基于MAF-GWO-LSTM算法的海浪有义波高预测模型

The significant wave height prediction model based on MAF-GWO-LSTM algorithm

由于复杂海况随机海浪对船舶航行及人命安全造成威胁,通过构建海浪波高预测模型实现高海况海浪预警对提升航行安全具有重要意义。针对海浪波高预测问题,本文提出一种MAF-GWO-LSTM预测模型。首先利用滑动平均滤波器(Moving Average Filter,MAF)对实测海浪数据进行处理得到有效波高的光滑趋势序列,作为预测模型的输入训练集;再选用长短时记忆神经网络LSTM作为预测浪模型,依据灰狼优化算法(Grey Wolf Optimization,GWO)对滑动窗口MA及神经网络训练过程中的参数进行自适应寻优,并以南海实测有效波高数据进行验证。研究结果表明,采用MAF滤波有利于提取海浪有效波高特征,再通过GWO-LSTM预测模型优化神经网络参数,最优参数下波高预报精度达到R^(2)=0.991 0。论文研究可为高海况下海浪有效波高预报预警提供一种有效手段。

As random waves in complex sea conditions pose a threat to ship navigation and human life safety,it is of great significance to build a wave height prediction model to realize high sea state wave warning for improving navigation safety.Aiming at the problem of wave height prediction,a prediction model of MAF-GWO-LSTM is proposed in this paper.Firstly,the Moving Average Filter(MAF)is used to process the measured wave data to obtain the smooth trend sequence of the significant wave height,which is used as the input training set of the prediction model.Then,LSTM was selected as the wave prediction model.Grey Wolf Optimization(GWO)was used to optimize the parameters of the sliding window MA and the neural network in the training process,and the data of significant wave height measured in the South China Sea was used to verify the results.The results show that MAF filtering is conducive to extracting significant wave height features of ocean waves,and then MAF-GWO-LSTM prediction model is used to optimize the neural network parameters.Under the optimal parameters,the prediction accuracy of wave height reaches R^(2)=0.9910.This paper provides an effective method for wave height prediction and early warning under high sea conditions.