Data driven method for transient stability prediction of power systems considering incomplete measurements

This paper presents a novel data pre-processing for data-based method of transient stability prediction considering incomplete measurements. Firstly, the statistical features are utilized as the input features, of which number is independent from the power system's scale. Secondly, the dataset is expanded considering the situations when some generator measurements are unavailable randomly. Next, a k-difference-neighbor method is used to reduce the number of instances. After that, a new dataset is generated as the training set to train a more robustness classifier. Case studies are conducted on the New England 10-machine 39-bus system and Northeast Power Coordinated Council 48-machine 140-bus system respectively to verify the effectiveness of the proposed method.