基于小波变换与Elman神经网络的短期风速组合预测

Short-term combination forecasting of wind speed based on wavelet transform and Elman neural networ

风速的准确预测对风电场发电系统的经济和安全运行有着重要的作用。为了克服风速随机性强的缺点,提高短期风速预测的精度,提出了一种将小波变换与Elman神经网络相结合的短期风速组合预测模型。该模型由小波预处理模块和神经网络预测模块组成。首先利用小波预处理模块将风速序列作多尺度分解,重构得到不同频段的子序列,然后利用Elman神经网络模块分别对其训练和预测。实际风速预测结果表明,与单一的Elman和ARMA法相比,该组合预测模型的预测精度有较大的改善,可以用于风电场短期风速的预测。

Accurate forecasting of wind speed is important for the economic and secure operation of wind power generation systems. In order to overcome the randomness of wind, improve the accuracy of short-term wind speed forecasting, a combination forecasting model of short-term wind speed based on wavelet transform and Elman neural network is presented in this paper. The model consists of a wavelet pre-processing module and a neural network prediction module. First, using wavelet transform, the wind speed time series is decomposed and reconstructed into the sub-sequences at different frequent band, then these sub-sequences are input into Elman networks for training and prediction, respectively. Results of the actual wind speed forecasting show, in comparison with single Elman network and ARMA method, the prediction accuracy of the combination forecasting model has greatly improved, which can be used as short-term wind speed prediction.