基于误差修正和LSTM的海浪高度预测模型

Wave height prediction model based on error correction and long and short term memory network

准确可靠的海浪高度预测是海洋工程和沿海工程应用的一项重要任务,如海洋渔业捕捞和近海勘探工程。提出一种基于误差修正和长短期记忆(LSTM)网络的海浪高度预测模型,采用自适应噪声完备集合经验模态(CEEMDAN)分解误差序列,产生误差模态分量,根据斯皮尔曼(Spearman)产生的相关系数划分每个模态分量的权重,利用长短期记忆网络对误差模态分量进行预测,将权重和预测模态分量相结合,融合到未来对应点位的预测值中,提高预测精度。实验结果表明,在均方根误差(RMSE)、拟合优度(R^(2))等评价指标上,与极限学习机(ELM)、融合注意力机制LSTM(A-LSTM)等模型进行比较,CSLM模型的评价指标较好,验证了CSLM模型的有效性和可行性。

Accurate and reliable wave height prediction is an important task for ocean engineering and coastal engineering applications,such as ocean fishing and offshore exploration engineering.In this paper,a wave height prediction model based on error correction and long and short-term memory(LSTM)network is proposed.Complete empirical Mode decomposition with adaptive noise(CEEMDAN)is used to decompose error sequences and generate error modal components.The weight of each modal component is divided according to the correlation coefficient generated by Spearman.The long and short-term memory network is used to predict the error modal components,and the weight and the predicted modal components are combined into the predicted values of corresponding points in the future to improve the prediction accuracy.The experimental results show that the CSLM model has better evaluation indexes than ELM and LSTM with attentional mechanisms(A-LSTM)in terms of root mean square error(RMSE)and goodness of fit(R^(2)),which verifies the effectiveness and feasibility of the proposed method.