PREDICTION OF DAM INFLOWS DURING SNOWMELT SEASON USING DEEP LEARNING

Abstract

 Accurate prediction of dam inflow during snowmelt is extremely important for disaster prevention and water use. Currently, AI is being utilized in the hydrological field, and research is being conducted on predicting river water levels and dam inflows. In this study, we used deep learning to predict dam inflows during the snowmelt season on an hourly basis. The results showed that reproducibility was high up to 24 hours ahead, but decreased after 24 hours. Therefore, it is considered that the practical limit of the forecast is 24 hours ahead. In addition, the effect of the number of intermediate layers was more significant than the effect of the normalization and standardization processes.