Deep Learning-Based Fault Detection in Hospital Microgrid Networks: A Comparative Model Evaluation

Authors

  • Blanchard H Department of Biomedical Engineering, University of Lyon, Lyon, France Author
  • Bakker A Department of Computer Science, Erasmus University Rotterdam, Rotterdam, Netherlands Author
  • Nowak T Department of Electrical Engineering, Jagiellonian University, Krakow, Poland Author

DOI:

https://doi.org/10.5281/zenodo.22644206

Keywords:

deep learning fault detection, hospital microgrid resilience, convolutional neural network fault classification, recurrent neural network fault diagnosis, comparative model evaluation

Abstract

Background: Hospital microgrids combining distributed generation, battery storage, and critical loads are increasingly proposed as
a resilient alternative to conventional backup power, but their multi-source, bidirectional architecture complicates fault
protection relative to traditional radial feeders. Deep learning has become the dominant approach to microgrid fault detection,
classification, and location, yet individual studies rarely compare architectures under matched conditions. Objective: This narrative
review performs a comparative model evaluation of deep learning architectures reported for microgrid fault diagnosis, and
examines what this comparative evidence implies for hospital microgrid protection specifically. Methods: Peer-reviewed studies
reporting convolutional, recurrent, attention-based, ensemble, and transfer-learning architectures for microgrid or closely related
power-network fault detection were synthesized narratively, alongside the hospital microgrid resilience and optimization
literature, with every cited finding traced to a verified original source. Findings: Convolutional architectures (including
Taguchi-optimized and transfer-learning variants), bidirectional LSTM and GRU networks, attention-augmented hybrids, and
ensemble classifiers each achieve peak reported accuracies in the mid-90s to high-90s percentage range on generic microgrid test
systems, with no single architecture family dominating simultaneously on accuracy, detection latency, noise robustness, and
labeled-data efficiency. Studies reporting performance across multiple operating scenarios show accuracy degrading unevenly,
with the largest drops for high-impedance and sensor/communication-fault conditions. Separately, hospital microgrid resilience
and optimization frameworks model storage and generation sizing against outage risk in considerable detail but treat fault
protection as an idealized, instantaneous function, without incorporating the detection latency or residual misclassification risk
that comparative deep-learning studies show real models to exhibit. Conclusions: The evidence base supports strong deep-learning
fault-diagnosis performance on generic microgrid topologies but has not yet been validated on hospital-representative feeder
architectures or integrated into hospital resilience metrics; closing this gap, rather than further optimizing accuracy on generic
test systems, is the priority for translating this technology into dependable hospital microgrid protection.

References

Eissa MA, Samy MM, Said M (2026) Techno-economic analysis and resilience enhancement of a hospital microgrid under grid outage scenarios. Scientific Reports, 16(1):18358, DOI:10.1038/s41598-026-56638-8.

Nazartalab P, Alavi-Rad H (2026) Resilience-oriented optimization of hospital microgrids with critical load support using ESS and PV under grid outage conditions. Scientific Reports, 16(1):5475, DOI:10.1038/s41598-026-34992-x.

Zaben MM, Worku MY, Hassan MA, Abido MA (2024) Machine Learning Methods for Fault Diagnosis in AC Microgrids: A Systematic Review. IEEE Access, 12:20260-20298, DOI:10.1109/access.2024.3360330.

Alhanaf AS, Balik HH, Farsadi M (2023) Intelligent Fault Detection and Classification Schemes for Smart Grids Based on Deep Neural Networks. Energies, 16(22):7680, DOI:10.3390/en16227680.

Sahu R, Panigrahi PK, Lal DK, Pradhan R, Mahanty C (2026) A Deep Recurrent Learning Framework for Multi-Class Microgrid Fault Classification Using LSTM and Bi-LSTM Models. Eng, 7(3):143, DOI:10.3390/eng7030143.

Khandelwal D, Anand P, Ray M, Sangeetha RG (2025) Fault detection in electrical power systems using attention-GRU-based fault classifier (AGFC-Net). Scientific Reports, 15(1):24133, DOI:10.1038/s41598-025-06493-w.

Bu Q, Lyu P, Sun R, Jing J, Lyu Z, Hou S (2025) Fault Diagnosis Method Using CNN-Attention-LSTM for AC/DC Microgrid. Modelling, 6(3):107, DOI:10.3390/modelling6030107.

Arévalo P, Cano A, Fedoseienko O, Jurado F (2025) A data-driven approach to microgrid fault detection and classification using Taguchi-optimized CNNs and wavelet transform. Applied Soft Computing, 170:112667, DOI:10.1016/j.asoc.2024.112667.

Anwar T, Mu C, Yousaf MZ, Khan W, Khalid S, Hourani AO, Zaitsev I (2025) Robust fault detection and classification in power transmission lines via ensemble machine learning models. Scientific Reports, 15(1):2549, DOI:10.1038/s41598-025-86554-2.

Tiwari SP (2025) An efficient protection scheme for critical fault detection in microgrid under uncertain scenarios using deep learning algorithm. Electrical Engineering, 107(7):9009-9023, DOI:10.1007/s00202-024-02861-3.

Wang T, Zhang C, Hao Z, Monti A, Ponci F (2023) Data-driven fault detection and isolation in DC microgrids without prior fault data: A transfer learning approach. Applied Energy, 336:120708, DOI:10.1016/j.apenergy.2023.120708.

Lee B, Kim Y, Lee H, Kang C (2025) Bidirectional Gated Recurrent Unit Neural Network for Fault Diagnosis and Rapid Maintenance in Medium-Voltage Direct Current Systems. Sensors, 25(3):693, DOI:10.3390/s25030693.

Esmaeilbeigi S, Kazemi Karegar H, Najar A (2025) A fault classification scheme based on protective agents for microgrid with parameters impact analysis. Scientific Reports, 15(1):44316, DOI:10.1038/s41598-025-28178-0.

Mamodiya U, Sinha D, Kishor I, Jawarneh M, Haj Qasem M, Almegren A (2026) Hybrid deep learning-driven explainable AI framework for fault detection and classification in smart power grids. Scientific Reports, 16:online ahead of print, DOI:10.1038/s41598-026-62771-1.

Aqamohammadi AR, Niknam T, Shojaeiyan S, Siano P, Dehghani M (2023) Deep Neural Network with Hilbert-Huang Transform for Smart Fault Detection in Microgrid. Electronics, 12(3):499, DOI:10.3390/electronics12030499.

Prashanth Kumar HK, Dash R, Jyotheeswara Reddy K, Atyam NR, Mohapatra B (2025) High resistance fault detection in DC microgrid using Hilbert Huang transform and vector-based ensemble optimized LSTM networks. Scientific Reports, 15(1):16039, DOI:10.1038/s41598-025-00977-5.

Dai Y, Wang M, Zhang L, Wang S, Jiang Z (2026) Sliding sampling and successive variational mode decomposition CNN-BiLSTM-attention based fault detection and early warning method for DC microgrid. Scientific Reports, 16:17545, DOI:10.1038/s41598-026-48188-w.

Taheri B, Hosseini SA, Hashemi-Dezaki H (2025) Enhanced Fault Detection and Classification in AC Microgrids Through a Combination of Data Processing Techniques and Deep Neural Networks. Sustainability, 17(4):1514, DOI:10.3390/su17041514.

Hübner RU, Küsel C, Oestmann JW (2023) Resilience of the critical infrastructure in hospitals: Categorization and quantification as a basis for optimization. Die Anaesthesiologie, 72(10):710-718, DOI:10.1007/s00101-023-01318-9.

Khalil M, Ravaghi H, Samhouri D, Abo J, Ali A, Sakr H, Camacho A (2022) What is "hospital resilience"? A scoping review on conceptualization, operationalization, and evaluation. Frontiers in Public Health, 10:1009400, DOI:10.3389/fpubh.2022.1009400.

Downloads

Published

2024-12-14