摘要
研究提出了一种基于变分模态分解(VMD)、排列熵(PE)和多层双向长度记忆网络(Multibilstm)的高精度预测模型,旨在提高短期风功率的预测精度。通过对风力发电历史数据进行VMD分解,将其分解成多个子模态分量。根据各子模态分量的能量特征,按PE值进行分类重组,以优化提取关键信息。利用MultibiLSTM模型优化的特征注意力机制和深度残差级联网络,预测重组后的子序列。实验结果表明:该模型在预测风功率时,精确度和鲁棒性明显提高,预测误差有效降低,为避免风电行业“两个细则”考核及电力系统的稳定运行提供了有力支撑。
This study introduces a high-precision prediction model that leverages Variational Mode Decomposi-tion(VMD),Permutation Entropy(PE),and a Multi-layer Bidirectional Long Short-Term Memory Network(MultiBiLSTM)to address the challenges of short-term wind power prediction accuracy.Historical wind power data is decomposed into multiple sub-modal components using VMD.These components are then classified and reorganized based on their PE values to achieve optimal refinement.The MultiBiLSTM model employs an optimized feature attention mechanism and a deep residual cascading network to predict the reordered subse-quences.Experimental results demonstrate that this model significantly enhances prediction accuracy and ro-bustness for wind power,effectively reducing prediction errors.It provides strong support for satisfying the as-sessment according to the"Two Implementation Rules"for the wind power industry and ensuring the stable operation of power systems.
作者
张家榕
ZHANG Jiarong(Inner Mongolia Branch,CHN Energy Guoyuan Electric Power Co.,Ltd.,Hohhot,Inner Mongolia,010010)
出处
《能源科技》
2025年第3期52-56,共5页
Energy Science and Technology
关键词
风功率预测
高精度预测模型
变分模态分解
排列熵
多层双向长短时记忆网络
Wind power prediction
high-precision prediction model
variational mode decomposition
permu-tation entropy
multi-layer bidirectional long short-term memory network
