基于WOA-VMD-SSA-LSTM的中长期风电预测  

MEDIUM AND LONG TERM WIND POWER FORECAST BASED ON WOA-VMD-SSA-LSTM

作　　者：胡锐 乔加飞 李永华[1] 孙亚萍 王兵兵 Hu Rui;Qiao Jiafei;Li Yonghua;Sun Yaping;Wang Bingbing(Department of Power Engineering,North China Electric Power University,Baoding 071066,China;CHN Energy New Energy Technology Research Institute Co.,Ltd.,Beijing 102209,China)

机构地区：[1]华北电力大学动力工程系,保定071066 [2]国家能源集团新能源技术研究院有限公司,北京102209

出　　处：《太阳能学报》2024年第9期549-556,共8页Acta Energiae Solaris Sinica

基　　金：国家能源集团科技项目(GJNY2030XDXM-19-10.1)。

摘　　要：针对风速预测中由于随机性和波动性使得风速预测精度不高和模型泛化性不佳的问题,提出一种基于变分模态分解(VMD)、鲸鱼优化算法(WOA)、长短期记忆神经网络(LSTM)和麻雀优化算法(SSA)的组合预测模型。首先利用WOA对VMD的核心参数(K值和惩罚系数α)进行自动寻优。经对风速时间序列进行分解之后,引入SSA优化LSTM的核心学习参数,最后整合各子分量的预测风速数据得到最终风速预测值,经过多项模型评价指标的验证,模型的RMSE、MAE、MAPE、R^(2)分别为0.0758 m/s、0.0578 m/s、1.492%和0.979,与其他单一优化预测模型WOA-VMD-LSTM和VMD-SSA-LSTM相比较,相关模型评价指标均有较显著的改观。Aiming at the problems of low accuracy and poor generalization of wind speed forecast due to randomness and volatility,a combined prediction model based on variational mode decomposition(VMD),whale optimization algorithm(WOA),long short-term memory neural network(LSTM)and sparrow search algorithm(SSA)was proposed.Firstly,WOA is used to automatically optimize the core parameters of VMD(K value and penalty coefficientα).After decomposing the wind speed time series,SSA is introduced to optimize the core learning parameters of LSTM,and finally,the predicted wind speed data of each subcomponent is integrated to obtain the final predicted wind speed,which is verified by a number of model evaluation indicators.The RMSE,MAE,MAPE and R^(2) of the model are 0.0758 m/s,0.0578 m/s,1.492%and 0.979,respectively.Compared with other single optimization prediction models WOA-VMD-LSTM and VMD-SSA-LSTM,the relevant evaluation indicators have significantly improved.

关 键 词：风速 预测分析 变分模态分解 长短期记忆神经网络 鲸鱼优化算法 

分 类 号：TM614[电气工程—电力系统及自动化]