[1]夏东升,张艺涵,王延峰,等.基于TOC-VMD 和CNN BiLSTM-Attention的风电功率预测研究[J].机械与电子,2026,44(09):81-88.
　XIA Dongsheng,ZHANG Yihan,WANG Yanfeng,et al.Wind Power Prediction Based on TOC-VMD and CNN-BiLSTM-Attention[J].Machinery & Electronics,2026,44(09):81-88.

点击复制

基于TOC-VMD 和CNN BiLSTM-Attention的风电功率预测研究()

分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:

44

期数:

2026年09期

页码:

81-88

栏目:

电力控制

出版日期:

2026-09-25

文章信息/Info

Title:

Wind Power Prediction Based on TOC-VMD and CNN-BiLSTM-Attention

文章编号:

1001-2257(2026)09-0081-08

作者:

夏东升¹; 张艺涵²; 王延峰²; 陈 扬²; 冯进通¹

1.南京南瑞继保电气有限公司,江苏 南京 211102;

2.郑州轻工业大学电气信息工程学院,河南 郑州 450002

Author(s):

XIA Dongsheng¹; ZHANG Yihan²; WANG Yanfeng²; CHEN Yang²; FENG Jintong¹

(1.NR Electric Co.,Ltd.,Nanjing 211102,China;

2.School of Electrical and Information Engineering,Zhengzhou University of Light Industry,Zhengzhou 450002,China)

关键词:

风电功率预测; 龙卷风优化算法; 变分模态分解; 卷积神经网络; 双向长短期记忆网络; 注意力机制

Keywords:

wind power prediction; tornado optimization algorithm; variational mode decomposition; convolutional neural network; bidirectional long short-term memory network; attention mechanism

分类号:

TM614

文献标志码:

A

摘要:

为解决风电功率信号存在显著非平稳性与剧烈波动,致使预测效果不佳的难题,构建一种结合科氏力龙卷风优化算法(TOC)、变分模态分解(VMD)以及卷积双向长短期记忆融合注意力网络(CNN-BiLSTM-Attention)的短期风电功率预测模型。首先,采用TOC算法优化VMD的分解模态数K 和惩罚因子α 这2个核心参数,利用优化后的VMD对原始风电功率序列进行自适应分解;然后,搭建CNN-BiLSTM-Attention预测框架,分别求解各分解子序列的预测值,叠加全部分量输出风电功率最终预测结果;最后,以华中某风电场实测数据为算例,分别采用CNN-BiLSTM-Attention、PSO-VMD-CNN-BiLSTM-Attention等模型和所提模型进行对比预测。测试结果显示,所提模型具备更优的预测性能,在均方根误差、平均绝对误差和平均绝对百分比误差等多项量化评价指标上均优于各类对照模型,验证了所提模型的有效性。

Abstract:

To address the poor prediction performance caused by the pronounced non-stationarity and severe fluctuations of wind power signals,this paper proposes a short-term wind power prediction model that integrates the Tornado Optimizer with Coriolis Force (TOC),Variational Mode Decomposition (VMD),and a Convolutional Neural Network-Bidirectional Long Short-Term Memory network with an attention mechanism (CNN-BiLSTM-Attention). First,the TOC algorithm is employed to optimize two core parameters of VMD,namely the number of decomposition modes K and the penalty factor α.The optimized VMD is then used to adaptively decompose the original wind power sequence.Subsequently,a CNN-BiLSTM-Attention forecasting framework is constructed to generate predictions for each decomposed sub-series individually,and the final wind power prediction is obtained by superimposing all component forecasts.Finally,using measured data from a wind farm in Central China as a case study,the proposed model is compared against several benchmark models,including CNN-BiLSTM-Attention,PSO-VMD-CNN-BiLSTM-Attention.Experimental results demonstrate that the proposed model achieves superior predictive performance,outperforming all comparison models in multiple quantitative evaluation metrics, including root mean square error (RMSE),mean absolute error (MAE),and mean absolute percentage error(MAPE),thereby verifying its effectiveness.

备注/Memo

备注/Memo:

收稿日期:2026-06-01

基金项目:河南省科技攻关项目(252102241057)

作者简介:夏东升 (1990-),男,江苏新沂人,硕士,工程师,研究方向为新能源发电与并网、电力二次设备研发。

常用功能