Desarrollo y aplicación de una metodología para clasificación de fallas en Sistemas Eléctricos de Potencia utilizando Deep Learning.

Abstract

The transmission and distribution of energy are areas of vital importance for the electrical power system that require high levels of reliability, so a failure in the system becomes a significant problem; for this reason, in recent years, methods based on artificial intelligence techniques, neural networks and others, have been proposed for analysis and solution papers of these. In this paper, a method for the classification of faults in Electrical Power Systems using Deep Learning implemented in an application created through App Designer is proposed. The classification of faults in transmission lines and the determination of the type of fault uses a method of feature vectors and categories using LSTM neural networks, in addition, the classification of the most common types of faults allows decisions to be made to reduce as much as possible the faults in the lines. The measurements that form the data used to train the LSTM neural network were obtained in DigSilent; furthermore, the creation, training and use of multi-layer Long-Short-Term-Memory (LSTM) networks was implemented in Matlab R2021a. The strong learning capability of LSTM networks is suitable for many time series in power transmission and distribution. The results obtained are promising for fault classification in power systems.