatic generation control (AGC) system of a power grid penetrated with photovoltaic (PV) units under cyber-attacks. A model-based intrusion detection and mitigation strategy are proposed to detect and eliminate the cyber-attacks targeting the AGC loop. Unlike most previous model-based algorithms, the proposed detection algorithm is designed by accounting for various system uncertainties. Additionally, the proposed technique is non-iterative in the prediction phase making them much more compatible with online operation in a large-scale power system. Further, in addition to basic attack templates (such as step, ramp, and denial of service), the proposed defense system can detect intelligent replay and optimal stealthy attacks. The effectiveness of the detection method is discussed using a confusion matrix. The simulation results reveal the algorithm's capability in identifying cyber-attacks on the AGC loop.