Regression algorithm-based prediction of burnup history in pebble-bed high-temperature reactors
The pebble-bed high-temperature reactor operates with continuous refueling, featuring a vast number of fuel spheres within the core that are challenging to locate and track, leading to an unknown burnup history of the nuclear fuel. This poses difficulties for nuclear material accounting, source term...
Saved in:
| Published in: | Nuclear engineering and design Vol. 433; p. 113907 |
|---|---|
| Main Authors: | , , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
Elsevier B.V
01.03.2025
|
| Subjects: | |
| ISSN: | 0029-5493 |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The pebble-bed high-temperature reactor operates with continuous refueling, featuring a vast number of fuel spheres within the core that are challenging to locate and track, leading to an unknown burnup history of the nuclear fuel. This poses difficulties for nuclear material accounting, source term analysis, and the establishment of experimental benchmarks in pebble-bed high-temperature reactors. Currently, the online burnup measurement system in HTR-PM can measure the activity of radioactive nuclides, which are highly correlated with the burnup history and can be utilized to construct the loss function in burnup history inversion.
To reliably provide anticipated optimal solutions, this paper conducts an in-depth study on the construction of the above loss function through regression on a large dataset for burnup history prediction. The following three aspects of work are undertaken: 1) Simulating the core pebble flow using a discrete element model, simplifying the burnup history into an irradiation history sequence that includes neutron flux rates and irradiation durations, and demonstrating the rationality of this simplification; 2) Utilizing the above irradiation history sequence as input into a nuclear inventory software to generate a substantial dataset of known burnup histories and nuclide activities. Constructing weighted regression and deep neural network models, with online measurable nuclides as input data and fuel sphere burnup history as output data, to quantify the data potential of solving the burnup history from online measurable nuclides and to evaluate the contribution of different nuclides to the solution; 3) Analyzing the contribution to the solution of online measurable nuclides from the perspectives of burnup chains and nuclide characteristics, and establishing criteria for selecting input data for the burnup history prediction model.
The research outcomes of this paper represent a crucial step in predicting the burnup history of pebble-bed high-temperature reactors, achieving a mathematical description of the high-amplitude, high-frequency variations in fuel sphere burnup history. This paper also provides the computational limits of regression models for burnup history based on online measurable nuclides and conducts a detailed analysis of the contribution of input data to the solution, breaking through the bottleneck of unknown burnup history of fuel spheres in pebble-bed high-temperature reactors. |
|---|---|
| ISSN: | 0029-5493 |
| DOI: | 10.1016/j.nucengdes.2025.113907 |