Development of a Method of Selecting Important Alarms for Nuclear Power Plants

Abstract

Presenting important alarms selected from a large number of activated alarms provides useful operational support under a transient status in a nuclear power plant. We have developed an alarm processing method which selects and presents important alarms depending on plant status. In this method, important alarms are selected, first, based on physical relationships between alarms and component status including alarms themselves and second, even more important alarms are selected from the previously selected alarms according to the identified initial event causing the transient. Identification of the initial event is implemented by a neural network. The identified initial event and selected important alarms are presented to show the cause and influence of the transient. A prototype based on the above alarm processing method was validated during the start-up test at Kashiwazaki Kariwa Nuclear Power Plant Unit Number 4 of Tokyo Electric Power Co. The initial events, which were load rejection, turbine trip and main steam isolation valve closure, were correctly identified and about 30% of all activated alarms were selected as important. It was verified by an operating expert that the presentation of the identified initial event and the selected important alarms were effective to understand rapidly and correctly the transient status of the plant.