한국수자원학회논문집 (Journal of Korea Water Resources Association)

  • 제57권2호
  • /
  • Pages.127-137
  • /
  • 2024
  • /
  • 2799-8746(pISSN)
  • /
  • 2799-8754(eISSN)
한국수자원학회 (Korea Water Resources Association)
권호 표지
DOI QR코드

DOI QR Code

LSTM - MLP 인공신경망 앙상블을 이용한 장기 강우유출모의

Long-term runoff simulation using rainfall LSTM-MLP artificial neural network ensemble

  • 안성욱 (강원대학교 방재전문대학원 도시환경재난관리전공) ;
  • 강동호 (강원대학교 방재전문대학원 도시환경재난관리전공) ;
  • 성장현 (강원대학교 도시환경재난관리전공) ;
  • 김병식 (강원대학교 AI소프트웨어학과)
  • An, Sungwook (Department of Urban Environmental Disaster Management, Kangwon National University) ;
  • Kang, Dongho (Department of Urban Environmental Disaster Management, Kangwon National University) ;
  • Sung, Janghyun (Department of Urban Environmental Disaster Management, Kangwon National University) ;
  • Kim, Byungsik (Department of AI Software, Kangwon National University)
  • 투고 : 2023.10.12
  • 심사 : 2024.02.27
  • 발행 : 2024.02.29
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

수자원 관리를 위해 주로 사용되는 물리 모형은 입력자료의 구축과 구동이 어렵고 사용자의 주관적 견해가 개입될 수 있다. 최근 수자원 분야에서 이러한 문제점을 보완하기 위해 기계학습과 같은 자료기반 모델을 이용한 연구가 활발히 진행되고 있다. 본 연구에서는 관측자료만을 이용하여 강원도 삼척시 오십천 유역의 장기강우유출모의를 수행했다. 이를 위해 기상자료로 3개의 입력자료군(기상관측요소, 일 강수량 및 잠재증발산량, 일강수량 - 잠재증발산량)을 구성하고 LSTM (Long Short-term Memory)인공신경망 모델에 각각 학습시킨 결과를 비교 및 분석했다. 그 결과 기상관측요소만을 이용한 LSTM-Model 1의 성능이 가장 높았으며, 여기에 MLP 인공신경망을 더한 6개의 LSTM-MLP 앙상블 모델을 구축하여 오십천 유역의 장기유출을 모의했다. LSTM 모델과 LSTM-MLP 모형을 비교한 결과 두 모델 모두 대체적으로 비슷한 결과를 보였지만 LSTM 모델에 비해 LSTM-MLP의 MAE, MSE, RMSE가 감소했고 특히 저유량 부분이 개선되었다. LSTM-MLP의 결과에서 저유량 부분의 개선을 보임에 따라 향후 LSTM-MLP 모델 이외에 CNN등 다양한 앙상블 모형을 이용해 물리적 모델 구축 및 구동 시간이 오래 걸리는 대유역과 입력 자료가 부족한 미계측 유역의 유황곡선 작성 등에 활용성이 높을 것으로 판단된다.

Physical models, which are often used for water resource management, are difficult to build and operate with input data and may involve the subjective views of users. In recent years, research using data-driven models such as machine learning has been actively conducted to compensate for these problems in the field of water resources, and in this study, an artificial neural network was used to simulate long-term rainfall runoff in the Osipcheon watershed in Samcheok-si, Gangwon-do. For this purpose, three input data groups (meteorological observations, daily precipitation and potential evapotranspiration, and daily precipitation - potential evapotranspiration) were constructed from meteorological data, and the results of training the LSTM (Long Short-term Memory) artificial neural network model were compared and analyzed. As a result, the performance of LSTM-Model 1 using only meteorological observations was the highest, and six LSTM-MLP ensemble models with MLP artificial neural networks were built to simulate long-term runoff in the Fifty Thousand Watershed. The comparison between the LSTM and LSTM-MLP models showed that both models had generally similar results, but the MAE, MSE, and RMSE of LSTM-MLP were reduced compared to LSTM, especially in the low-flow part. As the results of LSTM-MLP show an improvement in the low-flow part, it is judged that in the future, in addition to the LSTM-MLP model, various ensemble models such as CNN can be used to build physical models and create sulfur curves in large basins that take a long time to run and unmeasured basins that lack input data.

키워드

과제정보

본 연구는 행정안전부 지능형 상황관리 기술 개발사업의 연구비지원(과제번호2021-MOIS37-001)에 의해 수행되었습니다.

참고문헌

  1. Choi, J.R., An, S.W., Choi, J.Y., and Kim, B.S. (2021a). "A study on the derivation and evaluation of flow duration curve (FDC) using deep learning with a long short-term memory (LSTM) networks and soil water assessment tool (SWAT)." Journal of Korea Water Resources Association, Vol. 54, No. S-1, pp. 1107-1118. doi: 10.3741/JKWRA.2021.54.S-1.1107
  2. Choi, J.R., Yoon, H.C., Won, C.H., Lee, B.H., and Kim, B.S. (2021b). "A study on the estimation and evaluation of ungauged reservoir inflow for local government's agricultural drought forecasting and warning." Journal of Korea Water Resources Association, Vol. 54, No. 6, pp. 395-405. doi: 10.3741/JKWRA.2021.54.6.395.
  3. Han, H.C., Choi, C.H., Jung, J.W., and Kim, H.S. (2021). "Application of sequence to sequence learning based LSTM model (LSTM-s2s) for forecasting dam inflow." Journal of Korea Water Resources Association, Vol. 54, No. 3, pp. 157-166. doi: 10.3741/JKWRA.2021.54.3.157
  4. Han, J.H., Ryu, T.S., Lim, K.J., and Jung, Y.H. (2016). "A review of baseflow analysis techniques of watershed-scale runoff models." Journal of the Korean Society of Agricultural Engineers, Vol. 58, No. 4, pp. 75-83. doi: 10.5389/KSAE.2016.58.4.075
  5. Hochreiter, S., and Schmidhuber, J. (1997). "Long short-term memory." Neural Computation, Vol. 9, No. 8, pp. 1735-1780. doi: 10.1162/neco.1997.9.8.1735
  6. Jiang, L., Wang, X., Wang, L., Shao, M., and Zhuang, L. (2021). "A hybrid ANN-LSTM based model for indoor temperature prediction." 2021 IEEE 16th Conference on Industrial Electronics and Applications (ICIEA), Chengdu, China, doi: 10.1109/ICIEA51954.2021.9516151
  7. Kalin, L., Isik, S., Schoonover, J.E., and Lockaby, B.G. (2010). "Predicting water quality in unmonitored watersheds using artificial neural networks." Journal of Environmental Quality, Vol. 39, No. 4, pp. 1429-1440. https://doi.org/10.2134/jeq2009.0441
  8. Korea Research Institute for Human Settlements (KRIHS) (2020). A proposal to establish the reference discharges for rational river management.
  9. Lee, D.E., Lee, G.H., Kim, S.W., and Jung, S.H. (2020). "Future runoff analysis in the mekong river basin under a climate change scenario using deep learning." Water, Vol. 12, No. 6, 1156. doi: 10.3390/w12061556
  10. Lee, G.H., and Jung, S.H. (2018). "Comparison of physics-based and data-driven models for streamflow simulation of the mekong river." Journal of Korea Water Resources Association, Vol. 51, No. 6, pp. 503-514. doi: 10.3741/JKWRA.2018.51.6.503
  11. Ministry of Land, Infrastructure and Transportation of Korea (MOLIT) (2000). The long-term comprehensive water resource plan.
  12. Mok, J.Y., Choi, J.H., and Moon, Y.I. (2020). "Prediction of multipurpose dam inflow using deep learning." Journal of Korea Water Resources Association, Vol. 53, No. 2, pp. 97-105. doi:10.3741/JKWRA.2020.53.2.97
  13. Sahoo, B.B., Jha, R., Singh, A., Kumar, D. (2019). "Long short-term memory (LSTM) recurrent neural network for low-flow hydrological time series forecasting." Acta Geophys, Vol. 67, pp. 1471-1481. doi: 10.1007/s11600-019-00330-1
  14. Xu, W., Jiang, Y., Jhang, X., Li, Y., Zhang, R., and Fu, G. (2020). "Using long short-term memory networks for river flow prediction." Hydrology Research, Vol. 51, No. 6, pp. 1358-1376. doi: 10.2166/nh.2020.026