Real time nuclear power plant operating state cognitive algorithm development using dynamic Bayesian network

Abstract

In a reactor, various reactor instruments inform the operator of changes in the operating conditions. However, until now, the identification of the reactor state from interpretations of the instrumentation reading has been mostly left to the human operators, which sometimes can be flawed depending on circumstances. Therefore, an artificial cognitive system that can recognize the nuclear reactor state is suggested in this paper to aid the operator under severe environmental stress. Two reactor operating states are preliminarily considered: Normal state and loss of coolant accident state. The artificial cognitive system predicts the state of the reactor in real time, and it determines the type of LOCA if the reactor state is determined to be in an accident state. The proposed system uses only reactor protection system monitoring parameters among various available measurements. The proposed system is composed of two dynamic Bayesian models: Reactor State Determination (RSD) Model, and Accident Type Categorization (ATC) Model. When an accident occurs, the RSD model took 0.3 s to recognize it with 100% accuracy. The average accuracy of the ATC model is about 88%. The average accuracy increased more when the model was refined which suggests the model can be further improved in the future.