Power System Transient Stability Assessment Based on SVM-SMOTE and One-Dimensional Convolutional Neural Network

The huge amount of measurement data provides the basis for the data-driven transient stability assessment (TSA) method. However, the unstable samples are usually much less than the stable samples in the TSA, this inherent attribute affects the performance of such methods. Therefore, this paper proposes a TSA method based on SVM-SMOTE and 1D-CNN. The support vector machine-synthetic minority oversampling (SVM-SMOTE) is used for sample equalization, to improve the negative impact of sample imbalance. The One-dimensional convolutional neural network (1D-CNN) is used to extract the temporal information of input features from the phase measurement units (PMUs). The simulation results performed on the New England 39-bus test system demonstrate that the proposed method improves the problem of a high missing rate of unstable samples in unbalanced data sets, realizes end-to-end temporal information extraction and transient stability assessment, which can meet the requirements of accuracy, rapidity, and reliability of online assessment.