基于聚类经验模态分解和最小二乘支持向量机的短期风速组合预测

A Hybrid Model for Short-Term Wind Speed Forecasting Based on Ensemble Empirical Mode Decomposition and Least Squares Support Vector Machines

从分析风速序列的非线性和非平稳性特征出发,将一种基于聚类经验模态分解(EEMD)和最小二乘支持向量机(LSSVM)的组合预测模型引入到风速预测中。首先使用聚类经验模态分解将风速序列分解为一组相对平稳的子序列,以减轻不同趋势信息间的相互影响;然后运用最小二乘支持向量机对各子序列分别建模预测,为降低预测风险,使用自适应扰动粒子群算法(ADPSO)和模型学习效果反馈机制对LSSVM预测模型的输入维数和超参数进行联合优化;最后将各个子序列的预测结果叠加得到预测风速。实例研究表明,本文所提的组合预测模型可以有效挖掘风速序列特性,具有较高的预测精度。

This paper introduces a combination forecasting model which is based on ensemble empirical mode decomposition and least squares support vector machines(LSSVM) into forecasting short-term wind speed, in the view of excavating the nonstationarity and nonlinearity of wind series. Firstly, wind series are decomposed into a group of relatively stable subsequence by ensemble empirical mode decomposition to reduce mutual influences among diverse trend information. Secondly, build forecasting models respectively for each subsequence adopting least squares support vector machines. Adaptive disturbance particle swarm optimization and model learning feedback mechanism is used to jointly optimize the dimension of the learning sample input and the hyper parameters of LSSVM forecasting model to lower the predicting risk. Finally, the predicting results of each subsequence are superposed to obtain wind speed forecasting results. The case study shows that the proposed combination forecasting model is able to excavate wind series features effectively and has relatively high predicting accuracy.