Title: Walrus optimisation algorithm based bidirectional long short term memory for DC microgrid protection

Authors: Ramaprasanna Dalai; Sarat Chandra Swain

Addresses: School of Electrical Engineering, KIIT University, Patia, Bhubaneswar, Odisha, 751024, India ' School of Electrical Engineering, KIIT University, Patia, Bhubaneswar, Odisha, 751024, India

Abstract: Protection for direct current microgrids (MG) is critical, especially when they are subjected to renewable resources. When treated to several distributed generators, the MG experiences various renewable generator events, as previously reported by researchers. However, the MG also suffers due to the problem with load, cables and converters used in distributed systems. The existing deep learning algorithms failed to detect different faults with high accuracy; thus, the protection mechanism used violates safe operation. Hence, to detect and identify various events on MG with high accuracy, a fault detection and identification system concerning the deep learning algorithm is adopted. The voltage and current signal from the system is observed and used for detection using bidirectional long short-term memory (BLSTM). For enhanced accuracy and ease in detection performance, the voltage and current signal are processed by using Hilbert hung transform; based on the decomposed features, the BLSTM accurately identifies the faults in the system. As a result, the proposed model yielded improved performance with 98.71% precision, 99.89% F1-score, 99.93% recall and 99.99% accuracy under 17.36 ms of detection time.

Keywords: microgrid; shunt arc fault; ShAF; series arc fault; SAF; line to line fault; LLF; pole to pole fault; PPF; Hilbert-Huang transform; HHT; fault detection; FD.

DOI: 10.1504/IJPELEC.2024.139483

International Journal of Power Electronics, 2024 Vol.20 No.1, pp.62 - 92

Received: 22 Sep 2023
Accepted: 15 Jan 2024
Published online: 02 Jul 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article