基于WRF模式输出统计的逐时太阳总辐射预报初探

Preliminary research on prediction of hourly solar radiation based on WRF model output statistics

摘要:

基于WRF(weather research and forecasting model)模式逐时输出结果,设计了逐时太阳总辐射的模式输出统计(model output statistics,MOS)预报流程。主要包括:对逐时观测序列进行低通滤波再除以天文辐射,对模式输出因子的筛选和降维,以及建立MOS预报方程,并对2009年1月、4月、8月和10月武汉站逐时太阳总辐射进行预报试验。结果表明,该方案在各月预报相对稳定,拟合和预报效果均较为理想,可使平均绝对百分比误差控制在20%~30%,相对均方根误差控制在30%~40%,相对模式直接预报辐射改进了50%左右。由此可见,通过对模式输出进行解释应用,可以有效提高辐射预报的准确率。此外,客观分析所得的气温、云量、露点、比湿、相对湿度、地面气压等13个模式输出因子可以作为各地区建立MOS辐射预报方程的参考因子。

Abstract:

Based on the hourly output of WRF(weather research and forecasting model), MOS(model output statistics) prediction process of hourly solar radiation was designed, including low-pass filtering of hourly observation series which was then divided by astronomical radiation, selection and dimensionality reduction of output factors and foundation of MOS prediction functions.Judging from the results of experimental prediction of hourly solar radiation at Wuhan station in January, April, August and October 2009, it shows that this method is relatively stable in monthly forecasts and that both fitting and forecast results are satisfactory, mean absolute percentage error was controlled between 20% and 30%.Relative mean square root error was controlled between 30% and 40%.Radiation forecast was improved by 50% compared with model direct prediction.It is clear that accuracy rate of model radiation prediction can be greatly improved through interpretation and implement of model output results.Besides, 13 output factors including temperature, cloud amount, dew point, specifi ed humidity, relative humidity and surface pressure analyzed here can provide reference for other regions in fo undation of MOS functions of hourly solar radiation.