DEVELOPMENT OF AN AI-BASED MODEL FOR ESTIMATING SNOWFALL ON THE HOKURIKU SHINKANSEN

Abstract

In the winter season, we carry out snow removal work to prevent snow from falling on the railway cars. The necessity of the snow removal operation is determined based on the snowfall forecast in advance and the experience of the employees, but there were some cases where there was less snow than expected when the snow removal operation was conducted. On the other hand, it is clear that there is a correlation between weather information and snow accretion, so we examined whether it would be possible to create a highly accurate snow accretion estimation model using AI based on weather information and train running data. In order to create the estimation model, we held an open innovation competition, and considered the intro-duction of a support tool to determine whether or not snow removal operations should be activated by combining the top models in the competition. In this paper, we introduce our efforts from the conception of the solution to the practical implementation.