Development of Long-Term Prediction Algorithm Based on Component States Using BiLSTM and Attention Mechanism

Appropriate situation awareness (SA) of operators is important in managing nuclear power plants (NPPs), especially, in accident situations. Among the three levels of SA suggested by Ensley, Level 3 SA is one of the more difficult tasks due to the complexity of NPP as well as the uncertainty of accidents phenomena. Several prediction models using artificial intelligence techniques have been suggested to help the operator’s prediction in the case of accidents. However, those models are capable of predicting the short-term plant status and do not quantify the uncertainty of the prediction. This study suggests an algorithm for predicting the long-term status of plant parameters for 2 hours according to component states and quantifying the uncertainty of the prediction.