A State-Observer-Based Protection Scheme for AC Microgrids with Recurrent Neural Network Assistance

The microgrids operate in tie-up (TU) mode with the main grid normally, and isolation (IN) without during faults. In a dynamic operational regime, protecting is highly challenging. This article proposes new microgrid protection scheme based on state observer (SO) aided by recurrent neural network (RNN). Initially, particle filter (PF) serves as SO to estimate measured current/voltage signals from corresponding bus. Then, natural log of difference between estimated current signal taken per-phase deviation (PFD). If PFD any single phase exceeds preset threshold limit, proposed successfully detects classifies Finally, RNN implemented SO-estimated voltage retrieve non-fundamental harmonic features, which are then utilized compute RNN-based observation energy (SOE). directional attributes SOE employed for localization faults microgrid. tested using Matlab® Simulink 2022b an International Electrotechnical Commission (IEC) test bed. results indicate efficacy method TU IN operation regimes radial, loop, meshed networks. Furthermore, can detect both high-impedance (HI) low-impedance (LI) 99.6% accuracy.