Prediction of radon exhalation rate based on VMD-LSTM-ELMAN

Abstract

Uranium tailings reservoir is a huge radon source. Aiming at the complexity and uncertainty of radon exhalation law, a prediction method of radon exhalation based on variational mode decomposition (VMD) is proposed. The uranium tailings reservoir model was constructed by the shrinkage method, and the natural environment simulation and radon exhalation rate monitoring were carried out for 180 days in a uranium tailings reservoir in South China. The monitoring value of radon exhalation rate is decomposed into three components with different physical meanings by VMD, and the long short-term memory neural network model (LSTM) is established to predict the trend of radon exhalation. The ELMAN neural network model is established with external environmental factors as input, and the lag effect of environmental indicators on radon exhalation rate is specially considered. The results show that the VMD-LSTM-ELMAN method can accurately reflect the precipitation law of radon exhalation and has better prediction accuracy.