Prediction Of Wave Height And Disastrous Waves Caused By Typhoon

Hainan is often attacked by disastrous wave,and prediction of disastrous wave height is an important part of disaster prevention and mitigation.Typhoons are the main source of disastrous waves around Hainan.Using machine learning methods,the prediction and research of disastrous wave heights caused by typhoons plays an important guiding role in disaster prevention and mitigation in Hainan's offshore waters.Domestic and foreign research work on the prediction of catastrophic ocean wave height mainly focuses on the development of physical numerical models.The disadvantages are low computational efficiency and poor timeliness.Aiming at this practical problem,this paper analyzes the influence of typhoon on the wave height of waves,builds a wave height prediction model,and further combines the typhoon discrete data in practical applications to build a disastrous wave prediction model to judge whether it will cause disastrous waves,and remind the coastal people to take disaster prevention measures as soon as possible.The main work and contributions are as follows:1.Taking the prediction of wave height in the waters near Sanya as an entry point,the typhoon path,wind field and the temporal and spatial distribution characteristics of the wave field in the waters near Hainan Island were studied.Analysis of the correlation between typhoon parameters and wave heights is also practiced.The data visualization of the average wind speed,wind direction,wave height.wave direction and the calculation of the correlation between the wave height and wind speed in the South China Sea in each season lays the foundation for the subsequent prediction of the wave height in the waters near Hainan Island.2.Carry out data preprocessing on the typhoon,wind and wave height datasets in the sea area near Hainan Island,and use the Deep Neural Network(DNN)model to build a wave height prediction model in the sea area near Hainan Island.A Stacking ensemble model based on multiple sets of feature inputs is proposed to improve the Deep Neural Network wave height prediction model based on typhoon features.Four different feature combinations(typhoon feature,typhoon feature + local wind speed,typhoon feature + local wind speed + geographic information,typhoon feature +geographic information)are respectively input into the Deep Neural Network model.In order to reduce the misjudgment of abnormal situations,the results of four sets of Deep Neural Network models with different characteristics are stacked and integrated,and performance of the final model is significantly improved.3.Adopt the method of discretizing typhoon characteristics to compare the uncertainty of typhoon prediction,and input the discretized characteristic data into various machine learning classification algorithms to predict the level of disastrous wave caused by typhoon.Considering the problem of unbalanced ocean wave datasets with a certain amount of noise,a DBSCAN-SMOTE hybrid sampling algorithm is proposed to remove noise and perform oversampling.Finally,the balanced data is input into the machine learning model.Experiments show that hybrid-sampling models outperform unsampled or SMOTE-only models in classification.