EEMD-NNBR模型在降水预测中的应用

为解决降水资源预测复杂的问题,建立了具有物理意义的新预测模型,即利用集合经验模态分解(EEMD)方法,分解降水资源并识别其演变模式,获得各本征模函数(IMF),然后结合最近邻抽样回归模型(NNBR)对数据进行预测分析,汇总相应的计算结果,从而构成了EEMD-NNBR降水预测模型。以无锡市惠山区的降水序列资料为例,采用EEMD-NNBR模型预测降水资源,并与单一的NNBR模型预测值进行对比分析。结果表明,所建模型稳定性较好,能合理预测水资源演变趋势,提高降水资源预测精度,具有一定的应用价值。

Application of EEMD NNBR Model in Rainfall Forecasting

For solving complicated rainfall prediction, this paper proposes a new physical model. Ensemble empirical model decomposition (EEMD) is used to decompose rainfall series and identify its evolution mode. And then each inherent components function (IMF) is obtained. Meanwhile, an analysis of each IMF has been made by nearest neighbor bootstrapping regressive (NNBR) model. Finally, the prediction of each IMF is aggregated as the final result of EEMD NNBR model. Taking Huishan district in Wuxi City for an example, the annual rainfall is predicted by the EEMD NNBR method and the results are compared with those of single NNBR model. The results show that the model has better stability and it can predict evolution trend of water resources as well as improve prediction accuracy, which has a certain application value.