EVENT DOSSIER
The researchers proposed an efficient parametric state estimation method based on a shallow circulation decoder network, which is applied to the fourth-generation molten salt fast reactor (MSFR). This method requires only three neutron flux time series data from non-core regions to infer the complete reactor state vector, including neutron flux, precursor concentration, temperature, pressure, and velocity. The study extended the standard framework to handle parametric time series data and verified its accurate estimation capabilities under various operating conditions. Thanks to its extremely low training cost, this method can also quantify the uncertainty of state estimation, and its ability to real-time reconstruct various characteristic fields makes it suitable for monitoring and control in the reactor digital twin framework.
本研究利用浅层循环解码器架构，仅凭三个非核区时间序列中子通量测量数据，即可推断包含中子通量、先驱物浓度、温度、压力及速度在内的反应堆完整状态向量。该工作以第四代熔盐快堆（MSFR）为测试案例，扩展了标准架构以处理参数化时间序列数据，支持不同事故场景的探究并验证了其在多种工况下的准确状态估计能力。得益于极低的训练成本，该方法能够量化状态估计的不确定性，其实时重构各特征场的能力使其适用于反应堆数字孪生框架下的监控与控制。