A combination of Convolutional Neural Network (CNN), Long-Short Term Memory (LSTM), and Convolutional LSTM (ConvLSTM) is constructed in this work for the fault diagnosis and post-accident prediction for Loss of Coolant Accidents (LOCAs) in Nuclear Power Plants (NPPs). The advantages of ConvLSTM, such as effective feature determination and extraction, are applied to the classification of LOCA cases. The prediction accuracy is enhanced via the collaborative work of CNN and LSTM. Such a hybrid model is proved to be functional, accurate, and adaptive, offering quick accident judgment and a reliable decision basis for the emergency response purpose. It then allows NPPs to have an Artificial Intelligence (AI)-based solution for fault diagnosis and post-accident prediction.
A deep learning-based multi-node framework is constructed in this work to provide a data-driven platform that provides predictions for the operation condition of the primary heat transfer (PHT) loop in nuclear power plants (NPPs). Several deep learning models that have been verified and demonstrated in previous researches, such as Long-Short Term Memory (LSTM), Convolutional Neural Network (CNN), and zigmoid-based LSTM (zLSTM), are applied to modeling critical system parameters at three important nodes in the PHT loop. The feature extraction and process memory are enhanced via the collaborative work of CNN and LSTM. zLSTM, on the other hand, is successfully utilized to strengthen the long-term memory, especially for predictions of a node with multivariate inputs such as the steam generator. The node prediction results are also adopted for a polynomial fitting that generates an additional input to the next node, allowing each node to select a more accurate input. According to the verification experiments based on Loss of Coolant Accident (LOCA), the Mean Squared Error (MSE) result (1.29 × 10−3) and the Mean Absolute Error (MAE) result (1.37 × 10−2) of 0.7 cm2 LOCA case demonstrate the functionality and accuracy of the proposed framework. It is found that the fitting error (MSE) in the outlet node at 0.7 cm2 case is 38.5% lower than the prediction, showing the advantage of applying both deep learning and fitting methods. The best performance, in term of MSE, is obtained at SG node in the 0.7 cm2 case, where its processing error (0.001285) is 93.2% lower than that of the baseline models. Both the validation and verification experiments successfully proved the feasibility and advantages of the proposed framework, which offers an alternative option for the operation analysis of PHT performance.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.