Abstract: Wind power exhibits significant intermittency and stochastic characteristics, with its output power experiencing severe fluctuations due to wind speed variations. This poses substantial challenges for power system management, particularly in grid dispersion, penetration, and grid connection. As a core means to mitigate wind power uncertainty, wind power prediction technology plays a crucial role in enhancing grid stability, reducing wind curtailment rates, optimizing electricity market transactions, and promoting the sustainable development of wind power. This study systematically analyzed the classification, fundamental principles, and mainstream methodologies of wind power prediction, conducting an in-depth comparisons of the application scenarios, advantages, limitations, and evaluation index systems of various wind power prediction methods, such as physical modeling, statistical analysis, machine learning, and hybrid methods. Based on this, a comprehensively review was presented about the key technological pathways for improving prediction accuracy, covering cutting-edge research directions such as multi-source data fusion, deep learning algorithm optimization, and error correction mechanisms, and the latest research findings were also summarized. Finally, the prospect of the future development trends in wind power prediction technology was provided, proposing innovative solutions based on digital twins, reinforcement learning, and meteorological coupled modeling.