Abstract
Artificial intelligence (AI) and machine learning (ML) are rapidly reshaping operations and maintenance (O&M) in renewable energy systems (RES), facilitating early detection of faults, strong diagnostics and predictive maintenance (PdM). This review combines new developments in the field of wind energy, solar energy, hydro energy, geothermal, and hybrid energy systems making use of 15 basic peer-reviewed articles published between 2021 and 2025 within a broader context of 54 references. The analysis shows the rising dominance in time of deep architectures, Transformer and convolutional neural network- long short-term memory (CNN-LSTM) architectures, as compared to classic machine learning frameworks, with reported performance improvements that vary across studies for fault detection (FD) and remaining useful life (RUL) prediction, especially when applied to high-frequency supervisory control and data acquisition (SCADA) and thermal imagery. Hybrid physics-informed methodologies and digital twin technologies ensure even greater interpretability, stability and generalization in data-scarce domains. However, research still faces considerable challenges such as prominent data imbalance, the lack of publicly accessible fault data sets, the inconsistent benchmarking practice and weak cross-site transferability. Addressing these gaps through standardized evaluation protocols, FAIR-aligned data governance, and physically grounded hybrid models is essential for accelerating the deployment of reliable and trustworthy AI driven maintenance solutions in RES.