Transient Voltage Prediction Method Based on Spatial-Temporal Graph Convolutional Network

Research on data-driven transient voltage forecasting method is of great significance to transient voltage stability evaluation, but the existing transient voltage forecasting methods can’t effectively incorporate the temporal and spatial characteristics of transient voltage into the model. To solve this problem, a transient voltage forecasting method based on the Spatial-temporal graph convolutional network model is proposed. Firstly, the transient voltage prediction is modeled as a sample space mapping problem. It trains the neural network model by data-driven method and establishes the mapping from the spatial structure and time series data of power grid in transient process to the predicted value of transient voltage. Then, the model can quickly and accurately predict the transient voltage of the system by extracting the spatial structure characteristics and time series characteristics of the power grid before, during and after the fault in the transient process. Finally, the test results based on New England 39-bus system verify the feasibility, effectiveness and superiority of the proposed method.