The rapid expansion of renewable energy sources and their integration into modern power grids has intensified the need for accurate, reliable, and interpretable forecasting methods. While numerous review articles have examined machine learning (ML) applications in renewable energy forecasting, a comprehensive synthesis that systematically maps the evolution from classical models to emerging paradigms—while explicitly analyzing their trade-offs, deployment constraints, and domain-specific suitability—remains lacking. This review addresses this gap by providing a structured and up-to-date analysis of ML techniques for solar, wind, and hydropower forecasting, covering developments from 2018 to 2024. Unlike prior reviews that focus narrowly on model taxonomies or single energy sources, this work uniquely integrates three analytical dimensions: (i) a quantitative comparison of model families (regression, tree-based ensembles, deep learning, and hybrid frameworks) with respect to accuracy, interpretability, and computational efficiency; (ii) a critical evaluation of data processing strategies, feature engineering, and uncertainty quantification methods; and (iii) a forward-looking discussion of emerging paradigms—including physics-informed ML, federated learning, edge computing, and explainable AI—that address persistent challenges such as data scarcity, overfitting, and real-time deployment. By synthesizing findings from over 90 studies and presenting a comparative framework that links methodological choices to operational requirements, this review offers actionable insights for researchers and practitioners. Key challenges and future research directions are outlined to guide the development of more resilient, scalable, and cost-effective forecasting systems for next-generation renewable energy grids.
Matin Malakouti (Thu,) studied this question.