Session: 02-13: Structural Evaluation, Performance Assessment, Multiphysics Coupling - III

Paper Number: 136183

136183 - Method Research on the Model Optimization of Whole-Core Fuel-Assembly Bowing Based on 3d Variational Algorithm and the Measurement Values 

Abstract: 

During the operation of pressurized water reactor (PWR), the fuel assemblies would inevitably occur bowing phenomenon due to the axial radiation growth, lateral flow of coolant, loaded forces and so on. And this phenomenon is intuitively manifested as the power quadrant tilt. In order to quantify the impact of fuel-assembly bowing on the core key parameters and the safe operation, it is necessary to accurately establish model and perform simulation for the fuel-assembly bowing. However, the fuel-assembly bowing cannot be measured during power-operation period. Therefore, the challenge lies in obtaining the actual fuel-assembly bowing, which is a prerequisite for ensuring simulation accuracy.

In this paper, a newly method for the model optimization of whole-core fuel–assembly bowing was proposed. The measurement values of power distribution and theoretical whole-core model are combined in this method based on the 3D variational (3DVAR) and artificial neutral network (ANN) algorithm. In 3DVAR, the cost function is constructed to characterize the difference between the actual fuel–assembly bowing and theoretical model. Furthermore, the observations were taken as the measured power-distributions. The error covariance matrix of the model was approximated by the second-order autoregressive (SOAR) function. The error covariance matrix of observations was approximated by a diagonal matrix, and the diagonal element was taken a percentage of the corresponding measured assembly power. The observation operator was established by the ANN algorithm, which characterizes the complex relation between the whole-core fuel-assembly bowing and corresponding power distribution. Finally, the minimum value of the cost function would be solved by the Adam algorithm and the optimized model of fuel-assembly bowing would be obtained.

For method verifications, the power-distribution measurements of a commercial NPP in China has been applied, in which the fuel-assembly bowing was measured at the end of cycle (EOC) in the pool. The optimized model was taken as the bowing amplitude of each fuel-assembly, assuming that the bowing shapes of fuel assemblies were consistent between the pool and the core. The performance of ANN is verified with power-distribution measurements, and the maximum value of relative errors for predicted power distribution is about 0.7%, which indicates the observation operator obtained by the ANN has satisfactory accuracy. Applying the optimized model of fuel-assembly bowing to the operation history simulation, the relative errors of the power distribution in entire cycle could be reduced notably to meet the industrial criteria (10.0%), and the maximum value is reduced from about 13.6% to 7.7%. As conclude, the proposed method based on 3DVAR and the power-distribution measurements is efficient to optimize the model of fuel-assembly bowing, and provides an approach to optimize the core model combined with measurement values.

Presenting Author: Lin Guo Xi'an Jiaotong University

Presenting Author Biography: Name: Lin Guo(1997-)
Gender: male
Major: reactor physics
Email: gl1125@stu.xjtu.edu.cn

Authors:

Lin Guo Xi'an Jiaotong University
Kai Zhang Xi'an Jiaotong University
Chenghui Wan Xi'an jiaotong University
Hongchun Wu Xi'an Jiaotong University