Parameter-Free False Data Injection Attack Model on Power System’s Automatic Generation Control

Abstract

Automatic generation control (AGC) plays a crucial role in frequency control and economic dispatch of electric power systems. However, AGC systems, relying heavily on frequency and tie-line power measurements, are vulnerable to cyber-attacks. A novel false data injection attack (FDIA) algorithm targeting AGC is proposed, which requires no model information and parameter values. To this end, we first derive the maximum likelihood estimation of the multivariate Ornstein-Uhlenbeck (OU) process, based on which AGC parameters, topology information, and the conditional variance of states can be extracted purely from eavesdropped sensor data (frequency, tie-line power, reference power). Then we will exploit the extracted information to design FDIA vectors by solving an optimization problem, which can bypass conventional AGC defense mechanisms. Numerical studies in 2-area and 3-area systems show that the proposed FDIA algorithm can deteriorate the system’s frequency within several minutes when measurement noise, transmission delay, and computational time are considered.

Keywords

• Area control error
• automatic generation control
• false data injection attacks
• maximum likelihood estimation
• Ornstein-Uhlenbeck process