Online Identification of Transient Stability and Unstable Generators Based on Deep Learning

Machine learning methods have been considered as a promising method for online transient stability analysis. This paper presents a deep learning approach for transient stability prediction and unstable generator identification based on convolutional neural networks (CNNs). The post-disturbance dynamic responses of all the generators are utilized as the input features. A binary classification CNN model is generated for transient stability prediction. And a multi-class classification CNN model is built for unstable generator identification. Moreover, the multi-class CNN model can provide the unstable probability of each generator using the probability outputs and the label information. Comprehensive studies are conducted on the New England 10-machine 39-bus system to verify the effectiveness of the proposed approach.