The primary aim of this work is to utilize a Kalman filter (KF) to predict reaching the trip set-point for a trip parameter in a nuclear power plant (NPP). To address uncertainty in the predicted measurements, prediction bounds are calculated by propagating the state error covariance. These predicted bounds enable the calculation of levels of confidence in making trip decisions. Further, to address uncertainty in the estimation model, the observed prediction error is used to offset the predicted measurements.
The predictive trip detection routines are evaluated through simulations of a single NPP sub-system. More specifically, the water level parameter in a steam generator of a NPP is considered. The model of this sub-system is represented by the Irving linear parameter varying (LPV) model with inverse response characteristics. The simulations include a level low postulated initiating event (PIE) made to occur during two different common power transients for various estimation models. The results of this paper are a proof of concept for KF-based predictive trip detection which is demonstrated to achieve reduced time-to-trip when applied to a single sub-system.