| 基于小波变换和LSTM神经网络的格陵兰岛近海海域海平面变化预测 |
| |
| 引用本文： | 赵健,刘仁强.基于小波变换和LSTM神经网络的格陵兰岛近海海域海平面变化预测[J].海洋科学,2023,47(8):7-16. |
| |
| 作者姓名： | 赵健 刘仁强 |
| |
| 作者单位： | 中国石油大学(华东) 海洋与空间信息学院, 山东 青岛 266580 |
| |
| 基金项目： | 国家重点研发计划项目（2016YFA0600102） |
| |
| 摘 要： | 海平面变化包含多种不同时间尺度信息，传统的预测方法仅对海平面变化趋势项、周期项进行拟合，难以利用海平面变化的不同时间尺度信号，使得预测精度不高。本文基于深度学习的预测模型，提出一种融合小波变换（wavelet transform，WT）与LSTM （long short-term memory，LSTM）神经网络的海平面异常组合预测模型。首先利用小波分解得到反映海平面变化总体趋势的低频分量和刻画主要细节信息的高频分量；然后通过LSTM神经网络对代表不同时间尺度的各个分量预测和重构，实现海平面变化的非线性预测。基于该模型的海平面变化预测的均方根误差、平均绝对误差和相关系数分别为12.76 mm、9.94 mm和0.937，预测精度均优于LSTM和EEMD-LSTM预测模型，WT-LSTM组合模型对区域海平面变化预测具有较好的应用价值。
|
| 关 键 词： | 海平面异常 小波变换 长短时记忆网络 海平面变化 预测 |
| 收稿时间： | 2022/4/13 0:00:00 |
| 修稿时间： | 2022/10/1 0:00:00 |
Prediction of sea level change based on wavelet transform and LSTM neural network near Greenland |
| |
| ZHAO Jian,LIU Ren-qiang.Prediction of sea level change based on wavelet transform and LSTM neural network near Greenland[J].Marine Sciences,2023,47(8):7-16. |
| |
| Authors: | ZHAO Jian LIU Ren-qiang |
| |
| Institution: | College of Oceanography and Space Informatics, China University of Petroleum (East China), Qingdao 266580, China |
| |
| Abstract: | The information on sea-level change data spans several time scales. The traditional prediction methods only fit the trend and periodic terms of sea level change, making it difficult to decompose the signals of different time scales, thereby resulting in low prediction accuracy. This paper proposes a combined prediction model of sea level anomalies based on deep learning prediction models that integrate wavelet transform (WT) and long short-term memory (LSTM) neural networks. Firstly, wavelet decomposition is performed to obtain the low-frequency component reflecting the overall trend of sea level change and the high-frequency component reflecting the main features; each component is then predicted and reconstructed by an LSTM neural network to realize the nonlinear prediction of sea level change. The root mean square error, mean absolute error, and correlation coefficient of sea-level change prediction based on this model are 12.76 mm, 9.94 mm, and 0.937, respectively, and the prediction accuracy is better than that of the LSTM and ensemble empirical mode decomposition-LSTM prediction models. Therefore, WT-LSTM combined model has better application potential for regional sea-level change prediction. |
| |
| Keywords: | sea level anomaly wavelet transform long short-term memory network sea level change prediction |
|
| 点击此处可从《海洋科学》浏览原始摘要信息 |
| 点击此处可从《海洋科学》下载免费的PDF全文 |
