摘要
提出一种基于整体平均经验模态分解（EEMD）去噪和改进秩次集对分析的风电功率实时预测模型。首先利用EEMD将风电功率时间序列进行分解,得到有限个本征模态分量（IMF）和一个趋势分量;然后将高频IMF分量中与原始序列相关性较小的分量作为噪声滤除,将余下的分量进行重构得到消噪序列;最后利用秩次集对分析法对消噪序列进行预测。在建立秩次集对分析预测模型时,考虑风电功率幅值在建立联系度中的作用,改进联系度的建立过程。采用3个不同装机容量的风电场的实测风电功率数据进行仿真实验,结果表明：该文所提出的预测模型具有优越的预测性能,并显示出良好的普适性。
The nature of wind power prediction is to grasp the change law of wind power accurately. However,mostforecasting methods at present have ignored the impact of noise on grasping the variation of wind power. Therefore,thispaper presents a method of real-time prediction for wind power based on EEMD denoising and improved rank and setpair. At first,EEMD is used to decompose the wind power time series,and the wind power sequence is decomposed intofinite number of IMF component and a trend component. Then,the high frequency IMF component who have lesscorrelation with the original sequence is regarded as the noise,the rest of non-noise IMF component and trend componentis reconstructed as new sequence who hasn＇t noise. Finally,the rank and set pair prediction model is used to predict thenew sequence. When we establish the rank and set pair prediction model,the influence of amplitude of wind power forthe connection degree building is considered,so the connection degree built is improved. The data of three different windfarms with different installed capacity is used to implement simulation experiment,the results show that the proposedapproach has superior prediction performance and good universality.
作者
杨茂
陈郁林
魏治成
Yang Mao;Chen Yulin;Wei Zhicheng(School of Electronic Engineering, Northeast Electric Power University, Jilin 132012, China;State Grid Anhui Maintenace Company, Hefei 230061, China)
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2018年第5期1440-1448,共9页
Acta Energiae Solaris Sinica
基金
国家重点基础研究发展(973)计划(2013CB228201)
国家自然科学基金(51307017)
吉林省科技发展计划(20140520129JH)
吉林省产业技术与专项开发项目(2014Y124)
关键词
风电功率
实时预测
EEMD去噪
秩次集对分析
联系度
wind power
real-time prediction
EEMD de-noising
rank and set pair analysis
connection degree
作者简介
通信作者：杨茂（1982-），男，博士、副教授，主要从事风力发电技术方面的研究。yangma0820@163．com
