A hybrid machine learning and ied-based fault detection scheme for microgrids

Microgrids face significant challenges in fault detection due to the integration of distributed energy resources (DERs), dynamic operational conditions, and diverse fault scenarios. Existing methods often struggle to achieve accurate and timely fault identification, necessitating the development of an efficient fault detection framework. This paper proposes a new intelligent fault detection approach that leverages advanced signal processing techniques, including modified Variable Mode Decomposition (MVMD) for feature extraction, combined with a hybrid machine learning (ML) model. The proposed method relies solely on current signals, which are processed through a two-cycle buffer and analyzed by Intelligent Electronic Devices (IEDs) installed at both ends of the feeder. Various statistical features are extracted from the processed data and used to train multiple ML models, including the proposed LVS hybrid model. Performance evaluations conducted through simulations on a modified CIGRE microgrid confirm that the LVS algorithm outperforms other ML models in terms of fault detection accuracy, dependability, and security across all IEDs. Additionally, the proposed algorithm achieves higher fault detection speed and improved computational efficiency compared to other methods. By enhancing fault detection capabilities, this framework improves the reliability of microgrids protection system.