A New Attack Recovery Approach and H∞ Performance Analysis for LFC Systems With FDI Attacks and Uncertainties

This paper focuses on the attack recovery and $H_{\infty } $ performance analysis for load frequency control (LFC) systems with false data injection (FDI) attacks and uncertainties. Firstly, considering the impact of FDI attacks, a new Informer-based attack recovery (IAR) model is proposed for recovering FDI attacks. The IAR model is capable of extracting rich attack signal features from historical measurement data by leveraging multi-task learning. Compared with previous models, including LSTM and LSTM-AE, the IAR model improves the efficiency of recovering attacks for a future period and improves the recovery accuracy. Then, the recovered attacks can be used to regulate the control input signals of LFC systems to mitigate the impact of FDI attacks. Secondly, the uncertainties are considered in the LFC systems, which mainly consists of process and measurement noise. The $H_{\infty } $ performance of LFC systems with uncertainties and FDI attacks is analyzed. Finally, two datasets are generated for validating the effectiveness of the proposed attack recovery model and the analysis of $H_{\infty } $ performance.