In this study, an inflow prediction model was developed for Shiobara Dam in Tochigi Prefecture, Japan, which underwent an emergency discharge due to heavy rainfall in October 2019, using machine learning from past basin rainfall and dam inflow data.
A prototype system for predicting dam inflows up to 6 hours ahead was developed and applied to the October 2019 flood. The results confirm the possibility of estimating the time when rainfall in the upstream area begins to flow into the dam and reaches 250 m3/s as early as 6 hours before. Finally, we considered how to use this system, including during normal times, for efficient dam operation.
