EUVIMEDEUVIMED entdecken
Crossref

SMART SOLAR PV GRID FAULT DIAGNOSIS AND PREDICTIVE MAINTENANCE USING HYBRID AI MODELS

Ogar-Abang Michael Obi, Julie C. Ogbulezie, Gertrude Austin Fischer, Samuel Okon Essang, Peter O. Ohiero, Moses Ede Aigberemhon

Far East Journal of Experimental and Theoretical Artificial Intelligence · 24.04.2026 · 8 (1) · 43-61

DOI: 10.17654/0974325126003 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Far East Journal of Experimental and Theoretical Artificial Intelligence, online erschienen am 24.04.2026. Verfasst von Ogar-Abang Michael Obi und 5 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid expansion of grid-connected solar Photovoltaic (PV) installations requires fault-diagnosis and maintenance frameworks that are accurate, timely, and operationally deployable. This study presents a hybrid CNN-LSTM-CatBoost architecture for smart fault diagnosis and predictive maintenance in PV systems. The model was developed using a combined dataset of 38,640 labelled multivariate samples obtained from PVlib-based simulations and real Supervisory Control and Data Acquisition (SCADA) measurements. Four operating states were considered: normal operation, partial shading, inverter fault, and grid-sag condition. Input sequences were standardized and segmented before training with the Adam optimizer for 80 epochs, batch size of 64, early stopping, and stratified five-fold cross-validation on the development set. The framework extracts temporal features automatically, classifies faults, separates environmental variability from true anomalies, and estimates Remaining Useful Life (RUL) for maintenance planning. Reproducibility was strengthened through explicit reporting of dataset composition, model hyperparameters, random-seed control, and validation procedures. On the independent test set, the proposed model outperformed conventional baselines in fault classification and achieved an RUL Mean Absolute Error (MAE) of 5.7%, with a 95% bootstrap confidence interval of 5.1%-6.3%. The results indicate that the hybrid framework can support reliable predictive maintenance, reduced downtime, and improved operational resilience in PV-enabled smart grids.

Zitieren

APA

Obi, O. M., Ogbulezie, J. C., Fischer, G. A., Essang, S. O., Ohiero, P. O., & Aigberemhon, M. E. (2026). SMART SOLAR PV GRID FAULT DIAGNOSIS AND PREDICTIVE MAINTENANCE USING HYBRID AI MODELS. Far East Journal of Experimental and Theoretical Artificial Intelligence, 8(1), 43-61. https://doi.org/10.17654/0974325126003

Vancouver

Obi OM, Ogbulezie JC, Fischer GA, Essang SO, Ohiero PO, Aigberemhon ME. SMART SOLAR PV GRID FAULT DIAGNOSIS AND PREDICTIVE MAINTENANCE USING HYBRID AI MODELS. Far East Journal of Experimental and Theoretical Artificial Intelligence. 2026;8(1):43-61. doi:10.17654/0974325126003

BibTeX
@article{Obi2026,
  author = {Obi, Ogar-Abang Michael and Ogbulezie, Julie C. and Fischer, Gertrude Austin and Essang, Samuel Okon and Ohiero, Peter O. and Aigberemhon, Moses Ede},
  title = {SMART SOLAR PV GRID FAULT DIAGNOSIS AND PREDICTIVE MAINTENANCE USING HYBRID AI MODELS},
  journal = {Far East Journal of Experimental and Theoretical Artificial Intelligence},
  year = {2026},
  volume = {8},
  number = {1},
  pages = {43--61},
  doi = {10.17654/0974325126003}
}
Kostenlos starten →