Development of Dam Operation Support Technology for Minimizing Flood Damage Using Machine Learning Methods

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K04698
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 22040:Hydroengineering-related
• Research Institution: Muroran Institute of Technology
• Principal Investigator: Nakatsugawa Makoto 室蘭工業大学, 大学院工学研究科, 教授 (10344425)
• Co-Investigator(Kenkyū-buntansha): 小林 洋介 室蘭工業大学, 大学院工学研究科, 助教 (10735103) ; 一言 正之 日本工営株式会社中央研究所, 先端研究開発センター, 研究員 (40463559)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: ダム流入量予測 / 機械学習 / スパースモデリング / Elastic Net / 未経験事例 / 予測手法の一般化 / 気候変動 / 事前放流

Outline of Final Research Achievements

This research aims to improve the flood control function of dams by utilizing the large amount of accumulated and disclosed meteorological and hydrological information and based on the remarkable progress of machine learning methods. Specifically, the primary mission is to improve the accuracy of inflow prediction to develop technology that can be implemented in flood control measures that can respond to floods that are becoming more severe due to climate change.
Therefore, the prediction accuracy for "non-experienced cases" can be improved by applying "the Elastic Net", a representative method of sparse modeling, and considering the soil moisture state. In addition, by grouping the catchment area and using a representative model, it is possible to generalize the inflow prediction method that can be applied to an unspecified number of dams.

Free Research Field

水文学

Academic Significance and Societal Importance of the Research Achievements

気候変動によって頻発・激甚化する水害に対し，あらゆる関係者が防災・減災に取り組む「流域治水」の推進を念頭に，ダムの能力の最大限の活用が求められている。本研究は，蓄積公開が進む大量の気象水文情報および進歩の目覚ましい機械学習法を活用し，流入量予測の精度向上手法に取り組んだ。
この結果，未経験の洪水への対処でき，手法の一般化を図ることで，今後の気候変動で予想される大規模水害に，不特定多数のダムに適用できる予測手法を提案できたことが学術的意義となる。また，その技術が治水能力の向上を促進し，今後推進が図られる流域治水に貢献できることが社会的意義となる。