Research on the Detection of False Data Injection Attacks in DC Microgrids Based on Attention Mechanism and DAE-CNN-BiLSTM
Zhuoyu Guan, Gengwu Zhang, Xiaoqing Han · 2024
Distributed DC microgrid is a typical power information physical system (CPS) that is vulnerable to cyber-attacks. False Data Injection Attack (FDIA) on the information side can change the system's operational state and even destabilize the microgrid system. Most of the existing microgrid cyber attack detection methods in view of linear DC microgrid models. However, there are constant power loads (CPLs) in DC microgrids, which present negative resistive and nonlinear characteristics and are generally nonlinear. For this reason, in this paper, a distributed DC microgrid model with constant power loads is developed and a nonlinear DC microgrid false data injection attack detection method based on the attention mechanism and convolutional bi-directional recurrent neural network is proposed. After adding a known constant power load during the DC microgrid stabilization process, the distributed units suffering from FDIA are identified by comparing the discrepancy between the estimated value of the offline-trained neural network and the actual measured value to distinguish between the load change and the false data injection attack. Finally, a 4-node DC microgrid model is built in MATLAB/Simulink for simulation experiments to validate the effectiveness of the proposed assay.