Performance evaluation of neural network topologies for online state estimation and fault detection in pressurized water reactor

Abstract

The nuclear reactor is a multi-rate nonlinear system in which the state variables progress with widely varying dynamics. It has state variables such as reactivity and delayed neutron precursor densities that cannot be measured directly via sensors. Reactivity signifies the criticality of the reactor core. Delayed neutron precursors are the source for delayed neutrons which plays a vital role in the change of neutron densities. Besides, the other states which are measured are also corrupted by measurement noise and are susceptible to sensor faults. Thus, estimation of these state variables becomes critical. As traditional estimators like EKF, UKF, and Particle filers require a close model of the process, Data-driven Neural networks architectures like Feed-Forward Neural networks, Dynamic NARX Neural networks, and Recurrent Neural networks are designed to estimate the reactor core states. The performance of the selected neural estimator is also compared with the Unscented Kalman estimator.