📄 中文摘要
加速器驱动的亚临界反应堆（ADS）是当前核废料流中快速转化长寿命次要锕系元素的最成熟平台。然而，质子束传输的随机特性和反应堆核心中复杂的中子输运限制了可实现的转化速率，并增加了运行成本。提出了一种基于闭环强化学习的束控制系统，该系统在持续监测不断变化的中子谱和锕系元素浓度的同时，学习最佳的束电流-时间调度。控制器参数化束能量、电流和脉冲模式，动态调整这些参数以优化转化过程。
加速器驱动转化反应堆中的 AI 优化质子束控制
出处: **Title**
发布: 2026年2月18日
📄 English Summary
**Title**
Accelerator-driven subcritical reactors (ADS) represent the most advanced platform for the rapid transmutation of long-lived minor actinides present in current nuclear waste streams. However, the stochastic nature of proton beam delivery and the intricate neutron transport within the reactor core impose limitations on the achievable transmutation rates and elevate operational costs. A closed-loop, reinforcement-learning-based beam control system is proposed, which learns optimal beam current-time schedules while continuously monitoring the evolving neutron spectrum and actinide concentrations via high-resolution in-core detectors. The controller parametrizes beam energy, current, and pulsing patterns, dynamically adjusting these parameters to enhance the transmutation process.