Detailed Information

Cited 0 time in webofscience Cited 0 time in scopus
Metadata Downloads

강수-일유출량 추정 LSTM 모형의 구축을 위한 자료 수집 방안Data collection strategy for building rainfall-runoff LSTM model predicting daily runoff

Other Titles
Data collection strategy for building rainfall-runoff LSTM model predicting daily runoff
Authors
김동균강석구
Issue Date
Oct-2021
Publisher
한국수자원학회
Keywords
Long short-term memory; Rainfall-runoff model; Deep learning; Artificial neural network; Artificial intelligence; Flow discharge; Soyang river dam; Flow prediction; LSTM; 강우-유출 모형; 딥러닝; 인공신경망; 인공지능; 유량; 소양강댐; 유량예측
Citation
한국수자원학회 논문집, v.54, no.10, pp 795 - 805
Pages
11
Indexed
KCI
Journal Title
한국수자원학회 논문집
Volume
54
Number
10
Start Page
795
End Page
805
URI
https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/140749
DOI
10.3741/JKWRA.2021.54.10.795
ISSN
1226-6280
2287-6138
Abstract
본 연구는 소양강댐 유역을 대상으로 LSTM 기반의 일유출량 추정 딥러닝 모형을 개발한 후, 모형구조 및 입력자료의 다양한 조합에 대한 모형의 정확도를 살폈다. 첫 12년(1997.1.1-2008.12.31) 동안의 유역평균 일강수량, 일기온, 일풍속 (이상 입력), 일평균 유량 (출력)으로 이루어진 데이터베이스를 기반으로 모형을 구축하였으며, 이후 12년(2009.1.1-2020.12.31) 동안의 자료를 사용하여 Nash-Sutcliffe Model Efficiency Coefficient (NSE)와 RMSE를 살폈다. 가장 높은 정확도를 보인 조합은 64개의 은닉유닛을 가진 LSTM 모형 구조에 가능한 모든 입력자료(12년치의 일강수량, 일기온, 일풍속)를 활용한 경우로서 검증기간의 NSE와 RMSE는 각각 0.862와 76.8 m3/s를 기록하였다. LSTM의 은닉유닛이500개를 초과하는 경우 과적합으로 인한 모형의 성능 저하가 나타나기 시작했으며, 1000개를 초과하는 경우 과적합 문제가 두드러졌다. 12년치의 일강수만 입력자료로 활용한 경우에도 매우 높은 성능(NSE=0.8~0.84)의 모형이 구축되었으며, 한 해의 자료만을 활용하여 학습한 경우에도 충분히 활용 가능한 정확도(NSE=0.63~0.85)를 가진 모형을 구축할 수 있었다. 특히 유량의 변동성이 큰 한 해의 자료만을 활용하여 모형을 학습한 경우 매우 높은 정확도(NSE=0.85)의 모형이 구축되었다. 학습자료가 중유량과 양극한의 유량을 모두 포함한 경우라면 5년 이상의 입력자료는 모형의 성능을 크게 개선시키지 못했다.
In this study, after developing an LSTM-based deep learning model for estimating daily runoff in the Soyang River Dam basin, the accuracy of the model for various combinations of model structure and input data was investigated. A model was built based on the database consisting of average daily precipitation, average daily temperature, average daily wind speed (input up to here), and daily average flow rate (output) during the first 12 years (1997.1.1-2008.12.31). The Nash-Sutcliffe Model Efficiency Coefficient (NSE) and RMSE were examined for validation using the flow discharge data of the later 12 years (2009.1.1-2020.12.31). The combination that showed the highest accuracy was the case in which all possible input data (12 years of daily precipitation, weather temperature, wind speed) were used on the LSTM model structure with 64 hidden units. The NSE and RMSE of the verification period were 0.862 and 76.8 m3/s, respectively. When the number of hidden units of LSTM exceeds 500, the performance degradation of the model due to overfitting begins to appear, and when the number of hidden units exceeds 1000, the overfitting problem becomes prominent. A model with very high performance (NSE=0.8~0.84) could be obtained when only 12 years of daily precipitation was used for model training. A model with reasonably high performance (NSE=0.63-0.85) when only one year of input data was used for model training. In particular, an accurate model (NSE=0.85) could be obtained if the one year of training data contains a wide magnitude of flow events such as extreme flow and droughts as well as normal events. If the training data includes both the normal and extreme flow rates, input data that is longer than 5 years did not significantly improve the model performance.
Files in This Item
There are no files associated with this item.
Appears in
Collections
서울 공과대학 > 서울 건설환경공학과 > 1. Journal Articles

qrcode

Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.

Related Researcher

Researcher Kang, Seok koo photo

Kang, Seok koo
COLLEGE OF ENGINEERING (DEPARTMENT OF CIVIL AND ENVIRONMENTAL ENGINEERING)
Read more

Altmetrics

Total Views & Downloads

BROWSE