Ultra-short-term global horizontal irradiance forecasting in China based on satellite and numerical weather prediction models

To address deficiencies in cloud initialization for numerical weather prediction, this study integrates Fengyun-4B (FY-4B) geostationary satellite products with the Rapid-refresh Multi-scale Analysis and Prediction System – Short Term version 3.0 (RMAPS-ST3.0) to develop an ultra-short-term (0–6 h) global horizontal irradiance (GHI) forecasting system (USI) for China based on the RMAPS WIND model, version 3.0 (RMAPS-WINDV3). Results demonstrate that both the passive mode (USI_passive, incorporating advection and diffusion of hydrometeors) and active mode (USI_active, introducing hydrometeor nudging) effectively mitigate the false dissipation of initial clouds by assimilating real-time satellite data, thereby significantly enhancing the GHI prediction accuracy. Statistical analysis reveals a greater than 7% error reduction at the 15-min forecast lead time compared to the baseline WINDV3_setup (without satellite correction), with peak improvement seen in summer (mean 9.41%) and minimal improvement in winter (2.81%). Regionally, Sichuan exhibits a greater than 20% improvement at the 15-min forecast lead time. However, the RMSE of USI_active tends toward WINDV3_setup with extended lead time, while USI_passive shows increased RMSE at nearly half of stations at the 6-h forecast lead time compared to WINDV3_setup. Notably, USI_active exhibits significant RMSE growth after 1 h in summer, indicating limitations for warm-season applications and necessitating further optimization. Our study establishes a satellite–numerical-model-coupled ultra-short-term GHI forecasting framework, thereby enriching operational prediction products and supporting grid scheduling and dual-carbon strategies. 为解决数值天气预报中云初始化不足导致的地表太阳辐照度(GHI)预报不准问题, 本研究开发了适用于中国区域的超短期(0-6小时)GHI预报系统.其核心创新在于将风云四号B星实时产品“同化”至快速更新预报系统(RMAPS-ST3.0)中, 并设计了被动与主动两种云初始化模式以改善初始云场.结果表明, 该系统有效缓解了云的初始虚假消散, 显著提升了短临预报准确性, 在15分钟预报时效内误差降低超7%.然而, 预报改进效果随时效延长而减弱, 且主动模式在暖季存在局限性, 表明该系统仍需进一步优化.此项研究建立的卫星与数值模型耦合框架, 为业务预报提供了重要支撑.