A Deep Learning Method to Predict Loss of Coolant Accident in Nuclear Power Plants

Abstract

Nuclear safety is crucial for the reliable operation of nuclear power plants in order to produce electricity. It is very difficult for operators of nuclear power plants to foresee and recognize one of the most catastrophic accident scenarios, often known as LOCA, as they have to pay attention to the non-linearity as well as the complex system factors involved. Artificial intelligence-based applications, such as deep learning, are put to use in this research in order to predict the time series evolution of a simulation-based nuclear power plant incident that involves a loss of coolant accident. A neural network known as NARX was employed in this research. In order to provide an accurate LOCA prediction, the NARX model was employed with a wide range of system parameters, all of which were aggregated into a twelve-dimensional input vector. Mean Squared Error was used to evaluate the model efficacy. Furthermore, it has also been observed that NARX performs much better than other leading-edge research.