60 / 2021-11-15 20:38:16

An EMD-IPSO-LSTM hybrid model for landslide displacement prediction

landslide displacement prediction,Data fusion,empirical mode decomposition(EMD),improved particle swarm optimization,Long short-term memory

Since the impoundment of the Three Gorges Reservoir area, many existing historical landslides have been reactivated, which poses a threat to human life safety in the reservoir area. An accurate and effective landslide displacement prediction method is needed to reduce the disaster caused by landslides. At present, the traditional machine learning method commonly used to predict landslide displacement cannot accurately predict landslide displacement. Therefore, to improve the prediction accuracy of landslide displacement, an EMD-IPSO-LSTM prediction model based on hybrid algorithm is proposed in this study. Firstly, an adaptive spatio-temporal analysis method called Empirical Mode Decomposition (EMD) is introduced to deal with non-stationary nonlinear sequences. The nonlinear landslide field monitoring data are decomposed to extract the relevant characteristics of landslide displacement in the reservoir area. Secondly, the improved particle swarm optimization (IPSO) combined with Long short-term memory (LSTM) neural network is used to establish the prediction model. Finally, this prediction model is applied to Jiuxianping landslide in the Three Gorges Reservoir area of China. The calculation results show that EMD can effectively extract the characteristics of nonlinear landslide data. In addition, compared with the traditional single algorithm landslide prediction model, the EMD-IPSO-LSTM prediction model based on the hybrid algorithm has better prediction accuracy, which can provide important reference for the landslide displacement early warning and risk assessment in the Three Gorges Reservoir area.