An Adaptive Optimization Regulation Method for Neutron Spectrum Based on Reinforcement Learning
Lists
Only you see this
Abstract
The regulation and control of neutron transport behavior is crucial to the design and development of nuclear energy systems and the application of nuclear technology. The accurate and efficient regulation of neutrons to satisfy the specific needs of neutron experimental environments is an urgent challenge to be solved. The traditional forward neutron spectrum regulation method based on artificial designing could hardly meet the demands of high accuracy and high efficiency at the same time. In this research, an adaptive optimization regulation method for neutron spectrum based on reinforcement learning was developed. The proposed method designed different types of standardized response modules and calibrated it with their corresponding multi-dimensional response matrix calculated by the Monte Carlo program. Through the proximal policy optimization (PPO) algorithm outputs the selection and layout of the response module, and generates the optimal scheme that conforms to the target neutron spectrum by the process of continuous interaction with the environment and self-game. The 14MeV single-energy neutron source was used to verify the regulated results of the proposed method on the tritium breeder blanket of a fusion reactor (ITER) whose results showed that the accuracy of the proposed method was improved by 39.31% compared to the forward manual experiment.
BibTeX
@article{cao2025adaptive,
author = {Chenglong Cao and Pei Cao and Xin Xu and Hongdou Yao},
title = {{An Adaptive Optimization Regulation Method for Neutron Spectrum Based on Reinforcement Learning}},
journal = {IEEE Transactions on Nuclear Science},
year = {2025},
doi = {10.1109/tns.2025.3634625},
}Loading related content...