JIITA, vol.9 no.2, p.1075-1080, DOI: 10.22664/ISITA.2025.9.2.1075
Jung-seok Kwon, Tong-kyu Park, Sung-kyun Zee
Abstract. This study presents a data-driven neural network surrogate model for predicting assembly-wise power distribution changes in the core of an i-SMR. The neural network was trained using a dataset generated by ASTRA, a nodal diffusion code.
A convolutional neural network (CNN)-based architecture was designed to predict power distributions at target load, based on current power distributions, ramp rate, depletion time, current load and target load value. The model achieved a mean relative error of 0.833% and a peak power prediction error of 0.768%, demonstrating its ability to effectively predicting power distributions under varying load.
Keywords; Neural network; power distribution; i-SMR; surrogate model; load variation
Fullpaper: