The application of Kriging model for three-dimensional core power distribution calibration
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
The safety operation of a nuclear reactor relies on knowing the core power distributions as accurately as possible. PWR adapt the calculated power distributions to obtain accurate power distributions, which requires knowing the calibration factors for the calculated powers of each fuel bundle in the core. The calibration factors are obtained by comparing the measured reaction rates and the calculated reaction rates at the instrumented locations in the core. For the fuel bundles position that do not have measurement data, their calibration factors shall be inferred by using the calibration factor of the instrumented bundles nearby. In this study, we propose using Kriging model to fit the calibration factors of the instrumented bundles and then predict the calibration factors for the un-instrumented bundles. To verify the correctness of the model, we employed a method of repeated random sampling from the dataset to validate the model. After expanding the samples in each group, which originally contained only one plane, to include two or three planes, it was observed that the average RMSE value, as defined by cross-validation, not only significantly decreased but also showed a marked improvement in the predicted calibration factor for most of the validated samples.
Abstract
The safety operation of a nuclear reactor relies on knowing the core power distributions as accurately as possible. PWR adapt the calculated power distributions to obtain accurate power distributions, which requires knowing the calibration factors for the calculated powers of each fuel bundle in the core. The calibration factors are obtained by comparing the measured reaction rates and the calculated reaction rates at the instrumented locations in the core. For the fuel bundles position that do not have measurement data, their calibration factors shall be inferred by using the calibration factor of the instrumented bundles nearby. In this study, we propose using Kriging model to fit the calibration factors of the instrumented bundles and then predict the calibration factors for the un-instrumented bundles. To verify the correctness of the model, we employed a method of repeated random sampling from the dataset to validate the model. After expanding the samples in each group, which originally contained only one plane, to include two or three planes, it was observed that the average RMSE value, as defined by cross-validation, not only significantly decreased but also showed a marked improvement in the predicted calibration factor for most of the validated samples.