Abstract

The rapid advancement in global energy demand, driven by industrialization and technological expansion, underscores the necessity for efficient renewable energy integration into smart grids. Power generation forecasting (PGF) plays a significant role in improving and supporting microgrids, energy storage, and smart grid planning, particularly due to the intermittent nature of renewable resources. While AI and deep learning (DL) models have emerged as effective solutions for real-time PGF, existing reviews often lack a comprehensive analysis of methodologies, datasets, and evaluation strategies. This survey makes several key contributions: first, it presents a new taxonomy of AI-driven PGF methods, categorizing the latest literature across machine learning, DL, fuzzy logic, and hybrid approaches. Second, it integrates an in-depth discussion of edge intelligence in PGF, highlighting its potential to enable real-time decision-making at the point of energy generation. Third, this survey provides a comprehensive overview of the datasets used in PGF, addressing gaps in existing reviews by offering a detailed analysis of the challenges and best practices. Finally, we propose an integrated framework for evaluating PGF methods, comparing performance metrics and identifying key challenges for researchers. By consolidating these areas, this work offers a unique and detailed perspective on the current state and future directions of intelligent PGF, guiding researchers and practitioners in advancing PGF solutions.