장단기 메모리를 기반으로 한 원자력 발전소의 전출력 운전과 기동 운전 시의 사고진단 알고리즘 개발

Abstract

Currently running nuclear power plants (NPPs) are managed under the goal of electricity power generation with safety. According to the safety purpose of NPPs, any action (e.g., control of components, maintenance, monitoring) taken at the plant are carried out based on procedures. Hence, in order to ensure the safety of NPP, operators must follow procedures and any other requirements described in the technical specification. However, in case of emergency operation, it requires that the appropriate accident diagnosis must be done first so that operators are possible to cope with the situation properly, nevertheless, it is the extremely strenuous task, which can give rise to the probability of operator errors. Furthermore, there are various operation modes. Concerning the issue of such diagnostic activities, a variety of operator supporting systems and diagnostic algorithms have been introduced to reduce the burden of operators. Owing to nuclear power plant data characteristics such as nonlinear, time sequential and multivariate, in this light, this thesis aims at proposing an accident diagnosis algorithm for the diagnosis of accidents considering the characteristics of NPP startup operation and full power operation based on LSTM. It is expected that the safety of NPP during startup operation can be improved by application of algorithm for diagnosis of accidents.