摘要
海表温度(SST)是海洋水文的重要参数,准确预测SST对海洋经济发展与极端天气的预防都有重大意义。首先,针对SST序列数据的多噪声特点,采用变分模态分解方法(VMD)预处理,以减少噪声对预测结果的影响。其次,将卷积神经网络(CNN)与长短时记忆网络(LSTM)结合,同时提取SST序列的空间与时间特征,以提高预测精度。最后,本文提出了一种基于深度学习并融合了去噪模块的SST预测模型,选取我国东海海域的SST进行实证研究。通过与基线模型、现有模型的对比,证明了本文模型不但在SST的预测精度方面提升明显,而且具有较好的鲁棒性。
Sea surface temperature(SST)is an important parameter of ocean hydrology,and accurate prediction of SST is of great significance for ocean economic development and extreme weather prevention.For the characteristics of SST series with multiple noises,variational model decomposition(VMD)is used to preprocess the SST series and reduce the influence of noise on the prediction results.Furthermore,the convolutional neural network(CNN)is combined with the long short-term memory network(LSTM)extracting both spatial and temporal features in SST sequences to improve the prediction accuracy.Finally,a SST prediction model based on deep learning with incorporating the denoising module is proposed in this paper.The SST of China's East China Sea waters is selected for empirical study.Through comparison and analysis with the baseline models and existing models,it is proved that the model in this paper not only improves the SST prediction accuracy significantly,but also has better robustness.
作者
韩莹
孙凯强
张栋
王乐豪
谈昊然
HAN Ying;SUN Kai-qiang;ZHANG Dong;WANG Le-hao;TAN Hao-ran(School of Automation,Nanjing University of Information Science and Technology,Nanjing 210044,China)
出处
《海洋环境科学》
CAS
CSCD
北大核心
2022年第5期791-798,共8页
Marine Environmental Science
基金
南方海洋科学与工程广东省实验室(珠海)基金项目(SML2020SP007)
国家自然科学基金项目(62076136)。
关键词
海表温度
变分模态分解
长短时记忆网络
卷积神经网络
sea surface temperature
variational modal decomposition
long short-term memory network
convolutional neural network
