摘要
高频地波雷达(HFSWR)海面回波谱中包含海态信息,通常基于一阶谱和二阶谱特征信息分别建立拟合模型来反演有效波高,但是单独利用一阶和二阶谱信息来反演波高,会分别存在一阶谱能量饱和和二阶谱信噪比低的问题。本文基于集成在线顺序极限学习机(EOS-ELM)的方法,利用高频地波雷达数据,综合考虑一阶谱和二阶谱的特征信息来进行有效波高的反演。学习机能够有效选择一阶谱和二阶谱信息,使结果达到最优化,从而提高有效波高的反演精度。针对低海况的数据,本文通过分析确定波高分类阈值,将数据分段进行波高反演,进一步提高了波高反演的精度。
The high frequency surface wave radar(HFSWR)sea surface echo spectrum contains sea state information.Usually,based on the first-order spectrum and second-order spectrum characteristic information,the fitting model is established to invert the significant wave height.However,if the first-order or second-order spectral information is used alone to invert the wave height,there is a problem that the first-order spectral energy saturation or second-order spectral signal-to-noise ratio is low.In this paper,the ensemble of online sequential extreme learning machine(EOS-ELM)is adopted to carry out significant wave height inversion by using the high frequency surface wave radar data and comprehensively utilizing the characteristic information of the first-order spectrum and the second-order spectrum.The learning machine can effectively select the first-order spectrum and the second-order spectrum information to optimize the result,and improve the inversion precision of effective wave height.Aiming at the data of low sea conditions,this paper determines the classification threshold of wave height through analysis,and the wave height inversion is performed on the data segment,and further improves the accuracy of wave height inversion.
作者
张晓愉
楚晓亮
王曙曜
ZHANG Xiao-Yu;CHU Xiao-Liang;WANG Shu-Yao(College of Information and Engineering,Ocean University of China,Qingdao 266100,China;CSIC PRIDE(Nan-Jing)Atmospheric and Oceanic Information System Co.Ltd,Nanjing 211106,China)
出处
《中国海洋大学学报（自然科学版）》
CAS
CSCD
北大核心
2019年第S1期163-169,共7页
Periodical of Ocean University of China
基金
国家重点研发计划“海洋环境安全专项”项目(2017YFC1405202)
国家自然科学基金项目(61671166)资助~~
关键词
高频地波雷达
有效波高反演
集成在线顺序极限学习机(EOS-ELM)
high frequency surface wave radar(HFSWR)
significant wave height inversion
ensemble of online sequential extreme learning machine(EOS-ELM)
