基于互补集合经验模态分解-模糊熵-深度信念网络的短期风速预测

针对原始风速序列具有非线性、非平稳性和不可控性的问题,提出基于互补集合经验模态分解(complementary ensemble empirical mode decomposition,CEEMD)-模糊熵(fuzzy entropy,FE)-深度信念网络(deep belief network,DBN)的短期风速预测模型。首先,利用CEEMD方法将原始风速序列分解为一系列不同尺度的本征模态分量(IMF)以降低其非平稳性;其次,利用模糊熵方法将多个IMF分量进行重组以避免分量数目过多给预测精度造成的影响;最后,利用深度信念网络其强大的深度特征提取能力和非线性映射学习能力的优点,分别对新的分量进行预测和叠加获得最终预测值。实验表明,较BP神经网络模型和DBN模型,组合模型提高了预测精度,具有可行性和有效性。

Short-term wind speed prediction based on complementary set empirical mode decomposition-fuzzy entropy-depth belief network

In view of the non-linear, non-stationary and uncontrollable nature of the original wind speed sequence, a complementary ensemble empirical mode decomposition (CEEMD) -fuzzy entropy (FE) -deep belief network (DBN) based short-term wind speed prediction model for wind farms is proposed. Firstly, CEEMD method was used to decompose the original wind speed sequence into a series of eigenmode components (IMF) of different scales to reduce its non-stationarity. Secondly, the fuzzy entropy method was used to restructure several IMF components to avoid the influence of excessive number of components on the prediction accuracy. Finally, using the advantages of deep depth belief network and its strong mapping feature, the new components are predicted and superimposed to obtain the final predicted value. Experiments show that compared with BP neural network and DBN model, the combined prediction model improves the prediction accuracy, and is feasible and effective.