부산대학교 도서관

In order to solve the security and stability problems occurred in the cyber-attack power system, this paper proposes a controller based on graph generation adversarial network algorithm, and simulates and verifies it in the load frequency control and automatic voltage regulation model. The generator of the graph generation adversarial network algorithm collects normal data and data that have been attacked and tampered with, and the discriminator compares the generated data with the historical data to determine whether it is under cyber attack; then, the discriminator outputs normal and safe data; finally, the attention network aggregates the output data of the discriminator and selects the data most similar to the historical signal for effective defense against the attack data. Simulation results show that the controller has better robustness than the PID controller and the generative adversarial network controller, can better suppress the frequency and voltage fluctuations caused by the network attack, and can maintain the safe and stable operation of the power system.