시뮬레이션 데이터 생성을 통한 기계학습 기반 원자로 노심 이상 탐지 방법론

Abstract

Anomaly Detection in the nucleareactor is a crucial technology to prevent malfunctions or unplaned shutdownsand to enhance eficient operations. Despite its importance, it remains at the level of relying on rule-baseddiagnostic or expert judgment due to the lack of training data. It is chalenging to obtain real data from the nuclearreactor because of safety and security isues. To overcome those chalenges, this paper proposes anew simulation based anomaly detection methodology in nucleareactors. We investigate asignable causes of abnormal behaviorsin the nuclear reactor, generate simulation data using nuclear core analysis code, RAST-K, and aply theclasifcation models for detecting abnormal behaviors. The proposed method is validated by the control rodpositonal anomaliesimulation data generated by RAST-K.