Artificial neural network-based enhanced distance protection sensitivity in microgrids

The integration of distributed energy resources (DERs) into microgrids introduces dynamic operational challenges that conventional distance relays struggle to address, particularly under variable network topologies, load fluctuations, and DER intermittency. This paper proposes an artificial neural network (ANN)-enhanced distance protection scheme to improve fault detection accuracy, classification, and localization in DER-rich microgrids. A 20-layer ANN model, trained on 50 fault scenarios encompassing 11 fault types (including phase-to-phase, phase-to-ground, and high-impedance faults up to 50 Ω) and non-fault conditions, processes raw three-phase and ground impedance measurements directly. The ANN achieves a mean squared error (MSE) of 0.0143 at epoch 21, with binary outputs enabling rapid fault identification (within two power cycles) and classification. Validated under grid-connected and islanded modes with DER penetration levels of 20–80 %, the scheme demonstrates 98.7 % accuracy, 97 % noise resilience at 20 dB SNR, and precise localization of faults. Comparative analysis against traditional relays and AI-based methods (CNNs, DTs, and SVMs) reveals superior fault coverage, adaptability to DER variability, and elimination of preprocessing delays. By mitigating false tripping and DER-induced impedance errors, this ANN-based approach significantly enhances microgrid reliability, offering a robust solution for evolving power systems.