太阳辐射预报滚动订正方法研究

Rolling Correction Methods for Global Radiation Forecasting

利用2013年1月至2014年12月北京南郊观象台逐时观测总辐射以及BJ-RUC模式系统预报的该站未来24h逐时总辐射、云量、水汽混合比、云水、云冰含量等14个气象要素数据,运用多种线性订正方案对总辐射预报值进行订正,重点分析了不同方法、不同季节、不同样本数的订正效果差异。结果表明:1不同季节辐射订正的影响因素不尽相同,需采用不同的组合订正因子,其中总辐射、2m比湿、2m相对湿度、低云量、中云量、高云量、总云量、云水雨冰雪霰混合比、水汽混合比可作为推荐因子;2最优样本数选取时需考虑季节差异;3逐时滚动订正方案的订正效果较好,明显优于非滚动方案。订正后总辐射误差显著减小,而且79%的时刻有改进,明显减小了预报偏大的系统误差;4冬春季订正效果优于夏秋季,这与云的宏观和微观物理量预报效果的季节差异有关。本文研究结果可为太阳能资源评估、总辐射和光伏电站发电量预报提供有效的改进方法。

Based on the hourly global radiation data collected at Beijing Weather Observatory and 24-hour forecasting data of 14 conventional meteorological elements using the BJ-RUC model, three linear correction methods are used to correct the forecasted global radiation data. Seasonal differences of corrected effects using different correction schemes and sample numbers are discussed in detail. The best rolling correction schemes are investigated. Results show： （1） Influencing factors for correcting global radiation are different in 4 seasons, and primary factors such as global radiation, specific humidity and relative humidity at 2 m, low and middle cloud cover, water vapor mixing ratios, total water mixing ratios （cloud, rain, ice, snow, and graupel） are recommended to for correction of global radiation. （2） Seasonal differences of the best sample number for different correction schemes should be considered. （3） The corrected effects using the hourly rolling scheme are better than those by using other models. By means of the hourly roiling scheme, the average absolute error, mean relative error, and relative root mean are obviously reduced, and 79 % of total data are improved after correction. Furthermore, absolute errors of the corrected global radiation present the typical normal distribution pattern. The scatter distribution of corrected and observed values is more concentrated, and larger positive errors are improved after correction. （4） The correction effects in spring and winter are better than those in autumn and summer. It may be caused by seasonal differences of forecasting errors of macroscopic and microscopic cloud physical parameters.