[1]胡雅娇,李培强.基于改进型变分模态分解和集成学习的超短期风功率预测[J].福建理工大学学报,2025,23(06):584-590.[doi:10.3969/j.issn.2097-3853.2025.06.011]
 HU Yajiao,LI Peiqiang.Ultra-short-term wind power prediction based on improved variational mode decomposition and ensemble learning[J].Journal of Fujian University of Technology;,2025,23(06):584-590.[doi:10.3969/j.issn.2097-3853.2025.06.011]
点击复制

基于改进型变分模态分解和集成学习的超短期风功率预测()
分享到:

《福建理工大学学报》[ISSN:2097-3853/CN:35-1351/Z]

卷:
第23卷
期数:
2025年06期
页码:
584-590
栏目:
出版日期:
2025-12-25

文章信息/Info

Title:
Ultra-short-term wind power prediction based on improved variational mode decomposition and ensemble learning
作者:
胡雅娇李培强
福建理工大学电子电气与物理学院
Author(s):
HU Yajiao LI Peiqiang
School of Electronic, Electrical Engineering and Physics, Fujian University of Technology
关键词:
超短期风功率预测本征模态分量分类分类复合指标改进型变分模态分解集成学习
Keywords:
ultra-short-term wind power predictionclassification of intrinsic mode componentsclassification composite indicatorsimproved variational mode decompositionensemble learning
分类号:
TM614
DOI:
10.3969/j.issn.2097-3853.2025.06.011
文献标志码:
A
摘要:
准确高效的风功率预测是保障电力系统安全经济运行的关键技术。然而,风功率数据具有非线性、非平稳性的特征导致预测建模困难,提出了一种基于改进型变分模态分解(variational mode decompo-sition,VMD)和Stacking集成学习的超短期风功率预测模型。首先,通过VMD将历史风功率序列分解为相对简单的本征模态分量(intrinsic mode function,IMF),有效降低原始风功率序列的复杂度。其次,利用由C0 复杂度和Hurst 指数组成的分类复合指标将IMF 分为简单趋势分量和复杂变化分量,对简单趋势类分量采用长短期记忆网络(long short-term memory,LSTM)建模,对复杂变化类分量采用带注意力机制的增强的LSTM-Attention 作为基学习器进行预测。设计了基于卷积神经网络(convolutional neural network,CNN)的Stacking 集成学习策略,输入各个基学习器的预测值、未分解的原始风功率序列组作为元特征学习得到最终的预测结果。采用俄罗斯某风电场的实测数据构建超短期风功率预测场景,实验结果表明,结合改进型VMD 能够实现更加精准的IMF分类,采用Stacking集成学习组合各IMF分量预测值,可以有效降低最终预测误差。与传统组合方法相比,所提模型的预测精度提高了近8%、效率提高了近40%。
Abstract:
Accurate and efficient wind power prediction is a key technology for ensuring the safe and economic operation of the power system. However, the non-linear and non-stationary characteristics of wind power data make prediction modeling difficult. An ultra-short-term wind power prediction model based on improved variational mode decomposition (VMD) and Stacking ensemble learning was proposed. Firstly, the historical wind power sequence was decomposed into relatively simple intrinsic mode functions (IMF) through VMD, effectively reducing the complexity of the original wind power sequence. Secondly, the IMF was divided into simple trend components and complex change components using a classification composite index composed of C0 complexity and Hurst exponent. The simple trend components were modeled using a long short-term memory (LSTM) network, while the complex change components were predicted using an enhanced LSTM Attention with attention mechanism as the base learner. A Stacking ensemble learning strategy based on convolutional neural network (CNN) was designed. The predicted values of each base learner and the original wind power sequence group without decomposition were input as meta features to learn the final prediction results. Using measured data from a wind farm in Russia to construct an ultra-short-term wind power prediction scenario, and experimental results show that combining the improved VMD can achieve more accurate IMF classification. Using Stacking ensemble learning to combine the predicted values of each IMF component can effectively reduce the final prediction error. Compared with traditional combination methods, the proposed model has improved prediction accuracy by nearly 8% and efficiency by nearly 40%.

参考文献/References:

[1] 姚良忠,朱凌志,周明,等. 高比例可再生能源电力系统的协同优化运行技术展望[J]. 电力系统自动化,2017,41(9):36-43.[2] 钟吴君,李培强,涂春鸣. 基于尾流关联的动态超图风电功率超短期预测方法[J/OL]. 中国电机工程学报,2024:1-18. (2024-05-28)[2024-09-15]. https∥kns.cnki.net/kcms/detail/11.2107.tm.20240528.0953.003.html.[3] LIU H Y,HAN H,SUN Y,et al. Short-term wind power interval prediction method using VMD-RFG and Att-GRU[J]. Energy,2022,251:123807.[4] 胡威,张新燕,郭永辉,等. 基于游程检测法重构CEEMD的短时风功率预测[J]. 太阳能学报,2020,41(11):317-325.[5] 任东方,马家庆,何志琴,等. 基于AVMDCNNGRUAttention的超短期风功率预测研究[J]. 太阳能学报,2024,45(6):436-443.[6] 赵征,汪向硕. 基于CEEMD和改进时间序列模型的超短期风功率多步预测[J]. 太阳能学报,2020,41(7):352-358.[7] 文博,陈芳芳,胡道波,等. 基于完全集成经验模态分解和模糊熵分频的短期风电功率预测[J]. 科学技术与工程,2023,23(25):10835-10845.[8] 刘纯,范高锋,王伟胜,等. 风电场输出功率的组合预测模型[J]. 电网技术,2009,33(13):74-79.[9] 张亚超,刘开培,秦亮. 基于VMD-SE和机器学习算法的短期风电功率多层级综合预测模型[J]. 电网技术,2016,40(5):1334-1340.[10] 崔树银,汪昕杰. 基于特征工程的集成学习短期光伏功率预测[J]. 科学技术与工程,2022,22(2):532-539.[11] 王彪,吕洋,陈中,等. 考虑信息时移的分布式光伏机理-数据混合驱动短期功率预测[J]. 电力系统自动化,2022,46(11):67-74.[12] 江南,王昌,杜永峰. 分形的成因及影响探析:纪念芒德勃罗诞辰100周年[J]. 自然辩证法研究,2024,40(9):115-122.[13] 何鹏,孙瑞山. 中国民航典型征候的趋势研究[J]. 中国安全科学学报,2023,33(7):133-139.[14] 杨晶显,张帅,刘继春,等. 基于VMD和双重注意力机制LSTM的短期光伏功率预测[J]. 电力系统自动化,2021,45(3):174-182.[15] 沈恩华,蔡志杰,顾凡及. C0复杂度的数学基础[J]. 应用数学和力学,2005,26(9):1083-1090.[16] 王启云,郑中团. CEEMDAN-HURST算法在新冠疫情预测中的应用[J]. 计算机工程与应用,2023,59(7):261-268.

更新日期/Last Update: 2025-12-25