基于RIME-VMD-RIME-BiLSTM的短期风电功率预测  

Short-term wind power prediction based on RIME-VMD-RIME-BiLSTM

作　　者：王秀云[1] 祝宏斌 WANG Xiuyun;ZHU Hongbin(School of Electrical Engineering,Northeast Electric Power University,Jilin 132012,China)

机构地区：[1]东北电力大学电气工程学院,吉林吉林132012

出　　处：《电气应用》2025年第4期85-95,共11页Electrotechnical Application

基　　金：吉林省教育厅科研项目(JJKH20230118KJ)。

摘　　要：针对风电功率时间序列的随机性和波动性,为提高风电预测准确度,提出了一种结合霜冰优化算法(RIME)、变分模态分解(Variational Mode Decomposition,VMD)与双向长短期神经网络(Bidirectional Long Short-Term Memory,BiLSTM)的短期风电功率预测组合模型。首先,利用RIME算法对VMD的分解层数和惩罚因子寻优;然后,使用VMD对风电序列进行分解,得到不同频率且平稳的固有模态分量(Intrinsic Mode Function,IMF);接着,将各IMF输入至经RIME算法完成超参数寻优的BiLSTM中进行预测;最后,将各输出值进行叠加重构,得到最终结果。实验结果表明,所提预测模型在测试集上的预测误差指标分别为0.584、0.489和3.26%,均为最低值,验证了RIMEVMD-RIME-BiLSTM混合预测模型在风电功率预测领域具有较好的预测准确度和鲁棒性。To improve the accuracy of wind power forecasting and address the randomness and fluctuation of wind power time series,a hybrid short-term wind power forecasting model combining the Rime Optimization Algorithm(RIME),Variational Mode Decomposition(VMD),and Bidirectional Long Short-Term Memory(BiLSTM)is proposed.First,the RIME algorithm is used to optimize the number of decomposition layers and the penalty factor of the VMD.Then,VMD is applied to decompose the wind power time series into different frequency,stationary intrinsic mode functions(IMF).Next,each IMF is input into a BiLSTM model whose hyperparameters are optimized using the RIME algorithm for prediction.Finally,the predicted values of each IMF are aggregated and reconstructed to obtain the final forecasting result.The case study results demonstrate that the proposed forecasting model achieves prediction error metrics of 0.584,0.489,and 3.26%on the test set,all of which are the lowest values.This verifies that the RIME-VMD-RIME-BiLSTM hybrid forecasting model has high accuracy and robustness in the field of wind power forecasting.

关 键 词：风电功率 霜冰优化算法 变分模态分解 BiLSTM 

分 类 号：TM614[电气工程—电力系统及自动化]