首页 | 官方网站   微博 | 高级检索  
     

基于小波神经网络的畸形波预报
引用本文:赵勇,杨秋爽,苏丹,邹丽,王爱民.基于小波神经网络的畸形波预报[J].哈尔滨工业大学学报,2021,53(6):112-117.
作者姓名:赵勇  杨秋爽  苏丹  邹丽  王爱民
作者单位:大连海事大学船舶与海洋工程学院,辽宁大连116026;工业装备结构分析国家重点实验室(大连理工大学),辽宁大连116024;高技术船舶与深海开发装备协同创新中心,上海200240;工业装备结构分析国家重点实验室(大连理工大学),辽宁大连116024
基金项目:国家自然科学基金(1,2,51939003);国防基础科研计划(SXJQR2018WDKT02);中央高校基本科研业务费专项资金资助(3132019115);青岛海洋科学与技术国家实验室开放基金(QNLM2016ORP0402);工业装备结构分析国家重点实验室自主研究课题资助(S18408);大连海事大学重点科研培育项目(3132019320)
摘    要:为对畸形波这类突发性事件进行较为准确的预报,避免畸形波对海上建筑物和人员安全产生的巨大危害.采用紧致型小波神经网络模型,根据某岛礁地形实测数据建立的岛礁三维模型中测得的波高试验数据,选取试验数据中3种典型波高时间序列分别实现了包含畸形波的波浪数据对常规波浪的预报、包含近似畸形波的波浪数据对畸形波的预报以及常规波浪对包含畸形波的波浪数据的预报.为验证小波神经网络模型精度,同时采用常规神经网络BP模型在相同条件下对3种典型波高时间序列进行预报,最后将两种神经网络预报结果精度进行对比.研究结果表明:小波神经网络能较好的捕捉畸形波突发事件,对于3种工况中的波面整体预报精度以及畸形波处的预报精度,小波神经网络预报模型均高于BP神经网络预报模型,预报的波高曲线也与实际波高曲线拟合效果更好.在神经网络训练样本中若存在畸形波特征,也将进一步提高对未来畸形波的预报精度.该项研究对船舶或海洋工程的畸形波风险预警具有一定的应用价值.

关 键 词:畸形波  小波神经网络  BP神经网络  时间序列  预报
收稿时间:2019/12/22 0:00:00

Rogue wave prediction based on wavelet neural network
ZHAO Yong,YANG Qiushuang,SU Dan,ZOU Li,WANG Aimin.Rogue wave prediction based on wavelet neural network[J].Journal of Harbin Institute of Technology,2021,53(6):112-117.
Authors:ZHAO Yong  YANG Qiushuang  SU Dan  ZOU Li  WANG Aimin
Affiliation:Naval Architecture and Ocean Engineering College, Dalian Maritime University, Dalian 116026, Liaoning, China;State Key Laboratory of Structural Analysis for Industrial Equipment Dalian University of Technology, Dalian 116024, Liaoning, China ;Collaborative Innovation Center for Advanced Ship and Deep-Sea Exploration, Shanghai 200240, China
Abstract:To accurately forecast rogue waves and avoid the great harm to the safety of the buildings and people on the sea, by utilizing the compact wavelet neural network model, combined with the wave height test data in a three-dimensional model of reefs established based on the measured data of the topography of a reef, time series of three typical wave heights were selected from experimental data to realize the prediction of wave data containing rogue waves against conventional waves, the prediction of wave data containing approximate rogue waves against rogue waves, and the prediction of conventional waves against wave data containing rogue waves. In order to verify the accuracy of the wavelet neural network model, the conventional neural network BP model was used to predict the time series of the three typical wave heights under the same conditions. Finally, the accuracy of the two neural network prediction results was compared. Results show that the wavelet neural network could capture the rogue wave emergencies better. For the overall prediction accuracy of wave surface and the prediction accuracy of rogue waves in three working conditions, the prediction model of wavelet neural network was higher than that of BP neural network, and the predicted wave height curves had better fitting effect with the actual wave height curves. If there were rogue wave features in the neural network training samples, the prediction accuracy of future rogue waves would be further improved. This research has certain application value for the risk warning of rogue waves in ships or marine engineering.
Keywords:rogue wave  wavelet neural network  BP neural network  time series  prediction
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《哈尔滨工业大学学报》浏览原始摘要信息
点击此处可从《哈尔滨工业大学学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号