A STUDY ON THE GENERALIZATION OF DAM INFLOW PREDICTION METHOD USING THE ELASTIC NET

Abstract

This study aims to propose a generalized method for predicting the inflow of dams with various catchment basin scales. In response to the large floods that have been occurring frequently throughout Japan due to climate change and are expected to continue, measures have been taken to strengthen flood control by pre-discharging not only dams with flood control functions, but also water-utilizing dams. To perform pre-discharge, it is necessary to establish a method for predicting the inflow rate. However, it is impractical to create prediction models for each of the many dams because of the extensive labor required. Therefore, it is desirable to establish a generalized prediction method that can be applied to all dams. In this study, Elastic Net, a machine learning-based model, was applied to dams in Hokkaido, and a generalization of the inflow prediction method was attempted. As a result, a generalized prediction method was proposed by performing a cluster analysis of the geographical information of dam basins and classifying them by basin area.