機械学習

[II-207] Prediction of dam inflow using deep learing from river water level and precipitation information

〇Hiroyuki Nakano¹, Takashi Miyamoto², Natsu Miura², Masazumi Amakata¹, Takato Yasuno¹, Akira Ishii¹ (1.Yachiyo Engineering CO.Ltd, 2.Yamanashi University)

Keywords:dam inflow, deep learing, spatio-temporal data, Conv-LSTM

The prediction of dam inflow is generally based on simulations of infiltration and runoff in the basin using physical models with the actual and predicted rainfall, and more recently, deep learning has been used. Authors have constructed a river water level sensor and built a real-time information acquisition system for the purpose of increasing the amount of information for predicting dam inflows, and used this information as input data for deep learning model in the upstream area of Miyagase Dam. In this paper, we introduce our efforts to design the dam inflow prediction model by comparing the accuracy between several deep learning models.