基于集合经验模态分解和深度神经网络模型的天津港风速预测研究  

Research on Wind Speed Prediction of Tianjin Port Based on Ensemble Empirical Mode Decomposition and LSTM Deep Neural Network

作　　者：卜清军 侯敏 王国松 常春辉[1] 王彩霞 BU Qingjun;HOU Min;WANG Guosong;CHANG Chunhui;WANG Caixia(Meteorological Bureau of Binhai New Area,Tianjin 300457,China;Hohai University,Nanjing 210098,Jiangsu Province,China;National Marine Data and Information Service,Tianjin 300171,China)

机构地区：[1]天津市滨海新区气象局,天津300457 [2]河海大学,江苏南京210098 [3]国家海洋信息中心,天津300171

出　　处：《天津科技》2021年第10期77-81,84,共6页Tianjin Science & Technology

基　　金：天津市海洋气象重点实验室基金“基于骨架识别和多源数据融合的近岸线状对流系统强度演变临近预警技术研究”(2020TKLOMYB02)。

摘　　要：风是天津港区域重点关注的气象要素之一,风速的预测准确性是港口安全生产的关键,为此提出了一种基于改进的经验模态分解(EMD)和神经网络模型(LSTM)的风速预测方法。EMD已广泛应用于分析非线性随机信号,集成经验模态分解(EEMD)是EMD的一种改进方法,可以有效处理模态混叠问题,并将原始数据分解为具有不同频率的更平稳的信号,每个信号均作为LSTM神经网络模型的输入数据,最终的预测风速数据则是通过汇总单个信号的预测数据而获得。以天津港区域2017年的站点实测风速为例,研究表明该混合方法比欧洲中心数据集(ERA)模式预报数据更加准确,该方法可以提高预报的准确性,适用于天津港区域的短期风速预报。Wind is one of the most important meteorological elements in the Tianjin Port area,and the accuracy of wind speed prediction is the key to safe production in the port.A wind speed predictionmethod based on improved empirical mode decomposition(EMD)and LSTM neural network is proposed.EMD has been widely used to analyze nonlinear random signals.Ensemble Empirical Mode Decomposition(EEMD)is an improved method of EMD,which can effectively deal with the problem of model aliasing and decompose the original data into more stable signals with different frequencies.Each signal is used as the input data of the LSTM neural network model.The final predicted wind speed data is obtained by summarizing the predicted data of a single signal.Taking the actual wind speed in the Tianjin Port area in 2017 as an example,the study shows that the hybrid method is more accurate than the European Central ERA model forecast data.This method can improve the accuracy of the forecast and is suitable for short-term wind speed forecasting in the Tianjin Port area.

关 键 词：天津港 EEMD LSTM 神经网络 风速预测 

分 类 号：N945[自然科学总论—系统科学]