A SMOTE and XGboost Based Method for Power System Transient Stability Assessment

Accurate and fast power system transient stability assessment (TSA) becomes more and more important considering power system operates at conditions more and more close to stability boundary. To realize fast TSA, artificial intelligence (AI) is highly concerned. However, this kind of methods encounter two challenges when applied in practical power systems, one is the inherent imbalance of samples, which limits the transient stability prediction accuracy. The other is the result interpretability, which blocks their further applications in engineering practice. Hence, a synthetic minority over-sampling technique (SMOTE) and extreme gradient boosting (XGboost) based TSA is proposed in this paper. SMOTE is selected to solve the imbalanced sample problems, based on which principal component analysis (PCA) is utilized to reduce feature dimension. XGboost is then adopted for constructing transient stability assessment model considering its good interpretability. The proposed method is validated with New England 39-bus test system and results show effectiveness.