Author(s): Shoma Wakasaya; Makoto Nakatsugawa; Yosuke Kobayashi

Linked Author(s): Makoto Nakatsugawa, Shoma Wakasaya

Keywords: Dam reservoir level prediction; Sparse modeling method; Elastic net; Ensemble rain; MEPS

Abstract: This study aims to propose a method for predicting dam inflows, reservoir levels, and releases that considers the uncertainty of predicted rainfall. In recent years, it has become increasingly important to predict reservoir levels, as it is utilized for effective dam operation, in response to the frequent large floods that have been occurring throughout Japan. In this study, a regression model for predicting dam inflow was developed using Elastic Net, a sparse modeling technique, for the Satsunaigawa Dam in Hokkaido. Rainfall by the error variance of the predicted rainfall and a meso-scale ensemble prediction system (MEPS), ensemble prediction system of 21 ensemble predicted rainfalls provided by the Japan Meteorological Agency from 2019, were used as input variables of the model, and the reservoir level and discharge of the dam were predicted based on the predicted inflow. The results showed that the use of MEPS produced predictions that encompassed the actual measurements. Furthermore, a comparison of the inflows using Elastic Net and the storage function method was conducted, and the results showed that the former can attain the same or better accuracy than the latter in the prediction using MEPS.

DOI: https://doi.org/10.3850/978-90-833476-1-5_iahr40wc-p0237-cd

Year: 2023