Renmin Zhujiang (Jan 2022)
Prediction of Landslide Displacement Based on EMD-TAR Combined Model
Abstract
This study aims to more accurately predict the displacement changes of landslides with nonlinear volatility development.The empirical mode decomposition is first employed to process the time series of monitoring surface displacement of a landslide,and then the irregularly changing displacement series is converted into modal components with regular changes,which generates displacement components at different frequencies.Each component is predicted separately so that the mutual influence of errors can be avoided.The comprehensive prediction of the changing trend of displacement series is based on the prediction of the changing trends of all components.The improved threshold autoregressive model able to well describe non-stationary harmonics is used to predict the landslide displacement components.Finally,the modal superposition yields the final predicted displacement.In this way,a combined prediction model based on empirical mode decomposition and threshold autoregressive model is established,and its prediction accuracy is verified with Baishuihe landslide data.Compared with a BP neural network model and a long short-term memory network model,the proposed model has a high prediction accuracy,which provides a new method for the prediction of landslide displacement.
