Research On Intelligent And Refined Forecast Of Ocean Waves Based On Machine Learning

Wave prediction has always been an indispensable part of marine environment prediction.The prediction of ocean waves is an important means to prevent ocean wave disasters.In terms of application,through the prediction of ocean waves,it can safely and effectively guide the production and living activities such as aquaculture,marine operation and coastal tourism,ensure the safety of cities and human life and property in coastal areas,help the development of marine resources and promote the construction of a strong marine country with Chinese characteristics.This paper deeply studies the wave prediction model based on machine learning and deep learning.In view of the low accuracy of current general prediction models,this paper studies wave prediction from two aspects: model input optimization and model improvement,and proposes two different wave prediction algorithm models,which show strong prediction performance and high accuracy in wave prediction.The main research work of this article are as follows:(1)Data preprocessing and correlation analysis.Aiming at the problem of complex and redundant data sets,the feature correlation and its characteristics are studied,and the correlation analysis of marine environment data features is carried out by using Spearman coefficient,so as to enhance the correlation of experimental target data sets and provide support for subsequent data prediction.(2)A wave prediction algorithm based on RF-ATBi LSTM model is proposed.Aiming at the problems of many and miscellaneous marine environment data and low prediction accuracy of the model,a wave prediction algorithm based on RF-ATBi LSTM is proposed.For redundant data,the characteristic importance measure of random forest is introduced,the model input is optimized,and the attention mechanism is introduced into the Bi LSTM network to improve the prediction ability of the network.The RF-ATBi LSTM neural network model is compared with common prediction models and ablation experiments.The results show that the prediction loss of RF-ATBi LSTM model is lower than that of other models,that is,the model can better fit the changes of wave data and has higher prediction accuracy.(3)A wave prediction algorithm based on PAL-GRU combined model is proposed.In order to simplify the data input and solve the problems of slow model training speed and inaccurate prediction accuracy caused by too many invalid inputs,a combined model based on PAL-GRU is proposed to predict the wave height of sea waves on the basis of gated cyclic neural network.The model uses principal component analysis to screen the data set and optimize the input.On the basis of GRU,attention mechanism and two-layer LSTM network are added to improve the prediction accuracy of the model.Experiments show that the PAL-GRU combined prediction model proposed in this paper has faster training speed and accurate prediction results in wave prediction,and has certain research value for wave prediction and other marine environment prediction.