Estudo Investigativo via o PyCaret de Ferramentas Inteligentes para a Classificação de Faltas em Linhas de Transmissão com Alta Penetração de Fontes Eólicas

Abstract

Abstract: The increasing integration of Inverter-Based Wind Resources (IBWRs) into the primary grid requires the adaptation of traditional short-circuit classification methods since the characteristics of faults in the presence of IBWRs vary according to the inverter control method used in this process. Such variations make traditional classification methods unfeasible, thus creating new challenges in accurately carrying out the classification task. In this context, the present work uses a database of different fault scenarios generated by computer simulation via software PSCAD, together with the PyCaret library, to compare fourteen intelligent methods in fault classification in systems with IBWRs. It investigates how the performance of these methods is affected by varying the number of training instances and the presence of different noise levels in the signals, in addition to their generalization capabilities for different measurement points in the transmission system. In this way, the studies highlight the potential of using machine learning to classify faults in systems with IBWRs, highlighting the methods that showed the best performance in this task and aspects that can impact such performance.