Research on Nuclear Power Plant Operational Task Complexity Assessment Method Based on Group Decision Making and Deep Learning of EEG Signals

Abstract

This study aims to explore the correlation between the operational task complexity of nuclear power plant (NPP) operators and electroencephalogram (EEG) features. Initially, we segmented EEG signals according to operational steps and extracted a total of 120 time domain, frequency domain, and time-frequency domain features. Subsequently, we applied an adaptive principal component analysis (PCA) dimensionality reduction method to process the features. On the other hand, three experts were invited to evaluate the complexity of the operational tasks, and their evaluation data were synthesized using a group decision-making approach. A correlation analysis was performed between these data and the PCA-reduced feature data, identifying the features with the highest correlation coefficient for each participant. Then we built a long short-term memory model with the data of the first group of participants to predict the task complexity value and tested it with the data of the second group of participants. Testing the model with data from the second group yielded favorable results, with a training set mean squared error (MSE) of 0.025 and a testing set MSE of 0.078. The results of this study indicate a significant correlation between specific EEG features and task complexity in the operational tasks of NPP operators. The model established through a combination of group decision making and machine learning methods effectively predicted the task complexity levels for operators in different operational tasks. This research provides a new perspective on NPP operators’ cognitive load and operational tasks, holding practical significance for operator training and workload management.