Fast Transient Voltage Stability Assessment for Complex Receiving-End Power Grid

The continuous increase of new energy generation and AC-DC interconnection make the transient voltage stability of the receiving-end power grid more and more complicated. In order to realize the rapid assessment of transient voltage stability, a one-dimensional Convolutional Neutral Network (CNN) for transient voltage stability assessment (TVSA) is constructed by selecting the features reasonably on the basis of using the Agglomerative Hierarchical Clustering (AHC) algorithm to partition the receiving-end grid. In order to enhance the assessment work, the Gradient-weighted Class Activation Mapping (Grad-CAM) algorithm is utilized to find out the concerns of the model to features, and the model knowledge of the neural network is analyzed on the basis of surveying on a large number of samples. Based on the transient voltage stability mechanism and the concerns of deep learning model, the factors that cause transient voltage instability are analyzed. The effectiveness of the proposed method is demonstrated by the case study of a provincial grid in eastern China.