Online analysis of transient stability state in power system based on multiscale spatiotemporal graph convolutional neural network

Electricity plays a crucial strategic role in contemporary society. To avoid significant harm to society caused by power system failures, it is necessary to accurately determine the stable state of the power system. In recent years, there have been many studies on the identification of transient stability states in power systems based on deep learning methods. Due to the increasing complexity of power systems, the full utilization of topological and temporal characteristics of power system information has gradually become a key point. To address this issue, this paper proposes a transient stability state identification model for power systems based on multi-scale spatiotemporal graph convolutional neural networks, which enables online evaluation of power system states. In the spatial module, different receptive field information is aggregated through residual connections and multi-scale Chebyshev graph convolution. In the temporal analysis module, based on the ability of LSTM to analyze temporal features, the temporal analysis of multi time step graph features are achieved. In the IEEE 39 node system experiment, accuracy, F1 score, recall, and precision were used as comprehensive evaluation indicators. The model proposed in this paper has high accuracy in state identification.