Convolutional Neural Network Based DC Microgrid Fault Classification

Authors: Khagendra Sharma Chapagain, Sujan Adhikari, Ansu Man Singh, Dipu Manandhara

Abstract— This research paper proposes a fault classification method for DC microgrid using convolutional neural network technique. As DC microgrid lack zero crossing and frequency information, the research applies undecimated wavelet transformation to preprocess the normal and faulty branch currents under various conditions. The preprocessed signals are then used to train and validate a convolutional neural network to perform fault classification task. The proposed scheme accurately classifies pole to pole and pole to ground DC faults, thereby providing fault information to the system caused by rapidly increasing fault currents. This research addresses a critical issue in the protection of DC microgrid and presents a promising solution that can improve the reliability and efficiency of these systems.

 Keywords— DC microgrid, DC microgrid protection, Undecimated wavelet transformation, Convolutional neural Network

Full Paper: to be purchased for $20 (contact us at [email protected])

Published In: International Conference on Role of Energy for Sustainable Social Development (RESSD-2023)
Date of Conference: 14th-15th May 2023                                
Conference Location: Kathmandu, Nepal
Publisher: IEEE Power and Energy Society Nepal Chapter

 Cite the paper as:

K. S. Chapagain, S. Adhikari, A. M. Singh, D. Manandhara, “Convolutional Neural Network Based DC Microgrid Fault Classification”, International Conference on Role of Energy for Sustainable Social Development, 14th-15th May 2023, Kathmandu, Nepal