基于EMD-ARIMA-LSTM的风速预测方法  

作　　者：崔胜秋 郜伊明 霍家豪 董子阳 吕方 白瑞琪 周家睦 高波 

机构地区：[1]河北水利电力学院电力工程系,河北沧州061001 [2]沈阳工业大学化工过程自动化学院,辽宁辽阳111003

出　　处：《科技创新与应用》2025年第5期50-53,共4页Technology Innovation and Application

基　　金：国家级大学生创新创业项目(202310085006,202410085006);河北省沧州市科技项目(222102001);河北省高等学校科学研究项目(ZC2023092)。

摘　　要：风速预测在风力发电中的准确性至关重要。为提高预测精度,该文提出一种基于EMD-ARIMA-LSTM组合模型的风速预测方法。首先,采用经验模态分解(EMD)技术对风速时间序列进行处理,并计算其固有模态分量的样本熵值。然后,分别利用长短期记忆神经网络(LSTM)和差分整合移动平均自回归模型(ARIMA)对高熵和低熵的分量序列进行建模预测。实验结果表明,所提组合模型相比单独使用EMD-ARIMA或EMD-LSTM模型,具有显著的预测优势,其均方根误差(RMSE)和平均绝对百分比误差(MAPE)大幅降低,R-squared评价指标达到98.21%,展示了较高的预测精度。Accurate prediction of wind speed plays an important role in wind power generation.In this paper,a wind speed prediction method based on EMD-ARMI-LSTM combined model is proposed.Using the advantages of EMD decomposition technology to process time series,the wind speed time series is decomposed,and the sample entropy of the inherent modal component after EMD decomposition is calculated.LSTM model and ARIMA model are respectively used to predict the sequences with higher and lower sample entropy,and the final prediction results are fitted to obtain the predicted wind speed.The experimental results show that compared with EMD-ARIMA or EMD-LSTM model alone,the root mean square error(RMSE)and mean absolute percentage error(MAPE)of the combined EMD-ARMI-LSTM model are significantly reduced,and the evaluation index R-squared reaches 98.21%.The prediction effect has a high precision.

关 键 词：风速预测 EMD LSTM ARIMA 样本熵 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]