An Anomaly Detection Method for Smart Grid Based on Convolutional Autoencoder and Gated Recurrent Unit

An anomaly detection method for smart grid based on convolutional autoencoders (CAE) and gated recurrent units (GRU) is proposed in this paper. Firstly, the CAE is used to reconstruct data, reducing the data dimension and extracting effective features. Subsequently, multi-head attention is employed to capture the relationships between different features, enhancing the model's ability to understand the data. Finally, anomaly detection is executed utilizing GRU. The power system attack dataset from Mississippi State University and Oak Ridge National Laboratory is used to conduct experiments. The experimental results show that the proposed method achieves an average accuracy of 96.66% on this dataset, outperforming the comparative models. Additionally, the precision, recall, and Fl-score of the proposed method are also superior to those of the comparison models.