PSAM 16 - Paper Overview

PSAM 16 Conference Paper Overview

Welcome to the PSAM 16 Conference paper and speaker overview page.

Lead Author: Sung-yeop Kim Co-author(s): Yun Young Choi (choi930121@nims.re.kr) Soo-Yong Park (sypark@kaeri.re.kr)

Application of Deep Learning Models to Estimate Source Release of NPP Accidents
In the event of nuclear power plant (NPP) accident, estimation of source release should be performed quickly and accurately in order to support the decision of public protection. In case of Fukushima Dai-ichi NPP accident, even though System for Prediction of Environmental Emergency Dose Information (SPEEDI) has been developed and prepared, it was not used to support the decision making of public protection due to the lack of source term information which should be provided from the system. In order to overcome the limitation of existing methods in aspect of quick and accurate source term estimation, deep learning approach using various NPP safety parameters as the learning input and releases of radioactive materials as the learning output is applied in this study. It was tried to search and apply variety of deep learning models such as ANN with pre-assigned function, encoder model of Transformer followed by fully connected layer, and multi-stage Transformer, in order to find and develop an optimized deep learning model to estimate the source release of NPP accidents.

Application of Deep Learning Models to Estimate Source Release of NPP Accidents

In the event of nuclear power plant (NPP) accident, estimation of source release should be performed quickly and accurately in order to support the decision of public protection. In case of Fukushima Dai-ichi NPP accident, even though System for Prediction of Environmental Emergency Dose Information (SPEEDI) has been developed and prepared, it was not used to support the decision making of public protection due to the lack of source term information which should be provided from the system. In order to overcome the limitation of existing methods in aspect of quick and accurate source term estimation, deep learning approach using various NPP safety parameters as the learning input and releases of radioactive materials as the learning output is applied in this study. It was tried to search and apply variety of deep learning models such as ANN with pre-assigned function, encoder model of Transformer followed by fully connected layer, and multi-stage Transformer, in order to find and develop an optimized deep learning model to estimate the source release of NPP accidents.

Paper SU171 Preview

Author and Presentation Info

Lead Author Name: Sung-yeop Kim (sungyeop@kaeri.re.kr)

Bio: PhD, Nuclear and Quantum Engineering, Korea Advanced Institute of Science and Technology (KAIST) Research Interest - Offsite consequence analysis - Multi-unit Level 3 PSA

Country: South Korea
Company: Korea Atomic Energy Research Institute
Job Title: Senior Researcher

PSAM 16 Conference Paper Overview

Welcome to the PSAM 16 Conference paper and speaker overview page.

Paper SU171 Preview

Author and Presentation Info

Download paper SU171.