Voltage Sags Characterization Using Fault Analysis and Deep Convolutional Neural Networks

Abstract

The power systems are transitioning to renewables. The power system infrastructure digitalization is the next stage in this evolution. A digitalized future distribution network (DN) has an opportunity to evolve into real-time analysis based on a huge volume of data. Such real-time analysis will be feasible through the integration of the theoretical background of fault analysis and machine learning techniques. Fault location is one of the major issues to improve reliability indices. The voltage sags characterization is used to locate faults in the DN. This article presents a methodology to characterize voltage sags using fault analysis and deep convolutional neural networks. The voltage divider model allows the characterization and the discrete wavelet transform is used in signal processing. The machine learning and deep learning models used allow estimating the sag magnitude, fault location, phases involved, duration, and impact of distributed generation (DG) in each event. The IEEE 13-node test feeder including DG was used to validate the effectiveness of the proposed methodology. This paper provides a way to handle a Big Data stream in the DN control center and to efficiently locate faults in several operational scenarios.