Physics-driven surrogate model for rapid ASI prediction during xenon transients
Physics-driven surrogate model for rapid ASI prediction during xenon transients
Yue-Yao Zhang 1Qian Zhang 1Shuai Qin1
作者信息
- 1. School of Physics, Zhejiang University, Hangzhou 310058, China
- 折叠
摘要
Core power capability analysis during xenon-oscillation transients is computationally demanding because a large number of burnup states, initial power levels, power-change magnitudes, and transient histories must be evaluated. This study develops a physics-driven surrogate framework for rapid prediction of the axial shape index (ASI). The framework couples a reduced-order xenon–iodine model, derived from one-dimensional, two-group neutron diffusion theory, with state-dependent parameter calibration. Seven physical parameters are identified from a sparse set of reference simulations generated by a nuclear design code. A two-stage optimization procedure combines the global-search capability of Differential Evolution with constrained local refinement by Sequential Least Squares Programming. The optimized parameters exhibit two convergence patterns: five relatively insensitive parameters approach common values across the calibration cases, whereas the xenon concentration and power-feedback coefficient remain dependent on the power-change magnitude. Regression and curve-transformation relationships are then constructed for these sensitive quantities, enabling interpolation-based prediction for operating conditions not included in calibration. The model is evaluated for different initial power levels, transient lengths, and burnup states. At low burnup (50 MWd/tU), it reproduces and predicts ASI trends accurately for large power reductions (ΔP ≥ 60%), although accuracy decreases for smaller perturbations and longer transient windows. At higher burnup (5000 MWd/tU), better performance is obtained for smaller power changes (ΔP ≤ 30%), demonstrating that the applicable power-change interval depends on the core state. The proposed framework reduces repeated high-fidelity calculations and provides an efficient method for preliminary core power capability and safety-margin evaluation. Future work should incorporate control-rod effects, burnup-dependent physics, and higher-fidelity reduced-order kernels.
Abstract
Core power capability analysis during xenon-oscillation transients is computationally demanding because a large number of burnup states, initial power levels, power-change magnitudes, and transient histories must be evaluated. This study develops a physics-driven surrogate framework for rapid prediction of the axial shape index (ASI). The framework couples a reduced-order xenoniodine model, derived from one-dimensional, two-group neutron diffusion theory, with state-dependent parameter calibration. Seven physical parameters are identified from a sparse set of reference simulations generated by a nuclear design code. A two-stage optimization procedure combines the global-search capability of Differential Evolution with constrained local refinement by Sequential Least Squares Programming. The optimized parameters exhibit two convergence patterns: five relatively insensitive parameters approach common values across the calibration cases, whereas the xenon concentration and power-feedback coefficient remain dependent on the power-change magnitude. Regression and curve-transformation relationships are then constructed for these sensitive quantities, enabling interpolation-based prediction for operating conditions not included in calibration. The model is evaluated for different initial power levels, transient lengths, and burnup states. At low burnup (50 MWd/tU), it reproduces and predicts ASI trends accurately for large power reductions (P 60%), although accuracy decreases for smaller perturbations and longer transient windows. At higher burnup (5000 MWd/tU), better performance is obtained for smaller power changes (P 30%), demonstrating that the applicable power-change interval depends on the core state. The proposed framework reduces repeated high-fidelity calculations and provides an efficient method for preliminary core power capability and safety-margin evaluation. Future work should incorporate control-rod effects, burnup-dependent physics, and higher-fidelity reduced-order kernels.关键词
Xenon oscillation/Surrogate model/Differential evolution algorithm/SLSQP method/Power capacity analysisKey words
Xenon oscillation/Surrogate model/Differential evolution algorithm/SLSQP method/Power capacity analysis引用本文复制引用
Yue-Yao Zhang,Qian Zhang,Shuai Qin.Physics-driven surrogate model for rapid ASI prediction during xenon transients[EB/OL].(2026-11-08)[2026-11-11].https://chinaxiv.org/abs/202611.00065.学科分类
核反应堆工程
