基于深度学习的有效波高预测

Effective Wave Height Prediction Based on Deep Learning

有效波高对于船舶航行安全、海洋能源开发以及海洋环境保护等领域具有重要意义,对于有效波高的预报有助于深入了解海洋信息。为全面评估深度学习模型对于有效波高预测的效果,文章对比了3种深度学习模型的预测效果,评估了不同优化算法的超参数搜索效果,对比了不同分解算法对模型精度提升的效果。结果表明,LSTM模型对于有效波高的预测效果最佳;蜣螂优化算法对于LSTM超参数的搜索效果最佳;VMD分解效果优于CEEMDAN分解;结合VMD分解和蜣螂优化的LSTM模型预测效果最好。对于不同深度学习模型、优化算法以及分解算法的讨论可以为深度学习模型在有效波高的预测中提供参考。

Effective wave height is important for ship navigation safety,marine energy development and marine environmental protection,and the prediction of effective wave height can help to understand marine information deeply.In order to comprehensively evaluate the effect of deep learning models for effective wave height prediction,the paper compares the prediction effect of three deep learning models,evaluates the hyperparameter search effect of different optimization algorithms,and compares the effect of different decomposition algorithms on model accuracy improvement.The results show that the LSTM model has the best prediction effect for the effective wave height;The dung beetle optimization algorithm has the best search effect for the LSTM hyperparameters;The VMD decomposition is better than the CEEMDAN decomposition;The LSTM model combining VMD decomposition and dung beetle optimization has the best prediction effect.The discussion of different deep learning models,optimization algorithms and decomposition algorithms can provide references for deep learning models in the prediction of effective wave heights.