In order to reduce the error of global radiation forecast and discuss the applicability of various correction methods,the correction effect of stepwise regression,support vector machine(SVM)and ana-log bias correction(AnBC)method were compared in four seasons,demonstrating that the correction re-sults obviously reduced the forecast error,and the situations that global radiation forecast was larger and its distribution more dispersed were improved to some extent.The correction effect of stepwise regres-sion model was best in autumn at Jiuquan Hongkun photovoltaic power station with a reduction of the mean absolute deviation(MAE)of 4.56 W/m2.While the effect of AnBC method was the best in winter,summer and spring,with a reduction of MAE of 10.33,12.71 and 44.27 W/m2.For Dunhuang Huineng photovoltaic power station,the correction effect of SVM was the best in autumn and winter with a reduc-tion of MAE of 4.73 and 7.85 W/m2.In spring and summer,the effect of AnBC method was the best,MAE decreased by 16.24 and 34.17 W/m2.Hourly error distribution of the two power stations showed that the AnBC method was superior.In terms of the best correction method in each season,spring had the highest percentage of times improvement,with 73.88%at Jiuquan Hongkun and 77.28%at Dunhuang Huineng.SVM correction method was most effective on sunny days,while the AnBC method on non-sunny days.
rapid updating of multi-scale analysis and prediction systemglobal radiationstepwise re-gressionsupport vector machineanalog bias correction
维普
万方数据
