Neural Network Algorithm-based Reactor Core Damage Assessment and Quick Source Term Calculation

• Mingliang Xie,
• Lei Yu,
• Yuqing Chen,
• Wei Wei,
• Guoyang Ma,
• Xiong Huang,
• Xuan Wang,

Vol. 19, No. 2, February 28, 2025
10.3837/tiis.2025.02.003, Download Paper (Free):

• Neural Network
• reactor core damage
• source term calculation

Abstract

In light of the severe accidents of some specific models of units in nuclear power plants (NPPs), an accident simulation model was established based on the data of nuclear power units; and numerous accident sequences were calculated by adopting the analysis procedures of severe accidents, there-by providing the data sources for training; meanwhile, combined with the emergency monitoring data in NPPs, a deep neural network algorithm was used to assess the reactor core damage under accident conditions and quickly estimate the accident source terms. Research on the CP300 reactor in QinShanⅠNuclear Power Plant was carried out, during which, 80% of the sample data were ran-domly selected as a training set (the other 20% as a test set), and some cases of the three types of initial accidents induced by loss of coolant accidents (LOCA), main steam line break (MSLB) and steam generator tube rupture (SGTR) were analyzed. The research results show that the reactor core damage fraction and the releasing trends of such major radioactive source terms as 133Xe, 131I, 134Cs, and 137Cs are consistent with the expected results, which proves that the neural network training model is reasonable and the methods of reactor core damage assessment and source term calculation are feasible.

Statistics

Cite this article