基于改进的LMD和GRNN组合风速预测  

Composite wind speed forecasting model based on improved LMD and GRNN

作　　者：雷庆坤 李生虎[1] 陈曦鸣[2] 王艳艳[1] 华玉婷[1] 

机构地区：[1]合肥工业大学电气与自动化工程学院,安徽合肥230009 [2]安徽省电力公司,安徽合肥230061

出　　处：《合肥工业大学学报（自然科学版）》2015年第7期891-896,共6页Journal of Hefei University of Technology：Natural Science

基　　金：国家自然科学基金资助项目(51277049)

摘　　要：传统局域均值分解(LMD)算法采用滑动平均法计算局域均值函数和局域包络函数,易造成过平滑,影响分解精度。文章提出采用Akima插值法分别计算上下极值点包络线,进而求出局域均值函数和局域包络函数,对LMD方法进行改进;针对风速的非线性和非平稳性,提出基于改进LMD和广义神经网络(GRNN)的组合预测模型,用改进LMD算法分解风速,然后用GRNN对各分量分别建模预测,最后将预测结果叠加得出最终预测值。算例结果表明,LMD分解预处理可以提高预测精度,相对于现有LMD算法,改进算法分解结果更为精确,相对于GRNN及LMD-GRNN模型,改进后LMD-GRNN组合模型预测精度更高。Traditional local mean decomposition（LMD） algorithm employs the moving average method to obtain the local mean function and local envelope function, which causes the over-smoothness easily and affects the precision of decomposition. Therefore, Aldma interpolation method is proposed to calculate the envelop lines of upper and down extreme point sets and then calculate the local mean function and local envelope function to improve LMD. Considering the non-linearity and non-stationary characteristics of wind speed, the combina- tion forecasting model based on the improved LMD and generalized regression neural network（GRNN） is built up. The wind speed time series are decomposed firstly by the improved LMD, then each component is predic- ted separately by GRNN, and the final forecasting value is obtained by adding forecasting results of each com- ponent up. The simulation results show that the decomposition of wind speed can improve the prediction accu- racy, the decomposition results of improved LMD are more accurate than those of traditional LIVID, and the improved LMD-GRNN combination forecasting model has higher accuracy than GRNN model and LMD- GRNN model.

关 键 词：风速预测 LMD算法 Akima插值 GRNN网络 组合预测 

分 类 号：TM614[电气工程—电力系统及自动化]