新型联合模型在多站点同步降水预报中的应用

SVM Considering Co-integration for Precipitation Forecasting

为了对半干旱区降水进行高精度预报,提出了一种新型预报模型,模型主要由协整理论和支持向量机两部分构成;相对于传统的单一站点预报,通过协整理论建立多站点时空关系,并运用支持向量机获得非线性关系进行联合预报;并以青海省海东和果洛月降水量数据为例,通过平均绝对误差、标准化平均方差误和纳什系数三指标,得出,所提出的多站点同步预报模型,相对于传统单站点预报模型具有更为突出的预报效果。

To obtain a high accuracy of precipitation prediction in semi-arid area,the paper proposes a new hybrid Support Vector Machine model considering co-integration theory. Based on the monthly rainfall date of Haidong and Guoluo,Qinghai Province,the proposed model can get outstanding predictive performance according to the three indicators MAE,NMSE and NS when compared with the traditional approaches.