Evaluación predictiva de estabilidad transitoria en tiempo real mediante aprendizaje automático y la identificación de patrones del máximo exponente de lyapunov.

Abstract

During the current Smart Grids development stage, the interest in deploying methodologies to assess the real time Power System vulnerability has increased. Such methodologies should focus on evaluating the system condition against Transient Stability issues. The aim is indeed to define intelligent control and protection strategies for timely mitigation of disturbances. A methodology for real time Transient Stability Predictive Assessment of Electric Power Systems, applying novel computational tools, is developed. The application of Data Mining techniques is included for pattern recognition in the Largest Lyapunov Exponent (LLE). In addition, the use of Machine Learning algorithms for predictive assessment, based on the identified patterns, is proposed; and the determination of optimal models, adapted to particular systems, is deployed. The proposed methodology is validated on the IEEE 39-Bus System, where its applicability is demonstrated with speed, precision, and flexibility features. In this way, the proposed methodology constitutes an innovative contribution since entails improvements to recent research, solves limitations while guarantees better results, and also exhibits the potential to be applied in the Ecuadorian National Interconnected System for the execution of appropriate control and protection actions.