STUDY ON ENHANCEMENT OF DIVERSION WEIR OPERATION BY DEEP LEARNING －TOWARDS ENHANCEMENT OF MANEGEMENT IN HOJO-RIVER－

Abstract

 The Hojo River, which flows through central Tottori Prefecture, frequently causes inundation damage due to increased flow during rainfall. In order to prevent flooding due to rising water levels, river administrators have been operating the diversion weirs installed at the diversion points of the Hojo River drainage canal to reduce the risk of flooding in the backwater sections and the middle and upper reaches of the river. However, continuous day and night monitoring is required for the decision making of weir over-turning, and it is difficult to allocate monitoring personnel and to understand the occurrence of flooding.

 In this study, we developed a water level prediction model to support the decision-making process for the operation of a diversion weir using deep learning, a type of artificial intelligence (AI) technology with advanced time series processing capabilities, and confirmed its usefulness. The method developed in this study can be applied to river management facilities such as weirs, sluices, gates, and flumes, where CCTV cameras are installed to measure the water level.