Agentic AI-driven framework for transformer fault prediction, diagnosis, and autonomous rectification planning
Bulletin of Electrical Engineering and Informatics · 01.12.2026 · 15 (6) · 5474-5485
DOI: 10.11591/eei.v15i6.13905 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bulletin of Electrical Engineering and Informatics, online erschienen am 01.12.2026. Verfasst von P. Lakshmi Narayana und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
Fault management in transformers is critical in ensuring the reliability in the power system and avoiding equipment failures. Nevertheless, the traditional monitoring methods tend to deal with fault prediction, diagnosis and corrective action independently and do not provide a combined decision making and post-action assessment. In this paper, an agentic artificial intelligence (AI) system is presented that combines transformer fault prediction, multi-class diagnosis, and autonomous rectification planning in a closed-loop system. An electrical, thermal and dissolved-gas variables reproducible synthetic dataset that includes 5,000 balanced operating records is created. Extra trees regressor is used to estimate the probability of fault and a random forest model is used to classify fault types. An expert-crafted policy based on a rule and operational and safety constraints is based on an automatic choice of corrective actions, without reinforcement learning (RL). A train-test split (80:20) and a five-fold cross-validation are used. In the held-out test set, the predictor results in MAE=0.0205, RMSE=0.0271, and R2=0.9881 and the diagnosis model results in 97.63% accuracy and macro F1=0.9762. The rule-based rectification mechanism generates an average simulated risk reduction (RR) of 72.67 with a range of 65.15-78.20 between the fault scenarios which shows integrated prediction, diagnosis, constrained action selection and feedback evaluation.
Zitieren
APA
Narayana, P. L., Rahim, R., Manoharan, K., & Prabu, A. V. (2026). Agentic AI-driven framework for transformer fault prediction, diagnosis, and autonomous rectification planning. Bulletin of Electrical Engineering and Informatics, 15(6), 5474-5485. https://doi.org/10.11591/eei.v15i6.13905
Vancouver
Narayana PL, Rahim R, Manoharan K, Prabu AV. Agentic AI-driven framework for transformer fault prediction, diagnosis, and autonomous rectification planning. Bulletin of Electrical Engineering and Informatics. 2026;15(6):5474-5485. doi:10.11591/eei.v15i6.13905
BibTeX
@article{Narayana2026,
author = {Narayana, P. Lakshmi and Rahim, Robbi and Manoharan, Kavitha and Prabu, A. V.},
title = {Agentic AI-driven framework for transformer fault prediction, diagnosis, and autonomous rectification planning},
journal = {Bulletin of Electrical Engineering and Informatics},
year = {2026},
volume = {15},
number = {6},
pages = {5474--5485},
doi = {10.11591/eei.v15i6.13905}
}
EUVIMED