A Review of Machine Learning Models and Evaluation Techniques for Single String Anomaly Detection in Malaysian Large-Scale Photovoltaic Systems
Journal of Renewable Energy and Smart Grid Technology · 03.07.2026 · 21 (2) · 178-191
DOI: 10.69650/rast.2026.265480 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Renewable Energy and Smart Grid Technology, online erschienen am 03.07.2026. Verfasst von Muhammad Ashraf ฺBin Ghazali und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
The rapid expansion of solar photovoltaic (PV) systems has increased the need for reliable and intelligent fault detection techniques to maintain system performance and operational efficiency. This paper presents a comprehensive review of machine learning (ML) models and evaluation techniques for single-string anomaly detection in large-scale PV systems, with particular emphasis on applications in Malaysia’s tropical environment. The review examines the transition from conventional monitoring approaches, such as I–V curve tracing and threshold-based diagnostics, to advanced data-driven methods. Various ML and deep learning models, including Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) and hybrid frameworks such as CNN–SVM and AE–GRU, are analysed in terms of their anomaly detection capabilities and performance. Common evaluation metrics, including RMSE, MAE, MAPE, R², and F1-score, are also reviewed to assess model effectiveness and reliability. In addition, recent Malaysian studies employing K-Means–LSTM clustering, Random Forest-based fault diagnosis, Vision Transformers (ViT) and Vertices Principal Component Analysis (VPCA) are discussed, highlighting their potential for PV anomaly detection under tropical operating conditions. The review identifies key challenges, particularly the dependence on labelled datasets, which limits scalability and early fault detection. Emerging opportunities in unsupervised and semi-supervised learning approaches, including autoencoders, Isolation Forests, and clustering-based reconstruction methods, are also explored as promising solutions for real-time and adaptive anomaly detection. The findings provide insights into current research trends, existing limitations, and future directions for enhancing the reliability and maintenance of large-scale PV systems in Malaysia.
Zitieren
APA
Ghazali, M. A. B., Rahiman, M. H. F., & Yasin, Z. M. (2026). A Review of Machine Learning Models and Evaluation Techniques for Single String Anomaly Detection in Malaysian Large-Scale Photovoltaic Systems. Journal of Renewable Energy and Smart Grid Technology, 21(2), 178-191. https://doi.org/10.69650/rast.2026.265480
Vancouver
Ghazali MAB, Rahiman MHF, Yasin ZM. A Review of Machine Learning Models and Evaluation Techniques for Single String Anomaly Detection in Malaysian Large-Scale Photovoltaic Systems. Journal of Renewable Energy and Smart Grid Technology. 2026;21(2):178-191. doi:10.69650/rast.2026.265480
BibTeX
@article{Ghazali2026,
author = {Ghazali, Muhammad Ashraf ฺBin and Rahiman, Mohd Hezri Fazalul and Yasin, Zuhaila Mat},
title = {A Review of Machine Learning Models and Evaluation Techniques for Single String Anomaly Detection in Malaysian Large-Scale Photovoltaic Systems},
journal = {Journal of Renewable Energy and Smart Grid Technology},
year = {2026},
volume = {21},
number = {2},
pages = {178--191},
doi = {10.69650/rast.2026.265480}
}
EUVIMED