• 2025-04-25
• A.O. Adeyeye, A.O. Amole, O.A. Bamido
• Volume 1, Issue 1

Abstract

Microgrids are modern small-scale versions of centralized electricity systems, and due to their complexity and the significant impact of financial loss or damage in the event of a fault, the need for an effective method of fault detection is crucial. This study addressed the critical need for effective fault detection and classification to ensure timely system restoration in the vent of fault. The investigation was based on design and simulation of a microgrid model, strategically engineered to manifest fault scenarios such as varying transient faults to different types of short circuit faults. The microgrid served the dual purpose of simulating real-world challenges and generating a robust dataset for the artificial intelligence-based fault detection models. The dataset was used for training and validating the long-short term memory (LSTM) and recurrent neural network (RNN) fault detection and classification models. The microgrid simulation served as a controlled yet representative environment for fault detection model assessment. A comparative analysis of the fault detection models was carried out by evaluating their performance using metrics like as precision, recall, F1-score, and accuracy across multiple fault classes. Notably, the LSTM model demonstrated a high accuracy of 93% while the RNN model excelled in achieving perfect precision and recall scores which resulted in the model’s 100% accuracy. This study has the potential to revolve the field of microgrid fault detection and classification thereby enhancing microgrid resilience. This study finds application in sustainable microgrids design and operation consequently, promoting the realization of SDG 7 and 11.