[2G14-21] AI Technology

[2G21] Integration of AI Technology and Thermal Hydraulics for the Development of a Data-Driven Methodology for Plant Safety Assessment

(6)Parameter analysis for LWR accident by deep neural networks using CNN and Transformer

*Kenji Aragane¹, Takumi Kitao¹, Shinsuke Tanaka¹, Masaaki Mori¹, Shuichiro Miwa² (1. NEL, 2. UTokyo)

Keywords:multivariate time series analysis, LWR, accident analysis, deep learning

Detailed plant behavior analysis based on accident information at the time of an accident at a nuclear power plant takes time. We conducted an analytical evaluation with the aim of quickly assessing the impact on accident progress caused by changes in the timing of operations by operators using AI. In order to deal with time-series data in which plant parameters fluctuate in a complex manner due to operator operations, we created models that incorporate structures in recent deep learning, such as CNNs and transformers, which handle complex data. Experimental results showed that these models enable fast learning of highly accurate estimation.