False Data Injection Attacks Detection in Power Grid Based on Deep Learning Multi-Model Fusion
Li Bo, Gong Ke Xu, Qi Jia Xiang · 2023
Aiming at FDIA in smart grid, this paper proposes a FDIA detection method for power grid set up on deep learning multi-model fusion. Firstly, two different Convolutional Neural Networks (CNN) were used to form a double-branch structure to extract and learn the global and local spatial features of normal data and unnormal data in the smart grid respectively. Then, Long Short-Term Memory (LSTM) network was used to extract and learn temporal features. Then, Multi-layer Perceptron (MLP) was used to further compress and extract features. Finally, the output of Softmax layer was used to determine whether there was unnormal data injection in the smart grid. The detection model was trained, verified and tested by IEEE14 and IEEE118 bus systems. The experimental shows that the proposed method can accurately detect false data in smart grid, which verifies the effectiveness of the proposed method.