黄清华,,
吴思弘
北京大学地球与空间科学学院地球物理学系, 北京 100871
基金项目: 国家自然科学基金项目(41874082)资助
详细信息
作者简介: 汪凯翔, 男, 1994年生, 北京大学硕士研究生.E-mail:kaixiang@pku.edu.cn
通讯作者: 黄清华, 男, 1967年生, 北京大学教授, 主要从事地球电磁学与地震物理学研究.E-mail:huangq@pku.edu.cn
中图分类号: P319收稿日期:2020-04-01
修回日期:2020-07-03
上线日期:2020-08-05
Application of long short-term memory neural network in geoelectric field data processing
WANG KaiXiang,HUANG QingHua,,
WU SiHong
School of Earth and Space Sciences, Peking University, Beijing 100871, China
More Information
Corresponding author: HUANG QingHua,E-mail:huangq@pku.edu.cn
MSC: P319--> Received Date: 01 April 2020
Revised Date: 03 July 2020
Available Online: 05 August 2020
摘要
摘要:作为深度学习方法的一种，长短时记忆神经网络(LSTM)是一种信号处理的重要方法.本文基于实际观测地电场数据来合成训练集，对特定结构的长短时记忆神经网络进行训练，将训练所得网络对测试集数据进行测试后，将网络应用至实际观测数据.结果显示，经过训练的网络很好地学到了训练集样本的特征，对测试集数据的信噪比压制了约20 dB，并过滤了人为添加的特定频率的干扰成分，对实际观测数据处理后得到明显的日变、半日变以及半月变、月变、半年变、年变等潮汐响应，表明长短时记忆神经网络可以有效应用于地电场数据处理研究.
关键词: 地电场/
长短时记忆神经网络/
信号处理/
潮汐响应
Abstract:As a method of deep learning,long short-term memory neural network (LSTM) is an important method for signal processing. In this paper,the training datasets are synthesized based on the characteristics of the observed geoelectric field data. The training datasets are used for training the LSTM neural network with a specific structure,while the test datasets are used to evaluate the performance of the trained network. The results indicate that the LSTM neural network works well on the synthetic training datasets. The trained network can learn the characteristics of the samples in training datasets very well,and suppress the signal-to-noise ratio by about 20 dB for the test datasets. The artificial noise with certain frequency can also be removed successfully by the trained network. Applying the trained network to the observed geoelectric data,we find clear tidal responses such as diurnal variation,semidiurnal variation and so on. This study demonstrates the effectiveness of applying the LSTM neural network to geoelectric data processing.
Key words:Geoelectric potential field/
Long-short-term memory neural network/
Signal processing/
Tidal response
PDF全文下载地址:
http://www.geophy.cn/data/article/export-pdf?id=dqwlxb_15549
