摘要
为了提高风电场输出功率的预测精度,提出一种基于经验模态分解(empiricalmode decomposition,EMD)与小波包分解(wavelet packet decomposition,WPD)的组合分解方法,与纵横交叉算法(crisscross optimization,CSO)优化后的Elman神经网络组成组合风电功率预测模型。该模型首先利用EMD将风电功率序列进行分解,然后利用样本熵计算EMD分解后序列的复杂度。对于高复杂度序列,利用WPD对序列进行二次分解,建立EMD-WPD-CSO-Elman预测模型;对于复杂度适中的序列,采用CSO优化Elman神经网络参数,建立EMD-CSO-Elman预测模型;对于低复杂度序列,直接建立EMD-Elman预测模型。最后叠加各个序列的预测结果,得到最终的风电预测功率。以某风电场实际采集数据为例,预测提前24 h的风电功率,并与EMD-WPD-CSO-BP、EMD-Elman及WPD-Elman预测模型比较,结果表明,本文提出的风电功率预测组合模型具有更好的精度。
In order to improve the prediction accuracy of wind power, put forward a combined decomposition method based on empirical mode decomposition (EMD) and wavelet packet decomposition (WPD), forming a combination forecasting method with the Elman neural network optimized by crisscross optimization algorithm (CSO). In this method, firstly, EMD is used to decompose wind power series. In order to overcome the model aliasing problems of EMD, using sample entropy to classify the EMD decomposed sequence. According to the characteristics of different categories, establish EMD-WPD-CSO-Elman, EMD-CSO-Elman, EMD-Elman prediction model, respectively. In the end, adding up the results of each predicting models, reach the final results of wind predicted power. An actual wind farm data are used as an example, forecast the pre-24-hour wind power in advance, then compare the prediction results with EMD-WPD-CSO-BP EMD-Elman and the WPD-Elman model, the results show that the proposed model has better accuracy.
出处
《内蒙古电力技术》
2017年第2期15-21,共7页
Inner Mongolia Electric Power
关键词
风电功率预测
经验模态分解
小波包分解
纵横交叉
神经网络
wind power forecast
empirical mode decomposition
wavelet packet decomposition
crisscross
neural network