随着风能和太阳能发电技术的迅猛发展，风能和太阳能发电在电力系统中的占比持续提升。然而，这两种能源的发电量直接受天气状况影响，具有显著的随机性、波动性和间歇性，这给电力调度和电网的安全稳定运行带来了严峻的挑战。高精度的功率预测技术是应对这一挑战、实现风能和太阳能发电高效利用的关键。系统回顾了风能和太阳能短中期气象预报技术的发展历程、核心技术及其最新研究进展。首先，对国内外相关文献的综合分析表明，传统的风能和太阳能预报方法，如数值天气预报方法、统计方法（包括时间序列分析和机器学习）以及二者结合的混合/统计后处理方法，已无法完全满足当前需求。而通过在高分辨率快速循环预报、能源专用物理过程优化、混合集合卡曼滤波与四维变分同化等关键领域开展协同优化，预报的准确性与可靠性得到了显著提升。此外，新一代动力框架与先进人工智能大模型的引入、多源数据的同化融合以及“数字孪生”等新兴技术的应用，为进一步优化预报结果提供了新的可能。最后，分析了当前预报技术面临的挑战，并对未来发展方向进行了展望，包括极端天气的预报挑战；通过嵌入物理规律发展“物理信息神经网络”；由传统物理模型与大模型成员构成的集合概率预报将得到普及；量子计算等颠覆性技术有望推动超高分辨率气象模拟成为现实；实现预报技术与电网自动发电控制等场景的深度耦合，构建闭环智能决策系统。旨在为能源气象领域的研究人员和工程技术人员提供技术参考。

Abstract：With the rapid advancement of wind and solar power generation technologies, the proportion of wind and solar energy within power systems continues to rise. However, the output of these two energy sources is directly influenced by weather conditions, exhibiting significant randomness, volatility, and intermittency. This poses severe challenges to power dispatch and the secure, stable operation of the grid. High-precision power forecasting technologies at short- and medium-term time scales are crucial for addressing these challenges and achieving efficient utilization of wind and solar power generation. This paper systematically reviews the developmental trajectory, core technologies, and latest research advancements in short- and medium-term meteorological forecasting for wind and solar energy. First, a comprehensive analysis of relevant domestic and international literature indicates that traditional wind and solar energy forecasting methods—such as numerical weather prediction techniques, statistical approaches (including time series analysis and machine learning), and hybrid/statistical post-processing methods combining both—no longer fully meet current demands. Significant improvements in forecast accuracy and reliability have been achieved through synergistic optimization in key areas: high-resolution rapid-cycle forecasting, energy-specific physical process optimization, and hybrid data assimilation method of ensemble Kalman filtering (EnKF) and four-dimensional variational assimilation (4DVar). Furthermore, the introduction of next-generation dynamic frameworks and advanced artificial intelligence (AI) large models, the assimilation and fusion of multi-source data, and the application of emerging technologies like “digital twins” offer new avenues for further refining forecast outcomes. Finally, the paper analyses the challenges confronting current forecasting techniques and outlines future development directions which include but not limited to challenges posed by forecasting extreme weather events， complex terrain physical and dynamical representations, enhancing model interpretability, and subseasonal to seasonal-scale forecasting; Deep integration of artificial intelligence with physical models to develop “physical-information neural networks” by embedding physical laws; Collective probabilistic forecasting that combines traditional physical models with large AI model ensembles will become widespread; Disruptive technologies like quantum computing are poised to advance ultra-high-resolution meteorological simulations; Achieving deep coupling between forecasting technologies and scenarios such as the grid-based automatic power generation control to build closed-loop intelligent decision-making systems. This review aims to provide a technical reference for researchers and engineering technicians in the field of energy meteorology.