预测有效波高的深度学习模型研究

Research on deep learning models for predicting significant wave height

中文关键词:  深度学习  海浪  有效波高  LSTM-Attention

英文关键词:deep learning  sea wave  significant wave height  LSTM-Attention

中文摘要:

      研究基于RNN、LSTM、GRU深度学习模型，针对NOAA浮标数据集中的44013、44014、44017浮标的数据，通过斯皮尔曼相关性分析提高模型预测效果。实验结果表明，在进行相关性分析后，S-RNN、S-LSTM、 S-GRU的预测效果均比原始RNN、LSTM、GRU模型预测效果好。此外，提出一种基于LSTM的LSTM-Attention 波高预测模型，并进行相关实验，量化LSTM-Attention模型的预测效果，实验结果表明LSTM-Attention模型有更好的预测效果。为评估模型的泛化能力，研究还提出了一种采用邻近浮标数据进行学习，预测浮标缺失数据的方 法。实验结果表明，该方法的预测精度可以达到97.93%。本研究为海浪预测提供了新的方法和思路，也为未来深 度学习模型在海浪预测中的应用提供了参考。

英文摘要:

      Based on deep learning models RNN, LSTM, and GRU, this study aims to improve the predictive performance of the model for 44013, 44014, and 44017 buoys in the NOAA buoy dataset through Spearman correlation analysis. The experimental results show that after conducting correlation analysis, the prediction performance of S-RNN, S-LSTM, and S GRUmodels is better than that of the original RNN, LSTM, and GRU models. In addition, an LSTM Attention wave height prediction model based on LSTM was proposed and relevant experiments were conducted to quantify the predictive performance of the LSTM Attention model. The experimental results showed that the LSTM Attention model had better predictive performance. To evaluate the generalization ability of the model, a learning method using neighboring buoy data was proposed to predict missing data buoys. The experimental results show that the prediction accuracy of this method can reach 97.93%. This study provides new methods and ideas for wave prediction, and also provides reference for the application of deep learning models in wave prediction in the future.