RESEARCH ON MODEL ORDER REDUCTION FOR PREDICTING CORE DAMAGE FREQUENCY BASED ON RISMC METHOD

抄録

Risk informed Safety Margin Characteristic Analysis (RISMC) combines the advantages of two existing safety analysis techniques, using discrete dynamic event tree (DDET) and Monte Carlo (MC) methods to quantify safety margin and risk response to optimize the design of power plant critical safety systems. The existing RISMC method needs to simulate a large number of accident scenes and power plant state, which has high time and economic cost. Due to the limited sample size of the best estimation program relying on high fidelity, the reliability of the prediction results of safety parameters such as core damage frequency (CDF) is poor. In this paper, the safety analysis of SBO accidents of CPR1000 unit is carried out by RISMC method, and support vector machine (SVM) is used to reduce the order of the original core failure frequency prediction model under typical sequence. By comparing with the simulation results of the original optimal estimation program, it is found that the accuracy of the reduced order model to predict the core damage frequency is more than 90% under the two types of SBO accident conditions. It is proved that the study in this paper can replace the best estimation program and predict the core damage of different power plant conditions faster under the typical SBO accident sequence. It can effectively improve the reliability of core failure frequency prediction and save a lot of time cost and calculation requirements.