Prediction of Landslide Displacement Based on GreyRelational Analysis and VMD-SES-BP Model

Aiming at the “stepped” landslide displacement in the Three Gorges area,this paper proposes a new landslide displacement time series prediction model,namely VMD-SES-BP prediction model by combining variational mode decomposition (VMD),second exponential smoothing (SES),and BP neural network (BPNN);conducts the VMD of the GPS monitoring displacement data of landslide at Baishuihe River of the Three Gorges through this model to obtain the trend component and other sub-sequence components;makes rolling predictions of trend components by the SES,determines the influencing factors of other displacement components of the landslide through gray relational analysis (GRA),and learns and predicts by considering it as the training sample of BPNN.Comparing the prediction results of each component with the true value,the average relative error of prediction is 0.78%,the mean square error is 3.14 cm,and the correlation coefficient is 0.986.The experimental results show that the model is well applicable to the prediction of “stepped” landslide displacement,with high prediction accuracy,which provides a certain reference value for landslide displacement prediction.