State Estimation in Smart Grids Using Spatio-Temporal Graph Convolutional Networks

State estimation (SE) in smart grids relies on accurate models and measurement data. Traditional methods struggle with computational complexity and adaptability to topological changes. As smart grids evolve, the nodes and transmission lines form a natural graph structure, enabling the use of graph convolutional networks (GCN). Recent advancements in artificial intelligence, such as recurrent neural networks (RNN), convolutional neural networks (CNN), and long short-term memory net- works (LSTM), enhance state estimation by utilizing time series data. This paper presents a multi-layer spatio-temporal graph convolutional network (STGCN) model that integrates spatial and temporal convolution to capture spatiotemporal characteristics. The model includes multiple spatiotemporal convolution blocks, each with two time-gated convolution layers and one spatial graph convolution layer, facilitating high-precision state estimation. Experiments on the IEEE 39-bus system demonstrate that the STGCN model improves estimation accuracy and reduces computation time compared to traditional methods.