Detection and reconstruction of measurements against false data injection and DoS attacks in distribution system state estimation: A deep learning approach

Abstract

In a smart grid, the presence of advanced measurement devices and communication channels is significantly vulnerable due to cyberattacks such as false data injection attacks (FDIAs), and denial-of-service (DoS) attacks. To tackle these cyberattacks, a data-driven approach namely, the attention-based temporal convolutional denoising autoencoder is proposed which combines the advantages of the attention mechanism and temporal convolutional network to capture spatio-temporal information. This model identifies the FDIA location and also replaces the corresponding measurements with the reconstructed values. The robustness of the proposed technique is evaluated with different levels of FDIAs as well as missing data caused by DoS attacks. The simulations have been performed on IEEE 13-bus and IEEE 37-bus distribution systems and their reconstruction results along with the classification metrics are presented. Finally, the model’s effectiveness is compared with other denoising autoencoder approaches and ML/DL approaches. From the simulations, the results show that the proposed model outperforms in reconstruction and classification.