摘要
海冰管理是抵御寒区海洋资源开发海冰威胁的有效手段,海冰风险的准确、快速预测是海冰管理系统的关键组成部分。文中面向海冰管理中的冰情短时预测需求,明确了基于现场监测的海冰风险预测模式,开展了应用机械学习理论的海冰风险短时预测方法研究,并以渤海辽东湾海冰管理为例,讨论了神经网络与小波分解等非线性预测方法在冰情短时预测中的适用性。结果表明,时间序列小波神经网络在短时(6 h)冰厚预测中的预测精度与Elman神经网络相仿,而在24~48 h预测中的精度偏差较大;Elman神经网络在6 h、24 h与48 h的冰厚预测中均能保持较好的预测精度,在冰流速与来冰方向预测中,模型预测精度达到80%左右。
Sea ice management is an effective means to ensure the safety of offshore resource exploitation in frigid regions.The accurate and rapid forecast of sea ice risks is the key component of the sea ice management system.In this paper,a sea ice risks forecast mode based on field monitoring has been established aiming at the short-term risk warning of sea ice,and a short-term forecast research of sea ice risks has been carried out with the machine learning methods.It also takes the Bohai sea ice management as an example to discuss the application of nonlinear forecast methods in sea ice short-term forecast.The results show that the accuracy of wavelet neural network is similar to the Elman neural network in the next 6 h ice thickness forecast while it shows a relatively large error in the next 24 h and 48 h forecasts;the Elman neural network keeps high accuracy in the next 6 h,24 h and 48 h ice thickness forecasts,and the accuracies of the sea ice flow velocity and direction forecasts are closing to 80%.
作者
于嵩松
李思茵
张大勇
王刚
岳前进
李刚
YU Song-song;LI Si-yin;ZHANG Da-yong;WANG Gang;YUE Qian-jin;LI Gang(School of Ocean Science&Technology,Dalian University of Technology,Panjin 124221,Liaoning Province,China;Faculty of Vehicle Engineering&Mechanics,Dalian University of Technology,Dalian 116024,Liaoning Province,China)
出处
《海洋技术学报》
2020年第1期32-38,共7页
Journal of Ocean Technology
基金
国家重点研发计划资助项目(2017YFF0210700,2016YFC0303402)
国家自然科学基金资助项目(51679033)
基本科研业务费理科基础科研专项资助项目(DUT18LK49)。
关键词
海冰管理
冰情预测
ELMAN神经网络
小波分解
sea ice management
ice conditions forecast
Elman neural network
wavelet neural network