利用深度學習方法檢測震前電離層異常;Using deep learning to detect pre-earthquake ionospheric anomalies

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/93323

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/93323

Title:	利用深度學習方法檢測震前電離層異常;Using deep learning to detect pre-earthquake ionospheric anomalies
Authors:	郭家暉;Kuo, Chia-Hui
Contributors:	資訊工程學系
Keywords:	全電子含量;深度學習;地震;TEC;Deep learning;Earthquake
Date:	2023-07-27
Issue Date:	2024-09-19 16:53:50 (UTC+8)
Publisher:	國立中央大學
Abstract:	本篇論文使用深度學習模型 ConvLSTM 來預測具有震前電離層異常之地震，使用的資料為 GIMTEC 公開資料。有別於地震預警系統只能在地震發生前幾秒或是幾分鐘才能收到通知，對於地震發生前的準備，明顯是時間不足的，如果能早在前一天甚至是數天前掌握地震即將到來的資訊，便可以及早預防、疏散，才能夠大幅度減少地震所帶來的災害。為了預測出那些具有明顯的地震電離層異常之規模6以上的地震，有別於使用傳統 LSTM 模型，本篇論文使用之ConvLSTM模型能夠獲得數值圖象二維空間的訊息，相較於LSTM，ConvLSTM 模型能夠更大的利用相鄰幾天的資訊來訓練模型，並得到更可靠的結果。;In this paper, we use the deep learning model ConvLSTM to detect earthquakes with pre-seismic ionospheric anomalies, using the publicly available GIMTEC data. Unlike earthquake early warning systems that can only receive notifications a few seconds or minutes before an earthquake occurs, it is obviously that there is insufficient time for preparation before an earthquake. If we can obtain information about an impending earthquake a day or even several days in advance, we can take early preventive measures and evacuation, thus significantly reducing the disasters caused by earthquakes. In order to predict earthquakes of magnitude 6 or above with significant ionospheric anomalies, this paper utilizes the ConvLSTM model instead of the traditional LSTM model. The ConvLSTM model can capture spatial information in two-dimensional numerical images, enabling it to better utilize information from adjacent days for training and obtain more reliable results compared to LSTM.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	15	View/Open

社群 sharing

Loading...