基于FY-4A地面辐射产品的太阳能资源短临预报方法研究

Research on the Nowcasting Method of Solar Energy Resources Base on FY-4A Ground Radiation Products

利用FY-4A的地面入射辐射数据和地面气象站的辐射数据,采用基于均生函数的短临预报方法获得预报辐射,并对其进行滚动订正。以山西太原站为例,对2019年进行试验,结果表明:新一代静止气象卫星FY-4A提供的高时空分辨率辐射相关产品可以为短临预报提供比较可靠的初始数据。基于卫星辐射产品的短临预报能抓住天气变化带来的影响,经过订正的短临预报辐射能得到有效改进。晴天情况下,短临预报总辐射从订正前的最大偏差120 W·m^(-2)左右改进到订正后的40 W·m^(-2)左右。多云天情况下,平均绝对误差下降到14 W·m^(-2)。阴天情况下,平均绝对误差下降到56 W·m^(-2)。从不考虑时间变化的16个预报时间步长的综合统计分析看,订正后的短临预报总辐射和观测总辐射在每个预报时间步长都吻合得非常好,相对误差从44.7%~66.4%下降到21.3%~54.3%,均方根误差从166.4~198.8 W·m^(-2)下降到132.2~189.9W·m^(-2);从整点预报时刻的全年综合统计分析看,订正后的相关系数总体比订正前提高,从0.77~0.84提高到0.79~0.88,仅在15:00(北京时)略有下降,均方根误差从162.7~197.1W·m^(-2)下降到129.3~180.9 W·m^(-2),平均相对误差从46.0%~76.8%下降到24.0%~46.5%;本方法可做全圆盘任意时刻的短临预报。

Based on the ground incident radiation data of FY-4A and the radiation data of ground weather station,the nowcasting method based on mean generation function is used to obtain the forecast radiation,as well as its rolling correction.Taking Taiyuan station in Shanxi Province as an example,the experiment in 2019 shows that the high spatial-temporal resolution radiation related products provided by the new generation geostationary meteorological satellite FY-4A can provide relatively reliable initial conditions for nowcasting prediction;The nowcasting forecast based on satellite radiation products captures the impact of weather changes,and the revised nowcasting forecast radiation has been effectively improved;In sunny days,the maximum deviation of nowcasting forecast global radiation is improved from about 120 W·m^(-2) before correction to about 40 W·m^(-2) after correction.In cloudy days,the average absolute error decreases to 14 W·m^(-2).In cloudy days,the average absolute error decreases to 56 W·m^(-2).According to the comprehensive statistical analysis of 16 prediction time steps without considering temporal variation,the corrected nowcasting global radiation and observed global radiation are coincided very well at each prediction time step,the relative error decreased from 44.7%~66.4%to 21.3%~54.3%,and the RMSE decreased from 166.4~198.8 W·m^(-2) to 132.2~189.9 W·m^(-2);From the comprehensive statistical analysis of the whole year,the revised correlation coefficient is generally higher than that before the revision,increased from 0.77~0.84 to 0.79~0.88,only slightly decreased at 15 hour,the RMSE decreased from 162.7~197.1 W·m^(-2) to 129.3~180.9 W·m^(-2),and the average absolute relative error decreased from 46.0%~76.8%to 24.0%~46.5%.This method can be used for nowcasting prediction of the whole disk at any time.