基于长短期记忆神经网络的潮位缺测值填充方法研究
Research on tidal missing data imputing with long and short term memory neural network
苗庆生 1刘玉龙 1韦广昊 1杨锦坤 1杨扬 1徐珊珊1
作者信息
摘要
潮位数据反映了沿海海平面变化,在许多领域都有着十分重要的作用,潮位数据的缺测给使用带来不便.本文基于2017年崇武和晋江海洋站潮位数据,提出了一种基于LSTM模型(长短期记忆神经网络模型)的数据缺测值填充方法,同线性插值、样条插值等传统插值方法相比,LSTM方法表现稳定,精度较高,实现方便.尤其是在缺测时间较长时,LSTM方法明显优于传统插值方法,同时该方法同样适用于海洋站其它要素如水温缺测数据的填充.
Abstract
Tidal data reflects changes in coastal sea level,and plays a very important role in various fields.The lack of tidal data brings inconvenience to the use of tide level data.Based on the tidal data of the Chongwu and Jinjiang ocean stations in 2017,this paper proposed a method of filling missing data based on the LSTM model(long short-term memory neural network model).Compared with tradition-al interpolation methods such as linear interpolation and spline interpolation,the LSTM method had stable performance,high accuracy and easy implementation.Especially when the lack of measurement time was long,the LSTM method was obviously better than the traditional interpolation method.At the same time,this method was also suitable for filling other missing data including water temperature.
关键词
潮位/缺测/LSTM/填充Key words
tidal data/data missing/LSTM/imputing引用本文复制引用
基金项目
国家重点研发计划(2023YFC2808800)
国家自然科学基金(42206226)
出版年
2024
NETL
NSTL
万方数据