基于MEEMD-KELM的短期风电功率预测  

Short-term prediction of wind power based on MEEMD-KELM

作　　者：赵睿智 丁云飞 Zhao Ruizhi;Ding Yunfei(School of Electrical Engineering,Shanghai Dianji University,Shanghai 201306,China)

机构地区：[1]上海电机学院电气学院,上海201306

出　　处：《电测与仪表》2020年第21期92-98,共7页Electrical Measurement & Instrumentation

基　　金：国家自然科学基金资助项目(11302123);上海市浦江人才计划(15PJ1402500)。

摘　　要：风电功率时序信号是间歇性、波动性的非平稳信号,信号的平稳化处理是风电功率预测的关键。针对EEMD在分解风功率时序信号时存在模态混淆、伪分量和较大的重构误差等问题,将MEEMD用于风功率信号分解并与KELM模型相结合,提出了基于MEEMD-KELM的风电功率短期预测方法。该方法采用CEEMD将原始信号按频率高低依次分解,检测分量的排列熵值,通过熵值判断异常分量信号并将其从原始信号中剔除,再对分离后的信号进行EMD分解,得到的若干个IMF分量分别通过KELM模型进行组合预测。以上海某风场为例进行仿真实验,并与传统方法进行对比,结果表明该方法预测精度更优且更具稳定性。Wind power time-sequence signal is a non-stationary signal with intermittent and fluctuating characteristics,and the stationary processing of the signal is the key to wind power prediction.Aiming at the problems of mode confusion,pseudo-component and large reconstruction error in decomposition of time-sequence wind power signals by EEMD,a short-term wind power prediction method based on MEEMD-KELM is proposed by combining MEEMD applied to wind power signal decompositionwith kernel extreme learning machine(KELM)model.Firstly,the original signal is decomposed by CEEMD in order of frequency,and then,the permutation entropy value of components is detected.Then,the abnormal component signal is judged by the entropy value and removed from the original signal.Finally,the separated signal is decomposed by EMD,and several IMF components are obtained and predicted respectively by KELM model.Taking a wind farm in Shanghai as an example,the simulation experiments are carried out and compared with the traditional methods.The results show that the prediction accuracy of this method is better and more stable.

关 键 词：MEEMD KELM 风电功率预测 排列熵 模态混淆 

分 类 号：TM614[电气工程—电力系统及自动化]