Exploration on the comprehensive data reconciliation framework for unknown parameter inference in the nuclear power plant system

Abstract

Nuclear power plants have been considered as a promising future energy source. As for maintenance of the nuclear power plants, it is crucial to acquire complete and accurate measurement, and even make precise inference for the unknown parameters all around the system. In this paper, a new comprehensive data reconciliation framework based on a 310 MW schematic nuclear power plant system is proposed, which is capable for calculation, uncertainty reduction and gross error detection. A new method combined with heat balance calculations is developed based on the classical method, which can make more accurate inferences for the unknown parameters. An interative algorithm is also applied to effectively eliminate the residual of constraints. The proposed framework is evaluated over the simulated datasets, and it proves that the data reconciliation process can effectively capture and reduce the errors in the measurements, as all the sensors with gross errors are detected and have their errors reduced by at least 66.7 %. The results suggest the great potential for this framework to be applied for measurement correction and error diagnostics in those practical nuclear power systems.