基于CEEMD-PSR-FOA-LSSVM的短期风电功率预测  

Short-term prediction of wind power based on CEEMD-PSR-FOA-LSSVM

作　　者：田丽[1] 凤志民 刘世林[1] 

机构地区：[1]安徽工程大学安徽省电气传动与控制重点实验室,安徽芜湖241000

出　　处：《可再生能源》2016年第11期1632-1638,共7页Renewable Energy Resources

基　　金：安徽省自然基金项目(1508085ME74)

摘　　要：为提高短期风电功率预测精度,针对风电功率波动性大、非周期性和非线性强的特点,提出基于互补集合经验模态分解(complementary ensemble empirical mode decomposition,CEEMD)-相空间重构(phase space reconstruction,PSR)-果蝇优化算法(fruit fly optimization algorithm,FOA)-最小二乘支持向量机(least squares support vector machine,LSSVM)的组合预测方法。首先,运用CEEMD算法把风电功率序列分解为若干个分量,并用PSR算法来确定LSSVM建模过程中各个分量的输入和输出;然后,采用FOA算法优化LSSVM建模中的参数,并用训练好的LSSVM对各个分量进行单独预测;最后,用某风电场的实测数据对该组合预测方法进行验证。结果表明,与单独的LSSVM方法和FOA-LSSVM方法预测结果相比,建立的组合模型预测方法精度更高,对风电功率的短期预测更为有效和适用。Wind power has the characteristics of fluctuations, non-periodic and non-linear which will bring about large predictive errors. In order to improve forecasting accuracy of short-term wind power, we proposed an improved combined wind power forecasting method based on complementary ensemble empirical mode decomposition（CEEMD）, phase space reconstruction（PSR）, fruit fly optimization algorithm（FOA）, least squares support vector machine（LSSVM）.Firstly, CEEMD algorithm is used to decompose wind power sequence into several components, and then PSR algorithm is applied to determine the input and output component of LSSVM. After establishing LSSVM forecasting model whose parameters are optimized by FOA, all components are forecasted respectively and then combined into final forecasted results. The proposed forecast technology is applied to forecast wind power for a wind farm in the paper. The simulation results proved that the proposed method can obtain higher forecasting accuracy and effectiveness compared with the lonely LSSVM model or FOA-LSSVM model.

关 键 词：短期风电功率预测 互补集合经验模态分解 相空间重构 果蝇优化算法 最小二乘支持向量机 

分 类 号：TK89[动力工程及工程热物理—流体机械及工程]