Intelligent Fault Diagnosis in Electrical Power Networks Using Multisensor Signal Fusion and Explainable Machine Learning
Computer and Decision Making: An International Journal · 01.11.2026 · 3 (6S) · 983-992
DOI: 10.1922/comdem.v3i6s.18766 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 01.11.2026. Verfasst von Kavitha P und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Reliable fault diagnosis is essential for maintaining the stability, safety, and continuity of electrical power networks. This study developed a two-stage intelligent fault-diagnosis framework integrating three-phase voltage and current measurements with machine-learning techniques and model explainability. The analysis considered 12,001 observations for binary fault detection and 7,861 observations distributed across six operating states for multiclass fault classification. Current-only, voltage-only, and fused voltage–current representations were evaluated to determine the contribution of signal fusion. Logistic Regression, Random Forest, Extra Trees, SVM-RBF, and HistGradientBoosting were comparatively assessed using multiple classification metrics. The fused representation achieved the strongest binary detection performance, with HistGradientBoosting and Extra Trees attaining 99.67% accuracy and a 99.64% F1-score. For multiclass diagnosis, Extra Trees achieved the highest accuracy of 87.67% and a macro F1-score of 85.48%, although greater ambiguity occurred between the 0111 and 1111 fault states. Permutation feature importance provided an interpretable assessment of model behaviour and identified phase-current measurements as the dominant diagnostic contributors, with Ic exhibiting the greatest importance. Overall, voltage–current fusion combined with nonlinear machine learning provided accurate fault recognition while retaining interpretable relationships between electrical measurements and diagnostic outcomes. The framework offers a practical basis for further development of transparent, data-driven monitoring systems for electrical power networks.
Zitieren
APA
P, K., Rathod, A., Hardas, D. R. S., Ranjan, D. R., Kumar, G., & Parmar, J. (2026). Intelligent Fault Diagnosis in Electrical Power Networks Using Multisensor Signal Fusion and Explainable Machine Learning. Computer and Decision Making: An International Journal, 3(6S), 983-992. https://doi.org/10.1922/comdem.v3i6s.18766
Vancouver
P K, Rathod A, Hardas DRS, Ranjan DR, Kumar G, Parmar J. Intelligent Fault Diagnosis in Electrical Power Networks Using Multisensor Signal Fusion and Explainable Machine Learning. Computer and Decision Making: An International Journal. 2026;3(6S):983-992. doi:10.1922/comdem.v3i6s.18766
BibTeX
@article{P2026,
author = {P, Kavitha and Rathod, Amit and Hardas, Dr Rajani Shreepad and Ranjan, Dr. Rajeev and Kumar, Gautam and Parmar, Jaydeep},
title = {Intelligent Fault Diagnosis in Electrical Power Networks Using Multisensor Signal Fusion and Explainable Machine Learning},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {3},
number = {6S},
pages = {983--992},
doi = {10.1922/comdem.v3i6s.18766}
}
EUVIMED