Determination of alarm templates for decision support in nuclear power plants alarm floods using evolutionary computation

Alarm floods are a recurring problem of industrial plants and a particular nuisance of nuclear power plants (NPPs), where it is possible, for example, to activate 1000 alarms in a period of 10 min, with the majority of alarms being activated in the very first minutes. In addition, the alarms are presented as sequences of events (SOE), making it almost impossible for the operator to understand the situation in a timely manner, often letting alarms of extreme importance pass unnoticed by the operator. In this work, the Quantum Evolutionary Algorithm (QEA) was adapted for the creation of a decision support system in cases of alarm floods, where the objective is the determination of the minimum alarms set capable of differentiating a failure from all others, so as to enable the operator of a control room to take mitigating measures for the diagnosed event. The algorithm proposed was tested with real data from Angra 1 and Angra 2 Brazilian NPPs and results show that the QEA was able to efficiently perform this task, with the important advantage of converging at a much faster rate when compared to a classical genetic algorithm applied to the same problem. It should be noted that the application of the methodology developed is not restricted to NPPs, and may be applied to any industrial plant that has an alarm system.