Comprehensive Performance Analysis for Transient Stability Assessment Machine Learning Models Activated by Features in Multi Temporal Sections

• Published in: 2021 IEEE 4th International Electrical and Energy Conference (CIEEC) pp. 1 - 6
• Main Authors: Ren, Junyu, Chen, Jinfu, Yang, Ruixiong, Yang, Yexin, Lin, Guihui
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28.05.2021
• Subjects: Adaptation models; Analytical models; comprehensive performance analysis; Computational modeling; Deep learning; interpretability; machine learning models; multi temporal sections; Power system stability; Stability analysis; Training; transient stability assessment
• DOI: 10.1109/CIEEC50170.2021.9510472

To meet the requirement of highly efficient transient stability assessment, we propose a concise but efficient feature set collected from multi temporal sections. Activated by the proposed feature set, some typical machine learning models are constructed and tested respectively on two benchmark systems of different scales. Comprehensive performance analysis of different machine learning models is implemented. The accuracy of most typical models based on the proposed feature set is greater than 98% both for two systems. Further, the interpretability performance is discussed. Rules of stability are very easy to acquire through data propagation in decision tree, showing that the decision tree has a strong interpretability, which is conducive to preventing the system from transient instability.