基于改进局部自注意力机制的VMD-GRU模型短期风电功率预测  

Short-Term Wind Power Prediction Based on VMD-GRU Model with Improved Local Self-Attention Mechanism

作　　者：徐武[1] 刘洋 沈智方 范鑫豪 刘武 XU Wu;LIU Yang;SHEN Zhifang;FAN Xinhao;LIU Wu(School of Electrical and Information Engineering,Yunnan Minzu University,Kunming 650504,Yunnan,China;The Water Supply and Power Supply Company of Xinjiang Dushanzi Petrochemical Company,Karamay 824000,Xinjiang,China)

机构地区：[1]云南民族大学电气信息工程学院,云南昆明650504 [2]新疆独山子石化公司供水供电公司,新疆克拉玛依824000

出　　处：《电网与清洁能源》2023年第3期83-92,共10页Power System and Clean Energy

基　　金：国家自然科学基金项目(U1802271)。

摘　　要：较高的随机波动性使得风电功率的预测十分困难。为改善风电功率预测的效果,建立了一种基于变分模态分解(variational mode decomposition,VMD)、改进局部自注意力机制(Improved Local Self-Attention,ILSA)和门控循环单元网络(gated recurrent unit,GRU)的短期风电功率预测方法。使用VMD分解将原始风电功率序列分解为中心频率不一的子模态;对各子模态的中心频率分别建立具有不同高斯偏置优化窗口大小的ILSA模型,并改进其注意力分数公式以更有效地提取信息;采用GRU模型进行风电功率预测,并对各预测序列进行重组,得到最终的预测结果。实验结果表明,相比于各传统模型,所提改进方法能有效提高风电功率预测精度,且对于低频分量有更高的拟合度。High random volatility makes wind power prediction very difficult.In order to improve the effect of wind power prediction,a short-term wind power prediction method based on variational mode decomposition(VMD),Improved Local Self-attention(ILSA)mechanism and gated recurrent unit network(GRU)is established in this paper.First,using VMD decomposition the original wind power sequence is decomposed into sub0mioses of different center frequencies,then for the center frequency of each mode,the ILSA model with different Gaussian bias optimal window size is established,and its attention score formula is improved to extract information more effectively.The GRU is used to forecast the wind power,and each forecasting sequence is restructured to get the final prediction result.The experimental results show that compared with the traditional method,the proposed method can effectively improve the prediction accuracy of wind power,and has higher fitting degree for low-frequency components.

关 键 词：风电功率预测 变分模态分解 自注意力机制 门控循环单元 

分 类 号：TM614[电气工程—电力系统及自动化]