DAM INFLOW PREDICTION USING DEEP NEURAL NETWORK AND COMPARISON WITH RUNOFF MODELS

Abstract

 DX in the infrastructure section has been gotten attention by recent understaffing problem in the field of civil engineering and the changes of social situation. AI has been widely used in facilities inspection and flood disaster prevention. Although the utilization case of AI increases in the infrastructure section, the method which uses it in the flood and dam inflow prediction has not still been established. Therefore, the purpose of this study is to develop the dam inflow prediction model by using Deep Neural Network (DNN) for the Shiokawa Dam where was constructed in the Shiokawa River of the Fuji River system in Yamanashi prefecture. Furthermore, in order to evaluate the reproducibility of the model, we compared the result of DNN with storage function model and Rainfall-Runoff-Inundation model (RRI) by some evaluation indexes. As a result, it is found that DNN model for the Shiokawa Dam can well reproduce and estimate stably the flood discharge relative to the other runoff models. Thus, it is suggested that DNN model is one of available dam inflow prediction method.