Diagnostic & Prognostic System for Identification of Accident Scenarios and Prediction of “Source Term” in Nuclear Power Plants under Accident Conditions

Nuclear power plant experiences a number of transients during its operations. These transients may be due to equipment failure, malfunctioning of process support systems etc. In such a situation, the plant may result in an abnormal state which is undesired. In case of such an undesired plant condition, the operator has to carry out diagnostic and corrective actions. When an event occurs starting from the steady state operation, instruments’ readings develop a time dependent pattern and these patterns are unique with respect to the type of the particular event. Therefore, by properly selecting the plant process parameters, the transients can be distinguished. In this connection, a computer based tool known as Diagnostic and Prognostic System has been developed for identification of large pipe break scenarios in 220 MWe Pressurised Heavy Water Reactors (PHWRs) and for prediction of expected “Source Term” and consequence for a situation where Emergency Core Cooling System (ECCS) is not available or partially available. Diagnostic and Prognostic System is essentially a transient identification and expected source term forecasting system. The system is based on Artificial Neural Networks (ANNs) that continuously monitors the plant conditions and identifies a Loss Of Coolant Accident (LOCA) scenario quickly based on the reactor process parameter values. The system further identifies the availability of injection of ECCS and in case non-availability of ECCS, it can forecast expected “Source Term”. The system is a support to plant operators as well as for emergency preparedness. The ANN is trained with a process parameter database pertaining to accident conditions and tested against blind exercises. In order to see the feasibility of implementing in the plant for real-time diagnosis, this system has been set up on a high speed computing facility and has been demonstrated successfully for LOCA scenarios.