Data Reconstruction of Faulty Sensors for the Nuclear Power Plants Control System: A Strong Tracking Filter Approach

Abstract

As an indispensable component of the control system, the sensor usually works in high temperature, high pressure, strong vibration and other failure-prone environment. The complex, changeable, and high intensity working environment makes the sensors work unstable, which eventually leads to the sensor fault. Once the sensor fault occurs, the output signal will deviate from the normal value, which seriously decreases the accuracy, stability and reliability of the control system. To solve this problem, the researchers tried to locate the faulty sensors and recover the missing data by the expertise, which results in low accuracy and long-time consuming. In this paper, a novel data reconstruction method for padding the missing data of the faulty sensor is proposed. Based on the extensive redundancy and complementarity of the information obtained from various types of sensors, the data reconstruction of the target sensor is realized by combining the thermal system model with the strong tracking filter algorithm. Simulation data of the single pressure condenser system show that, under different sensor fault modes, the reconstruction accuracy deviation of the signal reconstructed by the algorithm is less than 5% compared with the normal signal. The application of the proposed algorithm can significantly improve the safety and reliability of the control system.