A set of transient correlations for fast and unprotected loss of flow accident in VVER-1000 reactor using single-heated channel approach and Gene Expression Programming

Abstract

Artificial intelligence methodologies along with human observations in the main control room of nuclear power plants can be applied for predictive analysis and accident detection in the early phase of accidents. This study set out with the aim of assessing the importance of accumulated information in the early phase of a fast and unprotected Loss of Flow Accident (LOFA) for exploring the behavior of significant operating parameters in the reactor. In this study, various fast and unprotected LOFA scenarios are numerically simulated using a single-hearted approach to generate a comprehensive dataset in a forward direction. The nondimensional transient dataset including inlet mass flux and outlet temperature of the core has been given to the Gene Expression Programming (GEP) algorithm for developing a set of correlations at different moments after initiating the accident. The Multi Time-Step Data (MTSD) and Single Time-Step Data (STSD) approaches have been used to extract the required correlations from the generated dataset. Mean Square Error (MSE) and coefficient of determination are used to measure the error of each approach. The most striking result to emerge from the computed results is that the MTSD approach based on trigonometric functions has exceptionally high prediction accuracy (MSE<10-5). Overall, the obtained results show that the GEP algorithm can be used as a powerful tool for generating a set of transient correlations with high accuracy to identify the progress in a fast and unprotected LOFA.