基于EEMD和GRNN的降水量序列预测研究  

Prediction Analysis of Rainfall Series Based on EEMD and GRNN Model

作　　者：黄春艳[1,2] 韩志伟[1,3] 畅建霞[2] 王志良[1] 柳闪 丁皖豫 王影[1] 

机构地区：[1]华北水利水电大学数学与信息科学学院,河南郑州450046 [2]西安理工大学西北水资源与环境生态教育部重点实验室,陕西西安710048 [3]电子科技大学航空航天学院,四川成都611731

出　　处：《人民黄河》2017年第5期26-28,共3页Yellow River

基　　金：国家自然科学基金重大项目(51190093);国家自然科学基金资助项目(51179148);国家"973"计划项目(2011CB403306);河南省高等学校重点科研项目(15A110050)

摘　　要：为提高非平稳性、非线性降水序列的预测精度,利用基于集合经验模态分解和广义神经网络的预测模型对郑州市1951—2011年的年降水量序列进行了分析。结果表明:集合经验模态分解减弱了经验模态分解IMF分量的模态混叠现象,提高了广义神经网络的预测精度;相对于经验模态分解的广义神经网络和传统的ARMA方法,基于集合经验模态分解的广义神经网络的预测结果更加精确,具有收敛速度快及预测精度高等特点。It is very important to improve the forecasting accuracy of the future changes inhydro logicalrainfall time series, which is non-sta- tionary and non-linear. So ,using generalized regression neural netw-ork based on ensemble empirical mode decomposition to make prediction analysis for precipitation series of Zhengzhou City in the period of 1951-2011. The results show that the ensemble empirical mode decomposi-tion has restrained intrinsic mode function of empirical mode decomposition, and improved the prediction accuracy of generalized regression neural network. Comparing with ARMA method, the prediction results by generalized regression neural network based on ensemble empirical mode decomposition are more accurate, which has faster convergence and higher prediction accuracy.

关 键 词：ARMA 广义神经网络 集合经验模态分解 降水预测 

分 类 号：P457.6[天文地球—大气科学及气象学]