Aplicação de Inteligência Artificial na detecção de falhas em sistemas HVDC com conversores modulares multiníveis

Abstract

The modular multilevel converter (MMC) is a modern topology for power converters. Due to its modular characteristics, ease of expansion, and high-quality output voltage waveform, the MMC has been especially recommended for applications in high-voltage power systems, particularly in HVDC (High Voltage Direct Current) systems. Like all electrical power systems, the MMC is subject to open-circuit and short-circuit faults. To this day, there are few studies focused on fault identification/location in submodules (SM), especially related to open-circuit faults. This work proposes a fault detection methodology using a convolutional neural network, LSTM (Long Short-Term Memory), to compile simulated data from two MMCs forming an HVDC system and create an artificial intelligence capable of classifying and providing their location based on the time-series analysis of arm currents.