Estimation of Axial and radial in-core power peaking in PWR plant using artificial neural network technique

Abstract

Axial and radial power peaking factors (Fq, Fah) were estimated in Chashma Nuclear Power Plant Unit-1 (C-1) core using artificial Neural Network Technique (ANNT). Position of T4 control bank, axial offsets in four quadrants and quadrant power tilt ratios were taken as input variables in neural network designing. Power Peaking Factors (PPF) were calculated using computer codes FCXS, TWODFD and 3D-NB-2P for 52 core critical conditions made during C-1 fuel cycle-7. A multilayered Perceptron (MLP) neural network was trained by applying a set of measured input parameters and calculated output data for each core state. Training average relative errors between targets and ANNT estimated peaking factors were ranged from 0.018% to 0.054%, implies that ANNT introduces negligible error during training and exactly map the values. For validation process, PPF were estimated using ANNT for 36 cases devised at the time when power distribution measurement test and in-core/ex-core detectors calibration test were performed during fuel cycle. ANNT Results were compared with C-1 peaking factors measured with in-core flux mapping system and INCOPW computer code. Results showed that ANNT estimated PPF deviated from C-1 measured values within ±3%. The results of this study indicate that ANNT is an alternate technique for PPF measurement using only ex-core detectors signals data and independent of in-core flux mapping system. It might increase the time interval between in-core flux maps to 180 Effective Full Power Days (EFPDs) and reduce usage frequency of in-core flux mapping system during fuel cycle as present in Advanced Countries Nuclear Power Plants.