CVaR-Optimized Distributionally Robust Stacked Ensemble With Range-Based Current Signatures for Reliable Fault Detection in Photovoltaic Farms
Keywords:
Photovoltaic farm; fault detection; stacked ensemble learning; distributionally robust optimization (CVaR); range-based current signatures; multi-class fault classification.Abstract
Reliable fault detection in photovoltaic (PV) farms is critical to prevent energy losses, ensure safety, and reduce operational costs. However, the inherent variability of operating conditions and overlapping fault signatures pose significant challenges for conventional diagnostic methods. This study proposes a robust machine learning framework for multi-class string-level fault diagnosis. A compact, interpretable feature set is engineered from string current statistics and plant-level DC measurements, emphasizing range-based signatures that capture current imbalance. A heterogeneous stacked ensemble is developed, integrating Extra Trees, histogram-based gradient boosting, and an RBF support vector machine via probability-level fusion to leverage complementary decision boundaries. To enhance reliability under realistic operating shifts, a DRO approach is employed, using a Conditional Value-at-Risk (CVaR) criterion during hyperparameter tuning to prioritize performance in worst-case scenarios. Evaluated on a simulated 250-kW PV plant dataset under variable irradiance, temperature, and fault resistance, the proposed CVaR-optimized stacked ensemble achieves superior performance, with an AROC of 99.3 compared to individual base learners. The method maintains 98.3 specificity, indicating a low false-alarm rate. while ensuring strong sensitivity to fault classes. It demonstrates that combining physically informed feature engineering, stacked ensemble learning, and robustness-aware tuning provides an effective and practical approach to enhancing fault-detection reliability in grid-connected PV systems.
Downloads
References
[1] IEAPVPS.SnapshotoftheGlobalPVMarkets2020; InternationalEnergyAgency(IEA):Paris, France, 2020.
[2] S. N.Syahirah, N. Amin, M. Safar, R. Zulkafli, M. S. A. Majid, M. Afendi, and I. Zaman, “Performance factors of the photovoltaic system: A review,” MATEC Web Conf., vol. 225, p. 03020, 2018.
[3] D. S. Pillai et al., “A comprehensive review on protection challenges and fault diagnosis in PV systems,” Renew. Sustain. Energy Rev., vol. 151, p. 111561, 2021.
[4] P. Haidari, A. Hajiahmad, A. Jafari, and A. Nasiri, “Deep learning-based model for fault classification in solar modules using infrared images,” Sustain. Energy Technol. Assess., vol. 52, p. 102110, 2022.
[5] J. A. Arenas-Prado, A. H. Rangel-Rodriguez, J. P. Amezquita-Sanchez, D. Granados-Lieberman, G. Tapia-Tinoco, and M. Valtierra-Rodriguez, “Automated detection of shading faults in photovoltaic modules using convolutional neural networks and I–V curves,” Processes, vol. 13, no. 9, p. 2999, 2025.
[6] A. Sabati, R. Bayindir, and J. Rahebi, “Photovoltaic panels fault detection with convolutional neural network and bitterling fish optimization algorithm,” Int. J. Comput. Intell. Syst., vol. 18, p. 239, 2025.
[7] M. Hamad and A. Sheikholeslami, “Comparative analysis of machine learning algorithms for fault detection and localization in photovoltaic systems,” in Proc., pp. 187–193, 2024.
[8] M. A. O. Ahmed, Y. Abdelsatar, R. Alotaibi, and O. Reyad, “Enhancing Internet of Things security using performance gradient boosting for network intrusion detection systems,” Alexandria Eng. J., vol. 116, p. 472–482, 2025.
[9] Z. Wang, R. Cao, D. Tang, C. Wang, X. Liu, and W. Hu, “Distributionally robust energy optimization with renewable resource uncertainty,” Mathematics, vol. 13, no. 6, p. 992, 2025.
[10] M. Dhimish et al., “A comprehensive review of catastrophic faults in PV arrays: Types, detection, and mitigation techniques,” IEEE J. Photovoltaics, vol. 11, no. 2, p. 512–527, 2021.
[11] N.MuhammadandN.Ridzuan,“A review of fault detection and diagnosis approaches for photovoltaic systems using voltage and current analysis,” in IEEE Int. Conf. Power Electron. Appl., p. 25–30, 2024.
[12] R. Guejia-Burbano, G. Petrone, and M. Piliougine, “Impedance spectroscopy for diagnosis of photovoltaic modules under outdoor conditions,” IEEE J. Photovoltaics, 2022.
[13] J. Zwirtes, F. Bastos Líbano, L. A. Lima Silva, and E. P. Freitas, “Fault detection in photovoltaic systems using a machine learning approach,” IEEE Access, vol. 13, p. 41406–41421, 2025.
[14] Y. Zhao, L. Yang, B. Lehman, J.-F. Palma, J. Mosesian, and R. Lyons, “Decision tree-based fault detection and classification in solar photovoltaic arrays,” in Appl. Power Electron. Conf. Expo., p. 1–6, 2012.
[15] R. Janarthanan, U. Maheshwari, P. Shukla, P. Shukla, S. Mirjalili, and M. Kumar, “Intelligent detection of the PVfaults based on artificial neural network and type 2 fuzzy systems,” Energies, vol. 14, p. 6584, 2021.
[16] F. Harrou, B. Taghezouit, S. Khadraoui, A. Dairi, Y. Sun, and A. Hadj Arab, “Ensemble learning techniques-based monitoring charts for fault detection in photovoltaic systems,” Energies, vol. 15, no. 18, p. 6716, 2022.
[17] I. S. Samanta, S. Panda, P. K. Rout, M. Bajaj, M. Piecha, V. Blazek, and L. Prokop, “A comprehensive review of deep-learning applications to power quality analysis,” Energies, vol. 16, no. 11, p. 4406, 2023.
[18] N.-C.YangandH.Ismail,“Robust intelligent learning algorithm using random forest and modified independent component analysis for PV fault detection,” IEEE Access, vol. 10, p. 41119–41130, 2022.
[19] M.Bougoffa,S.Benmoussa,M.Djeziri,andO.Palais,“Hybrid deep learning for fault diagnosis in photovoltaic systems,” Machines, vol. 13, no. 5, p. 378, 2025, doi: 10.3390/machines13050378.
[20] S. F. Castellanos-Buitrago, P. Maya-Duque, W. M. Villa-Acevedo, N. Muñoz-Galeano, and J. M. López-Lezama, “Enhancing energy microgrid sizing: A multiyear optimization approach with uncertainty considerations for optimal design,” Algorithms, vol. 18, no. 2, p. 111, 2025, doi: 10.3390/a18020111.
[21] P. Geurts, D. Ernst, and L. Wehenkel, “Extremely randomized trees,” Mach. Learn., vol. 63, no. 1, p. 3–42, 2006.
[22] G.Keetal., “LightGBM: Ahighly efficient gradient boosting decision tree,” in Adv. Neural Inf. Process. Syst., vol. 30, pp. 3146–3154, 2017.
[23] C. Cortes and V. Vapnik, “Support-vector networks,” Mach. Learn., vol. 20, no. 3, p. 273–297, 1995.
[24] D. Levy, Y. Carmon, J. C.Duchi,andA.Sidford,“Large-scale methods for distributionally robust optimization,” in Adv. Neural Inf. Process. Syst., vol. 33, p. 8847–8860, 2020.
[25] R. T. Rockafellar and S. Uryasev, “Optimization of conditional value-at-risk,” J. Risk, vol. 2, p. 21–42, 2000
Downloads
Published
Issue
Section
Categories
License
Copyright (c) 2026 Engineering Systems and Intelligent Technologies (ESIT)
This work is licensed under a Creative Commons Attribution 4.0 International License.
Engineering Systems and Intelligent Technologies (ESIT) content is published under a Creative Commons Attribution License (CCBY). This means that content is freely available to all readers upon publication, and content is published as soon as production is complete.
Engineering Systems and Intelligent Technologies (ESIT) seeks to publish the most influential papers that will significantly advance scientific understanding. Selected articles must present new and widely significant data, syntheses, or concepts. They should merit recognition by the wider scientific community and the general public through publication in a reputable scientific journal.
