摘要
由于风、浪、潮等环境因素的影响,传统的潮位平差和分析方法无法准确捕捉潮汐时间序列数据的复杂特征。为了解决这个问题,文章提出了一种基于长短期记忆(LSTM)神经网络的方法来预测连云港站点的潮位。通过设置不同的参数如LSTM层数、批处理大小、隐藏层节点数、初始学习率和序列长度,构建了LSTM模型,并使用了2022年1月以来连云港验潮站的小时级潮汐数据组成的数据集,进行模型的训练,并评估了该模型在不同网络参数设置下的性能,最后选择最优的模型参数对连云港未来潮汐数据进行了预测,预测结果分析表明该模型可以较好地完成预测任务。
Owing to the influence of environmental factors such as wind,wave and tide,traditional methods of tidal range and analysis are incapable of accurately capturing the complex features of tidal time series data.To solve this problem,this paper proposes a method based on the LSTM neural network to predict the tide level at the Lianyungang station.The LSTM model is constructed by setting different parameters,including the number of LSTM layers,the batch size,the number of hidden layer nodes,initial learning rate,and sequence length.A dataset composed of hourly tidal data from the Lianyungang tide gauge station since January 2022 is used to train the model,and the performance of the model under different network parameter settings is evaluated.Finally,the optimal model parameters are selected to predict the future tidal data of Lianyungang.The analysis of the prediction results demonstrates that the model can perform the prediction task quite well.
作者
凌鑫辉
尚玉杰
李小平
LING Xinhui;SHANG Yujie;LI Xiaoping(School of Communication and Artificial Intelligence,School of Integrated Circuits,Nanjing Institute of Technology,Nanjing 211167,China)
出处
《现代信息科技》
2025年第4期38-42,共5页
Modern Information Technology
