<p>The global transition to renewable energy has underscored the critical role of solar power, which offers both environmental and economic benefits while addressing climate change. However, the inherent variability of solar energy due to atmospheric conditions, seasonal fluctuations, and cloud cover poses significant challenges for its integration into power grids. Accurate solar power forecasting is critical for maintaining grid reliability, optimizing energy dispatch, reducing reserve requirements, and enhancing participation in energy markets. This study presents a comprehensive evaluation of solar power forecasting methods developed between 2021 and 2025, a period marked by the rapid advancement in artificial intelligence (AI) and a significant increase in hybrid deep learning models applied to this domain. The review covers traditional statistical models, machine learning techniques, deep learning architectures, and hybrid approaches, analyzing their strengths and limitations with a focus on prediction accuracy, computational complexity, scalability, and adaptability to different climatic and operational contexts. Despite progress, key challenges persist, including data quality issues, high computational demands, a lack of generalizability across regions, and limited explainability of AI-based models. Furthermore, the adaptability of current forecasting approaches to evolving weather patterns driven by climate change remains underexplored. By investigating the most recent literature, this review identifies critical research gaps and suggests future directions for enhancing forecasting models, including improving model transparency, optimizing hybrid designs, and developing adaptive prediction schemes robust to climate uncertainty. The findings aim to guide researchers and industry practitioners in developing robust, accurate, and interpretable solar power forecasting solutions suitable for real-world grid integration.</p>