An efficient LSTM network for predicting the tailing and multi-peaked breakthrough curves

Abstract

As typical characteristics of solute transport in the complex flow field, the tailing and the multi-peaked phenomena of the breakthrough curve (BTC) have been widely studied. Due to a large number of model parameters and the complex structure, traditional numerical models are difficult to carry out in practical applications. In this study, we proposed a prediction method of BTC time series based on long short-term memory (LSTM) network. The method was validated by the field tracer experiment data conducted in a subterranean river. The model performance using multivariable and univariate time series as input was compared, and the prediction of the tailing and multi-peaked phenomena of BTC was realized. The results showed that the proposed method can effectively and accurately predict and reproduce different tailing and multi-peaked BTC. In addition, the prediction in different seasons also showed that LSTM could well describe the tracer data in the natural flow with various discharges.